Citation
Insult, injury and impact

Material Information

Title:
Insult, injury and impact social movement impact on public policy in the context of recognition and redistribution
Creator:
Reeder, Brett J
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English
Physical Description:
1 electronic file. : ;

Subjects

Subjects / Keywords:
Social justice ( lcsh )
Social movements ( lcsh )
Social justice ( fast )
Social movements ( fast )
Genre:
non-fiction ( marcgt )

Notes

Review:
This study is motivated by a simple question that has a complex set of answers: Do social movements impact social justice? To answer this question, I draw on social movement theory to build a model with US Congressional bills from the 109th Congress as my dependent variable, and (a) social movement industry (SMI) strength, (b) interest group strength, (c) public opinion, (d) political elite support and (e) media coverage as my independent variables. I also draw on contemporary critical theory, utilizing Nancy Fraser's distinction between recognition and redistribution, to split my data into two distinct data sets: one built around redistribution-based bills and one built around recognition-based bills. I analyze these data using rare events logistic regression (relogit) to see if SMI strength is correlated with bill passage. The results suggest that SMIs do influence the passage of recognition-based bills but do not affect redistribution-based bills. This finding has profound implications that span from practical politics and political theory to social theory and moral philosophy.
Thesis:
Thesis (M.S.)--University of Colorado Denver. Humanities and social sciences
Bibliography:
Includes bibliographic references.
General Note:
Department of Humanities and Social Sciences
Statement of Responsibility:
by Brett J. Reeder.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
862818969 ( OCLC )
ocn862818969

Downloads

This item has the following downloads:


Full Text
INSULT, INJURY AND IMPACT:
SOCIAL MOVEMENT IMPACT ON PUBLIC POLICY IN THE CONTEXT OF
RECOGNTION AND REDISTRIBUTION
by
Brett J. Reeder
B.S. and B.A., University of Colorado, Boulder, 2006
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Master of Social Science
Humanities and Social Sciences
2012


This thesis for the Master of Social Science degree by
Brett J. Reeder
has been approved for the
Humanities and Social Sciences
by
Dr. Paul Stretesky, Chair
Dr. Akihiko Hirose
Dr. Lucy McGuffy
11/8/2012


Reeder, Brett, J. M.S.S., Master of Social Science Program
Insult, Injury and Impact: Social Movement Impact on Public Policy in the Context of
Recognition and Redistribution
Thesis directed by Professor Paul Stretesky.
ABSTRACT
This study is motivated by a simple question that has a complex set of answers: Do social
movements impact social justice? To answer this question, I draw on social movement
theory to build a model with US Congressional bills from the 109th Congress as my
dependent variable, and (a) social movement industry (SMI) strength, (b) interest group
strength, (c) public opinion, (d) political elite support and (e) media coverage as my
independent variables. I also draw on contemporary critical theory, utilizing Nancy
Frasers distinction between recognition and redistribution, to split my data into two
distinct datasets: one built around redistribution-based bills and one built around
recognition-based bills. I analyze these data using rare events logistic regression (relogit)
to see if SMI strength is correlated with bill passage. The results suggest that SMIs do
influence the passage of recognition-based bills but do not affect redistribution-based
bills. This finding has profound implications that span from practical politics and
political theory to social theory and moral philosophy.
The form and content of this abstract are approved. I recommend its publication.
Approved: Dr. Paul Stretesky
in


TABLE OF CONTENTS
CHAPTER
I. INTRODUCTION.................................................................1
Importance of the Study.......................................................2
Overview of Chapters..........................................................3
II. LITERATURE REVIEW...........................................................4
Social Movement Literature....................................................4
My Utilization of the Social Movement Literature..........................14
Gaps in the Social Movement Literature....................................16
Critical Theory Literature...................................................18
My Utilization of the Critical Theory Literature..........................21
Gaps in the Critical Theory Literature....................................23
III. METHODOLOGY...............................................................24
Dependent Variable: Social Justice Bills.....................................25
Bill Selection............................................................25
Database Distinction......................................................27
Independent Variable #1: Social Movement Industries (SMIs)...................28
Data Source: NCCS IRS Business Master Files...............................29
Data Reduction............................................................32
Differentiating SMOs and Interest Groups..................................38
Determining Directionality and Assigning Values...........................43
Independent Variable #2: Interest Groups.....................................44
Independent Variable #3: Public Opinion......................................44
IV


Independent Variable #4: Political Elite Allies
48
Independent Variable #5: Media Coverage.......................................53
Analysis: Rare Events Logistic Regressions (ReLogits).........................54
Overview of Methods...........................................................58
IV. RESULTS......................................................................59
Descriptive Statistics........................................................60
Dependent Variable: Social Justice Bills....................................60
Independent Variable #1: Social Movement Industries (SMIs)..................63
Independent Variable #2: Interest Groups....................................65
Independent Variable #3: Public Opinion.....................................68
Independent Variable #4: Political Elite Allies.............................68
Independent Variable #5: Media Coverage.....................................69
Uncontrolled Direct Effects: Single ReLogit Models............................70
Controlled Effect: Full ReLogit Models........................................72
Political Elite Allies......................................................80
Public Opinion..............................................................84
Interest Groups.............................................................88
Social Movement Industries (SMIs)...........................................91
SMI Impact in Context: Various ReLogit Models.................................93
SMIs Against the World......................................................93
SMIs Riding a Tide of Support...............................................95
People vs. Power............................................................96
SMIs with Key Elite Allies..................................................97
Comparison of Circumstances.................................................99
Overview of Results..........................................................101
v


V. DISCUSSION....................................................................102
Implications for the Social Movement Literature................................102
Support for Resource Mobilization Theory (RMT]...............................106
Political Elites, Public Opinion and Master Frames...........................106
Uncertain Effect of Media Coverage...........................................108
Distinguishing Between Interest Groups and SMOs..............................108
SMI Impact and Contextual Circumstance.......................................110
Implications for Critical Theory...............................................112
Practical Politics (SMI Tactics].............................................112
Political Theory.............................................................115
Social Theory................................................................117
Moral Philosophy.............................................................121
Overview of Discussion.........................................................123
VI. CONCLUSION...................................................................125
Study Limitations..............................................................126
Directions for Future Research.................................................129
REFERENCES........................................................................132
vi


LIST OF TABLES
Table
III. 1 Search Terms Used to Identify Bills by Major Social Categories.............26
111.2 Step 1 in Data Reduction for the BMF Master Files: IRS Code.................32
111.3 Step 2 in Data Reduction for the BMF Master Files: Major Group (12).........34
IIF4 Step 3 in Data Reduction for the BMF Master Files: Major NTEE Groups.......35
IV. 1 Number of Bills by Chamber..................................................61
IV.2 Number of Bills by Year and Chamber..........................................61
IV.3 Number of Bills..............................................................62
IV.4 Descriptive Statistics for SMI Revenue ($ in Billions).................63
IV. 5 Descriptive Statistics for SMI Assets ($ in Billions).......................64
IV.6 Descriptive Statistics for SMI Revenue + Assets ($ in Billions).............65
IV.7 Descriptive Statistics for Interest Groups Revenue ($ in Billions)..........66
IV. 8 Descriptive Statistics for Interest Groups Assets ($ in Billions)...........67
IV.9 Descriptive Statistics for Interest Groups Revenue + Assets ($ in Billions).67
IV. 10 Descriptive Statistics for Public Opinion (in Favor of Bill)..............68
IV. 11 Percent of Bills that Fall within each Political Elite Support (in Favor of Bill)
Category..........................................................................69
IV. 12 Descriptive Statistics for Media Coverage (in 100s of Articles)...........70
IV. 13 Significance and Direction of Uncontrolled ReLogit Estimates..............71
IV. 14 Corrected Logit Coefficient Estimates (Robust SE) for Revenue Model.......73
IV. 15 VIF and (Tolerance) for Revenue Model......................................74
IV. 16 Corrected Coefficient Estimates (Robust SE) for Assets Model..............76
vii


IV. 17 VIF and (Tolerance) for Assets Model.......................................77
IV. 18 Corrected Logit Coefficients (Robust SE) for Revenue + Assets Model........78
IV. 19 VIF and (Tolerance) for Revenue + Assets Model.............................79
IV.20 Relative Risk of Bill Passage for Partisan Progression for All Bills (n = 370). 81
IV.21 Relative Risk of Redistribution Bill Passage for Partisan Progression (n = 159). 82
IV.22 Relative Risk of Recognition Bill Passage for Partisan Progression (n = 211)....83
IV.23 Relative Risk of Bill Passage for Public Opinion for All Bills (n = 370)....... 85
IV.24 Relative Risk of Redistribution Bill Passage for Public Opinion (n = 159).......86
IV.25 Relative Risk of Recognition Bill Passage for Public Opinion (n = 211)..........87
IV.26 Relative Risk of Passage for Interest Groups for All Bills (n = 370)........... 88
IV.27 Percentage of Cases with Democratic Sponsorship Only (Low Political Elite
Support)..........................................................................91
IV.28 Relative Risk of Passage for SMIs for Recognition Bills (n = 211)...............93
IV.29 Relative Risk of Recognition Bill Passage for SMIs with IVs at p20 (n = 211)... 94
IV.30 Relative Risk of Bill Passage for SMIs with IVs at p80 (n = 211)................95
IV.31 Relative Risk of Recognition Bill Passage for SMIs with Public Opinion at p80
and All Other IVs at p20 (n = 211)................................................97
IV.32 Relative Risk of Recognition Bill Passage for SMIs with Elite Political Allies at
p80 and All Other IVs at p20 (n = 211)................................................98
viii


LIST OF FIGURES
Figure
III. 1 NTEE-CC Decile/Centile Code Eliminations.....................................36
III. 2 NTEE-CC Decile/Centile Code Eliminations.....................................37
IV. 1 Probability of Bill Passage by Political Elite Support for All Bills (n = 370). 81
IV.2 Probability of Redistribution Bill Passage by Partisan Progression (n = 159)....82
IV.3 Probability of Recognition Bill Passage by Partisan Progression (n = 211).......83
IV.4 Probability of Recognition Bill Passage by Public Opinion for All Bills (n = 370).
.....................................................................................84
IV. 5 Probability of Redistribution Bill Passage by Public Opinion (n = 159)........86
IV. 6 Probability of Recognition Bill Passage by Change in Public Opinion (n = 211). .87
IV.7 Probability of Bill Passage by Interest Groups for All Bills (n = 370)......... 88
IV.8 Probability of Recognition Bill Passage by SMIs (n = 211)......................92
IV. 9 Probability of Recognition Bill Passage for SMIs with IVs at p20 (n = 211).....94
IV. 10 Probability of Recognition Bill Passage for SMIs with IVs at p80 (n = 211)...95
IV. 11 Probability of Recognition Bill Passage for SMIs with Public Opinion at p80 and
All Other IVs at p20 (n = 211)......................................................96
IV. 12 Probability of Recognition Bill Passage for SMIs with Elite Political Allies at p80
and All Other IVs at p20 (n = 211)..................................................98
IV. 13 Probability of Recognition Bill Passage in the Revenue Model by SMI Strength in
Various Circumstances (n = 211)......................................................99
IV. 14 Probability of Recognition Bill Passage in the Assets Model by SMI Strength in
Various Circumstances (n = 211).....................................................100
IX


IV. 15 Probability of Recognition Bill Passage in the Revenue + Assets Model by SMI
Strength in Various Circumstances (n = 211).................................100
x


CHAPTER I
INTRODUCTION
This study is inspired by an overarching question: Do social movements impact
social justice? More specifically, Im interested in movement influence on the passage of
federal bills related to issues of social justice1 in the United States. To further clarify this
investigation, I utilize the concepts of social movement organizations (SMOs) and social
movement industries (SMIs). SMOs are formal organizations that set goals aligned with
social movements or counter-movements. SMIs are collections of all the SMOs that
share the broad goals of a movement or counter-movement (McCarthy & Zald, 1977).
I utilize these concepts to clarify the bounds of my research with my first research
question:
Within the contemporary United States, does the strength of a social movement
industry impact whether the U.S. Congress passes a bill that is important to that
industry?
1 Though there are competing conceptions of social justice, here I use the term in the
broad sense to mean justice (however defined) realized throughout society. Though I
discuss competing claims over the content of justice (i.e., what constitutes justice) below,
with this definition, I intentionally leave it open to contestation. I do, however, insist that
social justice must extend throughout all social institutions. As such, though formalized
legal justice is an important element of social justice, it is only one component.
Further, social justice must be social, not individual (though the social impacts
individuals). That is, an injustice done to an individual is only a matter of social justice if
the injustice is endemic to society itself. For example, a racist manager who only
promotes white employees to particular positions is (at least from most conceptions of
justice) committing an injustice at the individual level. However, a social structure that
only allows white people to hold particular positions (whether through legal mandate,
structural impediments or cultural bias) is social injustice. While the first example may
be a symptom of a social justice issue, in and of itself it is a matter of individual justice.
However, the second is a clear matter of social justice.
1


As a second layer to my project, I utilize Nancy Frasers (2003) distinction between
recognition and redistribution. Described in more detail in the literature section below,
this theory views recognition and redistribution as fundamentally separate, yet inter-
related, social realities. Using this concept, my second research question is:
If a relationship between social movement industry strength and bill passage
exists, is that relationship moderated by the focus of the bill on redistribution or
recognition?
Importance of the Study
This work is important for reasons that mirror my two layers of investigation.
First, my research adds to the social movement literature by further clarifying how a
social movement industry may influence public policy. Most of the existing social
movement literature takes social movements as the unit of analysis, selecting cases based
on the movements themselves. As a result, these investigations often look at how large
and well-known movements impact policy, ignoring cases with weak or absent
movements. In contrast, this study uses social justice bills as the unit of analysis, and
thus includes cases with a wide range of social movement presence and strength.
Second, this study expands the recognition/redistribution debate raging in
contemporary critical theory by injecting much needed empirical evidence into it. This
debate centers on a moral philosophical distinction Nancy Fraser makes between
recognition, redistribution and representation, contrary to Axel Honneths assertion that
redistribution and representation are forms of recognition (Fraser & Honneth, 2003).
From its moral philosophical roots, this debate extends through social theory, to political
theory and practical politics. Its treatment at the level of practical politics is largely
2


limited to policy prescriptions for a just society, but to date, there has been little
investigation as to whether this distinction is meaningful at the level of practical politics.
That is, while having a lot to say about the types of policy that would lead to social
justice, it is yet unclear whether the recognition/redistribution distinction matters when
investigating how social justice policy is actually formed. This study remedies that gap.
Overview of Chapters
This study is divided into six chapters. The first chapter, the introduction,
outlines the paper as a whole, including its purpose and importance. The second chapter
presents a review of two literatures relevant to this study: (1) the social movement
literature related to movement outcomes and (2) the critical theory literature on
recognition. It also describes how I utilize this literature, as well as the gaps within these
literatures. The third chapter describes the methods used, including the overall design,
how the databases were developed and the analyses used. The fourth chapter presents the
results of this analysis, including descriptive statistics for each variable and a variety of
rare events logistic regressions (relogits). I discuss the implications of these findings for
both the social movement literature and the critical theory literature in the fifth chapter.
The sixth chapter concludes with an overview of the study as a whole, an identification of
the studys limitations and potential directions for future research.
3


CHAPTER II
LITERATURE REVIEW
There are two major bodies of literature relevant to this study: social movement
literature and the critical theory literature on recognition.2 Below, I review major works
within each tradition, describe how I utilize the literatures and discuss gaps within each
literature.
Social Movement Literature
Until recently, social movement scholars largely neglected the impacts of social
movements, focusing instead on other aspects such as their emergence, growth and
operation. This has changed in recent years, but the range of social movement outcomes
investigated is staggering (see Snow & Soule, 2009 for an account of the different types
of outcomes studied). In addition to the influence movements have on states, scholars
have investigated the effect movements have on culture, their participants and even other
movements. Further, those who study movement impact on states look at a variety of
potential outcomes, from new policies to changes in the political structure. Finally,
movement scholars studying movement impact on policy focus on different aspects of
policy, including agenda setting, policy development and policy implementation. Thus,
my focus on movement impact on the passage of public policy is a subset (policy
passage/non-passage) of a subset (impact on policy) of a subset (influence on states) of a
subset (consequences of social movements) of the social movement literature (Amenta &
2 A third body of literature, the public policy process literature, is also arguably relevant
to this study. This rich literature includes complex accounts of how policy is passed and
the factors that influence passage. That said, due to its relative silence on social
movements, I have not included it as part of this review. I do, however, include several
references to it in footnotes throughout this section.
4


Caren, 2007; Snow et al., 2007). In other words, the portion of the social movement
literature I cover heresocial movement impact on public policyis a narrow section of
a large body of social movement literature.
Research in this narrow area has expanded substantially in the last decade, and
numerous reviews of social movement political outcome studies exist (Amenta & Caren,
2007; Amenta et al., 2010; Burstein & Linton, 2002; Earl, 2000; Giugni, 1998; Giugni,
2008; Meyer, 2005). These reviews generally suggest that social movements sometimes,
but certainly not always, influence the passage of public policy. For example, Burstein
and Lintons (2002) review of 53 articles published between 1990 and 2000 indicated
that the political organizations (their umbrella term for SMOs, interest groups and
political parties) influence on public policy is statistically significant for about half of the
sample studied. Further, they suggest that this effect is meaningful in policy terms (i.e., a
substantial impact on policy) only about one-fifth of the time (Burstein & Linton, 2002).
Similarly, Amenta et al.s (2010) review suggests that social movement impact on
public policy is incredibly episodic and nuanced. This review of 45 articles on the
political consequences of social movements published between 2001 and 2009 suggests
that, in general, scholars have moved away from William Gamsons (1990) approach of
defining success in terms of gaining new benefits or acceptance and toward their
influence on particular political outcomes. This work has indicated that movements
impact varies substantially based on the particular phase of the policy process in question
5


(i.e., agenda setting, adoption and implementation).3 In particular, the literature suggests
that movements are most influential during the agenda-setting stage (the stage that
determines the which policy areas are discussed) of the policy process, and often only
minimally influential in the adoption stage (the stage that involves the selection of
particular policies) (Amenta et al., 2010; Johnson, Agnone, & McCarthy, J. D., 2000;
Johnson, 2008; Snow & Soule, 2009).
To successfully influence policy, movements utilize resources ranging from the
tangible (i.e., money or human capital) to the intangible (i.e., legitimacy or cultural
artifacts) in an effort to realize goals. While the relative importance of various types of
resources is disputed, social movement scholars almost universally accept that
movements need sufficient resources to successfully attain their goals (Cress & Snow,
1996; Khawaja, 1994; Zald, 1992).
There is, however, more disagreement about the importance of organization.
Some scholars argue that as movements formalize their organizational structure and
professionalize, they simultaneously become less radical and confrontational
(Staggenborg, 1988). Additionally, a move toward formal organization can shift a
movements focus away from movement goals and toward organizational survival and
professional advancement (Michels, 1915). In other words, leaders of formal
3 Within the policy literature, it has long been customary to conceptualize policy
formation as a sequential process that proceeds through delineated stages. Indeed, this
Stages Heuristic dominated the field until the mid-1980s. In recent years, this view has
been criticized for a variety of reasons, and most contemporary scholars now recognize
that the stages of a policy process are not strictly sequential, nor are they unidirectional.
Further, many have pointed out that it is impossible to completely delineate separate
stages. However, most contemporary policy scholars continue to utilize some variant of
the stages model as a taxonomic device within their theories. For a discussion of the role
various stages models have played in policy theory, see deLeon, 1999 and Sabatier, 2007


organizations may begin to act in ways that benefit the organization and/or their own
professional advancement, to the detriment of the movement as a whole (e.g., going after
grants poorly aligned with the goals of the organization but that allow the organization to
make its payroll). Along these lines, Frances Fox Priven and Richard A. Cloward (1978)
famously argued that organization undermines movements ability to realize outcomes by
redirecting resources away from disruptive protest and toward organization building.
Taking the contrary position, scholars working out of the resource mobilization
tradition have stressed the importance of formal organizations in social movements.
Developed most comprehensively by John McCarthy and Mayer Zald in the mid-1970s,
this perspective argues that formal organizations, known as social movement
organizations (SMOs), are indispensible in a movements ability to acquire, and direct the
use of, resources (Cress & Snow, 1996; Gamson, 1990; McCarthy & Zald, 1973, 1977,
2001). From this perspective, movement success (such as influencing bill passage) is
thus dependent on a movements ability to organize through one or more SMOs, and for
these SMOs to acquire resources they direct at well-defined goals, such as promoting (or
opposing) the passage of federal bills (Edwards & McCarthy, 2007). From this
perspective, the strength of a movement as a whole is thus closely associated with the
aggregate strength of SMOs within italso known as a social movement industry (SMI).
The implication is, therefore, that well-resourced SMIs have the ability to influence
policy. In other words, movement mobilization is sufficient for impact on policy
(McCarthy & Zald, 2001). For example, Cress and Snow (2000) found in a study of 15
homeless SMOs that organizational viability (defined as the ongoing maintenance of a
7


formal organization and sustained protest activity over time) was the most important
indicator of SMO success.
Beyond resources and organization, much of the recent work on social
movements has focused on either the tactics utilized by movements or the contextual
factors that shape movement outcomes as well as the movements themselves (Snow &
Soule, 2009). With regard to tactics, scholars have found that the type of action that
movements engage in matters. Along these lines, scholars have long debated the effect of
the tactical choice to engage in disruptive or violent tactics. Some argue that
violence/disruption is necessary to realize political outcomes that upset the status quo,
while others believe violence undermines and delegitimizes movement efforts. Thus far,
the empirical evidence is mixed, with some support for both claims (Giugni, 1998).
Gamsons 1975 classic Strategy of Social Protest (second edition 1990) study, as
well as several re-analyses of these data (Mirowsky & Ross, 1981; Steedly & Foley,
1979) found evidence that disruptive tactics used by challenging groups is positively
correlated with both attaining legitimacy and obtaining new advantages (his two
measures of movement success). Similar findings appear in a variety of works on related
phenomenon such as strikes (Shorter & Tilly, 1971) and in Frances Fox Priven and
Richard A. Clowards (1977) Poor Peoples Movements. However, other works focused
on labor conflicts and urban riots find little evidence that violence is an effective tactic
for achieving desired outcomes (Kelly & Snyder, 1980; Snyder & Kelly, 1976; Taft &
Ross, 1969). Indeed, the literature seems to suggest that, like movements generally,
disruptive tactics sometimes successfully influence policy and sometimes they do not
(Giugni, 1998). This extends to tactics generally as some scholars suggest that different
8


tactics may be more or less effective at different phases in the policy process (Andrews &
Edwards, 2004), while others have found that the importance of a particular tactic can
vary by context (Amenta, 2006).
Another particularly fruitful line of research focuses on the act of meaning
construction, or framing. Movement scholars apply this concept to the meanings and
beliefs that inspire and legitimate social movement activities. These types of frames,
called collective action frames vary in scope from those employed by particular
organizations (organizational frames) to general conceptual frames that are employed by
entire cultures (Benford & Snow, 2000). Theorists working within the framing tradition
have argued that movement success depends largely on its ability to effectively frame the
problem (diagnostic frames) an attractive solution (prognostic frames) and the
motivation to act (motivational frames) (Benford & Snow, 2000; Cress & Snow, 2000).
Cress and Snow (2000) found in addition to organizational vitality, framing processes
were consistently one of the two most important factors in a social movement
organizations success. More recently, McCammon (2009) found that favorable
outcomes associated with womens jury rights mobilizations were tied specifically to
those that defined the problem (diagnostic frame) as serious and wide spread, contained
clear rationale and supported their position with concrete evidence (prognostic frame).
Scholars working within the framing tradition also have begun to investigate the
importance of the mass media as a master arena in which public discourse takes place
and meaning is established as frames compete for superiority (Gamson, 2007; Gamson
&Wolfsfeld, 1993). From this perspective, the media act as the primary site where the
contestants assume meaning is established, whether or not this is actually the case
9


(Gamson, 2007). Empirical work in this tradition suggests that the media do, in fact,
affect the political views of legislators and, as a result, impact public policy (Cook, Tyler,
Goetz, Gordon, Protess, Leff, &Molotch, 1983). Further, some scholars have
investigated the ways SMOs and other organizations strategically target the media, as
well as the relative success of these efforts (Rohlinger, 2002).
Beyond the role context plays in the relative influence of particular frames or
tactics, many scholars have found that contextual factors play an important role generally.
Within this line of research, scholars have seen the political structure as particularly
important. It should come as no surprise that the impact of social movements varies by
regime type (Goldstone, 1980), but even within Western democracies, the peculiarities of
particular political systems has been shown to affect the degree to which social
movements will have any influence on public policy (Schwartz, 2000). Scholars working
within this political process tradition focus on the political context movements are
embedded within. In particular, they argue that a political systems openness, and its
capacity to act, interact to establish a political opportunity structure. In its more pure
forms, these theorists view the opportunity structure as the primary explanatory variable
in both movement development and their potential for success (Kriesi, 1995).
The changing political context within regimes has also been shown to influence
the ability of social movements to impact policy. In particular, scholars have argued for
the importance of powerful allies within political institutions (Tarrow, 1993, 2011). For
example, Meyer and Minkoff (2004) found that the influence of left-leaning movements
is amplified when Democrats are in power. Similarly, Tarrow (1993) argues persuasively
that during the May 1968 student uprisings in France, social movement activity was only
10


able to affect university policy with the help of elite institutional allies. Thus, the
political process literature suggests that movement success is greatly associated with the
presence or absence of political elite allies.
Perhaps chief among the non-structural contextual factors is public opinion.
Several scholars have found evidence that public opinion influences public policy
directly, especially when it is widely held and strongly felt (Burstein, 2003; Costain &
Majstorovic, 1994; Hartley & Russett, 1992; Hill & Hinton-Andersson, 1995; Page &
Shapiro, 1983). Further, some have found that when public opinion is taken into account,
the influence of political organizations (i.e., SMOs, interest groups and political parties)
is greatly diminished or gone entirely (Burstein & Linton, 2002;Uba, 2009). However,
peoples attention span is limited to a finite number of topics, and thus, public opinion is
limited to a finite number of issues. Further, within the universe of topics the public is
generally aware of, many do not interest most people. As such, while public opinion is
an important factor where it exists, it is often not present in any meaningful form
(Burstein, 2006).
In recent years, scholars have started to combine the above variables into joint
and/or mediated approaches that look at the interactive effects of social movements,
political context and public opinion (Snow & Soule, 2009). Though the particularities of
these models vary somewhat, they have a common model of causation: Movement
impact is mediated through other institutions. This implies that much of a movements
influence on policy is indirect, through its influence on other relevant variables such as
elected officials or public opinion (Snow, Soule, & Kriesi, 2007). For example, in the
context of the struggle to pass the Equal Rights Amendment (ERA), Soule and Olzak
11


(2004) found that the presence of elite allies intensified the effect of the pro-ERA
movement on state ratification but that the presence of elite allies was not necessary for
state ratification.
In an influential mediated model, Edwin Amenta and his colleagues (1992, 2005)
argue that the consequences of social movement activities depend on (1) the level of
movement mobilization, (2) the favorability of the political context and (3) the strategies
implemented by the movement. Movement mobilization in this context is defined as the
ability of a movement to mobilize supporters in political actions with varying degrees of
assertiveness, with the chosen mix of assertive tactics representing the movements
strategy. Additionally, Amenta et al. (1992, 2005) identify four specific aspects of a
political context that influence movement impact: (1) the degree to which democratic
practice bounds democratic institutions, (2) the degree to which political parties rely on
patronage, (3) the presence/absence of bureaucrats with missions aligned with the
movement and (4) the partisan makeup of the regime (Amenta et al., 2005). Finally, the
causal mechanisms within this model, operating within the assumptions of bounded
rationality, assume elected officials consider the costs/benefits associated with supporting
a bill. From this perspective, the most important costs/benefits are those related to (1)
reelection, (2) standing within the party, (3) perceived efficacy of the policy and (4)
ideological commitments (Amenta, 2006).
Giugni and Passy (1998) explicitly tested some of the propositions discussed thus
far by setting up models with social movements, public opinion and political alliances as
independent variables, and public policy as a dependent variable. They then tested to see
if there was a direct effect between movements and policy passage, an effect mediated
12


through other variables or a joint effect in combination with other variables. Their
analysis found very little evidence to support the direct-effect and mediated-effect
models, but some support for the joint-effect model. However, this effect only held for
two of the three movements studied. Additionally, public opinion had a direct effect on
one of the issues areas, but it did not add to the statistically significant interactive terms.
These findings further support the now widely held view that social movement
impact is highly contextual, and that in some contexts movements influence the passage
of public policy while in others they do not. The literature discussed above suggests that
the following factors, internal to movements, affect their ability to influence the passage
of public policy: (1) the resources available to the movement, (2) the organization of the
movement, (3) the tactics employed by the movement and (4) the way the movement
frames its issues. Additionally, the literature suggests that several external factors affect
the ability of movements to influence the passage of public policy. These include: (5) the
political system movements are embedded within, (6) the presence or absence of elite
allies within this system and (7)public opinion related to the policy.4 Finally, the
literature suggests that while movements have only minimal direct effect on the passage
4 The policy process literature is largely silent on the impact of social movements, yet it
overlaps in many ways with the movement theory described in this section. Among the
varied policy process literature, the Advocacy Coalition Framework (ACF) is perhaps
best positioned to dovetail with social movement theory. In the ACF, the primary actors
are advocacy coalitions, or informal networks of individuals held together by common
beliefs, engaged in collective action directed at particular policy objectives. This concept
has much in common with the concept of a social movement: both are social collectives
engaged in collective action motivated by common beliefs. However, unlike advocacy
coalitions, social movements activities extend beyond efforts to influence policy.
Additionally, advocacy coalitions include actors generally not considered part of social
movements (e.g.. the media). For an overview of the ACF, see Sabatier and Weible
(2007). For an example of how the ACF has been applied to social movements, see
Turina (2009).
13


of public policy, they interact in complex ways with the factors above to, at times,
influence policy passage.5 There are, however, a variety of interactions and combinations
that appear to result in social movement influence on policy passage. Thus, the most
promising research in this area now includes many of the factors above and takes a
combinatorial or interactive approach.
My Utilization of the Social Movement Literature
As discussed above, contemporary social movement theory contains four general
arguments designed to account for social movement influence on states. Three of these
are closely related to well-known theories of social movement mobilization: (1) resource
mobilizationsuggests that outcomes result from formal organizations mobilizing
resources; (2) framing theorysuggests that outcomes are related to the ways in which
movements frame their claims; and (3) political process theorysuggests that outcomes
are largely associated with elements of the political structure. Additionally, many
theorists have combined elements of each of these theories into various (4) mediated
modelssuggesting that movement outcomes vary by context, with a special emphasis
5 Three prominent scholars who have played a large role in establishing this body of
literature (Doug McAdam, Charles Tilly and Sidney Tarrow) have developed an
approach that shifts the focus from correlation between variables to paired comparisons
of the processes and mechanisms that make up such correlations. This approach, termed
Dynamics of Contention (DOC), also seeks to shift the focus from social movement life
histories to episodes of contention (McAdam, Tarrow, & Tilly, 2001; Tilly &Tarrow,
2006). In this sense, it hopes to expand the field to include other forms of political
contention such as revolutions and democratization, while simultaneously narrowing it to
include episodic political activities (as opposed to movements, which can persist beyond
contentious episodes) (Buechler, 2011). Though this body of work provides important
insights in the social movement literature, it is not directly relevant to this research
because my focus remains social movements and correlations between variables, not
paired comparisons of mechanisms across a variety of episodes of contention.
14


on the movements themselves, political elites and public opinion (Amenta et al. 1992,
2005; Snow, Soule, & Kriesi, 2007).
In this study, I partially sidestep many of the arguments made by both political
process theorists and their framing counterparts. I avoid much of the political process
argument by holding the larger political context constantthe US Congress from 2005-
2006. As such, my findings are limited to the particular political context in which my
data is embedded (the US federal legislative system in 2005-2006), but I can also claim
that any variation I observe is not the result of different legislatures or different political
systems. This approach reduces, but does not eliminate, the effect of shifting
opportunity/threat within the political process.
Similarly, Im able to avoid large portions of the framing argument, by using a
large-N design. My sample is large enough to capture both effective and ineffective
frames and strategies (if such things exist), and thus particular organizational frames and
strategic positions should not systematically affect my claims about aggregate effect.
However, a large sample does not overcome the all-encompassing nature of master
frames. As such, my study is able to avoid the effect of micro level framing, but as
discussed in chapter 5, framing at the macro level may play an important role in these
data.
My study more directly confronts both resource mobilization theory and mediated
models. Most importantly, following RMT, I assume that movements need sufficient
resources to realize their goals. Further, I assume that because money is highly liquid
(i.e., it can easily be translated into other resources), it is the most important resource an
organization can possess. Additionally, in line with RMT and contra Fox-Piven (1978), I
15


assume that formal organization is a net positive in terms of movements achieving their
outcomes.
I recognize that movements extend beyond formal organizations, and money is
not the only organizational resource utilized by movements. However, following RMT, I
assume that movements with well-resourced SMOs can influence policy, whereas
movements that lack either organization or resources will not. By including a variety of
other possible influences on public policy (my independent variables), I can see the
degree to which SMO strength is correlated with bill passage when controlling for other
factors. Thus, a statistically significant correlation between SMI strength and bill passage
would validate much of RMT.
I also draw directly on mediated models and assume that SMI impact is at least
somewhat dependent upon the context they are embedded within. In particular, I assume
that public opinion, political elites and media coverage prominently factor into a
movements ability to realize political outcomes. That is, following mediated models, I
assume that SMI influence on bill passage is at least partially mediated by public opinion,
the presence of political elites and media coverage. Thus, in this study, I borrow from
RMT the presumed importance of organizations and resources in social movement
impact, and from mediated models I borrow the idea that movement impact is partially
mediated by public opinion, political elites and media coverage.
Gaps in the Social Movement Literature
The social movement literature largely developed within the disciplines of
sociology and political science, with disciplinary boundaries sometimes translating into
conceptual barriers. Political scientists tended to focus on the more institutionalized
16


interests groups and political parties, and sociologists tended to focus on less
institutionalized SMOs and the informal movement networks they are embedded within.
Since at least the 1990s, these disciplinary boundaries have faded somewhat, and a
community of social movement scholars working across disciplines has emerged
complete with their own specialty journals dedicated to social movements (Mobilization
and Social Movement Studies).
Additionally, in recent years, this scholarly community has made substantial
strides in building a respectable body of work on social movement impact (compare the
review in Guigni, 1998, with the review in Amenta et al., 2010). Most of these scholars
now acknowledge the importance of a wide range of factors, including: movement
resources, organization, tactics, framing, the political system, political elites, public
opinion and media coverage. Many have also looked carefully at the interaction between
these factors.
Nonetheless, this literature has several substantial gaps. Among the most
problematic is the tendency to study large movements that have seen their goals realized.
Most social movement studies first identify the movements they will study and then
examine their effect. This approach undermines the plausibility of claims of movement
impact because scholars rarely examine instances in which movements were weak or
non-existent. My study overcomes this deficiency by starting with the outcome (bill
passage) and then matching movements (through SMIs) to this outcome.
However, the most important contribution of this study is the bridging of two
literatures and two traditions: social movement theory and critical theory works on
recognition. By using the recognition literatures distinction between redistribution and
17


recognition to typologize political outcomes (bill passage), I add a new perspective to the
social movement literature.6
Critical Theory Literature
The concept of recognition has gained a lot of traction within contemporary
critical theory.7 Within this context, recognition typically involves acts of
acknowledging or respecting others. In this sense, critical theorists approaches to
recognition share a common model of both the individual and society. While these
theorists acknowledge that society is comprised of individuals, they also argue that the
individual (or subject) is largely a product of interaction with other people (or subjects).
Thus, these theorists have an intersubjective model of both the individual and society.
6 The distinction I make between recognition and redistribution is, in many ways, similar
to distinctions made by European scholars working within what has been termed the New
Social Movement tradition. New Social Movement (NSM) theories are diverse and often
expand well beyond social movements. Despite this diversity, NSM theories of all types
tend to stress that different social arrangements lead to different social movements. In
particular, they tend to argue that prior to the 1960s, social movements stressed class-
based material claims and since the 1960s have stressed symbolic claims based on
identities such as race or gender (Buechler, 2011). Though the critical theory recognition
literature I draw on makes a similar distinction, it differs in important ways that I discuss
in the 8th footnote below.
7 Within philosophy and the social sciences, Critical Theory has both a narrow and a
broad definition. In the narrow sense, Critical Theory refers to the Frankfurt School,
which consists of the theory produced by specific German philosophers and social
theorists associated with the Institute for Social Research at the University of Frankfurt
am Main. This tradition includes a variety of well-know theorists, including Max
Horkheimer, Theodor Adorno, Herbert Marcuse, Jurgen Habermas and, most recently,
Axel Honneth (who plays a central role in this paper). In the broad sense, Critical Theory
refers to any social theory that aims to both explain and actively transform instances of
social domination and human enslavement. Nancy Fraser (another central figure within
this paper) falls within this version of Critical Theory. Both versions differ from
traditional theory in that they explicitly embrace normative positions related to human
emancipation. For a brief but comprehensive history of Critical Theory in both senses,
see the Critical Theory entry in the Stanford Encyclopedia of Philosophy (Bohman,
2012).
18


Working from such a model, and drawing from what the renowned 19th Century
philosopher Georg Wilhelm Friedrich Hegel termed the struggle for recognition, many
contemporary philosophers have appropriated the term recognition in their examination
of the normative underpinnings of political claims. Yet the ubiquity of the term itself is
not matched by the ways in which it is used and defined. Indeed, the proper conception
of recognition is at the center of a major debate within contemporary critical theory
(Fraser &Honneth, 2003).
For the purposes of this project, two contemporary thinkers engaged in this debate
are particularly important: Axel Honneth and Nancy Fraser. From Honneths
perspective, recognition is a historically specific concept. According to Honneth, the
transition to bourgeois-capitalist society led to the differentiation of recognition into these
three independent spheres: (1) loverecognition of others through loving care for their
well-being; (2) lawrecognition of others as members of society afforded particular
rights; and (3) achievementrecognition of others as productive members of society.
Honneth claims that while these three spheres are separate concepts in modern society,
they were not independent of each other prior to modernity (Fraser &Honneth, 2003).
Borrowing heavily from Hegel, Honneth goes on to argue that human
development requires recognition. In Hegels account of modern society, this implies
that to become fully actualized beings, people need reciprocated love, legal recognition as
members of society and to be considered productive members of society (however
defined). Thus, for Honneth, political struggles for justice are ultimately the modern
embodiment of pre-political struggles for recognition in an individual effort to become
fully actualized human beings (Fraser &Honneth, 2003).
19


In contrast, Nancy Fraser has developed a status model of recognition in which
recognition is viewed as a matter of social justice. Frasers conception of justice revolves
around what she calls participatory parityor equity in the ability to participate fully in
society. From this perspective, recognition entails the cultural and symbolic
preconditions for participatory parity. That is, without social recognition, one is unable
to fully participate in society and thus the denial of social recognition is unjust.
However, for Fraser, recognition alone is insufficient to allow for full social participation
(Fraser & Honneth, 2003).
According to Fraser, in addition to recognition, justice entails equitable
distribution and political representation as separate, yet interrelated, concepts. In the case
of material distribution, Fraser argues that full participation in the social world requires at
least minimal resources, and thus inequitable distributions of resources impedes ones
ability to fully participate in society. In the case of political representation, full social
participation necessitates representation within all the political institutions under which
one is subject (Fraser, 2008).
Working from this moral philosophical level, Frasier develops a social theory
that, in quasi-Weberian fashion, differentiates class, status and party while carefully
attending to their interaction. She then turns to political theory to identify forms of
institutional change that correct misrecognition, maldistribution and misframing (of
political representation) (Fraser, 2008). Thus, for Frasier, there is a meaningful
distinction between recognition, distribution and representation that is important at the
philosophical, social and political levels.
20


This runs contrary to Axel Honneths theory of recognition (recognition monism),
which views recognition as matter of self-realization and distribution as a reflection of
recognition. According to Honneth, material distribution within a society reflects that
societys recognition patternsspecifically, its achievement-recognition structures. That
is, collective evaluation of the social contribution of individuals guides material
allocations. High levels of compensation are associated with the achievements socially
recognized as especially valuable, while low levels of compensation are associated with
the achievements that are not recognized by society as particularly valuable. Thus, for
Honneth, redistribution is a special case of recognition, and as such there is no
meaningful distinction (Fraser &Honneth, 2003).
My Utilization of the Critical Theory Literature
As discussed above, there are two major conceptions of recognition within
contemporary critical theory: (1) an identity model associated with Honneth and (2) a
status model associated with Fraser. The identity model presumes that distribution is a
form of recognition, and thus, that redistribution is not distinct from recognition. In
contrast, the status model views recognition as distinct from, but interrelated with,
redistribution.
Following Fraser, and contra Honneth, I believe the distinction between
redistribution and recognition is meaningful at the levels of practical politics, political
21


theory and social theory.8 At the level ofpractical politics, I expect there to be
differences in the efficacy of movement efforts based on whether their target is
redistribution or recognition. Further, at the political theoretical level, I believe that
institutional corrections to the recognition order will not directly translate into a more
equitable redistribution of material resources. Likewise, a redistribution of resources
through institutional channels will not directly translate into improved recognition.
Additionally, at the level of social theory, I believe the distributional structure (or class)
and the recognition order (or status group) are, at least partially, uncoupled from each
other in modern capitalist societies.
If these assertions are correct, we would expect to see differences in the way the
institutional rules (or laws) that govern distribution and recognition are established. We
would expect to see them established at different rates, and we would expect to see the
impact movements have on them to be different. On the other hand, if Honneth is correct
and there is no meaningful difference between redistribution and recognition, then we
would not expect to see any such difference.
This is, in many ways, similar to the distinctions made by New Social Movement
(NSM) theorists who tend to argue that there was a historical shift in which movements
switch their focus from class-based material claims to identity-based symbolic claims,
and that this shift corresponded with macro-social shifts. However, Frasers distinction
differs in a variety of key ways. Where NSM theorists view the key historical shift as
from class to identity, Fraser sees the key historical shift as a partial uncoupling of class
and status. This leads to two further distinctions, as NSM theorists tend to view claims
for material distribution and identity recognition as an either or proposition, whereas
Fraser views them as an interrelated and simultaneous social reality. Thus, while NSM
theorists see a shift from class to identity, Fraser sees a shift in emphasis on class or
status, though both exist simultaneously. Additionally, where NSM theorists are
concerned with identity, Fraser is concerned with status. That is, NSM theorists models
of recognition are largely identity models, while Fraser uses a status model (Buechler,
2011; Fraser &Honneth, 2003; Fraser, 2011).
22


To test this, I code bills (my dependent variable) as either mostly distribution or
mostly recognition based. I then use this coding to separate my data into two distinct
data sets: (1) a redistribution data set and (2) a recognition dataset. Next, I explore the
correlation between SMI strength and bill passage in each data set. A difference between
these two data sets adds empirical support for Frasers model of recognition, while the
absence of difference would add support for Honneths.
Gaps in the Critical Theory Literature
Like much of the literature in philosophy, the recognition literature suffers from a
relative lack of empirical evidence.9 In an attempt to contribute to remedying this issue,
this study extends the debate between Nancy Fraser and Alex Honneth to an empirical
exploration of social movement impact. Using Frasers redistribution/recognition
typology, I divide my data into two distinct data sets: one built around distribution bills
and one built around recognition bills. This allows me to see if there is an empirical
difference between SMI influence on the passage of redistribution and recognition bills.
The presence of a difference supports Frasers position, while the absence of difference
would be consistent with Honneths position. As such, while this study is not sufficient
to wholly accept or reject either Frasers or Honneths view, it is a significant step in this
direction.
9 Limited use of Frasers recognition/redistribution typology has been utilized by at least
one empirical study of social movements (see Carroll & Ratner, 1999). However, Carroll
& Ratner (1999) use of the distinction is limited to a description of how different types of
SMOs (recognition, redistribution and their addition: salvation) utilize media strategies.
23


CHAPTER III
METHODOLOGY
For this study, I developed data sets with federal social justice bills in the 109th
Congress (2005-2006) as my unit of analysis. After identifying social justice bills, I split
these bills into either recognition bills (focusing primarily on cultural justice) or
redistribution bills (focusing primarily on economic justice) to establish two distinct data
sets. Bills not clearly in either camp, or those solidly in both, were eliminated. This left
me with two data sets of social justice bills, distinguished by their relation to the
Fraserian typology of injustice claims.
I then built each data set out, connecting SMI strength, interest group strength,
political elite support, public opinion in favor and media coverage to each bill. With
these data sets built, I ran a rare event logistic regression on each data set as well as a
combined data set (the redistribution and recognition data sets combined). This allowed
me to statistically analyze each of my research questions:
Research Question #1: Within the contemporary United States, does the strength
of a social movement industry impact whether the U.S. Congress passes a bill that
is important to that industry?
Research Question #2: If a relationship between social movement industry
strength and bill passage exists, is that relationship moderated by the focus of the
bill on redistribution or recognition?
In the following sections, I discuss each variable, the development of these data sets and
this analysis in more detail.
24


Dependent Variable: Social Justice Bills
I operationalized this variable, which acted as my unit of analysis, as federal
social justice bills in the 109th Congress. Bills are potential laws introduced in the
legislature, and thus this variable holds a value of either pass (1), or fail to pass (0).
I chose the 109th Congress due to the relative lack of major events external to the
political system that might have independent effects on policy during this period (2005-
2006). For example, while I could have attained all the same data for the 110th Congress,
this time frame would have included the onset of the Great Recession, which
undoubtedly impacted bill passage. This is not to say there were not events external to
the political system that impacted policy during the 109th Congress. For example, in
2005, Hurricane Katrina clearly impacted policy, particularly in disaster relief. Likewise,
in 2006, the global outbreak of the bird flu impacted health policy. Indeed, at any
given time, global events affect public policy. However, I maintain that in comparison to
other years, during 2005 and 2006 there were relatively few major global events that
substantially impacted domestic social justice policy in the United States.
Bill Selection
To identify bills, I used key words representing several major social categories
around which social justice issues and industries cluster. I also explicitly excluded
several major areas of social justice, including: health, education, children, criminal
justice and international issues. I avoided these issues because each is associated with
rather complex and specific policy subsystems, which would substantially complicate my
analysis. Instead, I attempted to choose major social categories that cross policy
25


subsystems. Table III. 1 below displays the search terms used for each major social
category included in this study.
Table III.l Search Terms Used to Identify Bills by Major Social Categories.
Social Category Search Terms
Sexual Orientation Sexual orientation Homosexual LGBT Same sex GLBT One man and one woman Gay
Race Race Latina Ethnicity Hispanic Minority African American Latino Asian
Organized Labor Organized labor Freedom to work Labor union
Compensation Compensation Davis-Bacon* Universal wage Maximum compensation Minimum wage Limit compensation Living wage
Language Language English language Language minority Limited English Spanish
Gender Gender Mother Sex Men Women Man Woman Male Female Father
Poverty Poverty Homelessness Low income Unemployment Hunger TANF* Homeless WIC*
Wealth Windfall profit Estate tax Excess profit Death tax Capital gains Alternative minimum tax*
* Note: These search terms relate to specific policies that fall within the major social
category.
26


To identify the bills used for this study, I entered the search terms listed above into the
advanced search function of the Library of Congress THOMAS system, with the search
limited to bills and amendments in the 109th Congress. From these results, I eliminated
all bills focused on health, education, children, criminal justice and international issues,
since issues arent addressed by this study.
I then viewed the CRS Summary for each bill or amendment that showed up in
the search results to ensure the social category was a primary focus of the bill (when the
major social category was secondary to other issues, the bill was not included). For
example, using the search term women, the Mercury-Free Vaccines Act of 2005 (H.R.
881) was one of the results. This bill would have regulated vaccines containing mercury,
but because its text refers to specific regulations regarding the use of flu vaccines with
pregnant women, it appeared in my search results. However, because the bill doesnt
focus on women per-se, it was eliminated.
This process resulted in 460 bills. For these bills, I recorded whether the bill
passed, the date of the last action taken on the bill, and whether the bill enhanced or
detracted from social justice, in a Fraserian sense. That is, does the bill increase
individuals ability to participate in society as peers (enhance social justice) or does it
hinder their ability to do so (detract from social justice).
Database Distinction
To place each bill (and as a result each case) within one of my two data sets
(recognition and redistribution), I used the following criteria.
Redistribution Bills. To be considered a redistribution bill, the bill had to do at
least one of the following: (1) provide or deny material resources directly to people, (2)
27


alter an economic safety net, (3) alter the tax structure, (4) regulate/deregulate wages or
benefits or (5) regulate/deregulate access to work.
Recognition Bills. To be considered a recognition bill, the bill had to do at least
one of the following: (1) limit/open a social institution to certain types of people, (2)
define categories of people, (3) define categories of people as deserving/undeserving or
(4) signal a preference for a particular cultural artifact.
Using these definitions, I then carefully reviewed the CRS Summary for each bill,
recording whether each bill met each of the nine criteria listed above (five for
redistribution and four for recognition). When a bill met one or more redistribution
criteria, it was coded as a redistribution bill, and when it met one or more recognition
criteria, it was coded as a recognition bill. However, federal bills often include a wide
variety of provisions, and some of these bills included elements that cut across the
Fraserian typology laid out above (i.e., they met both recognition and redistribution
criteria). To maintain the conceptual clarity of this typology, I eliminated all bills that
clearly met both sets of criteria. I also, eliminated any bill that did not meet any of the
criteria. As such, I included only bills that clearly fall primarily within one of the two
categories. This process left me with 159 redistribution bills and 211 recognition bills,
for a total of370 bills (and thus cases).
Independent Variable #1: Social Movement Industries (SMIs)
This variable is meant to measure the influence of social movement industry
(SMI) strength on the probability of bill passage. To account for the fact that SMIs can
often be found on both sides of a bills (for and against), I made the value of these
variables the strength of those for, less the strength of those against. As this measure
28


increases, I expect the probability of bill passage to also increase. That is, as the
difference between the strength of an SMI in favor of a bill, less those against, increases,
I expect bill passage will be more likely.
I operationalized this variable in three ways: (a) the total annual revenue for
SMOs (aggregated into SMIs) likely to be for the bill, less the total revenue for SMOs
(aggregated into SMIs); (b) the same calculation using the total end of year assets, rather
than revenue; and (c) these two values (annual revenue and end-of-year assets) combined.
Thus, I ended up with three separate models: one that used revenue, one that used assets
and one that combined revenue and assets. In all three cases, I expect that as the variable
increases, bill passage is more likely.
I choose to use both revenue and assets because they measure different aspects of
an organizations capacity. Assets are a better measure for organizations that have
converted their financial resources into tangible assets that they utilize in their work (such
as buildings or supplies). Revenue is a better measure for organizations that spend most
of what they take in on nontangible assets (like services paid for). The revenue + assets
model, thus, combines the two. While money in any form (including assets and revenue)
is by no means a comprehensive measure of organizational strength, I take it to be a
reasonable proxy for this study.
Data Source: NCCS IRS Business Master Files
Data for this variable was derived from the Urban Institutes National Center for
Charitable Statistics (NCCS) IRS Business Master Files (BMF). The BMF is a
cumulative file that contains basic descriptive information for active (in this sense
active means currently providing services) tax-exempt organizations released twice a
29


year. Data for this study are derived from the first data set in 2004 (BMF 04/2004) and
2005 (BMF 07/2005).
Unfortunately, the BMF does contain a few inactive organizations. Every three
years, the IRS mails postcards to organizations to verify that they exist. This reduces the
number of defunct organizations included in the database, but those that have gone
defunct since the last postcard mailing may be included. However, because I
operationalized this variable using organizational revenue and assets, defunct
organizations shouldnt overly affect the data (as truly defunct organizations would have
reported no revenue or assets).
The data for the BMF is drawn from IRS forms 1023, 1024, 990 and 990-EZ for
all active and registered tax-exempt organizations. Form 1023 is used by 501(c)(3)
organizations, and Form 1024 is used by other 501(c) organizations, to apply for
recognition of their tax-exempt status. 501(c) refers to the section of the United States
Internal Revenue Code (26 U.S.C. § 501(c)) that defines tax-exempt organizations. Form
990 is the annual IRS return required for tax-exempt organizations, which includes a
variety of financial information. Form 990-EZ is the short form of the 990, filed by
organizations with less than $100,000 in gross receipts and less than $250,000 in total
assets at the end of the year.
Twenty-eight types of nonprofit organizations are exempt from some federal
income taxes (the type of exemption depends on the type of organization), the most
prevalent of which are 501(c)(3)s, which are charities or private foundations. Charities,
in this sense, fit a rather broad set of organizations, including those organized for the
relief of the poor, the distressed or the underprivileged; advancement of religion;
30


advancement of education or science; erecting or maintaining public buildings,
monuments or works; lessening the burdens of government; lessening neighborhood
tensions; eliminating prejudice and discrimination; defending human and civil rights
secured by law; and combating community deterioration and juvenile delinquency. Of
these organizations, those with gross receipts under $5,000 and religious congregations
are not required to register with the IRS. These organizations, unless they voluntarily
chose to register, are not included in this database (NCSS, 2006).
The other type of 501(c)(3) organization, private foundations, are all required to
register with the IRS. Of these, most are grantmaking foundations that were set up by
an individual or family to provide grants to other 501(c) organizations (though a minority
of grantmaking foundations also provide scholarships or support government activities).
Most of these (more than 97%) have no paid staff and exist solely to provide small grants
to operating nonprofits. Additionally, a small portion of private foundations are
operating foundations, which use their endowment to directly fund their own
programming.
The remainder of the BMF file consists of a variety of exempt organizations,
including: (a) civic leagues, social welfare organizations and local associations of
employees (501(c)(4)s); (b) labor, agricultural and horticultural organizations
(501(c)(5)s); (c) business leagues, chambers of commerce, real estate boards
(501 (c)(6)s); and (d) many organization not relevant to this study (discussed in detail
below).
31


Data Reduction
The BMF files contain a variety of exempt organizations that are not relevant for
my study and were thus removed. This process was done in several steps, the first of
which was to eliminate non-relevant organizations based on their IRS code. Table III.2
below shows the types of organizations that were removed and those that were kept.
Table III.2 Step 1 in Data Reduction for the BMF Master Files: IRS Code.
Removed Kept
501(c)(1) 501(c)(16) 501(c)(27) 501(c)(3)
501(c)(2) 501(c)(17) 501(d) 501(c)(4)
501(c)(7) 501(c)(18) 501(e) 501(c)(5)
501(c)(8) 501(c)(19) 501(f) 501(c)(6)
501(c)(9) 501(c)(20) 501(k)
501(c)(10) 501(c)(21) Charitable risk pools
501(c)(l 1) 501(c)(22) Farmers cooperative
501(c)(12) 501(c)(23) Qualified tuition program
501(c)(13) 501(c)(24) 527
501(c)(14) 501(c)(25)
501(c)(15) 501(c)(26)
Additionally, grantmaking organizations (a particular type of 501(c)(3)) were removed
from the database so as not to count dollars twice (i.e., not prior to, and after, a grant has
been distributed).
32


With the above removals, the following types of organizations remained in the
data set:
501(c)(3) charities and operating foundations only
501(c)(4) civic leagues, social welfare organizations and local associations of
employees
501(c)(5) labor, agricultural and horticultural organizations
501(c)(6) business leagues, chambers of commerce, real estate boards
I then further paired down these data by eliminating health and education organizations,
as well as those with purposes/missions that are not relevant to domestic social justice.
To do this, I used the National Taxonomy of Exempt Entities Core Codes (NTEE-CC)
classifications assigned to each organization in the database.
The NTEE-CC is a classification system used by the IRS and NCCS to organize
entities by purpose, type or major function. Organizations are divided into 12 major
groups and 26 major NTEE groups based on their broad topical subsector (i.e., education,
medical research, housing and shelter, etc.). The 26 major groups are further subdivided
by decile codes, which represent specific activity areas (e.g., Civil Liberties is a decile
subdivision of the Civil Rights, Social Action and Advocacy Group). Some of these
decile codes are again subdivided by centile codes that represent specific types of
organizations (e.g. Censorship, Freedom of Speech & Press is a centile subdivision of
Civil Liberties). Each of these is further subdivided by seven common codes:
01 alliance/advocacy organizations
02 management and technical assistance
03 professional societies/associations
33


05 research institutes and/or public policy analysis
11 monetary support single organization
12 monetary support multiple organizations
19 nonmonetary support not elsewhere classified (N.E.C.)
Using these NTEE-CC classifications, I was able to eliminate additional organizations
not relevant to my study. First, I eliminated the entirety of six of the 12 major groups that
were associated with health, education the environment or international work. Table III.3
shows the major groups that were removed and those that remained.
Table III.3 Step 2 in Data Reduction for the BMF Master Files: Major Group (12).
Removed Kept
Higher education (BH) Arts, culture & humanities (AR)*
Education (ED) Environment (EN)*
Hospitals (EH) Human services (HU)
Health (HE) Mutual benefit (MU)
International (IN) Public & societal benefit (PU)
Religion (RE) Unknown (UN)
* Note: Much of the arts, culture and humanities code is not relevant, but a few of these
organizations do engage in social justice through cultural means. The same goes for the
environmental code, as some organization focused on environmental justice overlap
substantially with social justice. As such, I retained these codes and thinned out the non-
relevant organizations in subsequent reductions.
Next, I eliminated the entirety of 14 major NTEE groups that were not relevant to
this study because they were either not relevant to social justice generally (e.g., sports
clubs) or the entire category fell within one of the social justice categories excluded from
34


this study: health, education, children, criminal justice or international work. Table III.4
shows the major NTEE groups that were eliminated and those that remained.
Table III.4 Step 3 in Data Reduction for the BMF Master Files: Major NTEE
Groups.
Removed Kept
Education (B)
Environment (C)
Animal-Related (D)
Health Care (E)
Mental Health & Crises Intervention
(F)
Diseases, Disorders & Medical
Disciplines (G)
Medical Research (H)
Public Safety, Disaster Preparedness
& Relief (M)
Recreation & Sports (N)
Youth Development (O)
International, Foreign Affairs &
National Security (Q)
Science & Technology (U)
Religion-Related (X)
Mutual & Member Benefit (Y)
Unknown (Z)
Arts, Culture & Humanities (A)
Crime & Legal-Related (I)
Employment (J)
Food, Agriculture & Nutrition
(K)
Housing & Shelter (L)
Human Services (P)
Civil Rights, Social Action &
Advocacy (R)
Community Improvement &
Capacity Building (S)
Philanthropy, Voluntarism &
Grantmaking Foundations (T)
Social Science (V)
Public & Societal Benefit (W)
Finally, I used decile and centile codes to eliminate additional organizations not
relevant to the study. Figures III. 1 and III.2 below show the decile and centile codes that
were removed in red, and those that were not in green.
35


NATIONAL TAXONOMY OP EXEMPT ENTITIES CORE CODES (NTEE-CC) CLASSIFICATION SYSTEM (rev. Mny 2009)
Figure III.l NTEE-CC Decile/Centile Code Eliminations.
U>
G\


NATIONAL TAXONOMY OP EXEMPT ENTITIES CORE CODES (NTEE-CC) CLASSIFICATION SYSTEM (r*v. May 2009)
Figure III.2 NTEE-CC Decile/Centile Code Eliminations.
U>


With these reductions, the remaining organizations include all the registered tax-
exempt organizations in 2004 and 2005 that are relevant for my study. The data set thus
includes larger formal organizations, but it does not include exceptionally small
organizations, nor does it include informally organized groups (i.e., loose networks of
individuals without a formal organization). Further, it does not fully capture the people
power of the organizations included in the data set (i.e., people who support the
organizations or their cause in non-financial terms). Despite these limitations, I take
these data to be an appropriate proxy for the SMIs active in the United States during 2004
and 2005.
Differentiating SMOs and Interest Groups
In this study, I take the position that interest groups are political insiders and
SMOs are political outsiders. From this perspective, interest groups are political actors
that are embedded within the political structure itself and as such pursue their policy
agendas through institutionalized means such as lobbying. In contrast, SMOs are
organizations external to the political structure that pursue their policy agendas through
extra-institutional means such as boycotts. Unfortunately, this distinction is rarely sharp,
as many organizations engage in both institutional and extra-institutional activities (Snow
& Soule, 2009).
To make this distinction, I used the Internal Revenue Code (IRC) 501(c)
designation. Generally, the legal parameters of 501(c) organizations' participation in
political activity are determined by the IRC, Treasury regulations and IRS guidance.
These parameters restrict the political activity of 501(c)(3) organizations in two ways: (1)
38


They are not allowed to intervene in political campaigns and (2) they can only conduct an
insubstantial amount of lobbying. Lobbying includes activities that attempt to
influence legislation by (a) directly contacting, or urging the public to contact, legislators
about proposing, supporting or opposing legislation; (b) advocating for or against
legislation; or (c) making contributions or loans to other entities that engage in these
activities. As such, these restrictions cover both formal lobbying (directly contacting
governmental officials) and grassroots lobbying (appeals to the public to contact their
public officials) (Lunder, 2007).
Though 501(c)(3)s are not prohibited from engaging in lobbying (as they are from
becoming involved in political campaigns), they are restricted in the amount they can
lobby. Defining this amount can happen one of two ways: (1) The organization can elect
adhere to the numerical standards provided by IRC § 501(h) or (2) they can decide to
have the government apply a "no substantial part" test, the standards of which are based
primarily in case law. Under IRC § 501(h), an organizations formal direct lobbying
efforts are limited to 20% of its first $500,000 of expenditures, 15% of its second
$500,000, 10% of its third $500,000 and 5% of its remaining expenditures, the total of
which cannot exceed $1 million in the year. Additionally, grassroots lobbying is limited
to 5% of its first $500,000 of expenditures, 3.75% of its second $500,000, 2.5% of its
third $500,000 and 1.25% of its remaining expenditures, not to exceed $250,000 on
grassroots lobbying in the year. Organizations that do not elect to be subjected to IRC §
are subjected to a no substantial part test, which takes each case as unique and makes
the determination through a broad examination of the organizations purpose and
39


activities. However, case law suggests that no substantial part typically amounts to
between 5% and 20% of the organizations expenditures (Lunder, 2007).
So organizations with a 501(c)(3) status are legally (1) prohibited from engaging
in campaign politics and are (2) legally required to limit their lobbying (both direct and
grassroots) to 20% or less of their expenditures. If they violate either of these provisions,
they are subject to the loss of their tax-exempt status. However, it is no doubt true that
laws are not always followed and that violators are not always caught. Thus, compliance
rates with these laws are just as important as the laws themselves. Though relatively
scant, the available data suggest that large-scale noncompliance involving substantial
expenditures is rare (Mayer, 2010). As such, the restrictions on political activity imposed
on 501(c)(3)s means that they are largely, though perhaps not entirely, external to the
political structure. Thus, by my definition, the IRCs 501(c)(3) designation is a
reasonable indicator that an organization is an SMO and not an interest group.
In contrast, the remaining organizations in my data set (501(c)(4)s, 501(c)(5)s and
501(c)(6)s) are not explicitly restricted in the amount of expenditures they can devote to
(1) intervening in political campaigns or to (2) lobbying (both direct and grassroots).
They are, however, implicitly limited by the restrictions on their primary purpose. That
is, these organizations can engage in unlimited campaign activity as long as it is
consistent with its primary purpose, but campaign activity itself cannot be its primary
purpose (organizations with this as their purpose are 527s). Additionally, these
organizations can engage in unlimited lobbying as long as the lobbying is related to the
organizations primary purpose. Further, lobbying can be these organizations sole
activity as long as the purpose of the lobbying is a qualified tax-exempt purpose.
40


So, 501(c)(4)s, 501(c)(5)s and 501(c)(6)s are interest organizations that face no
legal restrictions to lobbying for their interest. Thus, registered exempt interest
organizations directly engaged with the political structure through direct or grassroots
lobbying as a primary activity (or political insiders) fall within one of these categories.
However, despite their legal ability to do so, not all organizations within these categories
actively lobby.
Luckily, there is an additional reason to think that organizations within this
category fall closer to the political insider pole of the spectrum than do 501(c)(3)
organizationsthey are not 501(c)(3)s. Unlike the organizations within these categories,
contributions to 501(c)(3)s are tax deductible as charitable contributions. Thus, it stands
to reason that most 501(c)(4)s, 501(c)(5)s and 501(c)(6)s would choose to be 501(c)(3)s
if they could. However, these organizations cannot be 501(c)(3)s because their primary
purpose is not consistent with the legal definition of a charity. Rather, by the IRC
definitions, these organizations represent particular interests (as interest groups).
Because these organizations by definition represent particular interests and because they
are allowed to expend resources on political activities without restriction, I believe they
are largely political insiders. In other words, I take a 501(c)(4), 501(c)(5) or 501(c)(6)
designation to indicate than an organization is an interest group.
A few examples may help to illuminate the distinction I made here. First, here are
examples of 501(c)(3) organizations:
The American Red Cross
People for the Ethical Treatment of Animals Inc. (PETA)
Nature Conservancy
41


Each of the above organization may, at times, act politically. However, I contend that
they fit the idea of an SMO, which is not embedded within the political structure. In
contrast, I argue that each of the following organizations more closely represent interest
groups that are intimately intertwined with the political structure:
National Rifle Association (NRA) (a 501(c)(4))
International Brotherhood of Teamsters (a 501(c)(5))
Association of Trial Lawyers of America (a 501(c)(6))
Here it is important to point out that exempt organizations with different designations are
sometimes closely related to each other. For example, a 501(c)(3) might be paired with a
501(c)(6). Such is the case with the American Bar Association (a c6 trade association)
and the American Bar Association Endowment (a c3 charitable fund that supports the
public service and educational programs of the American Bar Association).
While such organizations are in a certain sense part of the same global entity, they
are technically and legally separate organizations. As part of this legal separation, their
activities and funds must be kept separate. As such, despite their relationship, related
organizations with separate designations are separate entities within my data sets.
Through the processes described above, I established two subsets of the BMF
data: (a) SMOs identified as organizations with a 501(c)(3) designation and (b) interest
groups identified as organizations with a 501(c)(4), 501(c)(5) or 501(c)(6) designation.
The data usedfor this variable (SMIs) include only 501(c)(3)s, with 501(c)(4)s,
501(c)(5)s or 501(c)(6)s included in the next variable (interest groups).
42


Determining Directionality and Assigning Values
To assign values for this variable, I used the NTEE-CC codes to identify
organizations for each major social category listed in Table III. 1 above. When possible, I
also used these codes to identify directionality. For example, I assumed that labor unions
(NTEE-CC code J40) were pro-organized labor and pro-wage increases. However, such
clean designations were not always possible. For these less clear NTEE-CC codes, I used
Guide Star and organizational websites to identify their primary purpose/mission. Using
this information, I attached each organization to all relevant social categories as either pro
or anti. I then calculated four values for each major and minor category: (a) 2004 pro-
bill, (b) 2004 anti-bill, (c) 2005 pro-bill and (d) 2005 anti-bill. These values were
calculated as follows:
W = Sum of organizational monetary values (in billions) for SMOs that
are in favor of the category in 2004
X = Sum of organizational monetary values (in billions) for SMOs that are
against of the category in 2004
Y = Sum of organizational monetary values (in billions) for SMOs that are
in favor of the category in 2005
Z = Sum of organizational monetary values (in billions) for SMOs that are
against of the category in 2005
2004 pro-bill value = W X
2004 anti-bill value = X W
2005 pro-bill value = Y Z
2005 anti-bill value = Z Y
43


These values are calculated in billions to ensure any changes in this variable that are
significantly correlated with bill passage lead to a perceptible corrected logit coefficients
within the rare events logistic regression models.10
Independent Variable #2: Interest Groups
This variable is meant to control for the influence of interest group strength on the
probability of bill passage. It is calculated in precisely the same manner as SMI strength
(independent variable #1 above), except it includes 501(c)(4)s, 501(c)(5)s and 501(c)(6)s
(interest organizations that face no legal restrictions to lobbying for their interest) rather
than 501(c)(3)s (interest organizations that face substantial legal restrictions on
lobbying). That is, using the NCCS BMF as a data source, each bill is associated with
three separate values for this variable: (a) the total annual revenue for interest groups
likely to be for the bill less the total revenue for interest groups likely to be against the
bill; (b) the same calculation, using the total end-of-year assets rather than revenue; and
(c) these two values (annual revenue and end-of-year assets) combined. For more detail
on this variable, see the calculations for SMIs (independent variable #1) above.
Independent Variable #3: Public Opinion
This variable is meant to measure the influence of public opinion on the
probability of bill passage. More specifically, it measures the effect of favorable opinion
on bill passage. As this measure increases, I expect the probability of bill passage will
10 If I were to use dollars, rather than billions of dollars, the corrected logit coefficients
would be so small they would be imperceptible. Put another way, the difference each
additional dollar makes is very small, but the difference each additional billion dollars
makes is rather large. Thus, to see the impact of SMI and interest group dollars, we need
units large enough to see the impactin this case, billions of dollars.
44


also increase. That is, as the percentage of the public in favor of a bill increases, I expect
bill passage will be more likely.
This variable was operationalized as public opinion in favor of the concept
underpinning the bill, with data derived from nine sources:
Associated Press/Ipsos poll, conducted:
o January 9-12, 2006
CNN Poll, conducted:
o August 30-September 2, 2006
o December 15-17, 2006
Gallop Poll, conducted:
o August 9-11, 2004
o March 18-20, 2005
o April 29, 2005
o August 8-11, 2005
o May 8-11, 2006
General Social Survey, conducted:
o 2004
o 2006
NBC News Poll, conducted:
o April 3-5, 2005
NBC News/Wall Street Journal, conducted:
o June 9-12, 2006
Pew Research Center for the People and the Press Survey, conducted:
45


o March 8-12, 2006
Rasmussen Reports, conducted:
o June 7-8, 2006
USA Today/Gallop Poll, conducted:
o November 19-21, 2004
o April 28-30, 2006
All opinion data were normalized by converting the scale to a percentage in favor of the
bill. Where public opinion data existed in multiple sources, the survey that was most
proximate to the final action date of the bill was used.
For example, for bills related to constitutional amendments around same-sex
marriage, I used data from the Gallop Poll question, Would you favor or oppose a
constitutional amendment that would define marriage as being between a man and a
woman, thus barring marriages between gay or lesbian couples? with three response
options: Favor, Oppose and Unsure. Thus, for bills to get a constitutional
amendment that bans gay marriage, the public opinion value corresponds to the
percentage of respondents who selected Favor. Conversely, any bill supporting an
amendment legalizing gay marriage would have a value corresponding with the
percentage of respondents who selected Opposed.
However, because this question was asked multiple times from 2004 to 2006, the
value varies slightly for virtually identical bills presented at different times. For example,
H.J.RES.39, which has the title Proposing an amendment to the Constitution of the
United States relating to marriage, received a value of 0.53, because the public opinion
data closest to its final action date (4/4/2005) indicated that 53% of the public would
46


support this bill. Meanwhile, a virtually identical billThe Marriage Protection
Amendment (S.J.RES.l)received a value of 0.5, because the public opinion data
closest to its final action date (6/7/2006) indicated that 50% of the public would support
this bill.
For bills with underlying concepts that lack opinion data, I assigned a value of 0.5
or 50%, representing evenly split public opinion relative to the bill. This is not to say that
when presented with the bill people in the public would not form an opinion. Rather, the
0.5 value indicates that the political representatives voting for or against a bill have no
reason to believe the public generally favors or generally opposes the bill. I believe this
is justified because, as Paul Burstein (2010) points out, public opinion polls tend to target
subjects that people have opinions about. That is, people dont know enough about many
subjects of legislation to have formed an opinion on them. Further, even if they have
developed an opinion in sufficient numbers for there to be a public opinion, if it is not
measured, its difficult to see how the opinion could influence lawmakers, as they would
have no sense of what the opinion was.
Conversely, public opinion is measured for major concepts that are in the public
discourse, and as such, about which people are likely to hold an opinion. So, concepts
that have public opinion data are likely to equate to (a) the major concepts on which
public opinion exists, and (b) by definition, are the concepts with explicitly visible public
opinion. As such, I believe it is reasonable to assign a value of 0.5 for this variable to all
bills with concepts that lack public opinion data.
47


Independent Variable #4: Political Elite Allies
This variable is meant to measure the influence of political elites (legislators) on
the probability of bill passage, within the partisan structure of the 109th Congress.
Because Republicans controlled both the House and the Senate in the 109th Congress,
bills favored by Republicans had a better chance of passing than those favored by
Democrats. Bills favored by both parties had the strongest chance to pass. Thus, as bills
move from Democrat-favored, to Republican favored, to bipartisan, I expect the
probability of bill passage will increase.
Following this logic, this variable was operationalized as one of four values:
0 = Democrat bill
1 = Republican bill
2 = Weak bipartisan bill
3 = Strong bipartisan bill
To assign these values, I first aligned each bill with the party of the sponsoring legislator.
I then recorded the party and count of all cosponsors. Next, I coded Bills without any co-
sponsors from the opposing party as either a Democratic bill (0) or a Republican bill (1).
For bills that had both Democratic and Republican co-sponsors, I used a set of formulas
based on the relative ratio of Democrats and Republicans in each chamber, to designate
the bill as either weakly or strongly bipartisan.
The Senate of the 109th Congress had 44 Democrats, 55 Republicans and 1
Independent (James Jeffords). However, because Representative Jeffords caucused with
the Democrats, I counted him as a Democrat for the purposes of this study. Thus, for a
bill to pass, at least six Republicans have to vote for the bill (45 Democrats + 6
48


Republicans = 51, or a simple majority). Conversely, a bill favored primarily by
Republicans did not need any Democrats to pass (because there were more than 51
Republicans).
However, Senate rules dont place time limits on debate, and thus any single
Senator can delay or entirely prevent a vote on legislation by extending debatea
parliamentary procedure known as the filibuster. A filibuster can be overcome if three-
fifths of the Senate agrees to do so by invoking cloture. So, while Senate bills can pass
with 51 votes, 60 votes are necessary to pass a bill that is being filibustered. Thus, a bill
favored by Democrats would require at least 15 Republicans to be filibuster proof (45
Democrats +15 Republicans = 60). Conversely, a bill favored primarily by Republicans
would need at least five Democrats to be filibuster proof (55 Republicans + 5 Democrats
= 60).
So a primarily Democratic bill could not pass without Republican support, while a
Republican bill could pass without any Democratic support. However, if a Republican
bill were filibustered, the Republicans would need Democratic support to overcome the
filibuster, and if a Democratic bill were filibustered, the Democrats would need even
more Republican support to overcome it. There are different levels of bipartisanshipa
weak bipartisan bill that has enough support from each party to pass but not to overcome
a filibuster, and a strong bipartisan bill that has enough support from each party to
overcome a filibuster. But the number of opposition support necessary depends on
whether the bill is primarily a Democratic or a Republican bill (because there were 10
more Republicans than Democrats in the 109th Senate). Thus, in my definition,
bipartisanship is defined relative to the level of support from the opposing party
49


necessary to either (a) pass (weak bipartisanship) or (b) overcome a filibuster (strong
bipartisanship).
To make these distinctions, I used co-sponsorship as an indicator of the level of
opposing party support. Though some studies have found that co-sponsorship generally
(i.e., the presence or absence of co-sponsorship) has only a minor effect on bill passage
(Wilson & Young, 1997), I take it to be an accurate measure of the partisan nature of a
bill. For example, bills sponsored by Democrats are considered weakly bipartisan when
they have enough Republican cosponsors to pass (six or more) or when the ratio of
Republican co-sponsors over Democratic co-sponsors is greater than or equal to the
number of Republicans needed for passage divided by the number of votes needed for
passage (6/51 or 0.12). Thus, the formulas for assigning the value to this variable for
Senate bills are as follows:
DS = Democratic Sponsor
DCS = Democratic Co-sponsor
RS = Republican Sponsor
RCS = Republican Co-sponsor
Strong Bipartisan =
DS & >15 RCS
or
DS & RCS/DCS > 0.25
or
RS & >5 DCS
or
50


RS & DCS/RCS > 0.08
Weak Bipartisan (2) =
DS & >6 RCS
or
DS & RCS/DCS > 0.12
or
RS & >1 DCS
or
RS & DCS/RCS > 0.01
Republican (1) =
RS & 0 DCS
Democratic (0) =
DS & <5 RCS
While the House rules do not allow filibustering, I used the same thresholds to assign the
values for House bills (though the numbers were based on the numbers in the House).
For a variety of reasons, the number of Democratic and Republican representatives varied
slightly over the two years the 109th Congress was in session. The numbers used here are
from December 31, 2006, to January 3, 2007. I chose to use these numbers because they
represent the smallest gap between Democrats and Republicans during the 109th, and thus
represent the point at which the partisan structure of the house should have had the
weakest effect. As such, any observed effect is more likely to be real.
During this time period, there were 229 Republicans, 202 Democrats and one
Independent. However, because the Independent (Bernie Sanders) caucused with the
51


Democrats, I included him with the Democrats. Using these numbers, the formulas for
House bills are as follows:
Strong Bipartisan(3) =
DS & >59 RCS
or
DS & RCS/DCS > 0.23
or
RS & >29 DCS
or
RS & DCS/RCS >0.11
Weak Bipartisan (2) =
DS & >16 RCS
or
DS & RCS/DCS > 0.07
or
RS & >1 DCS
or
RS & DCS/RCS > 0.01
Republican (1) =
RS & 0 DCS
Democratic (0) =
DS & <15 RCS
52


Here it is important to note that I use sponsorship and co-sponsorship as
indicators of the partisanship of the bill, but I do not consider the total number of
co-sponsors. I take this position because, as Wilson & Young (1997) have
demonstrated, the raw number of co-sponsors a bill has is not substantially related
to bill passage.
Independent Variable #5: Media Coverage
This variable is meant to measure the influence of media coverage on the
probability of bill passage. Due to the logistical difficulties involved, I only
measured the amount of coverage, not the type. That is, a qualitative evaluation
of many thousands of articles is far more time consuming and resource intensive
than a raw count. Thus, I expect the probability of bill passage to increase as
media coverage increases.
This variable was operationalized as the number of New York Times
articles in which the (a) bill number, (b) the name of the bill and/or (c) up to three
sets of keywords representing the concept underpinning the bill appeared for one
year prior to the last action taken on the bill. The value for this variable, then, is
the total number of articles, which met these criteria. To arrive at these values, I
entered the appropriate search criteria into LexisNexis Academic using Boolean
terms and connectors and limiting each search to the New York Times and a one-
year period prior to the last action taken on the bill.
The New York Times was selected because it is one of the largest nationally
distributed newspapers not located in Washington, D.C. As would be the case with any
53


single newspaper, the New York Times will not be perfectly representative on media
coverage generally, but I take it to be an acceptable proxy.
Analysis: Rare Events Logistic Regressions (ReLogits)
My analysis is designed to measure the effect of SMI strength (operationalized
through aggregate SMO revenue, assets or a combination of the two) on the probability
of bill passage, controlling for interest groups, public opinion, political elite support and
media coverage. Typically, when scholars attempt to run such an analysis, they would
use a logistic regression a statistical technique used to predict the probability that an
event will occur. It does so by fitting a logit function logistic curve to data provided by
several independent (or predictor) variables. This allows analysts to determine how
much each independent variable contributes to the probability of the event when
controlling for other variables (Agresti, 2007).
Sample size is an important consideration in logistic regression. Agresti (2007)
argues that a minimum of 20 cases per independent variable is necessary. Because I have
five independent variables, to meet this threshold I would need a minimum of 100 cases
(20*5) for each data set. However, following Peduzzi et al. (1996) and Agresti (2007), at
least 10 events (in this case bill passage) per independent variable are necessary. This
means I would need a minimum of 50 bills that passed (10*5 IVs) to run a standard
logistic regression.
This presents a problem for my study because bill passage is a relatively rare
eventroughly 20% of the bills introduced in the 109th Congress passed their respective
54


chamber.11 As such, my data sets would each need a minimum of 250 cases each to run a
standard logistic regression. Given the scope of this study, this would require a
substantial amount of work. However, I was able to reduce the size of my databases,
while at the same time improving my statistical rigor, by using the rare events logistical
regression, or relogit, technique developed by Gary King and Langche Zeng (2001a,
2001b).
King and Zeng developed the relogit technique by adapting econometric and
epidemiology techniques to handle exceptionally rare events in political sciencetens of
events in thousands of casesas well as small sample sizes (fewer than 200
observations). The technique is designed to overcome three major problems that arise in
logistic regression with rare events and/or small samples: (a) logit coefficients are biased
in small samples of fewer than 200 observations; (b) even with large sample sizes,
estimated event probabilities of rare events are always too small; and (c) for rare-events
data, the most common method of computing probabilities of events in logistic
regressions leads to errors in the same direction as biases in the coefficients (King
&Zeng, 2001b).
The relogit technique overcomes these problems with mathematical corrections12
and is performed in virtually the same way as a logistic regression. However, unlike a
logistic regression, a rare events logistic regression is an unbiased estimator, rather than a
likelihood technique. As such, the output of a relogit is slightly different than a logit.
11 In 2005, 19.3% of the bills introduced passed in the Senate and 22.9% in the House. In
2006, these numbers were 21.5% and 21.1%, respectively.
12 For the mathematical proofs of the ReLogit technique, see King and Zeng (2001a); for
a less technical account with examples, see King and Zeng (2001b).
55


Instead of logit coefficients and standard errors, relogits compute bias-corrected
coefficients and robust standard errors. Analysts calculate quantities of interest from
these by setting each independent variable (IV) at specific values. In particular, the
probability of the dependent variable (DV) can be calculated with each IV set at specific
values. Additionally, relative risks can be calculated. Relative risks are the likelihood
(or risk) of an event, given chosen values of the explanatory variables, relative to a
baseline of these values.13 Finally, because a relogit simultaneously reduces bias and fit,
pseudo-R2 is typically not reported as it is taken for granted that the reduction in bias
comes at the cost of a reduction in fit. However, because most readers will be more
familiar with logit models than relogit models, I report pseudo-R2 from standard logit
models with the caveat that goodness of fit is not a primary consideration in relogit.
In my analysis, I begin by providing descriptive statistics for each variable. I then
provide the results of relogits in each data set (redistribution, recognition and combined)
with bill passage as the dependent variable, and each independent variable alone. Next, I
display the results of full relogit models on each data set with all independent variables
included in the models. I then investigate how the probability of bill passage shifts with
changes in each independent variable. To do so, I graph the change in the probability of
bill passage as each independent variable shifts from its minimum value to its maximum
value, while holding all other independent variables at their mean. I also run relative
risks for each variable, from their minimum value to their maximum value and from their
13 Relative risk is, in some ways, similar to odds-ratio, which is more common in Logit
models. However, the two are distinct statistical concepts. Where p is the probability of
an event under one set of circumstances, and q is the probability of the event under a
second set of circumstances, relative risk is p/q, but odds-ratio is [p/(l-p)] / [q/(l-q)].
56


20th percentile to their 80thholding all other independent variables at their mean.
Finally, I further investigated how SMIs impact bill passage by constructing four ideal
typical scenarios:
SMIs Against the World where all IVs are set at their 20th percentile.
SMIs Riding a Tide of Support where all IVs are set at their 80th percentile.
People vs. Power where public opinion is set at its 80th percentile, while all
other IVs are set at their 20th percentile.
Key Elite Allies where political elites are set at their 80th percentile, while
all IVs are set at their 20th percentile.
These scenarios allow me to investigate the extent to which SMI influence on bill passage
is dependent upon context. In the first scenario, SMIs are operating in a context that is
not favorable to bill passage as interest group support is low, support from political elites
is weak, public opinion is unfavorable and media coverage is scant. The second scenario
flips these circumstances, with high interest group support, strong support from political
elites, favorable public opinion and substantial media coverage. The final two scenarios
investigate the relative importance of public opinion and political elite support when
combined with SMI strength. In the third scenario, public opinion is favorable, despite a
lack of interest group support, support from political elites and media coverage. In the
fourth scenario, support from political elites is high, despite a lack of interest group
support, unfavorable public opinion and very little media coverage. Thus, by
mathematically simulating a variety of scenarios, I am able to approximate the effect
context has on SMIs ability to impact bill passage.
57


Overview of Methods
For this study, I developed two distinct data sets, one associated with recognition
bills and one associated with redistribution bills, in order to determine if the relationship
between SMI and bill passage is moderated by this distinction. I then attached SMI
strength, interest group strength, political elite support, public opinion in favor and media
coverage to each bill. Next, I ran a rare event logistic regression on each data set as well
as a combined data set (the redistribution and recognition data sets combined). Finally, I
further investigated how SMIs impact bill passage by constructing four ideal typical
scenarios and calculating how the strength of SMIs correlates with the probability of bill
passage under a variety of circumstances.
58


CHAPTER IV
RESULTS
My analysis is designed to measure the effect of SMI strength on the probability
of bill passage, controlling for interest groups, public opinion, political elite support and
media coverage. To do so, I utilized the rare events logistical regression, or relogit,
technique developed by Gary King and Langche Zeng (2001a; 2001b). Relogit
overcomes the bias that occurs in logit models when the dependent variable is a rare
event, such as bill passage. For a more comprehensive account of the relogit technique,
see the methods section above.
In this section, I begin by providing basic descriptive statistics for each variable. I
then run relogits in each data set (redistribution, recognition and combined), with bill
passage as the dependent variable, and each independent variable alone. Next, I run full
relogit models in each data set with all independent variables included in the models. I
then investigate how the probability of bill passage shifts with changes in each
independent variable. To do so, I graph the change in the probability of bill passage as
each independent variable shifts from its minimum value to its maximum value while
holding all other independent variables at their mean. I also run relative risks for each
variable, from their minimum value to their maximum value and from their 20th
percentile to their 80thholding all other independent variables at their mean. Finally, I
further investigate how SMIs impact bill passage within various contexts by constructing
four ideal typical scenarios:
SMIs Against the World where all IVs are set at their 20th percentile, and
thus SMIs are operating in a context that is not favorable to bill passage.
59


SMIs Riding a Tide of Support where all IVs are set at their 80th percentile,
and thus SMIs are operating in a contest favorable to bill passage.
People vs. Power where public opinion is set at its 80th percentile, while all
other IVs are set at their 20th percentile, and thus public opinion is favorable
but all other variables are unfavorable for bill passage.
Key Elite Allies where political elites are set at their 80th percentile, while
all IVs are set at their 20th percentile, and thus political elite support is
favorable but all other variables are unfavorable for bill passage.
As such, I am able to approximate the effect of context on SMIs influence on bill
passage by mathematically simulating a variety of scenarios.
Descriptive Statistics
Dependent Variable: Social Justice Bills
In the 109th Congress, 13,072 bills were introduced (8,152 in the House and
4,920 in the Senate). Of these, 2,103 (16.1%) passed their respective chamber (upper
house/Senate, or lower house/House of Representatives). My data set contains a total of
370 bills (cases) from the 109th Congress, of which 58 (15.7%) passed their respective
chamber. As such, the rate of passage in my data set is comparable to, but slightly lower
than, for the 109th Congress as a whole.
Table IV. 1 displays how many bills in my data set were in each chamber and how
many of those passed.
60


Table IV.l Number of Bills by Chamber.
Legislature Number of bills Number that passed Percent that passed
House of Representatives 253 34 13.4
Senate 117 24 20.5
Combined 370 58 15.7
As shown in Table IV. 1, of the bills in this data set, those in the Senate passed at a higher
rate than those in the House. Given the different rules in each chamber, as well as their
different structures and ratios of Republicans to Democrats (1.13 in the House, 1.22 in
the Senate), it is plausible that this is a meaningful difference. Additionally, as
demonstrated by a biserial correlation, there is a significant correlation at the 0.1 level
(r(368) = 0.09,p = 0.082) between bill passage and the chamber in which the bill was
considered.
Another potentially important characteristic of these bills is the time frame in
which a bill is considered. It is plausible that the relative proximity to elections would
influence bill passage as legislators become more sensitive to the way their actions affect
their re-election efforts. Table IV.2 below displays the number of bills considered in
each year, by the chamber in which they were considered.
Table IV.2 Number of Bills by Year and Chamber.
Legislature Number of 2005 bills Number of 2006 2005 & 2006 Combined
House of Representatives 174 79 253
Senate 78 39 117
Combined 252 118 370
61


As shown above, about two-thirds of the bills in this data set were considered in 2005.
However, as demonstrated by a biserial correlation, within these data there does not
appear to be a correlation (r(368) = 0.06,p = 0.284) between bill passage and the year in
which the bill was considered. As such, the year in which a bill was considered does not
matter in and of itself relative to bill passage.
Yet a biserial correlation demonstrated that the distinction between redistribution
and recognition is correlated with bill passage (r(209) = 0.20, p = 0.003). Table IV.3
below displays the breakdown of bills by redistribution and recognition.
Table IV.3 Number of Bills.
Data set Number of bills Number that passed Percent that passed
Redistribution 159 10 6.3
Recognition 211 48 22.7
Combined 370 58 15.7
As shown above, this data set contains more recognition bills than redistribution bills
(43% are redistribution bills, 57% are recognition bills). Additionally, recognition bills
passed at a substantially higher rate, with just over one-fifth (22.7%) of recognition bills
passing through their chamber, compared with just 6.3% of the redistribution bills.
Further, as described above, there is a statistically significant correlation between a
redistribution/recognition designation and bill passage. This is the first evidence that the
redistribution/recognition distinction matters in terms of the likelihood of bill passagea
theme that we will see repeatedly played out in these findings.
62


Independent Variable #1: Social Movement Industries (SMIs)
As described in the Methods section above, I measure the strength of SMIs using
three different monetary measures: annual revenue, assets and revenue plus assets. For
each of these, I subtract the total value for all SMOs likely to be against the bill, from the
total value of SMOs likely to be in favor of the bill. I then divide this value by one
billion to arrive at the value for this variable (a more entailed description can be found in
the Methods section above). Thus, the SMI variable is a measure of financial resources
for all SMOs likely to be in favor of a bill, minus the financial resources for all SMOs
likely to be against a bill.
Table IV.4 below displays some basic descriptive statistics for SMO strength
measured by revenue.
Table IV.4 Descriptive Statistics for SMI Revenue ($ in Billions).
Data set Median Mean(SD) 20th 80th
Percentile Percentile
Redistribution (n= 159) 0.12 0.42(1.680) -0.68 0.77
Recognition (n =211) 0.29 0.35(0.781) 0.10 0.33
Combined (n = 370) 0.24 0.38(1.248) -0.06 0.44
Note: The values for this variable consist of the combined dollar value (in billions) of the
revenue for SMOs likely to be in favor of a bill, minus the combined dollar value (in
billions) of the revenue for SMOs likely to be against the bill. As such, negative values
occur when the revenue for the SMOs likely to be against a bill exceed those likely to be
for a bill.
As shown above, values in the redistribution dataset fall within a wider range than do
values in the recognition dataset (compare the standard deviations and the range between
the 20th and 80th percentile values). Additionally, while the mean value for the
redistribution dataset is higher than the mean value for the recognition dataset, an
63


independent samples t-test indicated that this difference is not statistically significant
(7(209) = 0.50,/) = 0.616). This suggests that the values within the recognition data set
are generally less extreme than those in the redistribution datasetthat is, the values are
generally closer to 0but that we cannot say that the difference their means is real.
These findings are mirrored by Table IV.5 below, which displays descriptive
statistics for SMI strength measured by assets.
Table IV.5 Descriptive Statistics for SMI Assets ($ in Billions).
Data set Median Mean(SD) 20th 80th
Percentile Percentile
Redistribution (n= 159) 0.12 0.88(4.567) -1.28 1.03
Recognition (n = 211) 0.51 0.41(0.518) 0.08 0.63
Combined (n = 370) 0.41 0.61(3.022) -0.17 0.73
Note: The values for this variable consist of the combined dollar value (in billions) of the
assets for SMOs likely to be in favor of a bill, minus the combined dollar value (in
billions) of the assets for SMOs likely to be against the bill. As such, negative values
occur when the assets for the SMOs likely to be against a bill exceed those likely to be
for a bill.
As shown above, the range for values in the redistribution data set is larger than the range
for the recognition data set. Additionally, though the mean value is higher in the
redistribution dataset, this difference is not statistically significant (7(161) = 1.28, p =
0.201).
A similar pattern is apparent for SMI strength measured by revenue + assets as
shown in Table IV.6 below.
64


Table IV.6 Descriptive Statistics for SMI Revenue + Assets ($ in Billions).
Data set Median Mean(SD) 20th 80th
Percentile Percentile
Redistribution (n= 159) 0.26 1.30(6.117) -1.96 1.91
Recognition (n = 211) 0.79 0.76(1.228) 0.18 0.93
Combined (n = 370) 0.68 0.99(4.117) -0.23 1.17
Note: The values for this variable consist of the combined dollar value (in billions) of the
revenue and assets for SMOs likely to be in favor of a bill, minus the combined dollar
value (in billions) of the revenue and assets for SMOs likely to be against the bill. As
such, negative values occur when the revenue for the SMOs likely to be against a bill
exceed those likely to be for a bill.
As with the previous models, values fall within a far larger range in the redistribution
data set, but the mean value is not significantly (t(168) = 1.10,p = 0.275) larger.
Additionally, for all models there is a substantial difference between the mean and
median values in the redistribution dataset, indicating that a few cases have substantially
higher values than most (and as a result, pull the mean up well beyond the median).
Together these data suggest that SMI strength varies substantially more within the
redistribution dataset, and that there are a few extreme values (values relatively far from
0) within it.
Independent Variable #2: Interest Groups
Interest group values are calculated in the same manner as the SMI variable, and
interest groups are distinguished from SMOs based on their IRS designation (for further
details on this distinction, see the Methods section above). Thus, the interest group
variable is a measure offinancial resources for all interest groups likely to be in favor of
a bill, minus the financial resources for all interest groups likely to be against a bill.
65


Basic descriptive statistics for the revenue measure of this variable are displayed in Table
IV. 7 below.
Table IV.7 Descriptive Statistics for Interest Groups Revenue ($ in Billions).
Data set Median Mean(SD) 20th 80th
Percentile Percentile
Redistribution (n= 159) 0.04 0.62(4.252) 0.00 7.04
Recognition (n = 211) 0.02 0.12(0.846) 0.00 0.04
Combined (n = 370) 0.02 0.34(2.866) 0.00 0.07
Note: The values for this variable consist of the combined dollar value (in billions) of the
revenue for interest groups likely to be in favor of a bill, minus the combined dollar value
(in billions) of the revenue for interest groups likely to be against the bill.
As shown above, there is generally more interest group support (as measured by revenue)
for bills in the redistribution data set than there is in the recognition data set.
However, the mean value is not significantly (/(167) = 1.46, p = 0.146) larger in the
redistribution data set. Additionally, interest group values in the redistribution data set
cover a much wider range than do those in the recognition data set (to see this, look at the
standard deviations and compare the difference between the 20th and the 80th percentile
values).
The assets measure of this variable lead to similar results.
66


Table IV.8 Descriptive Statistics for Interest Groups Assets ($ in Billions).
Data set Median Mean(SD) 20th 80th
Percentile Percentile
Redistribution (n= 159) 0.02 0.75(5.479) 0.00 8.20
Recognition (n = 211) 0.02 0.14(1.090) 0.00 0.02
Combined (n = 370) 0.02 0.40(3.691) 0.00 0.05
Note: The values for this variable consist of the combined dollar value (in billions) of the
assets for interest groups likely to be in favor of a bill, minus the combined dollar value
(in billions) of the assets for interest groups likely to be against the bill.
As shown above, the range for values in the redistribution data set is larger than the range
for the recognition data set. However, the mean value is not significantly (/(167) = 1.36,
p = 0.175) higher in the redistribution dataset.
Finally, the revenue + assets measure again produces similar results.
Table IV.9 Descriptive Statistics for Interest Groups Revenue + Assets ($ in
Billions).
Data set Median Mean(SD) 20th 80th
Percentile Percentile
Redistribution (n= 159) 0.06 1.37(9.719) 0.00 15.24
Recognition (n = 211) 0.04 0.27(1.936) 0.00 0.07
Combined (n = 370) 0.04 0.74(6.548) 0.00 0.12
Note: The values for this variable consist of the combined dollar value (in billions) of the
revenue and assets for interest groups likely to be in favor of a bill, minus the combined
dollar value (in billions) of the revenue and assets for interest groups likely to be against
the bill.
As before, the mean value is not significantly (t(167) = 1.41,/) = 0.161) higher in the
redistribution dataset, but they do fall within a wider range. These data suggest that,
similar to SMI strength, interest group support varies substantially more within the
67


redistribution dataset, and that there are a few extreme values (values relatively far from
0) within it.
Independent Variable #3: Public Opinion
Public opinion values represent the portion of the public in favor of the concept
associated with a bill. A value of 0.5 (indicating exactly half of the public is in favor of
the bill) was assigned to bills where public opinion data is unavailable (see the Methods
section for a more thorough description of this variable).
Table IV.10 Descriptive Statistics for Public Opinion (in Favor of Bill).
Data set Median Mean(SD) 20th 80th
Percentile Percentile
Redistribution (n= 159) 0.50 0.55(0.157) 0.48 0.77
Recognition (n = 211) 0.50 0.59(0.128) 0.50 0.73
Combined (n = 370) 0.50 0.57(0.142) 0.50 0.73
Note: Values for this variable represent the percentage of the public in favor of the
concepts that underpin the bill. When such data was not available, a value of 0.5 was
assigned.
As shown above, public opinion is slightly skewed toward public support (mean values
are slightly greater than 0.5). Additionally, an independent samples t-test indicated that
there is a statistically significant difference between mean public opinion in the data sets
(/(3 68) = -2.50, p = 0.013). This suggests that there is generally more public support for
the bills within the recognition data set.
Independent Variable #4: Political Elite Allies
This variable is operationalized as legislator support (identified by bill
sponsorship), where values progress from only minority party (Democrat) support to
strongly bipartisan support (0 = Democratic support only, 1 = Republican support only, 2
68


= weak bipartisan support and 3 = strong bipartisan support). Table IV. 11 below shows
the percent of bills that fall within each of these values in all three data sets.
Table IV. 11 Percent of Bills that Fall within each Political Elite Support (in Favor
of Bill) Category.
Data set Democrats Republicans Weak Strong
Only Only Bipartisan Bipartisan
Redistribution (n= 159) 51.6% 26.4% 4.4% 17.6%
Recognition (n = 211) 48.8% 8.5% 15.6% 27.0%
Combined (n = 370) 50.0% 16.2% 10.8% 23.0%
As shown above, roughly half of the bills in each data set have only Democratic
(minority party only) support. However, a much larger portion of recognition bills have
bipartisan support (M= 1.21, SD = 1.300) than do redistribution bills (M= 0.88, SD =
1.122), and an independent samples t-test demonstrated that this difference is statistically
significant (/(361) = -2.60, p = 0.01). This suggests that redistribution bills are generally
more partisan than recognition bills.
Independent Variable #5: Media Coverage
This variable is operationalized as the number of New York Times articles over
the previous 12 months that mention either the bill or several associated key words,
divided by 100. Thus, a value of 1 indicates that 100 articles mentioned either the bill or
keywords associated with it, in the past 12 months (see the methods section for more
detail). Table IV. 12 below displays basic descriptive statistics for this variable.
69


Table IV.12 Descriptive Statistics for Media Coverage (in 100s of Articles).
Data set Median Mean(SD) 20th 80th
Percentile Percentile
Redistribution (n= 159) 1.35 2.55(3.174) 0.19 4.81
Recognition (n = 211) 0.33 2.23(4.033) 0.02 3.12
Combined (n = 370) 0.65 2.37(3.687) 0.05 3.86
As shown above, bills in the redistribution data set generally had more media coverage
than their counterparts in the recognition data set, however this difference is not
statistically significant (t(368) = 0.82,p = 0.412).
Uncontrolled Direct Effects: Single ReLogit Models
This section contains the results of rare events logistic regressions (relogits) with
each independent variable (in all their forms) alone, such that each model contains only
two variables (bill passage as the dependent variable, and one of the independent
variables). In other words, these analyses are uncontrolled. These results are displayed
in Table IV. 13 below, broken out by results for the redistribution data set, the recognition
data set and the undifferentiated data set.
70


Table IV.13 Significance and Direction of Uncontrolled ReLogit Estimates.
Data set Redistribution (n = 159) Recognition (n = 211) Combined (n = 370)
SMI (revenue) No(+) Yes(+)** Yes(+)*
SMI (assets) Yes(+)** Yes(+)*** No(+)
SMI (assets + revenue) Yes(+)* Yes(+)*** No(+)
Interest Groups (revenue) No(-) Yes(+)*** Yes(-)*
Interest Groups (assets) No(-) Yes(+)*** Yes(-)*
Interest Groups (revenue + assets) No(-) Yes(+)*** Yes(-)*
Elite Political Allies Yes(+)*** Yes(+)*** Yes(+)***
Public Opinion Yes(-)*** Yes(+)* Yes(+)*
Media Coverage No(+) No(-) No(-)
* p< 0.10
** p< 0.05
*** p< 0.001
As Table IV.13 shows, the relationship between all but one of these variables is
statistically significant within at least one data set. In particular, the media coverage
variable is not statistically significant for any of these data sets. This suggests that the
raw quantity of media coverage associated with a bill is not correlated with that bills
passage. However, because this variable does not take directionality into account (i.e.,
whether a particular article was purely descriptive or takes a pro/anti stance on the bill), it
is possible that media coverage is correlated with bill passage, but that directionality
needs to be taken into account to measure this correlation.
Yet media coverage is alone in its lack of a significant correlation with bill
passage. Elite Political Allies and Public Opinion are significant for all data sets, while
SMIs and Interest Groups (in their various forms) are only significant in some of the data
sets. This suggests that each of these variables, when taken alone, is correlated with bill
71


passage and that the distinction between redistribution and recognition matters. While
these correlations are not controlled, this finding is consistent with my hypothesis that the
recognition/redistribution distinction has implications for the correlation between SMI
strength and bill passage.
Controlled Effect: Full ReLogit Models
To further investigate the correlation between SMI strength and bill passage, three
separate models were developed based on the three different monetary measures of SMI
and aggregate interest group strength. The general model holds bill passage as the DV,
and includes SMIs, interest groups, elite political allies, public opinion and media
coverage as IVs. The three variations on this general model use the same measures for
elite political allies, public opinion and media coverage, but different measures for SMOs
and interest groups. The models are as follows
Revenue model: Both SMI and aggregate interest group strength are measured
through revenue.
Assets model: Both SMI and aggregate interest group strength are measured
through assets.
Revenue + Assets model: Both SMI and aggregate interest group strength are
measured through the combination of revenue plus assets.
Relogits were run on each of these models within each of the three data sets
(redistribution, recognition and combined) to assess the potential for different
relationships on redistribution and recognition. Table IV. 14 below shows these results
for the Revenue Model.
72


Table IV.14 Corrected Logit Coefficient Estimates (Robust SE) for Revenue Model.
Variable Redistribution (n = 159) Recognition (n =211) Combined (n = 370)
SMI (revenue) -0.208 0.461* 0.028
(0.139) (0.274) (0.125)
Interest Groups (revenue) -0.019 -0.889 -0.062**
(0.034) (9.251) (0.029)
Elite Political Allies 0.746*** 0.481*** 0.582***
(0.209) (0.140) (0.123)
Public Opinion -2.335* 2.227* 1.609*
(1.392) (1.252) (0.949)
Media Coverage 0.164 -0.035 -0.033
(0.108) (0.082) (0.070)
Note: Unlike standard logit models, the relogit technique does not report measures of
pseudo R2 because the architects of the technique do not believe they are meaningful
measures (see the Methods section for a more in-depth description). As such, I ran a
standard logit model with these variables to calculate a Cox & Snell R2 (C&S R2) and a
Nagelkerke R2 (N R2) for this model in each data set: Redistribution C&S R2 = 0.077, N
R2 = 0.206; Recognition C&S R2= 0.123, NR2 = 0.187; Combined C&S R2= 0.083, N
R2 = 0.143. However, it should be kept in mind that the adjustments relogit makes to
reduce bias also reduce fit.
* p< 0.10
** p< 0.05
*** p< 0.01
As shown above, in the Revenue Model, both the public opinion and the elite
political allies variables are significantly correlated with bill passage in all three data sets.
However, public opinion is negatively correlated with bill passage in the redistribution
data set but positively correlated in both the recognition and combined data sets.
Additionally, interest groups are negatively correlated with bill passage in the combined
data set. Finally, SMI strength is correlated with bill passage in the recognition data set,
73


but not in the redistribution or combined data sets. I expand on these results further in
the variable specific sections below.
To ensure these results did not suffer from multicollinearity14,1 calculated
Variance Inflation Factors (VIFs)15 and tolerance16 values for this model. These are
presented in Table IV. 15 below.
Table IV. 15 VIF and (Tolerance) for Revenue Model.
Variable Redistribution (n = 159) Recognition (n = 211) Combined (n = 370)
SMI (revenue) 1.68 1.04 1.22
(0.594) (0.961) (0.823)
Interest Groups (revenue) Ml 1.04 1.12
(0.790) (0.965) (0.894)
Elite Political Allies 1.19 1.01 1.06
(0.839) (0.986) (0.940)
Public Opinion 1.29 1.06 1.09
(0.774) (0.946) (0.920)
Media Coverage 1.49 1.07 1.10
(0.673) (0.938) (0.906)
Mean VIF 1.38 1.04 1.12
14Multicollinearity is a problem that arises when independent variables are highly
correlated with each other, and as a result, it is difficult to tell which independent variable
is actually producing an effect in the dependent variable.
15 Variance Inflation Factors (VIFs) is a measure of the severity of multicolliniarity.
Many scholars consider VIFs of greater than 10 problematic (Hair, Anderson, Tatham&
Black 1995; Menard, 1995), and those greater than 4 to be potentially concerning
(Menard, 1995).
16 Tolerance is the inverse of VIF, and thus tolerances below 0.25 (or a VIF greater than
4) indicate a potential multicollinearity problem (Menard, 1995).
74


As shown above, all tolerances are well above 0.25 (or a VIF of 4), which is generally the
conservative cutoff point at which multicollinearity is considered potentially problematic
(Menard, 1995). Additionally, the highest mean VIF is 1.38, which is well below 4 (the
cutoff for concern about multicolliniarity). As such, multicolliniarity does not appear to
be a problem with these data.
The Assets Model produced similar results, though public opinion became
insignificant in both the redistribution and recognition data sets (though it remained
significant in the combined data set). The results of the relogit for the Assets Model are
displayed in Table IV. 16 below.
75


Table IV.16 Corrected Coefficient Estimates (Robust SE) for Assets Model.
Variable Redistribution (n = 159) Recognition (n = 211) Combined (n = 370)
SMI (assets) 0.048 1.363* -0.042
(0.053) (0.810) (0.032)
Interest Groups (assets) 0.002 -18.175 -0.059***
(0.025) (27.097) (0.032)
Elite Political Allies 0.697*** 0.493*** 0.617***
(0.196) (0.136) (0.120)
Public Opinion -2.056 1.746 1.984**
(1.345) (1.280) (0.949)
Media Coverage 0.151 -0.018 -0.006
(0.103) (0.066) (0.057)
Note: Unlike standard logit models, the relogit technique does not report measures of
pseudo R2 because the architects of the technique do not believe they are meaningful
measures (see the methods section for a more in-depth description). As such, I ran a
standard logit model with these variables to calculate a Cox & Snell R2 and a Nagelkerke
R2 for this model in each data set: Redistribution C&S R2 = 0.078, N R2 = 0.208;
Recognition C&S R2= 0.124, NR2 =0.188; Combined C&S R2= 0.088, NR2 =0.151.
However, it should be kept in mind that the adjustments relogit makes to reduce bias also
reduce fit.
* p< 0.10
** p< 0.05
*** p< 0.01
As shown above, the Assets Model closely mirrors the Revenue model, and once again
SMI strength is correlated with bill passage in the recognition data set but not in the
redistribution or combined data sets.
Like with the revenue model, I calculated VIFs and tolerance values for these data
to ensure these results did not suffer from multicollinearity. These are presented in Table
IV. 17 below.
76


Table IV.17 VIF and (Tolerance) for Assets Model.
Variable Redistribution (n = 159) Recognition (n = 211) Combined (n = 370)
SMI (assets) 1.56 1.32 1.18
(0.642) (0.758) (0.850)
Interest Groups (assets) 1.22 1.30 1.09
(0.820) (0.768) (0.917)
Elite Political Allies 1.16 1.01 1.05
(0.865) (0.994) (0.917)
Public Opinion 1.31 1.06 1.11
(0.761) (0.941) (0.898)
Media Coverage 1.44 1.08 1.10
(0.695) (0.924) (0.910)
Mean VIF 1.34 1.15 1.10
As shown above, all tolerances are well above the 0.2 threshold (Menard, 1995), and the
highest mean VIF is 1.34, so multicolliniarity does not appear to be a problem with these
data.
These findings are again similar in the Revenue + Assets Model, though in this
model, public opinion is significant in the recognition and combined data sets but not in
the redistribution data set. Table IV. 18 below displays the relogit results for the Revenue
+ Assets Model.
77


Table IV.18 Corrected Logit Coefficients (Robust SE) for Revenue + Assets Model.
Variable Redistribution (n =159) Recognition (n = 211) Combined (n = 370)
SMI (revenue + assets) -0.013 0.374* -0.023
(0.044) (0.218) (0.027)
Interest Groups (revenue + assets) -0.004 -3.088 -0.032***
(0.015) (7.091) (0.012)
Elite Political Allies 0.715*** 0.481*** 0.610***
(0.201) (0.140) (0.121)
Public Opinion -2.038 2.113* 1.878**
(1.363) (1.256) (0.951)
Media Coverage 0.158 -0.032 -0.012
(0.104) (0.080) (0.059)
Note: Unlike standard logit models, the relogit technique does not report measures of
pseudo R2 because the architects of the technique do not believe they are meaningful
measures (see the Methods section for a more in-depth description). As such, I ran a
standard logit model with these variables to calculate a Cox & Snell R2 and a Nagelkerke
R2 for this model in each dataset: Redistribution C&S R2 = 0.079, N R2 = 0.210;
Recognition C&S R2= 0.125, NR2 =0.190; Combined C&S R2= 0.085, NR2 =0.147.
However, it should be kept in mind that the adjustments relogit makes to reduce bias also
reduce fit.
* p< 0.10
** p< 0.05
*** p< 0.01
Once again, I calculated VIFs and tolerance values for these data as presented in
Table IV. 19 below.
78


Table IV.19 VIF and (Tolerance) for Revenue + Assets Model.
Variable Redistribution (n = 159) Recognition (n = 211) Combined (n = 370)
SMI (revenue + assets) 1.63 1.10 1.20
(0.613) (0.905) (0.830)
Interest Groups (revenue + assets) 1.24 1.11 1.10
(0.809) (0.902) (0.908)
Elite Political Allies 1.17 1.01 1.05
(0.853) (0.991) (0.948)
Public Opinion 1.31 1.06 1.11
(0.762) (0.947) (0.902)
Media Coverage 1.47 1.06 1.11
(0.679) (0.941) (0.901)
Mean VIF 1.36 1.07 1.12
As shown above, all tolerances are well above the 0.2 threshold (Menard, 1995), and the
highest mean VIF is 1.36, so multicolliniarity does not appear to be a problem with these
data.
As each of the relogit tables above demonstrate, the presence of elite political
allies is positively correlated with bill passage in all three models and across all three data
sets. Additionally, public opinion is positively correlated with bill passage for all three
models in the combined data set, and for two of the models in the recognition data set.
However, public opinion is negatively correlated with bill passage for the Revenue
Model in the redistribution data set. Similarly, interest groups are negatively correlated
with bill passage for all three models in the combined data set. Finally, in all three
79


models, SMI strength is positively correlated with bill passage in the recognition data
set, but not in the redistribution or combined data sets.
To get a better sense of how to appropriately interpret these results, I provide
graphs of probability of bill passage and relative risks17 for each independent variable
that is significantly correlated with bill passage. These results are presented in the
sections below, organized by independent variable. For each of the following sections,
only statistically significant results are displayed.
Political Elite Allies
The presence of political elite support is positively correlated with bill passage in
all three models, across all three data sets. That is, as bills progress from minority party
support only to strongly bipartisan support, the probability of bill passage increases. To
visualize this correlation, I graphed the probability of bill passage (absolute risk) as the
political elite variable increases, holding all other independent variables at their mean
levels. Figure IV. 1 below displays these results for the combined data set.
17 Relative risk is a ratio of probabilities at two different set values for the IVs. See the
Methods section for a more in-depth description of relative risk.
80


Figure IV. 1 Probability of Bill Passage by Political Elite Support for All Bills (n =
370).
Note: All other independent variables were set at their mean.
These results are further supported by Table IV.20 below, which shows the relative risk
of bill passage as political elite support progresses from its minimum to its maximum
level (RR min to max), as well as the relative risk of bill passage as political elite support
progresses from its 20th percentile value to its 80th percentile value (RR p20 to p80).
Table IV.20 Relative Risk of Bill Passage for Partisan Progression for All Bills (n =
370).
Model________________________________RR min to max____________RR p20 to p80
Revenue 4.240 4.182
Assets 4.537 4.522
Revenue + Assets 4.462 4.451
Note: All other independent variables were set at their mean.
As shown above, in the combined data set, all three models show a substantial increase in
the probability of bill passage as political elite support increases, even when controlling
for mean levels of SMI strength, interest group strength, public opinion and media
coverage. Indeed, in all three models, the probability of bill passage rises from under 8%
81


to over 32% as political elite support increases from minority party support only to
strongly bipartisan.
Similar results are found in the redistribution data set, as shown by Figure IV.2
below.
Figure IV.2 Probability of Redistribution Bill Passage by Partisan Progression (n =
159).
Note: All other independent variables were set at their mean.
Table IV.21 supports these findings.
Table IV.21 Relative Risk of Redistribution Bill Passage for Partisan Progression (n
= 159).
Model RR min to max RR p20 to p80
Revenue 7.302 4.032
Assets 6.739 3.624
Revenue + Assets 6.947 3.672
Note: All other independent variables were set at their mean.
Though the probability of bill passage is lower overall for redistribution bills, the
relationship between political elite support and bill passage remains the same. In all three
82


models, the probability of bill passage rises from under 4% to over 22% as political elite
support increases from minority party support only to strongly bipartisan.
We again see similar findings in the recognition data set.
0)
exi
re
VI
VI
re
Q.
CQ
C
re
O

a.
0 12 3
Partisian Progression
Revenue Model
Assets Model
Revenue + Assets Model
Figure IV.3 Probability of Recognition Bill Passage by Partisan Progression (n =
211).
Note: All other independent variables were set at their mean.
Table IV.22 Relative Risk of Recognition Bill Passage for Partisan Progression (n =
211).
Model RR min to max RR p20 to p80
Revenue 2.879 2.800
Assets 3.673 3.485
Revenue + Assets 3.078 3.061
Note: All other independent variables were set at their mean.
As shown above, for all three models in the recognition data set, the probability of bill
passage rises from under 16% to over 34% as political elite support increases from
minority party support only to strongly bipartisan.
83


As these data demonstrate, the probability of bill passage increases as the
presence ofpolitical elite support increases, even when controlling for mean levels of
SMI strength, interest group strength, public opinion and media coverage. Further, this
is tr ue for all three models, in all three data sets.
Public Opinion
The level of public opinion in favor of a bill is positively correlated with bill
passage in the combined data set for all three models. Figure IV.4 aids in the
visualization of this correlation by showing the probability of bill passage (absolute risk)
as the public opinion variable increases.
Figure IV.4 Probability of Recognition Bill Passage by Public Opinion for All Bills
(n = 370).
Note: All other independent variables were set at their mean.
Table IV.23 supports these data, showing the relative risk of bill passage as public
opinion in favor of a bill progresses from its minimum to its maximum level (RR min to
84


max), as well as the relative risk as public opinion progresses from its 20th percentile
value to its 80th percentile value (RR p20 to p80).
Table IV.23 Relative Risk of Bill Passage for Public Opinion for All Bills (n = 370).
Model RR min to max RR p20 to p80
Revenue 2.816 1.355
Assets 3.536 1.485
Revenue + Assets 3.297 1.434
Note: All other independent variables were set at their mean.
As shown above, in the combined data set, public opinion is positively correlated with
bill passage. That is, bills with more public support (a higher percentage of public
opinion in favor of them) are more likely to pass. Indeed, in all three models, the
probability of bill passage rises from under 9% to over 21% as public support for the bill
increases.
Surprisingly, these findings do not hold in the redistribution data set. Indeed,
within the redistribution data set, public opinion is significantly correlated with bill
passage in only one of the models (the revenue model). Further, this correlation is
negativeas public support for redistribution bills increases, the probability of bill
passage decreases. Figure IV.5 and Table IV.24 below aid in visualizing this correlation.
85


Figure IV.5 Probability of Redistribution Bill Passage by Public Opinion (n = 159).
Note: All other independent variables were set at their mean.
Table IV.24 Relative Risk of Redistribution Bill Passage for Public Opinion (n =
159).
Model RR min to max RR p20 to p80
Revenue 0.210 0.519
Note: All other independent variables were set at their mean.
In the redistribution data set, the probability of bill passage falls from over 12% to just
over 3% in the Revenue Model as public support for the bill increases. This finding is
somewhat surprising, and I discuss possible explanations in the discussion section.
However, given the p value (p = 0.094), there is a reasonable chance that a Type I error
occurredthat this relationship is not real.
However, the relationship between public opinion and bill passage returns to our
expected direction in the recognition data set. Here, as shown by Figure IV.6 and Table
IV.25 below, the probability of bill passage rises as public support increases in both the
Revenue Model and the Revenue + Assets Model (the relationship is not significant in
the Assets Model).


0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Portion of Public Opinion in Favor of Bill
'Revenue Model
Revenue + Assets Model
Figure IV.6 Probability of Recognition Bill Passage by Change in Public Opinion (n
= 211).
Note: All other independent variables were set at their mean.
Table IV.25 Relative Risk of Recognition Bill Passage for Public Opinion (n = 211).
Model RR min to max RR p20 to p80
Revenue 3.472 1.464
Revenue + Assets 3.375 1.434
Note: All other independent variables were set at their mean. Thus, in the Revenue Model
and the Revenue + Assets Model in the recognition data set, the probability of bill
passage rises from under 13% to over 34% as public support for the bill increases.
As demonstrated above, in both the combined and the recognition data sets, the
probability of bill passage increases as public support for the bill increases, even when
controlling for mean levels of SMI strength, interest group strength, political elite
support and media coverage. However, contrary to my expectations, in the redistribution
data set, public opinion in the revenue model is negatively correlated with bill passage.
87


Interest Groups
Though interest groups are not significantly correlated with bill passage in either
the redistribution or the recognition data sets, there is a significant negative correlation
between interest group strength and bill passage in the combined data set. Figure IV.7
and Table IV.26 help to illustrate this relationship.
0.9
0.8
0.7
O)
M
re
VI
Vi
re
a.
= 0.6
CQ
u-
O
0.5
5 0.4
-18 -14 -10 -6 -2 2 6 10 14
Interest Group $ in Billions
Revenue Model
Assets Model
Revenue + Assets Model
Figure IV.7 Probability of Bill Passage by Interest Groups for All Bills (n = 370).
Note: All other independent variables were set at their mean.
Table IV.26 Relative Risk of Passage for Interest Groups for All Bills (n = 370).
Model RR min to max RR p20 to p80
Revenue 0.444 0.996
Assets 0.363 0.998
Revenue + Assets 0.380 0.997
Note: All other independent variables were set at their mean.
As shown above, in all three modes in the combined data set, the probability of bill
passage falls from over 21% to under 10% as interest group strength increases.
88


Given the surprising nature of these findings, I investigated this variable further
by removing 501(c)(5)s (labor unions and agricultural organizations) and 501(c)(6)s
(business leagues and chambers of commerce) from the data set. I then removed the bills
(and thus cases) directly related to labor unions (501(c)(5)s) or business leagues and
chambers of commerce (501(c)(6)s). When I reran the analysis under these conditions,
the corrected coefficient for the interest group variable remained insignificantly
correlated with bill passage in both the redistribution and recognition data sets, but
significant and negative in the combined data set. As such, removing 501(c)(5)s and
501(c)(6)s did not alter this unexpected finding.
If replicated, these results would undermine the assumption that there is a direct
causal link between interest group strength and bill passage. Rather, perhaps interest
groups tend to target bills that are not likely to pass (as there is little sense in focusing on
bills that are likely to pass). If true, we would expect interest group strength to be
strongest relative to bills that are otherwise unlikely to pass. In such instances, we might
find that results similar to those aboveinterest group strength negatively correlated with
bill passage.
To test this proposition, I investigated the relationship between interest group
strength and both public opinion and political elites. If interest groups do, in fact, target
bills that are difficult to pass, we would expect there to be a negative correlation between
interest groups and one or both of these variables. That is, wed expect interest group
strength to be the strongest when public opinion was against the bill and/or when elite
support was low. The evidence, in this regard, is mixed.
89


A bi serial correlation demonstrated that there is a statistically significant positive
correlation between interest group strength and public opinion in all three models
(Revenue Model: r{368) = 0.25,/) = 0.000; Asset Model: r(368) = 0.26,p = 0.000);
Revenue + Asset Model: r(368) = 0.26, p = 0.000). This suggests that interest groups
actually coalesce around bills with higher public support. However, there is a negative
(though not statistically significant) correlation between interest group strength and
political elite allies (Revenue Model: r(368) = -0.06, p = 0.284; Asset Model: r(368) =
-0.03,p = 0.524); Revenue + Asset Model: r(368) = -0.04, p = 0.408). Though the lack
of significance means we cant conclude there is a real correlation between interest group
strength and elite support, this finding does indicate that, within these data, political elite
support is generally lower when interest group strength is higher.
Yet, at this point, it is important to note that the political elite support variable is
based on partisan sponsorship, with the relative strength of elite support based on the
party sponsorship of a bill within the context of the partisan divide during the 109th
Congress. That is, because Democrats were the minority party in the 109th Congress,
only Democratic sponsorship equates to low elite support (as opposed to Republican or
bipartisan sponsorship). As such, the negative correlation between interest group
strength and elite support could be interpreted another way: perhaps interest groups
interested in social justice target bills that are more likely to be supported by Democrats.
To investigate this possibility, I compared the political elite values for the 37
cases with the highest interest group strength values (the top 10%), to the political elite
values for the whole sample. In particular, I looked at the percentage of cases with a
value of 0, indicating only Democratic support. These results are displayed below.
90


Full Text

PAGE 1

INSULT, INJURY AND IMPACT : S OCIAL MOVEMENT IMPACT ON PUBLIC POLICY IN THE CONTEXT OF RECOGNTION AND REDISTRIBUTION by Brett J. Reeder B.S. and B.A., University of Colorado, Boulder, 2006 A thesis submitted to the Faculty of the Graduate School of t he University of Colorado in partial fulfillment of the requirements for the degree of Mas t er of Social Science Humanities and Social Sciences 2012

PAGE 2

ii This thesis for the Master of Social Science degree by Brett J. Reeder has been approved for the Humani ties and Social Sciences by Dr. Paul Stretesky, Chair Dr. Akihiko Hirose Dr. Lucy McGuffy 11/8 /2012

PAGE 3

iii Reeder, Brett, J. M.S.S., Master of Social Science Program Insult, Injury and Impact : Social Movement Impact on Public Policy in the Context of Recognition and Redistribution Thesis directed by Professor Paul Stretesky ABSTRACT This study is motivated by a simple question that has a complex set of answers: D o social movements impact social justice? To answer this question I draw o n social movement theory to build a model with US Congressional b ills from the 109 th Congress as my dependent variable, and (a) social movement industry (SMI) strength, (b) interest group strength, (c) public opinion, (d) political elite support and (e) me dia coverage as my independent variables. I also draw on contemporary critical theory, utilizing Nancy Fraser's distinction between recognition and redistribution to split my data into two distinct datasets: one built around re distribution based bills an d one built around recognition based bills. I analyze these data using rare events logistic regression (rel ogit) to see if SMI strength is correlated with bill passage. The results suggest that SMIs do influence the passage of recognition based bills but do not affect re distribution based bills. This finding has profound implications that span from practical politics and political theory to social theory and moral philosophy. The form and content of this abstract are approved I recommend its publicatio n. Approved: Dr. Paul Stretesky

PAGE 4

iv TABLE OF CONTENTS !"#$%&' ! "## !$%&'()*%!'$ # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """" # + # !,-./01234#.5#064#7089: # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """" # ; # '<4/<=4>#.5#*61-04/? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""" # @ # !! "## A!% B&C%)&B#&BD!BE # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""" # F # 7.3=1G#H.<4,420#A=04/108/4 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""""""""""""""""" # F # H:#)0=G=I10=.2#.5#064#7.3=1G#H.<4,420#A=04/108/4 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""" # +F # J1-?#=2#064#7.3=1G#H.<4,420#A=04/108/4 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""" # +K # */=0=31G#%64./:#A=04/108/4 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""" # +L # H:#)0=G=I10=.2#.5#064#*/=0=31G#%64./:#A=04/108/4 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""" # ;+ # J1-?#=2#064#*/=0=31G#%64./:#A=04/108/4 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" # ;@ # !!! "## HB%M'('A'JN # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""""""""""""""""""" # ;F # (4-429420#D1/=1OG4P#7.3=1G#Q8?0=34#R=GG? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" "" # ;S # R=GG#74G430=.2 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""""""" # ;S # (101O1?4#(=?0=230=.2 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """ # ;T # !294-429420#D1/=1OG4#U+P#7.3=1G#H.<4,420#!298?0/=4?#V7H!?W # """""""""""""""""""""""""""""""" """""""""""""""""""" # ;L # (101#7.8/34P#$**7#!&7#R8?=24??#H1?04/#X=G4? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""""" # ;Y # (101#&49830=.2 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""" # @; # (=554/420=10=2Z#7H'?#129#!204/4?0#J/.8-? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""" # @L # (404/,=2=2Z#(=/430=.21G=0:#129#C ??=Z2=2Z#D1G84? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""" # F@ # !294-429420#D1/=1OG4#U;P#!204/4?0#J/.8-? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""" # FF # !294-429420#D1/=1OG4#U@P#[8OG=3#'-=2=.2 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""" # FF #

PAGE 5

v !294-429420#D1/=1OG4#UFP#[.G=0=31G#BG=04#CGG=4? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""""""""" # FL # !294-429420#D1/=1OG4#USP#H49=1#*.<4/1Z4 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""" # S@ # C21G:?=?P#&1/4#B<420?#A.Z=?0=3#&4Z/4??=.2?#V&4A.Z=0?W # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """" # SF # '<4/<=4>#.5#H406.9? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""" # SL # !D "## &B7)A%7 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""" # SY # (4?3/=-0=<4#7010=?0=3? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""" # K\ # (4-429420#D1/=1OG4P#7.3=1G#Q8?0=34#R=GG? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""""""""""""""""" # K\ # !294-429420#D1/=1OG4#U+P#7.3=1G#H.<4,420#!298?0/=4?#V7H!?W # """""""""""""""""""""""""""""""" """"""""""""""" # K@ # !294-42942 0#D1/=1OG4#U;P#!204/4?0#J/.8-? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""""""""""" # KS # !294-429420#D1/=1OG4#U@P#[8OG=3#'-=2=.2 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""""""""""""" # KL # !294-429420#D1/=1OG4#UFP#[.G=0=31G#BG=04#CGG=4? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""" # KL # !294-429420#D1/=1OG4#USP#H49=1#*.<4/1Z 4 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""" # KY # )23.20/.GG49#(=/430#B55430?P#7=2ZG4#&4A.Z=0#H.94G? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""" # T\ # *.20/.GG49#B55430P#X8GG#&4A.Z=0#H.94G? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"" # T; # [.G=0=31G#BG=04#CGG=4? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""" # L\ # [8OG=3#'=2=.2 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""" # LF # !204/4?0#J/.8-? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""" # LL # 7.3=1G#H.<4,420#!298?0/=4?#V7H!?W # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"" # Y+ # 7H!#!,-130#=2#*.204]0P#D1/=.8?#&4A.Z=0#H.94G? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""" # Y@ # 7H!?#CZ1=2?0#064#E./G9 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""""""""""""""""" # Y@ # 7H!?#&=9=2Z#1#%=94#.5#78--./0 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""" # YS # [4.-G4#4/ # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""" # YK # 7H!?#>=06#^4:#BG=04#CGG=4? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""""""""""""" # YT # *.,-1/=?.2#.5#*=/38,?01234? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""" # YY # '<4/<=4>#.5#&4?8G0? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""" # +\+ #

PAGE 6

vi D ## (!7*)77!'$ # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""" # +\; # !,-G=310=.2?#5./#064#7.3=1G#H.<4,420#A=04/108/4 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""" # +\; # 78--./0#5./#&4?.8/34#H.O=G=I10=.2#%64./:#V&H%W # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""" # +\K # [.G=0=31G#BG=04?_#[8OG=3#'-=2=.2#129#H1?04/#X/1,4? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"" # +\K # )234/01=2#B55430#.5#H49=1#*.<4/1Z4 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """ # +\L # (=?0=2Z8=?6=2Z#R40>442#!204/4?0#J/.8-?#129#7H'? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """ # +\L # 7H!#!,-130#129#*.204]081G#*=/38,?01234 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""""""""""" # ++\ # !,-G=310=.2?#5./#*/=0=31G#%64./: # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""" # ++; # [/130=31G#[.G=0=3?#V7H!#%130=3?W # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""" # ++; # [.G=0=31G#%64./: # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""" # ++S # 7.3=1G#%64./: # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""" # ++T # H./1G#[6=G.?.-6: # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""" # +;+ # '<4/<=4>#.5#(=?38??=.2 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" # +;@ # D! "## *'$*A)7!'$ # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" "" # +;S # 7089:#A=,=010=.2? # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""" # +;K # (=/430=.2?#5./#X808/4#&4?41/36 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """"""""""""""""" # +;Y # &BXB&B$*B7 # """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""""""""""""""""""""""""" """""""""" # +@; #

PAGE 7

vii LIST OF TABLES Table III.1 Search Terms Used to Identify Bills by Major Social Categories. .......................... 26 # III.2 Step 1 in Data Reduction for the BMF Master Files: IRS Code. ............................. 32 # III.3 Step 2 in Data Reduction for the BMF Master Files: Major Group (12). ................ 34 # III.4 Step 3 in Data Reduction f or the BMF Master Files: Major NTEE Groups. ........... 35 # IV.1 Number of Bills by Chamber. ................................ ................................ .................. 61 # IV.2 Number of Bills by Year and Chamber. ................................ ................................ .. 61 # IV.3 Number of Bills. ................................ ................................ ................................ ...... 62 # IV.4 Descriptive Statistics for SMI Revenue ($ in Billions). ................................ .......... 63 # IV.5 Descriptive Statistics for SMI Assets ($ in Billions). ................................ .............. 64 # IV.6 Descriptive Statistics for SMI Revenue + Assets ($ in Billions). ............................ 65 # IV.7 Descriptive Statistics for Interest Groups Revenue ($ in Billions). ......................... 66 # IV.8 Descriptive Statistics for Interest Groups Assets ($ in Billions). ............................ 67 # IV.9 Descriptive Statistics for Interest Groups Revenue + Assets ($ in Billions). .......... 67 # IV.10 Descriptive Statistics for Public Opinion (in Favor of Bill). ................................ 68 # IV.11 Percent of Bills that Fall within each Political Elite Support (in Favor of Bill) Category. ................................ ................................ ................................ ........................... 69 # IV.12 Descriptive Statistics for Media Coverage (in 100s of Articles). .......................... 70 # IV.13 Significance and Direction of Uncont rolled ReLogit Estimates. .......................... 71 # IV.14 Corrected Logit Coefficient Estimates (Robust SE) for Revenue Model. ............. 73 # IV.15 VIF and (Tolerance) for Revenue Model. ................................ ............................. 74 # IV.16 Co rrected Coefficient Estimates (Robust SE) for Assets Model. .......................... 76 #

PAGE 8

viii IV.17 VIF and (Tolerance) for Assets Model. ................................ ................................ 77 # IV.18 Corrected Logit Coefficients (Robust SE) for Revenue + Assets Model. ............. 78 # IV.19 VIF and (Tolerance) for Revenue + Assets Model. ................................ ............... 79 # IV.20 Relative Risk of Bill Passage for Partisan Progression for All Bills (n = 370). .... 81 # IV.21 Relative Risk of Redistribution Bill Passage for Partisan Progression (n = 159). 82 # IV.22 Relative Risk of Recognition Bill Passage for Partisan Progression (n = 211). .... 83 # IV.23 Relative Risk of Bill Passage for Public Opinion for All Bills (n = 370). ............ 85 # IV.24 Relative Risk of Redistribution Bill Passage for Public Opinion (n = 159). ......... 86 # IV.25 Relative Risk of Recognition Bill Passage for Public Opinion (n = 211). ............ 87 # IV.26 Relative Risk of Passage for Interest Groups for All Bills (n = 370). ................... 88 # IV.27 Percentage of Cases with Democratic Sponsorship Only (Low Political Elite Support). ................................ ................................ ................................ ............................ 91 # IV.28 Relative Risk of Passage for SMIs for Recognition Bills (n = 211). ..................... 93 # IV.29 Relative Risk of Recognition Bill Passage for SMIs with IVs at p20 (n = 211). .. 94 # IV.30 Relative Risk of Bill Passage for SMIs with IVs at p80 (n = 211). ....................... 95 # IV.31 Relative Risk of Recognition Bill Passage for SMIs with Public Opinion at p80 and All Other IVs at p20 (n = 211). ................................ ................................ .................. 97 # IV.32 Relative Risk of Recognition Bill Passage for SMIs with Elite Political Allies at p80 and All Other IVs at p20 (n = 211). ................................ ................................ ........... 98 #

PAGE 9

ix LIST OF FIGURES Figure III.1 NTEE CC Decile/Centile Code Eliminations. ................................ ......................... 36 # III.2 NTEE CC Decile/Centile Code Eliminations. ................................ .......................... 37 # IV.1 Probability of Bill Passage by Political Elite Support for All Bills (n = 370). ........ 81 # IV.2 Probability of Redistribution Bill Passage by Partisan Progression (n = 159). ....... 82 # IV.3 Probability of Recognition Bill Passage by Partisan Progression (n = 211). .......... 83 # IV.4 Probability of Recognition Bill Passage by Public Opinion for All Bills (n = 370). ................................ ................................ ................................ ................................ ........... 84 # IV.5 Probability of Redistribution Bill Passage by Public Opinion (n = 159). ............... 86 # IV.6 Probability of Recognition Bill Passage by Change in Public Opinion (n = 211). 87 # IV.7 Probability of Bill Passage by Interest Groups for All Bills (n = 370). ................... 88 # I V.8 Probability of Recognition Bill Passage by SMIs (n = 211). ................................ .. 92 # IV.9 Probability of Recognition Bill Passage for SMIs with IVs at p20 (n = 211). ........ 94 # IV.10 Probability of Recognit ion Bill Passage for SMIs with IVs at p80 (n = 211). ...... 95 # IV.11 Probability of Recognition Bill Passage for SMIs with Public Opinion at p80 and All Other IVs at p20 (n = 211). ................................ ................................ ......................... 96 # IV.1 2 Probability of Recognition Bill Passage for SMIs with Elite Political Allies at p80 and All Other IVs at p20 (n = 211). ................................ ................................ .................. 98 # IV.13 Probability of Recognition Bill Passage in the Revenue Model by SMI Strength in Variou s Circumstances (n = 211). ................................ ................................ .................... 99 # IV.14 Probability of Recognition Bill Passage in the Assets Model by SMI Strength in Various Circumstances (n = 211). ................................ ................................ .................. 100 #

PAGE 10

x IV.15 Probability of Recogni tion Bill Passage in the Revenue + Assets Model by SMI Strength in Various Circumstances (n = 211). ................................ ................................ 100 #

PAGE 11

1 CHAPTER I INTRODUCTION This study is inspired by an overarching question: D o social movements impact social justice ? More specifically I'm interested in movement influence on the passage of federal bills related to issues of social justice 1 in the United States To further clarif y this investigation, I utilize the concepts of social movement organizations (SMOs) and social movement industries (SMIs). SMOs are formal organizations that set goals aligned with social m ovements or counter movements. SMIs are collection s of all the S MOs that sh are the broad goals of a movement or counter movement ( McCarthy & Zald 1977). I utilize these concepts to clarify the bounds of my research with my first research question: Within the contemporary United States does the strength of a social mo vement industry impact whether the U.S. Congress passes a bill that is important to that industry? 1 Though there are competing conceptions of social justice here I use the term in the broad sense to mean justice (however defined) realized throughout society. Though I discuss competing claims over the content of justice (i.e., what constitutes justice) below, with this definition, I intentionally leave it op en to contestation. I do, however, insist that social justice must extend throughout all social institutions. As such, though formalized legal justice is an important element of social justice, it is only one component. Further, social justice must be so cial, not individual ( though the social impacts individuals ). That is an injustice done to an individual is only a matter of social justice if the injustice is endemic to society itself. For example, a racist manager who only promotes white employees to particular positions is (at least from most conceptions of justice) committing an injustice at the individual level. However, a social structure that only allows white people to hold particular positions (whether through legal mandate, structural impedim ents or cultural bias) is social injustice. While the first example may be a symptom of a social justice issue, in and of itself it is a matter of individual justice. However, the second is a clear matter of social justice.

PAGE 12

2 As a second layer to my project, I utilize Nancy Fraser's (2003) distinction between recognition and redistribution Described in more detail in the literat ure section below, this theory views recognition and redistribution as fundamentally separate, yet inter related, social realities Using this concept my second research question is : If a relationship between social movement industry strength and bill pa ssage exists, is that relationship moderated by the focus of the bill on re distribution or recognition? Importance of the Study This work is important for reasons that mirror my two layers of investigation. First, my research add s to the social movement l iterature by further clarifying how a social movement industry may influence public policy. Most of the existing social movement literature takes social movements as the unit of analysis, selecting cases based on the movements themselves. As a result, th ese investigations often look at how large and well known movements impact policy ignoring cases with weak or absent movements In contrast, this study uses social justice bills as the unit of analysis, and thus includes cases with a wide range of social movement presence and strength. Second, this study expand s the recognition/redistribution debate raging in contemporary critical theory by inject ing much needed empirical evidence into it This debate centers on a moral philoso phical distinction Nancy Fraser makes between recognition, redistribution and representation, contrary to Axel Honneth's assertion that redistribution and representation are forms of recognition (Fraser & Honneth, 2003) From its moral philosophical roots, this debate extends thr ough social theory, to political theory and practical politics. Its treatment at the level of practical politics is largely

PAGE 13

3 limited to policy prescriptions for a just society, but to date there has been little investigation as to whether this distinction is meaningful at the level of practical politics That is while having a lot to say about the types of policy that would lead to social justice it is yet unclear whether the recogniti on/redistribution distinction matters when investigating how social j ustice policy is actually formed. This study remedies that gap. Overview of Chapters This study is divided into six chapters. The first c hapter, the introduction, outlines the paper as a whole, including its purpose and importance. The second c hapter pr esents a review of two literatures relevant to this study: (1) the social movement literature related to movement outcomes and (2) the critical theory literature on recognition. It also describes how I utilize this literature, as well as the gaps within t hese literatures. The third chapter describes the methods used, including the overall design, how the databases were developed and the analyses used The fourth c hapter presents the results of this analysis including descriptive statistics for each vari able and a variety of rar e events logistic regressions (rel ogit s ). I discuss t he implications of these findings for both the social movement literature and the critical theory literature in the fifth chapter. T he sixth chapter concludes with an overview of the study as a whole, an identification of the study's limitations and potential directions for future research.

PAGE 14

4 CHAPTER II LITERATURE REVIEW There are two m ajor bodies of literature relevant to this study: social movement literature and the critical theory literature on recognition 2 Below I review major works within each tradition, describe how I utilize the literatures and discuss gaps within each literat ure Social Movement Literature Until recently, social movement scholars largely neglected the impacts of social movements focusing instead on other aspects such as their emergence, growth and operation. This has changed in recent years, but the range o f social movement outcomes investigated is sta ggering (see Snow & Soule 2009 for an account of the different types of outcomes studied). In addition to the influence movements have on states, scholars have investigated the effect movements have on cultur e, their particip a nts and even other movements Further, those who study movement impact on states look at a variety of potential outcome s from new policies to changes in the political structure. Finally, movement scholars studying movement impact on po licy focus on different aspects of policy including agenda setting, policy development and policy implementation. Thus, my focus on movement impact on the passage of public policy is a subset (policy passage / non passage ) of a subset ( impact on policy) of a subset ( influence on states) of a subset (consequences of social movements) of the social movement literature ( Amenta & 2 A third body of literature, the public policy process literature, is also arguably relevant to this study. This rich literature includes complex accounts of how policy is passed and the factors that influence passage. That said, due to its relative silence on social movements, I hav e not included it as part of this review. I do, however, include several references to it in footnotes throughout this section.

PAGE 15

5 Caren, 2007; Snow et al. 2007). In other words, the portion of the social movement literature I cover here social movement impact o n public policy is a narrow section of a large body of social movement literature. Research in this narrow area has expanded substantially in the last decade, and numerous reviews of social movement political outcome studies exist ( Amenta & Caren, 2007; A menta et al., 2010; Burstein & Linton 2002; Earl, 2000; Giugni 1998; Giugni, 2008 ; Meyer 2005 ). These reviews generally suggest that social movements sometimes, but certainly not always, influence the passage of public policy. For example, Burstein an d Linton's (2002) review of 53 articles published between 1990 and 2000 indicated that th e political organizations (their umbrella term for SMOs, interest groups and political parties) influence on public policy is statistically significant for about half of the sample studied Further, they suggest that this effect is meaningful in policy terms (i.e. a substantial impact on policy ) only about one fifth of the time ( Burstein & Linton 2002). Similarly, Amenta et al.'s (20 10) review suggests that social m ovement impact on public policy is incredibly episodic and nuanced. This review of 45 articles on the political consequences of social movements published between 2001 and 2009 suggests that, in general, scholars have moved away from Wil liam Gamson's (199 0 ) approach of defining success in terms of gaining new benefits or acceptance and toward their influence on particular political outcomes. This work has indicated that movements impact varies substantially based on the particular phase of the policy pro cess in question

PAGE 16

6 (i.e. agenda setting, adoption and implementation). 3 In particular, the literature suggests that movements are most influential during the agenda setting stage (the stage that determines the which policy areas are discussed) of the polic y process, and often only minimally influential in the adoption stage (the stage that involves the selection of particular policies ) ( Amenta et al. 2010 ; Johnson, Agnone, & McCarthy, J. D., 2000 ; Johnson, 2008; Snow & Soule 2009 ). To successfully influen ce policy, movements utilize resources ranging from the tangible (i.e. money or human capital) to the intangible (i.e. legitimacy or cultural artifacts) in an effort to realize goals. While the relative importance of various types of resources is disput ed, social movement scholars almost universally accept that movements need sufficient resources to successfully attain their goals ( Cress & Snow, 1996; Khawaja 1994; Zald 1992). There is however, more disagreement about the importance of organization. Some scholars argue that as movements formal ize their organizational structure and "professionalize they simultaneously become less radical and confrontational (Staggenborg 1988). Additionally, a move toward formal organization can shift a movement's focus away from movement goals and toward organizational survival and professional advancement (Michels 1915 ). In other words, leaders of formal 3 Within the policy literature, it has long been customary to conceptualize policy formation as a sequential process that procee ds through delineated stages. Indeed, this Stages Heuristic dominated the field until the mid 1980s In recent years, this view has been criticized for a variety of reasons, and most contemporary scholars now recognize that the stages of a policy process are not strictly sequential, nor are they unidirectional. Further, many have pointed out that it is impossible to completely delineate separate stages. However, most contemporary policy scholars continue to utilize some variant of the sta ges model as a taxonomic device within their theories. For a discussion of the role various stages models have played in policy theory see deLeon 1999 and Sabatier 2007.

PAGE 17

7 organizations may begin to act in ways that benefit the organization and/or their own professional advanceme nt, to the detriment of the movement as a whole (e.g. going after grants poorly aligned with the goals of the organization but that allow the organization to make it s payroll). Along these lines, Frances Fox Priven and Ric hard A. Cloward (1978 ) famously argued that o rganization undermines movements' ability to realize outcomes by redirecting resources away from disruptive protest and toward organization building. Taking the contrary position, scholars working out of the resource mobilization tradition hav e stressed the importance of formal orga nizations in social movements. Developed most comprehensively by John McCarthy and Mayer Zald in the mid 1970s, this perspective argues that formal organizations known as social movement organizations (SMOs) are i ndispensible in a movement's ability to acquire, and direct the use of, resources ( Cress & Snow, 1996; Gamson 1990; McCarthy & Zald, 1973, 1977 2001 ). From this perspective, movement success (such as influencing bill passage) is thus dependent on a move ment's ability to organize through one or more SMOs and for these SMOs to acquire resources they direct at well defined goals, such as promoting (or opposing) the passage of federal bills (Edwar ds & McCarthy 2007 ). From this perspective, the strength of a movement as a whole is thus closely associated with the aggregate strength of SMOs within it also know n as a social movement industry (SMI). The implication is therefore, that well resourced SMI s have the ability to influence policy. In other words, movement mobilization is sufficient for impact o n policy ( McCarthy & Zald, 2001 ). For example, Cress and Snow (2000) found in a study of 15 homeless SMOs that "organizational viability" (defined as the ongoing maintenance of a

PAGE 18

8 formal organization and sust ained protest activity over time) was the most important indicator of SMO success. Beyond resources and organization, much of the recent work on social movements has focused on either the tactics utilized by movements or the contextual factors that shape movement outcomes as well as the movements themselves ( Snow & Soule, 2009 ). With regard to tactics, scholars have found that the type of action that movements engage in matters. Along these lines, scholars have long debated the effect of the tactical cho ice to engage in dis ruptive or violent tactics. Some argue that violence/disruption is necessary to realize political outcomes that upset the status quo, wh ile others believe violence undermines and delegitimizes movement efforts. Thus far, the empirical evidence is mixed with some support for both claims (Giugni 1998). Gamson 's 1975 classic Strategy of Social Protest (second edition 1990) study, as well as several re analyses of these data ( Mirowsky & Ross, 1981; Steedly & Foley, 1979 ) found evidence that disruptive tactics used by challenging groups is positively correlated with both attaining legitimacy and obtaining new advantages (his two measures of movement success). Similar findings appear in a variety of works on related phenomenon such as st rikes (Sh orter & Tilly 1971) and in Frances Fox Priven and Richard A. Cloward s (1977) Poor Peoples Movements However, other works focused on labor conflicts and urban riots find little evidence that violence is an effective tactic for achieving desired outcomes ( Kelly & Snyder, 1980; Snyder & Kelly 1976; Taft & Ross, 1969 ). Indeed, the literature seems to suggest that, like movements generally, disruptive tactics sometimes successfully influence policy and sometimes they do not (Giugni 1998). This e xtends to tactics generally as some scholars suggest that different

PAGE 19

9 tactics may be more or less effective at different phases in the policy process (Andrews & Edwards 2004), while others have found that the importance of a particular tactic can vary by co ntext (Amenta 2006). A nother particularly fruitful line of research focuses on the act of meaning construction or framing. Movement scholars apply this concept to the meanings and beliefs that inspire and legitima te social movement activities These ty pes of frames, called "c ollective action frames vary in scope from those employed by particular organizations (organizational frames) to general conceptual frames that are employed by entire cultures (Benford & Snow, 2000). Theorists working within the f raming tradition have argued that movement success depends largely on its ability to effectively frame the "problem" (diagnostic frames) an attractive "solution" (prognostic frames) and the motivation to act (motivational frames) ( Benford & Snow, 2000; Cre ss & Snow, 2000 ). Cress and Snow (2000) found in addition to organizational vitality, framing processes were consistently one of the two most important factors in a social movement organization's success. More recently, McCammon (2009) found that favorab le outcomes associated with women's jury rights mobilizations were tied specifically to those that defined the problem (diagnostic frame) as serious and wide spread, contained clear rationale and supported their position with concrete evidence (prognostic frame) Scholars working within the framing tradition also have begun to investigate the importance of the mass media as a master arena in which public discourse takes place and meaning is established as frames compete for superiority ( Gamson 2007; Ga mson &Wolfsfeld, 1993). From this perspective, the media act as the primary site where the contestants assume meaning is established, whether or not this is actually the case

PAGE 20

10 (Gamson, 2007) Empirical work in this tradition suggests that the media do in fact, affect the political views of legislators and as a result, impact public policy (Cook, Tyler, Goetz, Gordon, Protess, Leff, &Molotch, 1983). Further, some scholars have investigated the ways SMOs and other organizations strategically target the me dia, as well as the relative success of these efforts (Rohlinger, 2002). Beyond the role context plays in the relative influence of particular frames or tactics, many scholars have found that contextual factors pla y an important role generally. Within thi s line of research, scholars have seen the political structure as particularly important. It should come as no surprise that the impact of social movements varies by regime type (Goldstone 1980), but even within W estern democracies, the peculiarities of particular political systems has been shown to a ffect the degree to which social movements will have any influence on public policy (Schwartz 2000). Scholars working within this political process tradition focus on the political context movement s are emb edded within In particular, they argue that a political system's openness and its capacity to act, interact to establish a political opportunity structure. In its more pure forms, these theorists view the opportunity structure as the primary explanator y variable in both movement development and their potential for success (Kriesi 1995). The changing political context within regimes has also been shown to influence the ability of social movements to impact policy. In particular, scholars have argued fo r the importance of powerful allies within political institutions ( Tarrow 1993 2011 ). For example, Meyer and Minkoff (2004) found that the influence of left leaning movements is amplified when Democrats are in power Similarly, Tarrow (1993) argues pers uasively that during the May 1968 student uprisings in France social movement activity was only

PAGE 21

11 able to a ffect university policy with the help of elite institutional allies. Thus, the political process literature suggests that movement success is greatly associated with the presence or absence of political elite allies. Perhaps chief among the non structural contextual factors is public opinion. Several scholars have found evidence that public opinion influences public policy directly, especially when it is widely held and strongly felt ( Burstein, 2003; Costain & Majstorovic, 1994; Hartley & Russett 1992; Hill & Hinton Andersson 1995; Page & Shapiro 1983). Further, some have found that when public opinion is taken into account, the influence of politi cal organizations (i.e. SMOs, interest groups and political parties) is greatly diminishe d or gone entirely (Burstein & Linton 2002; Uba 2009). However, people's attention span is limited to a finite number of topics, and thus, public opinion is limited to a finite number of issues. Further, within the universe of topics the public is generally aware of, many do not interest most people. As such, while public opinion is an important factor where it exists, it is often not present in any meaningful form ( Burstein 2006 ). In recent years scholars have started to combine the above variables into joint and/ or mediated approach es that look at the interactive effects of social movements, political context and public opinion ( Snow & Soule, 2009 ). Though the particularities of these models vary somewhat, they have a common model of causation: M ovement impact is mediated through other institutions. This implies that much of a movement's influence on policy is indirect, th r ough its influence on other relevant variables such as elected officials or public opinion (Snow, Soule & Kriesi 2007). For example, in the context of the struggle to pass the Equal Rights Amendment (ERA), Soule and Olzak

PAGE 22

12 (2004) found that the presence of elite allies intensified the effec t of the pro ERA movement on s tate ratification but that the presence of elite allies was not necessary for s tate ratification. In an influential mediated model, Edwin Amenta and his colleagues (1992, 2005) argue that the consequences of social movement ac tivities depend on (1) the level of movement mobilization, (2) the favorability of the political context and (3) the strategies implemented by the movement. Movement mobilization in this context is defined as the ability of a movement to mobilize supporte rs in political actions with varying degrees of "assertiveness," with the chosen mix of assertive tactics representing the movement's strategy Additionally, Amenta et al. (1992, 2005) identify four specific aspects of a political context that influence m ovement impact : (1) the degree to which democratic practice bounds democratic institutions (2) the degree to which political parties rely on patronage (3) the presence/absence of bureaucrats with missions aligned with the movement and (4) the partisan ma keup of the regime (Amenta et al. 2005). Finally, the causal mechanisms within this model, operating within the assumptions of bounded rationality, assume elected officials consider the costs/benefits associated with supporting a bill. From this perspec tive, t he most important costs/ benefits are those related to (1) reelection, (2) standing within the party, (3) perceived efficacy of the policy and (4) ideological commitments (Amenta 2006). Giugni and Passy (1998) explicitly tested some of the propositi ons discussed thus far by setting up models with social movements, public opinion and political alliances as independent variables, and public policy as a dependent variable. They then tested to see if there was a direct effect between movements and polic y passage, an effect mediated

PAGE 23

13 through other variables or a joint effect in combination with other variables. Their analysis found very little evidence to support the direct effect and mediated effect models, but some support for the joint effect model. H owever, this effect only held for two of the three movements studied. Additionally, public opinion had a direct effect on one of the issues areas, but it did not add to the statistically significant interactive terms. These findings further support the now widely held view that social movement impact is highly contextual, and that in some contexts movements influence the passage of public policy while in others they do not The literature discussed above suggest s that the following factors, internal to movements a ffect their ability to influence the passage of public policy: (1) the resources available to the movement, (2) the organization of the movement, (3) the tactics employed by the movement and (4) the way the movement frames it s issues. Addition ally, the literature suggests that several external factors affect the ability of movements to influence the passage of public policy. These include: (5) the political system movements are embedded within (6) the presence or absence of elite allies withi n this system and (7) public opinion related to the policy. 4 Finally, the literature suggests that while movements have only minimal direct effect on the passage 4 The policy process literature is largely silent on the impact of social movements, yet it overla ps in many ways with the movement theory described in this section. A mong the varied policy process literature, the Advocacy Coalition Framework (ACF) is perhaps best positioned to dovetail with social movement theory. In the ACF, the primary actors are "a dvocacy coalitions," or informal networks of individuals held together by common beliefs, engaged in collective action directed at particular policy objectives. This concept has much in common with the concept of a social movement: both are social collecti ves engaged in collective action motivated by common beliefs. However, unlike advocacy coalitions, social movements' activities extend bey ond efforts to influence policy. Additionally, advocacy coalitions include actors generally not considered part of so cial movements (e.g. the media). For an overview of the ACF see Sabatier and Weible ( 2007 ) For an example of how the ACF has been applied to social movements see Turina ( 2009 )

PAGE 24

14 of public policy, they interact in complex ways with the factors above to, at times, influenc e policy passage 5 There are however, a variety of interactions and combinat ions that appear to result in social movement influence on policy passage. Thus, the most promising research in this area now includes many of the factors above and takes a comb inatorial or interactive approach. My Utilization of the Social Movement L iterature As discussed above, contemporary social movement theory contains four general arguments designed to account for social movement influence on states. Three of these are clo sely related to well known theories of social movement mobilization: (1) resource mobilization suggests that outcomes result from formal organizations mobilizing resources ; (2) framing theory suggests that outcomes are related to the ways in which movement s frame their claims ; and (3) political process theory suggests that outcomes are largely associated with elements of the political structure. Additionally, many theorists have combined elements of each of these theories into various (4) mediated models s uggesting that movement outcomes vary by context, with a special emphasi s 5 Three prominent scholars who have played a large role in establishing thi s body of literature (Doug McAdam, Charles Tilly and Sidney Tarrow) have developed an approach that shifts the focus from correlation between variables to paired comparisons of the processes and mechanisms that make up such correlations. This approach, ter med Dynamics of Contention (DOC), also seeks to shift the focus from "social movement life histories" to "episodes of contention" (McAdam, Tarrow & Tilly, 2001; Tilly &Tarrow, 2006) In this sense, it hopes to expand the field to include other forms of p olitical contention such as revolutions and democratization, while simultaneously narrowing it to include episodic political activities (as opposed to movements, which can persist beyond contentious episodes) (Buechler, 2011). Though this body of work prov ides important insights in the social movement literature, it is not directly relevant to this research because my focus remains social movements and correlations between variables, not paired comparisons of mechanisms across a variety of episodes of conte ntion.

PAGE 25

15 on the movements themselves, political elites and public opinion ( Amenta et al. 1992, 2005; Snow, Soule & Kriesi 2007). In this study, I partially sidestep many of the arguments ma de by both political process theorists and their framing counterparts. I avoid much of the political process argument by holding the larger political context constant the US Congress from 2005 2006. As such, my findings are limited to the particular poli tical context in which my data is embedded (the US federal legislative system in 2005 2006), but I can also claim that any variation I observe is not the result of different legislatures or different political systems This approach reduces, but does not eliminate, the effect of shifting opportunity/threat within the political process. Similarly, I'm able to avoid large portions of the framing argument by using a large N design My sample is large enough to capture both "effective" and "ineffective" fra mes and strategies (if such things exist) and thus pa rticular organizational frames and strategic positions should not systematically affect my claims about aggregate effect However, a large sample does not overcome the all encompassing nature of master frames As such, my study is able to avoid the effect of micro level framing, but as discussed in chapter 5 framing at the macro level may play an important role in these data My study more directly confront s both resource mobilization theory and media ted models. Most importantly, following RMT, I assume that movements need sufficient resources to realize their goals. Further, I assume that because money is highly liquid (i.e. it can easily be translated into other resources) it is the most importan t resource an organization can possess. Additionally, in line with RMT and contra Fox Piven (1978) I

PAGE 26

16 assume that formal organization is a net positive in terms of movements achieving their outcomes. I recognize that movements extend beyond formal organ izations, and money is not the only organizational resource utilized by movements. However, following RMT, I assume that movements with well resourced SMOs can influence policy, whereas movements that lack either organization or resources will not. By in cluding a variety of other possible influences on public policy (my independent variables) I can see the degree to which SMO strength is correlated with bill passage when controlling for other factors. Thus, a statistically significant correlation betwee n SMI strength and bill passage would validate much of RMT. I also draw directly on mediated models and assume that SMI impact is at least somewhat dependent upon the context they are embedded within In particular, I assume that public opinion, political elites and media coverage prominently factor into a movement's ability to realize political outcomes. That is following mediated models, I assume that SM I influence on bill passage is at least partially mediated by public opinion, the presence of politi cal elites and media coverage Thus, in this study I borrow from RMT the presumed importance of organizations and resources in social movement impact and from mediated models I borrow the idea that movement impact is partially mediated by public opinion political elites and media coverage Gaps in the Social M o vement L iterature The social movement literature largely developed within the disciplines of sociology and political science, with disciplinary boundaries sometimes translating into conceptual bar riers P olitical scientists tended to focus on the more ins titutionalized

PAGE 27

17 interests groups and political parties, and sociologists tended to focus on less institutionalized SMOs and the informal movement networks they are embedded within Since at least the 1990s, these disciplinary boundaries have faded somewhat and a community of social movement scholars working across disciplines has emerged complete with their own specialty journals ded icated to social movements ( Mobilization and Social Movement Stud ies ). Additionally, in recent years this scholarly community has made substantial strides in building a respectable body of work on social movement impact (compare the review in Guigni 1998 with the review in Amenta et al. 2010). Most of these schol ars now acknowledge the importance of a wide range of factors including: movement resources, organization, tactics, framing, the politica l system, political elites, public opinion and media coverage Many have also looked carefully at the interaction bet ween these factors. Nonetheless, this literature has several substantial gaps Among the most problematic is the tendency to study large movements that h ave seen their goals realized. Most social movement studies first identify the movements they will s tudy and then examine their effect This approach undermines the plausibility of claims of movement impact because scholars rarely examine instances in which movements were weak or non existent. My study overcomes this deficiency by starting with the out come (bill passage) and then matching movements (through SMIs) to this outcome. However, the most important contribution of this study is the bridging of two literatures and two traditions: social movement theory and critical theory works on recognition. By using the recognition literature's distinction between redistribution and

PAGE 28

18 recognition to typologize polit ical outcomes (bill passage), I add a new perspective to the social movement literature. 6 Critical Theory Literature The concept of "recognition" has gained a lot of traction within contemporary critical theory. 7 Within this context, recognition typically involves acts of acknowledging or respecting others. In this sense, critical theorists' approaches to recognition share a common model of both t he individual and society. While these theorists acknowledge that society is comprised of individuals, they also argue that the individual (or subject) is largely a product of interaction with other people (or subjec ts). Thus, these theorists have an in tersubjective model of both the individual and society. 6 The distinction I make between recognition and redistribution is, in many ways, similar to distinctions made by European scholars working within what has been termed the New Social Movement tradition. New Social Movement (NSM) theories are diverse and often expand well beyond social movements. Despite this diversity, NSM theories of all types tend to stress that different social arrangements lead to different social movements. In particular, they tend to argue that prior to the 1960s, social moveme nts stressed class based material claims and since the 1960s have stressed symbolic claims based on identities such as race or gender (Buechler, 2011). Though the critical theory recognition literature I draw on makes a similar distinction, it differs in i mportant ways that I discuss in the 8 th footnote below. 7 Within philosophy and the social sciences, "Critical Theory" has both a narrow and a broad definition. In the narrow sense, Critical Theory refers to the "Frankfurt School," which consists of the th eory produced by specific German philosophers and social theorists associated with the Institute for Social Research at the University of Frankfurt am Main. This tradition includes a variety of well know theorists, including Max Horkheimer, Theodor Adorno Herbert Marcuse, JŸrgen Habermas and most recently Axel Honneth (who plays a central role in this paper). In the broad sense, Critical Theory refers to any social theory that aims to both explain and actively transform instances of social domination and human enslavement. Nancy Fraser (another central figure within this paper) falls within this version of Critical Theory. Both versions differ from "traditional" theory in that they explicitly embrace normative positions related to human emancipation For a brief but comprehensive history of Critical Theory in both senses, see the "Critical Theory" entry in the Stanford Encyclopedia of Philosophy (Bohman, 2012)

PAGE 29

19 Working from such a model, and drawing from what the renowned 19 th Century philosopher Georg Wilhelm Friedrich Hegel termed "the struggle for recognition," many contemporary philosophers have appropr iated the term recognition in their examination of the normative underpinnings of political claims. Yet the ubiquity of the term itself is not matched by the ways in which it is used and defined. Indeed, the proper conception of recognition is a t the c enter of a major debate within contemporary critical theory ( Fraser &Honneth, 2003) For the purposes of this project, two contemporary thinkers engaged in this debate are particularly important : Axel Honneth and Nancy Fraser. From Honneth's perspective, recognition is a historically specific concept. According to Honneth, the transition to bourgeois capitalist society led to the differentiation of recognition into these three independent spheres: (1) love recognition of others through loving care for the ir well being ; (2) law recognition of others as members of society afforded particular rights ; and (3) achievement recognition of others as productive members of society." Honneth claims t hat while these three spheres are separate concepts in modern soci ety, they were not independent of each other prior to modernity ( Fraser &Honneth, 2003 ). Borrowing heavily from Hegel, Honneth goes on to argue that human development requires recognition. In Hegel's account of modern society this implies that to becom e fully actualized beings, people need reciprocated love, legal recognition as members of society and to be considered productive members of society (however defined). Thus, for Honneth, political struggles for justice are ultimately the modern embodiment of pre political struggles for recognition in an individual effort to become fully actualized human beings ( Fraser &Honneth, 2003 ).

PAGE 30

20 In contrast, Nancy Fraser has developed a status model of recognition in which recognition is viewed as a matter of socia l justice. Fraser's conception of justice revolves around what she calls participatory parity or equity in the ability to participate fully in society. From this perspective, recognition entails the cultural and symbolic preconditions for participatory p arity. That is without social recognition, one is unable to fully participate in society and thus the denial of social recognition is unjust. However, for Fraser, recognition alone is insufficient to allow for full social participation ( Fraser & Honneth 2003). According to Fraser, in addition to recognition, justice entails equitable distribution and political representation as separate, yet interrelated concepts. In the case of material distributi on, Fraser argues that full participation in the so cial world requires at least minimal resources and thus inequitable distributions of resources impedes one's ability to fully participate in society. In the case of political representation, full social participation necessitates representation within al l the political institutions under which one is subject ( Fraser, 2008 ). Working from this moral philosophical level, Frasier develops a social theory that, in quasi Weberian fashion, differentiates class, status and party while carefully attending to their interaction. She then turns to political theory to identify forms of institutional change that correct misrecognition, maldistribution and misframing (of political repr esentation) ( Fraser, 2008 ). Thus for Frasier, there is a meaningful distinction between recognition, distribution and representation that is important at the philosophical, social and political levels.

PAGE 31

21 This runs contrary to Axel Honneth's theory of recognition (recognition monism), which views recognition as matter of self realizati on and distribution as a reflection of recognition. According to Honneth, material distributi on within a society reflects that society's recognition patterns specifically, its achievement recognition structures. That is collective evaluation of the soci al contribution of individuals guides material allocations High levels of compensation are associated with the a chievements socially recognized as especially valuable while low levels of compensation are associated with the achievements that are not rec ognized by society as particularly valuable. Thus, for Honneth, redistribution is a special case of recognition and as such there is no meaningful distinction ( Fraser &Honneth, 2003). My Utilization of the Critical T heory L iterature As discussed above, there are two major conceptions of recognition within contemporary critical theory: (1) an identity model associated with Honneth and (2) a status model associated with Fraser. The identity model presumes that distribution is a form of recognition, and th us, that redistribution is not distinct from recognition. In contrast, the status model views recognition as distinct from, but interrelated with, red istribution. Following Fraser, and contra Honneth, I believe the distinction between redistribution and r ecognition is meaningful at the levels of practical politics, political

PAGE 32

22 theory and social theory 8 A t the level of practical politics I expect there to be differences in the efficacy of movement efforts based on whether their target is re distribution or recognition. Further, at the political theoretical level I believe that institutional corrections to the recognition order will not directly translate into a more equitable redistribution of material resources. Likewise, a redistribution of resources th rough institutional channels will not directly transl ate into improved recognition. Additionally, at the level of social theory I believe the distributional structure (or class) and the recognition order (or status group) are at least partially, uncoupl ed from each other in modern capitalist societies If these assertions are correct, we would expect to see differences in the way the institutional rules (or laws) that govern distribution and recognition are established. We would expect to see them estab lished at different rates, and we would expect to see the impact movements have on the m to be different. On the other hand, if Honneth is correct and there is no meaningful difference between redistribution and recognition, then we would not expect to see any such difference. 8 This is in many ways, similar to the distinctions made by New Social Movement (NSM) theorists who tend to argue that there was a historical shift in which movements switch their focus from class based material claims to identity based symbolic claims, and that this shift corresponded with macro social shifts. However, Fraser's distinction differs in a variety of key ways. Where NSM theorists view the key historical shift as from class to identity, Fraser sees the key historical shift as a partial uncoupling of class and status. This leads to two further distinctions, as NSM theorists tend to view c laims for material distribution and identity recognition as an either or proposition, whereas Fraser views them as an interrela ted and simultaneous social reality Thus, while NSM theorists see a shift from class to identity, Fraser sees a shift in emphasi s on class or status though both exist simultaneously. Additionally, where NSM theorists are concerned with identity, Fraser is concerned with status. That is NSM theorists' models of recognition are largely identity models, while Fraser uses a status m odel (Buechler, 2011; Fraser &Honneth, 2003 ; Fraser, 2011).

PAGE 33

23 To test this, I code bills (my dependent variable) as either mostly distribution or mostly recognition based. I then use this coding to separate my data into two distinct data sets: (1) a redistribution data set and (2) a recognition dataset. Next, I explore the correlation between SMI strength and bill passage in each data set. A difference between these two data sets adds empirical support for Fraser's model of recognition, while the absence of difference would add support for Hon neth's. Gaps in the Critical Theory L iterature Like much of the literature in philosophy, the recognition literature suffers from a relative lack of empirical evidence. 9 In an attempt to contribute to remedying this issue, this study extends the debate be tween Nancy Fraser and Alex Honneth to an empirical exploration of social movement impact Using Fraser's redistribution /recognition typology I divide my data into two distinct data sets: one built around distribution bills and one built around recogniti on bi lls. This allows me to see if there is an empirical difference between SMI influence on the passage of redistribution and recognition bills. The presence of a difference supports Fraser's position, while the absence of difference would be consistent with Honneth's position. As such, w hile this study is not sufficient to wholly accept or reject either Fraser 's or Honneth's view, it is a significant step in this direction. 9 Limited use of Fraser's recognition/redistribution typology has been utilized by at least one empirical study of social movements ( see Carroll & Ratner, 1999) However, Carroll & Ratner (1999) u se of the distinction is limited to a description of how different types of SMOs (recognition, redistribution and their addition: salvation) utilize media strategies.

PAGE 34

24 CHAPTER III METHODOLOGY For this study I developed data sets with f ederal social justice bills in the 109 th Congress (2005 2006) as my unit of analysis. After identifying social justice bills, I split these bills into either recognition bills (foc using primarily on cultural justice) or redistribution bills (focusing primarily on economic justice) to establish two distinct data sets Bills not clearly in either camp, or those solidly in both, were eliminated. This left me with two data sets of soc ial justice bills, distinguished by their relation to the Fraserian typology of injustice claims I then built each data set out, connecting SMI strength, interest group strength, political elite support, public opinion in favor and media coverage to each bill. Wi th these data sets built, I ran a rare event logistic regression on each data set as well as a combined data set (the redistribution and recognition data sets combined). This allowed me to statistically analyze each of my research questions: Rese arch Question #1: Within the contemporary United States does the strength of a social movement industry impact whether the U.S. Congress passes a bill that is important to that industry? Research Question #2: If a relationship between social movement indu stry strength and bill passage exists, is that relationship moderated by the focus of the bill on redistribution or recognition? In the following sections I discuss each variable, the development of these data sets and this analysis in more detail.

PAGE 35

25 Depe ndent V ariable: Social Justice Bills I operationalized this variable, which acted as my unit of analysis, as f ederal social justice b ills in the 109 th Congress Bills are potential laws introduced in the legislature and thus this variable holds a value o f either pass (1) or fail to pass (0) I chose the 109 th Congress due to the relative lack of major events external to the political system that might have independent effects on policy during this period (2005 2006). For example, while I could have atta ined all the same data for the 110 th Congress, this time frame would have included the onset of the "Great Recession," which undoubtedly impacted bill passage. This is not to say there were not events external to the political system that impacted policy during the 109 th Congress. For example, in 2005 Hurricane Katrina clearly impacted policy, particularly in disaster relief. Likewise, in 2006, the global outbreak of the "bird flu" impacted health policy. Indeed, at any given time, global events affect public policy. However, I maintain that in comparison to other years, during 2005 and 2006 there were relatively few major global events that substantially impacted domestic social justice policy in the United States. Bill Selection To identify bills I used key words representing several major social categories around which social justice issues and industries cluster I also explicitly excluded several major areas of social justice including : health, education, children, criminal justice and internat ional issues. I avoided these issues because each is associated with rather complex and specific policy subsystems, which would substantially complicate my analysis. Instead, I attempted to choose major social categories that cross policy

PAGE 36

26 subsystems. Ta ble III.1 below displays the search terms used for each major social category included in this study Table III.1 Search Terms Used to Identify Bills by Major Social Categories Social Category Search Terms Sexual Orientation Sexual orientation Homosexua l LGBT Same sex GLBT One man and one woman Gay Race Race Latina Ethnicity Hispanic Minority African American Latino Asian Organized Labor Organized l abor Freedom to work Labor u nion Compensation Compensation Davis Bacon* Universal wag e Maximum compensation Minimum wage Limit compensation Living wage Language Language English language Language minority Limited English Spanish Gender Gender Mother Sex Men Women Man Woman Male Female Father Poverty Poverty Homelessnes s Low income Unemployment Hunger TANF* Homeless WIC* Wealth Windfall profit Estate tax Excess profit Death tax Capital gains Alternative minimum tax * Note: These search terms relate to specific policies that fall within the major social categ ory.

PAGE 37

27 To identify the bills used for this study, I entered the search terms listed above into the advanced search function of the Library of Congress' THOMAS system, with the search limited to bills and amendments in the 109 th Congress. From these results I eliminated al l bills focused on health, education, children, criminal justice and international issues, since issues aren't addressed by this study I then viewed the CRS Summary for each bill or amendment that showed up in the search results to ensure the social category was a primary focus of the bill (when the major social category was secondary to other issues, the bill was not included). For example, using the search term "women the Mercury Free Vaccines Act of 2005 (H.R. 881) was one of the res ults. This bill would have regulated vaccines containing mercury, but because its text refers to specific regulations regarding the use of flu vaccines with "pregnant women it appeared in my search results. However, because the bill doesn't focus on wo men per se, it was eliminated. This process resulted in 460 bills. For the se bills I recorded whether the bill passed, the date of the last action taken on the bill, and whether the bill enhanced or detracted from social justice, in a Fraserian sense That is does the bill increase individuals ability to participate in society as peers (enhance social justice) or does it hinder their ability to do so (detract from social justice). Database Distinction To place each bill (and as a result each case) wit hin one of my two data sets (recognition and redistribution ) I used the following criteria. Redistribution B ills. To be considered a redistribution bill, the bill had to do at least one of the following: (1) provide or deny material resources directly t o people, (2)

PAGE 38

28 alter an economic safety net, (3) alter the tax structure, (4) regulate/deregulate wages or benefits or (5) regulate/deregulate access to work. Recognition B ills. To be considered a recognition bill, the bill had to do at least one of the f ollowing: (1) limit/open a social institution to certain "types" of people, (2) define categories of people, (3) define categories of people as deserving/undeserving or (4) signal a preference for a particular cultural artifact. Using these definitions, I then carefully reviewed the CRS Summary for each bill, recording whether each bill met each of the nine criteria listed above (five for redistribution and four for recognition). When a bill met one or more redistribution criteria it was coded as a redis tribution bill, and when it met one or more recognition criteria it was coded as a recognition bill. However, f ederal bills often include a wide variety of provisions and some of these bills include d elements that cut across the Fraserian typology laid out above (i.e. they met both recognition and redistribution criteria) To maintain the conceptual clarity of this typology, I eliminated all bills that clearly met both sets of criteria. I also, eliminated any bill that did not meet any of the criteria As such, I included only bills that clearly fall primarily within one of the two categories. This process left me with 159 redistribution bills and 211 recognition bills, for a total of 370 bills (and thus cases). Independent V ariable #1 : Social Movem ent Industries ( SMI s ) This variable is meant to measure the influence of social movement industry (SMI) strength on the probability of bill passage. To account for the fact that SMIs can often be found on both sides of a bills (for and against), I made th e value of these variables the strength of those for, less the stre ngth of those against. As this meas ure

PAGE 39

29 increase s I expect the probability of bill passage to also increase. That is as the difference between the strength of an SMI in favor of a bill, less those against, increases I expect bill passage will be more likely. I operationalized this variable in three ways: (a) the total annual revenue for SMOs (aggregated into SMIs) likely to be for the bill, less the total revenue for SMOs (aggregated in to SMIs) ; (b) the same calculation using the total end of year assets, rather than revenue ; and (c) these two values (annual revenue and end of year assets) combined. Thus, I ended up with three separate models: one that used revenue, one that used assets and one that combined revenue and assets. In all three cases, I expect that as the variable increases, bill passage is more likely. I choose to use both revenue and assets because they measure different aspects of an organization's capacity. Assets a re a better measure for organizations that have converted their financial resources into tangible assets that they utilize in their work (such as buildings or supplies). Revenue is a better measure for organizations that spend most of what they take in on nontangible assets (like services paid for). The revenue + assets model, thus, combines the two. While money in any form (including ass ets and revenue) is by no means a comprehensive measure of organizational strength, I take it to be a reasonable proxy for this study. Data Source: NCCS IRS Business Master Files Data for th is variable was derived from the Urban Institute's National Center for Charitable Statistics (NCCS) IRS Business Master Files (BMF). The BMF is a cumulative file that contains basic descriptive information for "active" (in this sense active means currently providing services) tax exempt organizations released twice a

PAGE 40

30 year. Data for this study are derived from the first data set in 2004 (BMF 04/2004) and 2005 (BMF 07/2005). Unfortun ately, the BMF does contain a few inactive organizations. Every three year s, the IRS mails postcards to organizations to verify that they exist. This reduces the number of defunct organizations included in the database, but those that have gone defunct s ince the last postcard mailing may be included. However, because I operationalized this variable using organizational revenue and assets defunct organizations shouldn't overly affect the data (as truly defunct organizations would have reported no revenue or assets). The data for the BMF is drawn from IRS f orms 1023, 1024, 990 and 990 EZ for all active and registered tax exempt organizations. Form 1023 is used by 501(c)(3) organizations, and Form 1024 is used by other 501(c) organizations, to apply for re cognition of their tax exempt status. 501(c) refers to the section of the United States Internal Revenue Code (26 U.S.C. ¤ 501(c)) that defines tax exempt organizations. Form 990 is the annual IRS return required for tax exempt organizations, which inclu des a variety of financial information. Form 990 EZ is the short form of the 990, filed by organizations with less than $100,000 in gross receipts and less than $250,000 in total assets at the end of the year. Twenty eight types of nonprofit organizatio ns are exempt from some federal income taxes (the type of exemption depends on the type of organization), the most prevalent of which are 501(c)(3)s which are charities or private foundations. Charities, in this sense, fit a rather broad set of organizat ions including those organized for the relief of the poor, the distressed or the underprivileged; advancement of religion;

PAGE 41

31 advancement of education or science; erecting or maintaining public buildings, monuments or works; lessening the burdens of governme nt; lessening neighborhood tensions; eliminating prejudice and discrimination; defending human and civil rights secured by law; and combating community deterioration and juvenile delinquency. Of these organizations, those with gross receipts under $5,000 and religious congregations are not required to register with the IRS. These organizations, unless they voluntarily chose to register, are not included in this database (NCSS 2006). The other type of 501(c)(3) organization, private foundations, are all required to register with the IRS. Of these, most are "grantmaking foundations" that were set up by an individual or family to provide grants to other 501(c) organizations (though a minority of grantmaking foundations also provide scholarships or s upport government activities). Most of these (more than 97%) have no paid staff and exist solely to provide small grants to operating nonprofits. Additionally, a small portion of private foundations are "operating foundations which use their endowment to dire ctly fund their own programming. The remainder of the BMF file consists of a variety of exempt organizations including: (a) civic leagues social welfare organizations and local associations of employees (501(c)(4)s) ; (b) labor, agricultural and horticul tu ral organizations (501(c)(5)s) ; (c) business leagues chambers of commerce r eal estate boards (501(c)(6)'s) ; and (d) many organization not relevant to this study (discussed in detail below).

PAGE 42

32 Data R eduction The BMF files contain a variety of exempt org anizations that are not relevant for my study and were thus removed This process was done in several steps the first of which was to eliminate non relevant organizations based on their IRS code. Ta ble III.2 below shows the types of organizations that w ere removed and those that were kept. Table III.2 Step 1 in Data Reduction for the BMF Master Files: IRS Code Removed Kept 501(c)(1) 501(c)(16) 501(c)(27) 501(c)(3) 501(c)(2) 501(c)(17) 501(d) 501(c)(4) 501(c)(7) 501(c)(18) 501(e) 501(c)(5) 501(c)(8) 501(c)(19) 501(f) 501(c)(6) 501(c)(9) 501(c)(20) 501(k) 501(c)(10) 501(c)(21) Charitable risk pools 501(c)(11) 501(c)(22) Farmers cooperative 501(c)(12) 501(c)(23) Qualified tuition program 501(c)(13) 501(c)(24) 527 501(c)(14) 501(c)(25) 501 (c)(15) 501(c)(26) A dditionally, grantmaking organizations (a particular type of 501(c)(3)) were removed from the database so as not to count dollars twice (i.e. not prior to, and after, a grant has been distributed).

PAGE 43

33 With the above removals, the fol lowing types of organizations remained in the data set : 501(c)(3) charities and operating foundations only 501(c)(4) civic leagues, social welfare organizations and local associations of employees 501(c)(5) labor, agricultural and horticultural organizatio ns 501(c)(6) business leagues, chambers of commerce, real estate boards I then further paired down these data by eliminating health and education organizations as well as those with purposes/ missions that are not relevant to domestic social justice. To d o this I used the National Taxonomy of Exempt Entities Core Codes (NTEE CC) classification s assigned to each organization in the database. The NTEE CC is a classification system used by the IRS and NCCS to organize entities by purpose, type or major func tion. Organizations are divided into 12 major groups and 26 major NTEE groups based on their broad topical subsector (i.e. education, medical research, housing and shelter, etc.). The 26 major groups are further subdivided by decile codes, which represe nt specific activity areas (e.g. Civil Liberties is a decile subdivision of the Civil Rights, Social Action and Advocacy Group). Some of these decile codes are again subdivided by centile codes that represent s pecific types of organizations (e.g. Censors hip, Freedom of Speech & Press is a centile subdivision of Civil Liberties). Each of these is further subdivided by seven common codes: 01 alliance/advocacy organizations 02 management and technical assistance 03 professional societies/associations

PAGE 44

34 05 research institutes and/or public policy analysis 11 monetary support single organization 12 monetary support multiple organizations 19 nonmonetary support not elsewhere classified (N.E.C.) Using these NTEE CC classifications, I was able to eliminate a dditional organizations not relevant to my study. First, I eliminated the entirety of six of the 12 major group s that were associated with health, education the environment or international work Table III.3 shows the major grou ps that were removed and th ose that remained. Table III.3 Step 2 in Data Reduction for the BMF Master Files: Major Group (12) Removed Kept Higher e ducation ( B H) Arts, culture & humanities (AR)* Education (ED) Environment (EN)* Hospitals (EH) Human services (HU) Health (HE) Mut ual benefit (MU) International (IN) Public & societal benefit (PU) Religion (RE) Unknown (UN) Note: Much of the a rts, culture and humanities code is not relevant, but a few of these organizations do engage in social justice through cultural means. The same goes for the environmental code, as some organization focused on environmental justice overlap substantially with social justice. As such I retained these codes and thinned out the non relevant organizations in subsequent reductions. Next, I elim inated the entirety of 14 major NTEE groups that were not relevant to this study because they were either not relevant to social justice generally (e.g. s ports clubs) or the entire category fell within one of the social justice categories excluded from

PAGE 45

35 th is study: health, education, children, criminal justice or international work Table III.4 shows the major NTEE groups that were eliminated and those that remained. Table III.4 Step 3 in Data Reduction for the BMF Master Files: Major NTEE Groups Removed Kept Education ( B ) Arts, Culture & Humanities ( A ) Environment ( C ) Crime & Legal Related ( I ) Animal Related ( D ) Employment ( J ) Health Care ( E ) Food, Agriculture & Nutrition ( K ) Mental Health & Crises Intervention ( F ) Housing & Shelter ( L ) Diseases, D isorders & Medical Disciplines (G) Human Services (P) Medical Research (H) Civil Rights, Social Action & Advocacy (R) Public Safety, Disaster Preparedness & Relief (M) Recreation & Sports (N) Community Improvement & Capacity Building (S) Youth Develop ment (O) International, Foreign Affairs & National Security (Q) Philanthropy, Voluntarism & Grantmaking Foundations (T) Science & Technology (U) Religion Related (X) Social Science (V) Mutual & Member Benefit (Y) Public & Societal Benefit (W) Unknow n (Z) Finally, I used decile and centile codes to eliminate additional organizations not relevant to the study Figures III.1 and III.2 below show the decile and centile codes that were removed in red and those that were not in green

PAGE 46

36 Figure III.1 NTEE CC Decile/ Centile Code Eliminations Figure III.1 NTEE CC Decile/Centile Code Eliminations

PAGE 47

37 Figure III.2 NTEE CC Decile/ Centile Code Eliminations F igure III.2 NTEE CC Decile/Centile Code Eliminations

PAGE 48

38 With these reductions, the remaining organizations include all the registered tax exempt organizations in 2004 and 2005 that are relevant for my s tudy. The data set thus includes larger formal organizations, but i t does not include exceptionally small organizations nor does it in clude informally organized groups (i.e. loose networks of individuals without a formal organization). Further, it does not fully capture the "people power" of the organizations included in the data set (i.e. people who support the organizations or their cause in non financial terms). Despite these limitations, I take these data to be an appropriate proxy for the SMI s active in the United States during 2004 and 2005. Dif ferentiating SMO s and Interest G roups I n this study I take the position that interes t groups are "political insiders" and SMOs are "political outsiders." From this perspective, interest groups are political actors that are embedded within the political structure itself and as such pursue their policy agendas through institutionalized mea ns such as lobbying. In contrast, SMOs are organizations external to the political structure that pursue their policy agendas through extra institutional means such as boycotts. Unfortunately, this distinction is rarely sharp as many organizations engag e in both institutional and extra institutional activities ( Snow & Soule, 2009). To make this distinction, I used the Internal Revenue Code (IRC) 501(c) designation Generally, the legal parameters of 501(c) organizations' participation in political ac tivity are determined by the IRC, Treasur y regulations and IRS guidance. These parameters restrict the political activity of 501(c)(3) organizations in two ways: (1)

PAGE 49

39 T hey are not allowed to intervene in political campaigns and (2) they can only conduct an "insubstantial" amount of lobbying. Lobbying includes activities that attempt to influence legislation by (a) directly contacting, or urging the public to contact, legislators about proposing, supporting or opposing legislation ; (b) advocating for or aga inst legislation ; or (c) making contributions or loans to other entities that engage in these activities. As such, these restrictions cover both formal lobbying (directly contacting governmental officials) and grassroots lobbying (appeals to the public to contact their public officials) (Lunder 2007). Though 501(c)(3)s are not prohibited from engaging in lobbying (as they are from becoming involved in political campaigns), they are restricted in the amount they can lobby. Defining this amount can happen one of two ways: (1) T he organization can elect adhere to the numerical standards provided by IRC ¤ 501(h) or (2) they can decide to have the government apply a "no substantial part" test, the standards of which are based primarily in case law. Under IRC ¤ 501(h), an organization s formal direct lobbying efforts are limited to 20% of its first $500,000 of expenditures, 15% of its second $500,000, 10% of its third $500,000 and 5% of its remaining expenditures, the total of which cannot exceed $1 million in the year. Additionally, grassroots lobbying is limited to 5% of its first $500,000 of expenditures, 3.75% of its second $500,000, 2.5% of its third $500,000 and 1.25% of its remaining expenditures, not to exceed $250,000 on grassroots lobbying in the yea r. Organizations that do not elect to be subjected to IRC ¤ are subjected to a "no substantial part" test which takes each case as unique and makes the determination through a broad examination of the organization's purpose and

PAGE 50

40 activities. However, case law suggests that "no substantial part" typically amounts to between 5% and 20% of the organization's expenditures (Lunder 2007). So organizations with a 501(c)(3) status are legally (1) prohibited from engaging in campaign politics and are (2) legally required to limit their lobbying (both direct and grassroots) to 20% or less of their expenditures. If they violate either of these provisions, they are subject to the loss of their tax exempt status. However, it is no doubt true that laws are not alway s followed and that violators are not always caught. Thus, compliance rates with these laws are just as important as the laws themselves. Though relatively scant, the available data suggest that large scale noncompliance involving substantial expenditure s is rare (Mayer 2010 ). As such, the restrictions on political activity imposed on 501(c)(3)s means that they are largely though perhaps not entirely, external to the political structure. Thus, by my definition, the IRC's 501(c)(3) designation is a rea sonable indicator that an organization is an SMO and not an interest group. In contrast, the remaining organizations in my data set (501(c)(4)s, 501(c)(5)s and 501(c)(6)s) are not explicitly restricted in the amount of expenditures they can devote to (1) i ntervening in political campaigns or to (2) lobbying (both direct and grassroots). They are however, implicitly limited by the restrictions on their primary purpose. That is these organizations can engage in unlimited campaign activity as long as it is consistent with its primary purpose, but campai gn activity itself cannot be it s primary purpose (organizations with this as their purpose are 527s). Additionally, these organizations can engage in unlimited lobbying as long as the lobbying is related to the organization's primary purpose. Further, lobbying can be these organization's sole activity as long as the purpose of the lobbying is a qualified tax exempt purpose.

PAGE 51

41 So, 501(c)(4)s 501(c)(5)s and 501(c)(6)s are interest organizations that face no l egal restrictions to lobbying for their interest. Thus, registered exempt interest organizations directly engaged with the political structure through direct or grassroots lobbying as a primary activity (or "political insiders") fall w ithin one of these c ategories. However, despite their legal ability to do so, not all organizations within these categories actively lobby. Luckily, there is an additional reason to think that organizations within this category fall closer to the "political insider" pole o f the spectrum than do 501(c)(3) organizations they are not 501(c)(3)s. Unlike the organizations within these categories, contributions to 501(c)(3)s are tax deductible as charitable contributions. Thus, it stands to reason that most 501(c)(4)s, 501(c)(5 )s and 501(c)(6)s would choose to be 501(c)(3)s if they could. However, these organizations cannot be 501(c)(3)s because their primary purpose is not consistent with the legal definition of a charity. Rather, by the IRC definitions, these organizations r epresent particular interest s (as interest groups). B ecause these organizations by definition represent particular interests and because they are allowed to expend resources on political activities without restriction, I believe they are largely political insiders. In other words, I take a 501(c)(4), 501(c)(5) or 501(c)(6) designation to indicate than an organization is an interest group. A few examples may help to illuminate the distinction I made here. First, here are examples of 501(c)(3) organizatio ns: The American Red Cross People for the Ethical Treatment of Animals Inc. (PETA) Nature Conservancy

PAGE 52

42 Each of the above organization may, at times, act politically. However, I contend that they fit the idea of an SMO, which is not embedded within the poli tical structure. In contrast, I argue that each of the following organizations more closely represent interest groups that are intimately intertwined with the political structure: National Rifle Association (NRA) (a 501(c)(4)) International Brotherhood of Teamsters (a 501(c)(5)) Association of Trial Lawyers of America (a 501(c)(6)) Here it is important to point out that exempt organizations with different designations are sometimes closely related to each other. For example, a 501(c)(3) might be paired wi th a 501(c)(6) S uch is the case with the American Bar Association (a c6 trade association) and the American Bar Association Endowment (a c3 charitable fund that supports the public service and educational programs of the American Bar Association). While such organizations are in a certain sense part of the same global entity, they are technically and legally separate organizations As part of this legal separation, their activities and funds must be kept separate. As such, despite their relationship, r elated organizations with separate designations are separate entities within my data sets. Through the processes described above, I established two subsets of the BMF data: (a) SMOs identified as organizations with a 501(c)(3) designation and (b) interes t groups identified as organizations with a 501(c)(4), 501(c)(5) or 501(c)(6) designation. The data used for this variable (SMIs) include only 501(c)(3)s, with 501(c)(4) s 501(c)(5) s or 501(c)(6) s included in the next variable (interest groups)

PAGE 53

43 Determin ing Directionality and Assigning V alues To assign values for this variable I used the NTEE CC codes to ident ify organizations for each major social category listed in T able III.1 above. When possible, I also used these codes to identify directionality. For example, I assumed that labor unions (NTEE CC code J40) were pro organized labor and pro wage increases. However, such clean designations were not always possible For these les s clear NTEE CC codes, I used Guide Star and organizational websites to i dentify their primary purpose/mission. Using this information, I attached each organization to all relevant social categories as either pro or anti. I then calculated four values for each major and minor category: (a) 2004 pro bill, (b) 2004 anti bill, ( c) 2005 pro bill and (d) 2005 anti bill. These values were calculated as follows: W = Sum of organizational monetary values (in billions) for SMOs that are in favor of the category in 2004 X = Sum of organizational monetary values (in billions) for SMOs t hat are against of the category in 2004 Y = Sum of organizational monetary values (in billions) for SMOs that are in favor of the category in 2005 Z = Sum of organizational monetary values (in billions) for SMOs that are against of the category in 2005 200 4 pro bill value = W X 2004 anti bill value = X W 2005 pro bill value = Y Z 2005 anti bill value = Z Y

PAGE 54

44 These values are calculated in billions to ensure any changes in this variable that are significantly correlated with bill passage lead to a perc eptible corrected logit coefficients within the rare events logistic regression models. 10 Independent Variable #2: Interest Groups This variable is meant to control for the influence of interest group strength on the probability of bill passage. It is cal culated in precisely the same manner as SMI strength (independent variable #1 above), except it includes 501(c)(4)s, 501(c)(5)s and 501(c)(6)s (interest organizations that face no legal restrictions to lobbying for their interest) rather than 501(c)(3)s (i nterest organizations that face substantial legal restrictions on lobbying) That is using the NCCS BMF as a data source, each bill is associated with three separate values for this variable: (a) the total annual revenue for interest groups likely to be for the bill less the total revenue for interest groups likely to be against the bill ; (b) the same calculation using the total end of year assets rather than revenue ; and (c) these two values (annual revenue and end of year assets) combined. For more de tail on this variable, see the calculations for SMIs (independent variable #1) above. Independent Variable #3: Public O pinion This variable is meant to measure the influence of public opinion on the probability of bill passage. More specifically, it meas ures the effect of favorable opinion on bill passage. As this measure increase s I expect the probability of bill passage will 10 If I were to use dollars, rather than billions of dollars, the corrected logit coeffic ients would be so small they would be imperceptible Put another way, the difference each additional dollar makes is very small, but the difference each additional billion dollars makes is rather large. Thus, to see the impact of SMI and interest group d ollars, we need units large enough to see the impact in this case, billions of dollars.

PAGE 55

45 also increase. That is as the percentage of the public in favor of a bill increases, I expect bill passage will be more likely This variable was operationalized as public opinion in favor of the concept underpinning the bill with data derived from nine sources: Associat ed Press/Ipsos poll, conducted: o January 9 12, 2006 CNN Poll, conducted: o August 30 September 2, 2006 o Decemb er 15 17, 2006 Gallop Poll, conducted: o August 9 11, 2004 o March 18 20, 2005 o April 29, 2005 o August 8 11, 2005 o May 8 11, 2006 General Social Survey, conducted: o 2004 o 2006 NBC News Poll, conducted: o April 3 5, 2005 NBC News/Wall Street Journal, conducted: o Jun e 9 12, 2006 Pew Research Center for the People and the Press Survey, conducted:

PAGE 56

46 o March 8 12, 2006 Rasmussen Reports, conducted: o June 7 8, 2006 USA Today/Gallop Poll, conducted: o November 19 21, 2004 o April 28 30, 2006 All opinion data were normalized by co nverting the scale to a percentage in favor of the bill W here public opinion data existed in multiple sources the survey that was most proximate to the final action date of the bill was used. For example, for bills related to constitutional amendments around same sex marriage, I used data from the Gallop Poll question "Would you favor or oppose a constitutional amendment that would define marriage as being between a man and a woman, thus barring marriages between gay or lesbian couples?" with three re sponse options: "Favor," "Oppose" and "Unsure." Thus, for bills to get a constitutional amendment that bans gay marriage, the public opinion value corresponds to the percentage of resp ondents who selected "Favor." Conversely any bill supporting an amend ment legalizing gay marriage would have a value corresponding with the percentage of respondents who selected "Opposed." However, because this question was asked multiple times from 2004 to 2006, the value varies slightly for virtually identical bills pr esented at different times. For example, H.J.RES.39, which has the title "Proposing an amendment to the Constitution of the United States relating to marriage," received a value of 0.53, because the pu blic opinion data closest to it s final action date (4/ 4/2005) indicated that 53% of the public would

PAGE 57

47 support this bill. Meanwhile, a virtually identical bill The Marriage Protection Amendment (S.J.RES.1) received a value of 0.5, because the pu blic opinion data closest to it s final action date (6/7/2006) indi cated that 50% of the public would support this bill. For bills with underlying concepts that lack opinion data I assigned a value of 0.5 or 50%, representing evenly split public opinion relative to the bill. This is not to say that when presented with t he bill people in the public would not form an opinion. Rather, the 0.5 value indicates that the political representatives voting for or against a bill have no reason to believe the public generally favors or generally opposes the bill. I believe this is justified because, as Paul Burstein (2010) points out, public opinion polls tend to target subjects that people have opinions about. That is people don't know enough about many subjects of legislation to have formed an opinion on them. Further, even if they have developed an opinion in sufficient numbers for there to be a public opinion, if it is not measured it's difficult to see how the opinion could influence lawmakers as they would have no sense of what the opinion was Conversely public opinio n is measured for major concepts that are in the public discourse, and as such, about which people are likely to hold an opinion. So, concepts that have public opinion data are likely to equate to (a) the major concepts on which public opinion exists, and (b) by definition, are the concepts with explicitly visible public opinion. As such, I believe it is reasonable to assign a value of 0.5 for this variable to all bills with concepts that lack public opinion data.

PAGE 58

48 Independent Variable #4: Political Elit e Allies This variable is meant to measure the influence of political elites (legislators) on the probability of bill passage, within the partisan s tructure of the 109th Congress. Because Republicans controlled both the House and the Senate in the 109th C ongress bills favored by Republicans had a better chance of passing than those favored by Democrats. Bills favored by both parties had the strongest chance to pass. Thus, as bills move from Democrat favored, to Republican favored, to bipartisan I expec t the probability of bill passage will increase. Following this logic, this variable was operationalized as one of four values: 0 = Democrat bill 1 = Republican bill 2 = Weak bipartisan bill 3 = Strong bipartisan bill To assign these values, I first align ed each bill with the party of the sponsoring legislator. I then recorded the party and count of all cosponsor s. Next, I coded Bills without any co sponsor s from the opposing party as either a Democratic bill (0) or a Republican bill (1). For bills that had both Democratic and Republican co sponsors, I used a set of formulas based on the relative ratio of Democrats and Republicans in each chamber, to designate the bill as either weakly or strongly bipartisan The Senate of the 109 th Congress had 44 Dem ocrats, 55 Republicans and 1 Independent (James Jeffords). However, because Representative Jeffords caucused with the Democrats I counted him as a Democrat for th e purposes of this study. Thus for a bill to pass at least six Republicans have to vote f or the bill (45 Democrats + 6

PAGE 59

49 Republicans = 51, or a simple majority). Conversely, a bill favored primarily by Republicans did not need any Democrats to pass (because there were more than 51 Republicans). However, Senate rules don't place time limits on debate, and thus any single Senator can delay or entirely prevent a vote on legislation by extending debate a parliamentary procedure known as the filibuster. A filibuster can be overcome if three fifths of the Senate agrees to do so by "invoking cloture ." So, while Senate bills can pass with 51 votes, 60 votes are necessary to pass a bill that is being filibustered. Thus, a bill favored by Democrats would require at least 15 Republicans to be filibuster proof (45 Democrats + 15 Republicans = 60). Conv ersely, a bill favored primarily by Republicans would need at least five Democrats to be filibuster proof (55 Republicans + 5 Democrats = 60). So a primarily Democratic bill could not pass without Republican support, while a Republican bill could pass with out any Democratic support. However, if a Republican bill were filibustered, the Republicans would need Democratic support to overcome the filibuster, and if a Democratic bill were filibustered, the Democrats would need even more Republic an support to ove rcome it. T here are different levels of bipartisanship a weak bipartisan bill that has enough support from each party to pass but not to overcom e a filibuster, and a strong bi partisan bill that has enough support from each party to overcome a filibuster. But the number of opposition support necessary depends on whether the bill is primarily a Democratic or a Republican bill (because there were 10 more Republicans than Democrats in the 109 th Senate). Thus, in my definition, bipartisan ship is defined relat ive to the level of support from the opposing party

PAGE 60

50 necessary to either (a) pass ( weak bipartisan ship ) or (b) overcome a filibuster ( strong bipartisan ship ) To make these distinctions, I used co sponsorship as an indicator of the level of opposing party su pport. Though some studies have found that co sponsorship generally (i.e. the presence or absence of co sponsorship) has only a minor effect on bill passage ( Wilson & Young, 1997 ), I take it to be an accurate measure of the partisan nature of a bill. Fo r example, bills sponsored by Democrats are considered weakly bipartisan when they have enough Republican cosponsor s to pass ( six or more) or when the ratio of Republican co sponsors over Democra tic co sponsors is greater than or equal to the number of Rep ublicans needed for passage divided by the number of votes needed for passage (6/51 or 0.12). Thus, the formulas for assigning the value to this variable for Senate bills are as follows: DS = Democratic Sponsor DCS = Democratic Co sponsor RS = Republican Sponsor RCS = Republican Co sponsor Strong Bipartisan = DS & !15 RCS or DS & RCS/DCS 0.25 or RS & !5 DCS or

PAGE 61

51 RS & DCS/RCS 0.08 Weak Bipartisan (2) = DS & !6 RCS or DS & RCS/DCS 0.12 or RS & !1 DCS or RS & DCS/RCS 0.01 Republican (1) = RS & 0 DCS Democratic (0) = DS & "5 RCS While the House rules do not allow filibustering, I used the same thresholds to assign the values for House bills (though the numbers were based on the numbers in the House). For a variety of reasons, the number of Democra tic and Republican r epresentatives varied slightly over the two years the 109 th Congress was in session. The numbers used here are from December 31, 2006 to January 3, 2007. I chose to use these numbers because they represent the smallest gap between De mocrats and Republicans during the 109 th and thus represent the point at which the partisan structure of the house should have had the weakest effect. As such, any observed effect is more likely to be real. During this time period, there were 229 Republ icans, 202 Democrats and one Independent. However, because the Independent (Bernie Sanders) caucused with the

PAGE 62

52 Democrats, I included him with the Democrats. Using these numbers, the formulas for House bills are as follows : Strong Bipartisan (3) = DS & !59 RCS or DS & RCS/DCS 0.23 or RS & !29 DCS or RS & DCS/RCS 0.11 Weak Bipartisan (2) = DS & !16 RCS or DS & RCS/DCS 0.07 or RS & !1 DCS or RS & DCS/RCS 0.01 Republican (1) = RS & 0 DCS Democratic (0) = DS & "15 RCS

PAGE 63

53 Here it is important to note tha t I use sponsorship and co sponsorship as indicators of the partisanship of the bill, but I do not consider the total number of co sponsors. I take this position because as Wilson & Young (1997) have demonstrated, the raw number of co sponsors a bill has is not substantially related to bill passage. Independent V ariable #5: Media C overage This variable is meant to measure the influence of media coverage on the probability of bill passage. Due to the logistical difficulties involved, I only measured th e amount of coverage, not the type. That is a qualitative evaluation of many thousands of articles is far more time consuming and resource intensive than a raw count. Thus, I expect the probability of bill passage to increase as media coverage increases This variable was operationalized as the number of New York Times articles in which the (a) bill number, (b) the name of the bill and/or (c) up to three sets of keywords representing the concept underpinning the bill appeared for one year prior to the l ast action taken on the bill. The value for this variable, then, is the total number of articles, which met t hese criteria. To arrive at the s e values, I entered the appropriate search criteria into LexisNexis Academic using Boolean terms and connectors a nd limiting each search to the New York Times and a one year period prior to the last action taken on the bill. The New York Times was selected because it is one of the largest nationally distributed newspapers not located in Washington D C. As would be the case with any

PAGE 64

54 single newspaper, the New York Times will not be perfectly representative on media coverage generally, but I take it to be an acceptable proxy. Analysis: Rare Events Logistic Regressions (ReLogits) My analysis is designed to measure the e ffect of SMI strength (operationalized through aggregate SMO revenue, assets or a combination of the two) on the probability of bill passage, controlling for interest groups, public opinion, political elite support and media coverage. Typically, when scho lars attempt to run such an analysis, they would use a logistic regression a statistical technique used to predict the probability that an event will occur. It does so by fitting a logit function logistic curve to data provided by several independent (or predictor) variables. This allows analysts to determine how much each independent variable contributes to the probability of the event when controlling for other variables (Agresti 2007). Sample size is an important consideration in logistic regression. Agresti (2007) argues that a minimum of 20 cases per independent variable is necessary. Because I have five independent variables, to meet this threshold I would need a minimum of 100 cases (20*5) for each data set. However, following Peduzzi et al. (1 996) and Agresti (2007), at least 10 events (in this case bill passage) per independent variable are necessary. This means I would need a minimum of 50 bills that passed (10*5 IV's) to run a standard logistic regression. This presents a problem for my s tudy because bill passage is a relatively rare event roughly 20% of the bills introduced in the 109 th Congress passed their respective

PAGE 65

55 chamber 11 As such, my data sets would each need a minimum of 250 cases each to run a standard logistic regression. Give n the scope of this study, this would require a substantial amount of work. However, I was able to reduce the size of my databases, while at the same time improving my statistical rigor, by using the rare even ts logistical regression, or rel ogit, techniqu e developed by Gary King and Langche Zeng (2001a, 2001b). King and Zeng developed the rel ogit technique by adapting econometric and epidemiology techniques to handle exceptionally rare events in political science tens of events in thousands of cases as w ell as small sample sizes (fewer than 200 observations). The technique is designed to overcome three major problems that arise in logistic regression with rare events and/or small samples: (a) logit coefficients are biased in small samples of fewer than 2 00 observations ; (b) even with large sample sizes, estimated event probabilities of rare events are always too small ; and (c) for rare events data, the most common method of computing probabilities of events in logistic regressions leads to errors in the s ame direction as bias es in the coefficients (King & Zeng 2001b). The rel ogit technique overcomes these problems with mathematical corrections 12 and is performed in virtually the same wa y a s a logistic regression. However, unlike a logistic regression, a rare events logistic regression is an unbiased estimator, rather than a likelihood technique. As such, the output of a rel ogit is slightly different than a l ogit 11 In 2005, 19.3% of the bills introduced passed in the Senate and 22.9% in the House. In 2006, these numbers were 21.5 % and 21.1% respectively. 12 For the mathematical proofs of the ReLogit technique, see King and Zeng (2001a); for a less technical account with examples, see King and Zeng (2001b).

PAGE 66

56 Instead of logit coefficients and standard errors rel ogits compute bias corrected coeffici ents and robust standard errors Analysts calculate quantities of interest from these by setting each independent variable ( IV ) at specific values. In particular, the probability of the dependent variable ( DV ) can be calculated with each IV set at specif ic values Additionally, relative risk s can be calculated. Relative risks are the likelihood (or risk) of an event, given chosen values of the explanatory variables relative to a baseline of these values. 13 Finally, because a rel ogit simultaneously redu ces bias and fit, pseudo R 2 is typically not reported as it is taken for granted that the reduction in bias comes at the cost of a reduction in fit. However, because most readers will be more familiar with logit models than rel ogit models, I report pseudo R 2 from standard logit models with the caveat that goodness of fit is not a primary consideration in r e logit In my analysis, I b egin by providing descriptive statistics for each variable. I then provide the results of r e logit s in each data set ( redistri bution recognition and combined) with bill passage as the dependent variable, and each independent variable alone. N ext, I display the results of full r e logit models o n each data set with all independent variables included in the models. I then investig ate how the probability of bill passage shifts with changes in each independent variable. To do so, I graph the change in the probability of bill passage as each independent variable shifts from its minimum value to it s maximum value, while holding all ot her independent variables at their mean. I also run relative risks for each variable, from their minimum value to their maximum value and from their 13 Relative risk is in some ways, similar to odds ratio, which is more common in Logit models. However, the two are distinc t statistical concepts. Where p is the probability of an event under one set of circumstances, and q is the probability of the event under a second set of circumstances, relative risk is p/q, but odds ratio is [p/(1 p)] / [q/(1 q)].

PAGE 67

57 20 th percentile to their 80 th holding all other independent variables at their mean. Finall y, I further i nvestigate d how SMI s impact bill passage by constructing four ideal typical scenarios: SMI s Against the World where all IVs are set at their 20 th percentile. SMI s Riding a Tide of Support where all IVs are set at their 80 th percentile. People vs. Power wh ere public opinion is set at it s 80 th percentile, while all other IVs are set at their 20 th percentile. Key Elite Allies where political elites are set at their 80 th percentile, while all IVs are set at their 20 th percentile. These scenarios allow m e to investigate the extent to which SMI influence on bill passage is dependent upon context. In the first scenario, SMIs are operating in a context that is not favorable to bill passage as interest group support is low, support from political elites is w eak, public opinion is unfavorable and media coverage is scant. The second scenario flips these circumstances, with high interest group support, strong support from political elites, favorable public opinion and substantial media coverage. The final two scenarios investigate the relative importance of public opinion and political elite support when combined with SMI strength. In the third scenario public opinion is favorable, despite a lack of interest group support, support from political elites and me dia coverage. In the fourth scenario, support from political elites is high, despite a lack of interest group support, unfavorable public opinion and very little media coverage. Thus, by mathematically simulating a variety of scenarios, I am able to appr oximate the effect context has on SMI's ability to impact bill passage.

PAGE 68

58 Overview of Methods For this study I developed two distinct data sets, one associated with recognition bills and one associated with redistribution bills in order to determine if th e relationship between SMI and bill passage is moderated by this distinction I then attached SMI strength, interest group strength, political elite support, public opinion in favor and media coverage to each bill. Next I ran a rare event logistic regres sion on each data set as well as a combined data set (the redistribution and recognition data sets combined). Finally, I further investigated how SMIs impact bill passage by constructing four ideal typical scenarios and calculating how the strength of SMI s correlates with the probability of bill passage under a variety of circumstances.

PAGE 69

59 CHAPTER IV RESULTS My analysis is designed to measure the effect of SMI strength on the probability of bill passage, controlling for interest groups, public opinion, political elite support and media coverage. To do so I utilized the rare ev ents logistical regression, or r e logit technique developed by Ga ry King and Langche Zeng (2001a; 2001b). R el ogit ove rcomes the bias that occurs in l ogit models when the dependent variable is a rare event, such as bill passage. For a mor e comprehensive account of the rel ogit technique, see the methods section above. In this section, I b egin by providing basic descriptive statistics for each variable. I then run rel ogits in each data set ( redistribution recognition and combined) with bill passage as the dependent variable, and each independent variable alone. Next, I run full rel ogit models in each data set with all independent variables included in the models. I then investigate how the probability of bill passage shifts with changes in each independent variable. To do so, I graph the change in the probability of bill passage as eac h independent variable shifts from its minimum value to it s maximum value while holding all other independent variables at their mean. I also run relative risks for each variable, from their minimum value to their maximum value and from their 20 th percent ile to their 80 th holding all other independent variables at their mean. Finally, I further investigate how SMI s impact bill passage within various contexts by constructing four ideal typical scenarios: SMI s Against the World where all IVs are set at th eir 20 th percentile, and thus SMIs are operating in a context that is not favorable to bill passage.

PAGE 70

60 SMI s Riding a Tide of Support where all IVs are set at their 80 th percentile, and thus SMIs are operating in a contest favorable to bill passage. Peopl e vs. Power where public opinion is set at it s 80 th percentile, while all other IVs are set at their 20 th percentile, and thus public opinion is favorable but all other variables are unfavorable for bill passage. Key Elite Allies where political elites are set at their 80 th percentile, while all IVs are set at their 20 th percentile, and thus poli tical elite support is favorable but all other variables are unfavorable for bill passage. As such, I am able to approximate the effect of context on SMI's i nfl uence on bill passage by mathematically si mulating a variety of scenarios Descriptive Statistics Dependent Variable: Social Justice Bills In the 109th Congress, 13,072 bills were introduced (8,152 in the House and 4,920 in the Senate). Of these, 2,103 (1 6.1%) passed their respective chamber (upper house/Senate, or lower house/House of Representatives). My data set contains a total of 370 bills (cases) from the 109th Congress, of which 58 (15.7%) passed their respective chamber. As such, the rate of pass age in my data set is comparable to, but slightly lower than for the 109th Congress as a whole. Table IV.1 displays how many bills in my data set were in each chamber and how many of those passed.

PAGE 71

61 Table IV 1 Number of Bills by Chamber Legislature Nu mber of bills Number that passed Percent that passed House of Representatives 253 34 13.4 Senate 117 24 20.5 Combined 370 58 15.7 As shown in Table IV.1 of the bills in this data set, those in the Senate passed at a higher rate than those in the Hou se. Given the different rules in each chamber, as well as their different structures and ratios of Republicans to Democrats (1.13 in the House, 1.22 in the Senate), it is plausible that this is a meaningful difference. Additionally as demonstrated by a biserial correlation, there is a significant correlation at the 0.1 level ( r ( 368) = 0.09 p = 0.082) between bill passage and the chamber in which the bill was considered Another potentially important characteristic of these bills is the time frame in wh ich a bill is considered. It is plausible that the relative proximity to elections would influence bill passage as legislators become more sensitive to the way their actions affect their re election efforts. Table IV.2 below displays the number of bills considered in each year, by the chamber in which they were considered. Table IV 2 Number of Bills by Year and Chamber Legislature Number of 2005 bills Number of 2006 2005 & 2006 Combined House of Representatives 174 79 253 Senate 78 39 117 Combined 2 52 118 370

PAGE 72

62 As shown above, about two thirds of the bills in this data set were considered in 2005. However, as demonstrated by a biserial correlation, within these data there does not appear to be a correlation ( r (368) = 0.06, p = 0.284) between bill pa ssage and the year in which the bill was considered. As such, the year in which a bill was considered does not matter in and of itself relative to bill passage. Yet a biserial correlation demonstrated that the distinction between redistribution and recog nition is correlated with bill passage ( r ( 209) = 0.20 p = 0.003). Table IV.3 below displays the breakdown of bills by redistribution and recognition. Table IV.3 Number of B ills Data set Number of bills Number that passed Percent that passed Redistribu tion 159 10 6.3 Recognition 211 48 22.7 Combined 370 58 15.7 As shown above, this data set contains more recognition bills than redistribution bills (43% are redistribution bills, 57% are recognition bills). Additionally, recognition bills passed at a substantially higher rate with just over one fifth (22.7%) of recognition bills passing through their chamber, compared with just 6.3% of the redistribution bills. Further, as described above, there is a statistically significant correlation between a redistribution /recognition designation and bill passage. This is the first evidence that the redistribution /recognition distinction matters in terms of the likelihood of bill passage a theme that we will see repeatedly played out in these findings.

PAGE 73

63 Inde pendent Variable #1: Social Movement Industries (SMI s) As described in the Methods section abov e, I measure the strength of SMI s using three different monetary measures: annual revenue, assets and revenue plus assets. For each of these, I subtract the tot al value for all SMOs likely to be against the bill, from the total value of SMOs likely to be in favor of the bill. I then divide this value by one billion to arrive at the value for this variable (a more entailed description can be found in the M ethods section above). Thus, the SMI variable is a measure of financial resources for all SMOs likely to be in favor of a bill, minus the financial resources for all SMOs likely to be against a bill. Table IV.4 below displays some basic descriptive statistics fo r SMO strength measured by revenue. Table IV 4 Descriptive Statistics for SMI Revenue ($ in Billions) Data set Median Mean(SD) 20 th Percentile 80 th Percentile Redistribution (n = 159) 0.12 0.42(1.680) 0.68 0.77 Recognition (n =211) 0.29 0.35(0.781) 0.10 0.33 Combined (n = 370) 0.24 0.38(1.248) 0.06 0.44 Note: The values for this variable consist of the combined dollar value (in billions) of the revenue for SMOs likely to be in favor of a bill, minus the combined dollar value (in billions) of th e revenue for SMOs likely to be against the bill. As such, negative values occur when the revenue for the SMOs likely to be against a bill exceed those likely to be for a bill. As shown above, values in the redistribution dataset fall within a wider rang e than do values in the recognition dataset (compare the standard deviations and the range between the 20 th and 80 th percentile values). Additionally, while the mean value for the redistribution dataset is higher than the mean value for the recognition da taset an

PAGE 74

64 independent samples t test indicated that this difference is not statistically significant ( t (209 ) = 0.50 p = 0 .616 ). This suggests that the values within the recognition data set are generally less extreme than those in the redistribution data set that is the values are generally closer to 0 but that we cannot say that the difference their means is real These findings are mirrored by Table IV.5 below which display s descriptive statistics for SMI strength measured by assets. Table IV 5 Des criptive Statistics for SMI Assets ($ in Billions) Data set Median Mean(SD) 20 th Percentile 80 th Percentile Redistribution (n = 159) 0.12 0.88(4.567) 1.28 1.03 Recognition (n = 211) 0.51 0.41(0.518) 0.08 0.63 Combined (n = 370) 0.41 0.61(3.022) 0.17 0.73 Note: The values for this variable consist of the combined dollar value (in billions) of the assets for SMOs likely to be in favor of a bill, minus the combined dollar value (in billions) of the assets for SMOs likely to be against the bill. As suc h, negative values occur when the assets for the SMOs likely to be against a bill exceed those likely to be for a bill. As shown above, the range for values in the redistribution data set is larger than the range for the recognition data s et. Additionall y, though the mean value is higher in the redistribution dataset this difference is not statistically significant ( t (161) = 1.28, p = 0.201) A sim ilar pattern is apparent for SMI strength measured by revenue + assets as shown in Table IV.6 below.

PAGE 75

65 Tabl e IV.6 Descriptive Statistics for SMI Revenue + Assets ($ in Billions) Data set Median Mean(SD) 20 th Percentile 80 th Percentile Redistribution (n = 159) 0.26 1.30(6.117) 1.96 1.91 Recognition (n = 211) 0.79 0.76(1.228) 0.18 0.93 Combined (n = 370) 0 .68 0.99(4.117) 0.23 1.17 Note: The values for this variable consist of the combined dollar value (in billions) of the revenue and assets for SMOs likely to be in favor of a bill, minus the combined dollar value (in billions) of the revenue and assets fo r SMOs likely to be against the bill. As such, negative values occur when the revenue for the SMOs likely to be against a bill exceed those likely to be for a bill. As with the previous models values fall within a far larger range in the redistribution data set, but the mean value is not significantly ( t ( 168 ) = 1.10 p = 0. 275 ) larger Additionally, for all models there is a substantial difference between the mean and median values in the redistribution dataset, indicating that a few cases have substant ially higher values than most (and as a result, pull the mean up well beyond the median). Together these data suggest that SMI strength varies substantially more within the r edistribution dataset, and that there are a few extreme values (values relatively far from 0) within it Independent Variable #2: Interest Groups Interest group values are calculated in the same manner as the SMI variable and interest groups are distinguished from SMOs based on their IRS designation (for further details on this distin ction, see the M ethods section above). Thus, the interest group variable is a measure of financial resources for all interest groups likely to be in favor of a bill, minus the financial resources for all interest groups likely to be against a bill.

PAGE 76

66 Basic descriptive statistics for the revenue measure of this variable are displayed in Table IV.7 below. Table IV.7 Descriptive Statistics for Interest Groups Revenue ($ in Billions) Data set Median Mean(SD) 20 th Percentile 80 th Percentile Redistribution (n = 159) 0.04 0.62(4.252) 0.00 7.04 Recognition (n = 211) 0.02 0.12(0.846) 0.00 0.04 Combined (n = 370) 0.02 0.34(2.866) 0.00 0.07 Note: The values for this variable consist of the combined dollar value (in billions) of the revenue for interest groups likely to be in favor of a bill, minus the combined dollar value (in billions) of the revenue for interest groups likely to be against the bill. As shown above, there is generally more interest group support (as measured by revenue) for bills in the redis tribution data set than there is in the recognition data set. However the mean value is not significantly ( t ( 167 ) = 1.46 p = 0. 146 ) larger in the redistribution data set. Additionally, interest group values in the redistribution data set cover a much w ider range than do those in the recognition data set (to see this, look at the standard deviations and compare the difference between the 20 th and the 80 th percentile values) The assets measure of this variable lead to similar results.

PAGE 77

67 Table IV 8 Descriptive Statistics for Interest Groups Assets ($ in Billions) Data set Median Mean(SD) 20 th Percentile 80 th Percentile Redistribution (n = 159) 0.02 0.75(5.479) 0.00 8.20 Recognition (n = 211) 0.02 0.14(1.090) 0.00 0.02 Combined (n = 370) 0.02 0.40(3.691) 0.00 0.05 Note: The values for this variable consist of the combined dollar value (in billions) of the assets for interest groups likely to be in favor of a bill, minus the combined dollar value (in billions) of the assets for interest groups likely to be against the bill. As shown above, the range for values in the redistribution data set is larger than the range for the recognition data set. However the mean value is not significantly ( t ( 167 ) = 1.36 p = 0.1 75 ) higher in the redistribution dataset. Finally, the revenue + assets measure again produces similar results. Table IV 9 Descriptive Statistics for Interest Groups Revenue + Assets ($ in Billions) Data set Median Mean(SD) 20 th Percentile 80 th Percentile Redistribution (n = 159) 0.06 1.37(9.719) 0.00 15.24 Recognition (n = 211) 0.04 0.27(1.936) 0.00 0.07 Combined (n = 370) 0.04 0.74(6.548) 0.00 0.12 Note: The values for this variable consist of the combined dollar value (in billions) of the revenue and assets for interest groups likely to be in favor of a bill, minus the combined dollar value (in billions) of the revenue and assets for interest groups likely to be against the bill. As before, the mean value is not significantly ( t ( 167 ) = 1.41 p = 0. 161 ) higher i n the redistribu tion dataset, but they do fall within a wider range. These d ata suggest that, similar to SMI strength, interest group support varies substantially more within the

PAGE 78

68 redistribution dataset, and that there are a few extreme values (values relatively far from 0) within it Independent Variable #3: Public Opinion Public opinion values represent the portion of the public in favor of the concept associated with a bill. A value of 0.5 (indicating exactly half of the public is in favor of the bill) was assigned to bills where public opinion data is unavailable (see the M ethods section for a more thorough description of this variable). Table IV.10 Descriptive Statistics for Public Opinion (in Favor of Bill) Data set Median Mean(SD) 20 th Percentile 80 th Percentile Redistribution (n = 159) 0.50 0.55(0.157) 0.48 0.77 Recognition (n = 211) 0.50 0.59(0.128) 0.50 0.73 Combined (n = 370) 0.50 0.57(0.142) 0.50 0.73 Note: Values for this variable represent the percentage of the public in favor of the concepts that underpin the bill. When such data was not available, a value of 0.5 was assigned. As shown above, public opinion is slightly skewed toward public support (mean values are slightly greater than 0.5) Additionally an independent samples t test indicated that there is a statistically significant difference between mean public opinion in the data sets ( t (368 ) = 2.50 p = 0 .0 1 3 ) This suggests that there is generally more public support for the bills within the recognition data set. Independent Variable # 4: Political Elite Allies This variable is operationalized as legislator support (identified by bill sponsorship) where values progress from only minority party (Democrat) support to s trongly b ipartisan support ( 0 = Democratic support only, 1 = Republican support only, 2

PAGE 79

69 = weak bipartisan support and 3 = strong bipartisan support ). Table IV.11 below shows the percent of bills that fall within each of these values in all three data sets. Table IV 11 Percent of Bills that Fall within each Political Elite S upport (in Favor of Bill) Category Data set Democrats Only Republicans Only Weak Bipartisan Strong Bipartisan Redistribution (n = 159) 51.6% 26.4% 4.4% 17.6% Recognition (n = 211) 48.8% 8.5% 15.6% 27.0% Combined (n = 370) 50.0% 16.2% 10.8% 23.0% A s shown above, roughly half of the b ills in each data set have only D emocratic (minority party only) support. However, a much larger portion of recognition bills have bipartisan support ( M = 1.21, SD = 1.300) than do redistribution bills ( M = 0.88, SD = 1 .122) and an independent samples t test demonstrated that this difference is statistically significant ( t (361) = 2.60, p = 0.01). This suggests that redistribution bills are generally more partisan than recognition bills Independent Variable #5: Media Coverage This variable is operationalized as the number of New York Times articles over the previous 12 months that mention either the bill or several associated key words divided by 100 Thus, a value of 1 indicates that 100 articles mentioned either th e bill or keywords associated with it, in the past 12 months (see the methods section for more detail). Table IV.12 below displays basic descriptive statistics for this variable.

PAGE 80

70 Table IV.12 Descriptive Statistics for Media Coverage (in 100s of Articl es) Data set Median Mean(SD) 20 th Percentile 80 th Percentile Redistribution (n = 159) 1.35 2.55(3.174) 0.19 4.81 Recognition (n = 211) 0.33 2.23(4.033) 0.02 3.12 Combined (n = 370) 0.65 2.37(3.687) 0.05 3.86 As shown above, bills in the redistribu tion data set generally had more media coverage than their counterparts in the recognition data set however this difference is not statistically significant ( t ( 368 ) = 0.82 p = 0. 412 ) Uncontrolled Direct Effects: Single ReLogit Models This section contai ns the results of rare events logistic r egressions (rel ogit s ) with each independent variable (in all their forms) alone such that each model contains only two variables (bill passage as the dependent variable, and one of the independent variables). In ot her words, these analyses are uncontrolled. These results are displayed in T able IV.13 below, broken out by results for the redistribution data set, the recognition data set an d the undifferentiated data set

PAGE 81

71 Table IV.13 Significance and Directio n of Uncontrolled ReLogit Estimates p< 0.10 ** p< 0.05 *** p< 0.001 As Table IV.13 shows, the relationship between all but one of these variables is statistically significant within at least one data set In particular, the media coverage variable is not statistically si gnificant for any of these data sets. This suggests that the raw quantity of media coverage associated with a bill is not correlated with that bill's passage However, because this variable does not take directionality into account (i.e. whether a parti cular article was purely descriptive or takes a pro/anti stance on the bill), it is possible that media coverage is correlated with bill passage, but that directionality needs to be taken into account to measure this correlation. Yet media coverage is alon e in its lack of a significant correlation with bill passage. Elite Political Allies and Public Opinion are significant for all data sets, while SMI s and Interest Groups (in their various forms) are only significant in some of the data sets. This suggest s that each of these variables when taken alone, is correlated with bill Data set Redistribution (n = 159) Recognition (n = 211) Combined (n = 370) SMI (revenue) No(+) Yes(+)** Yes(+)* SMI (assets) Yes(+)** Yes(+)*** No(+) SMI (assets + revenue) Yes(+)* Yes(+)*** No(+) Interest Groups (r evenue) No( ) Yes(+)*** Yes( )* Interest Groups (assets) No( ) Yes(+)*** Yes( )* Interest Groups (rev enue + assets ) No( ) Yes(+)*** Yes( )* Elite Political Allies Yes(+)*** Yes(+)*** Yes(+)*** Public Opinion Yes( )*** Yes(+)* Yes(+)* Media Coverage No (+) No( ) No( )

PAGE 82

72 passage and that the distinction between redistribution and recognition matters. While these correlations are not controlled this finding is consistent with my hypothesis that the recognition/ redistribution distinction has implications for the correlation between SMI strength and bill passage. Controlled Effect: Full ReLogit Models To further investigate the correlation between SMI strength and bill passage, three separate models w ere developed based on the three di fferent monetary measures of SMI and aggregate interest group strength. The general model holds bill pas sage as the DV, and includes SMI s, interest groups, elite political allies, public opinion and media coverage as IVs The three variations on this general model use the same measures for elite political allies, public opinion and media coverage, but different measures for SMOs and interest groups. The models are as follows Revenue model: B oth SMI and aggregate interes t group strength are measured through revenue. Assets model: B oth SMI and aggregate interest group strength are measured through assets. Revenue + Assets model: B oth SMI and aggregate interest group strength are measured through the combination of revenue plus assets. Rel ogits were run on each of these models within each of the three data sets ( redistribution recognition and combined) to asse ss the potential for different relationships on redistribution and recognition Table IV.14 bel ow shows these resul ts for the Revenue M odel

PAGE 83

73 Table IV 14 Corrected Logit Coefficient Estimates (Robust SE ) for Revenue Model Note: Unlike standard logit models, the rel ogit technique does not report measures of pseudo R 2 because the architects of the technique do not believe they are meaningful measures (see the M ethods section for a more in depth description). As s uch I ran a standard l ogit model with these variables to calculate a Cox & Snell R 2 (C&S R 2 ) and a Nagelkerke R 2 (N R 2 ) for this model in each data set: Redistribution C&S R 2 = 0.077, N R 2 = 0.206; Recognition C&S R 2 = 0.123, N R 2 = 0.187; Combined C&S R 2 = 0.083, N R 2 = 0.143. However, it should be kept in mind that the adjus tments rel ogit makes to reduce bias also reduce fit. p< 0.10 ** p< 0.05 *** p< 0.01 As shown above, in the Revenue Model, both the public opinion and the elite political allies variables are significantly correlated with bill passage in all three data sets However, public opinion is negatively correlated with bill passage in the redistribution data set but positively correlated in both the recognition and combined data sets. A dditionally, interest groups are negatively correlated with bill passage in the combined data set. Finally, SMI strength is correlated with bill passage in the recognition data set, Variable Redistribution (n = 159) Recognition (n =211) Combined (n = 370) SMI (revenue) 0.208 0.461* 0.028 (0.139) (0.274) (0.125) Interest Gr oups (revenue) 0.019 0.889 0.062** (0.034) (9.251) (0.029) Elite Political Allies 0.746*** 0.481*** 0.582*** (0.209) (0.140) (0.123) Public Opinion 2.335* 2.227* 1.609* (1.392) (1.252) (0.949) Media Coverage 0.164 0.035 0.033 (0.108) (0.08 2) (0.070)

PAGE 84

74 but not in the redistribution or combined data sets. I expand on these r esults further in the variable specific sections below. To ensure these results did not suffer from multicollinearity 14 I calculated Variance Inflation Factors (VIFs) 15 and tolerance 16 values for this model. These are presented in Table IV.15 below. Table I V .15 VIF and (Tolerance) for Revenue Model 14 Multicollinearity is a problem that arises when independent variables are highly correlated with each other, and as a result, it is difficult to tell which independent variable is actually producing an effect in the dependent variable. 15 Variance Inflation Factors (VIFs) is a measure of the severity of multicolliniarity. Many scholars consider VIFs of greater than 10 problematic ( Hair, Anderson, Tatham& Black 1995; Menard, 1995), and those greater than 4 to be potentially concerning (Menard, 1995). 16 Tolerance is the inverse o f VIF, and thus tolerances below 0.25 (or a VIF greater than 4) indicate a potential multicollinearity problem (Menard, 1995). Variable Redistribution (n = 159) Recognition (n = 211) Combined (n = 370) SMI (revenue) 1.68 1.04 1.22 ( 0.594 ) ( 0.961 ) ( 0.823 ) Interest Groups (revenue) 1.27 1.04 1.12 ( 0.790 ) (0. 965 ) ( 0.894 ) Elite Polit ical Allies 1.19 1.01 1.06 ( 0.839 ) (0. 986 ) ( 0.940 ) Public Opinion 1.29 1.06 1.09 ( 0.774 ) (0. 946 ) ( 0.920 ) Media Coverage 1.49 1.07 1.10 ( 0.673 ) (0. 938 ) (0. 906 ) Mean VIF 1.38 1 .0 4 1.12

PAGE 85

75 As shown above, all tolerances are well above 0.2 5 (or a VIF o f 4 ), which is generally the conservative cutoff point at which multicollinearity is considered potentially problematic (Menard, 1995) Additionally, the highest mean VIF is 1.38 which is well below 4 (the cutoff for concern about multicolliniarity). As such, multicolliniarity does not appear to be a problem with these data. The Assets Model produced similar results, though public opinion became insignificant in both the redistribution and recognition data sets (though it remained significant in the com bined data set). The results of the rel ogit for the Assets Model are displayed in Table IV.16 below.

PAGE 86

76 Table IV 16 Corrected Coefficien t Estimates (Robust SE) for Assets Model Note: Unlike standard logit models, the rel ogit technique does not report measures of pseudo R 2 because the architects of the technique do not believe they are me aningful measures (see the methods section for a more in depth description). As su ch, I ran a standard l ogit model with these variables to calculate a Cox & Snell R 2 and a Nagelkerke R 2 for this model in each data set: Redistribution C&S R 2 = 0.078, N R 2 = 0.208; Recognition C&S R 2 = 0.124, N R 2 = 0.188; Combined C&S R 2 = 0.088, N R 2 = 0.151. However, it should be kep t in mind that the adjustments rel ogit makes to reduce bias also reduce fit. p< 0.10 ** p< 0.05 *** p< 0.01 As shown above, the Assets Model closely mirrors the Revenue model, and once again SMI strength is correlated with bill passage in the recognition data set but not in the redistribution or combined data sets. Like with the revenue model, I calculated VIFs and tolerance values for th ese data to ensure these results did not suffer from multicollinearity. Th ese are presented in Table IV.17 below. Variable Redistribution (n = 159) Recognition (n = 211) Combined (n = 370) SMI (assets) 0.048 1.363* 0.042 (0.053) (0.810) (0.032) Interest Groups (assets) 0.002 18.175 0.059*** (0.025) (27.097) (0.032) Elite Political Allies 0.697*** 0.493*** 0.617*** (0.196) (0.136) (0.120) Public Opinion 2.056 1.746 1.9 84** (1.345) (1.280) (0.949) Media Coverage 0.151 0.018 0.006 (0.103) (0.066) (0.057)

PAGE 87

77 Table IV.17 VIF and (Tolerance) for Assets Model As shown above, all tolerances are well above the 0.2 threshold (Menard, 1995), and the highest mean VIF is 1.34 so multicolliniarity does not appear to be a problem with these data. These findin gs are again similar in the Revenue + Assets Model, though in this model public opinion is significant in the recognition and combined data sets but not in the redistribution data set. Table IV.18 below displays the rel ogit results for the Revenue + Asse ts Model. Variable Redistribution (n = 159) Recognition (n = 211) Combined (n = 370) SMI ( assets ) 1.56 1.32 1.18 (0. 642 ) (0. 758 ) ( 0.850 ) Interest Groups ( assets ) 1.22 1.30 1.09 (0. 820 ) (0. 768 ) ( 0.917 ) Elite Political Allies 1.16 1.01 1.05 (0. 865 ) (0. 994 ) (0 .917 ) Public Opinion 1.31 1.06 1.11 (0. 761 ) (0. 941 ) (0. 898 ) Media Coverage 1.44 1.08 1.10 (0. 695 ) (0. 924 ) (0. 910 ) Mean VIF 1.34 1.15 1.10

PAGE 88

78 Table IV 1 8 Corrected Logit C oefficients (Robust SE) for Revenue + Assets Model Note: Unlike standard logit models, the rel ogit technique does not report measures of pseudo R 2 because the architects of the technique do not believe they are meaningful measures (see the M ethods section for a more in depth descripti o n). As such, I ran a standard l ogit model with these variables to calculate a Cox & Snell R 2 and a Nagelkerke R 2 for this model in each dataset: Redistribution C&S R 2 = 0.079, N R 2 = 0.210; Recognition C&S R 2 = 0.125, N R 2 = 0.190; Combined C&S R 2 = 0. 085, N R 2 = 0.147. However, it should be kep t in mind that the adjustments rel ogit makes to reduce bias also reduce fit. p< 0.10 ** p< 0.05 *** p< 0.01 Once again, I calculated VIFs and tolerance values for these data as presented in Table IV.1 9 below. Variable Redistribution (n = 159) Recognition (n = 211) Combined (n = 370) SMI (revenue + assets) 0.013 0.374* 0.023 (0.044) (0.218) (0.027) Interest Gr oups (revenue + assets) 0.004 3.088 0.032*** (0.015) (7.091) (0.012) Elite Political Allies 0.715*** 0.481*** 0.610*** (0.201) (0.140) (0.121) Public Opinion 2.038 2.113* 1.878** (1.363) (1.256) (0.951) Media Coverage 0.158 0.032 0.012 (0. 104) (0.080) (0.059)

PAGE 89

79 Table IV.1 9 VIF and (Tolerance) for Revenue + Assets Model As shown above, all to lerances are well above the 0.2 threshold (Menard, 1995), and the highest mean VIF is 1.36 so multicolliniarity does not appear to be a problem with these data. As each of the rel ogit tables above demonstrate, the presence of elite political allies is po sitively correlated with bill passage in all three models and across all three data sets. Additionally, public opinion is positively correlated with bill passage for all three models in the combined data set, and for two of the models in the recognition d ata set. However, public opinion is negatively correlated with bill passage for the Revenue Model in the redistribution data set. Similarly, interest groups are negatively correlated with bill passage for all three models in the combined data set. Final ly, in all three Variable Redistribution (n = 159) Recognition (n = 211) Combined (n = 370) SMI (revenue + assets) 1.63 1.10 1.20 (0. 613 ) (0. 905 ) (0. 830 ) Interest Groups (revenue + assets) 1.24 1.11 1.1 0 (0. 809 ) (0. 902 ) (0. 908 ) Elite Political Allies 1.17 1.01 1.05 (0. 853 ) (0. 991 ) (0. 948 ) Public Opinion 1.31 1.06 1.11 (0. 762 ) (0. 947 ) (0. 902 ) Media Coverage 1.47 1.06 1.11 (0. 679 ) (0.941 ) (0. 901 ) Mean VIF 1.36 1.07 1.12

PAGE 90

80 models, SMI strength is positively correlated with bill passage in the recognition data set, but not in the redistribution or combined data sets. To get a better sense of how to appropriately interpret these results, I provide graphs of p robability of bill passage and relative risks 17 for each independent variable that is significantly correlated with bill passage These results are presented in the sections below, organized by independent variable. For each of the following sections, onl y statistically significant results are displayed. Political E lite Allies The presence of political elite support is positively correlated with bill passage in all three models, across all three data sets. That is as bills progress from minority party su pport only to strongly bipartisan support, the probability of bill passage increases. To visualize this correlation, I graphed the probability of bill passage (absolute risk) as the political elite variable increases, holding all other independent variabl es at their mean levels. Figure IV.1 below displays these results for the combined data set 17 Relative risk is a ratio of probabilities at two different set values for the IVs. See the Methods section for a more in depth description of relative risk.

PAGE 91

81 Figure IV.1 Probability of Bill Passage by Political Elite Support for All Bills (n = 370) Note: All other independent variables were set at their mea n. These results are further supported by Table IV.20 below, which shows the relative risk of bill passage as political elite support progresses from its minimum to its maximum level (RR min to max), as well as the relative risk of bill passage as politica l elite support progresses from its 20 th percentile value to it s 80 th percentile value (RR p20 to p80). Table IV 20 Relative Risk of Bill Passage for Partisan Progression for All Bills (n = 370) Note: All other independent variables were set at their mean. As shown above, in the combined data set, all three models show a substantial increase in the probability of bill passage as political elite suppor t increases, even when co ntrolling for mean levels of SMI strength, interest group strength, public opinion and media coverage. Indeed, in all three models the probability of bill passage rises from under 8% \# \"+# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# +# \# +# ;# @# $()*+*,-,./0)102,--0$+33+450 $+(.,3,+60$()4(533,)60 &4<4284#H.94G# C??40?#H.94G# &4<4284#`#C??40?#H.94G# Model RR min to max RR p20 to p80 Revenue 4.240 4.182 Ass ets 4.537 4.522 Revenue + Assets 4.462 4.451

PAGE 92

82 to over 32% as political elite support increas es from minority party support only to strongly bipartisan Similar results are found in the redistribution data set as shown by Figure IV.2 below. Figure IV. 2 Probability of Redistribution Bill Passage by Partisan Progression (n = 159) Note: All othe r independent variables were set at their mean. Table IV.21 supports these findings. Table IV 21 Relative Risk of Redistribution Bill Passage for Partisan Progression (n = 159) Note: All other independent variables were set at their mean. Though the probability of bill passage is lower overall for redistribution bills, the relationship between political elite support and bill passage remains the same. In all three \# \"+# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# +# \# +# ;# @# $()*+*,-,./0)102,--0$+33+450 $+(.,3,+60$()4(533,)60 &4<4284#H.94G# C??40?#H.94G# &4<4284#`#C??40?#H.94G# Model RR min to max RR p20 to p80 Revenue 7.302 4.032 Assets 6.739 3.624 R evenue + Assets 6.947 3.672

PAGE 93

83 models the probability of bill passage rises from under 4% to over 22% as political elite support increases from minority party support only to strongly bipartisan We again see similar findings in the recognition data set Figure IV.3 P robability of Recognition Bill Passage by Partisan Progression (n = 211) Note: All other independent variables were set at their mean. Table IV. 22 Relative Risk of Recognition Bill Passage for Partisan Progression (n = 211) Note: All other independent variables were set at their mean. As shown above, for all three models in the recognition data set the probability of bill passage rises from under 1 6% to over 34% as political elite support increases from minority party support only to strongly bipartisan \# \"+# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# +# \# +# ;# @# $()*+*,-,./0)102,--0$+33+450 $+(.,3,+60$()4(533,)60 &4<4284#H.94G# C??40?#H.94G# &4<4284#`#C??40?#H.94G# Model RR min to max RR p20 t o p80 Revenue 2.879 2.800 Assets 3.673 3.485 Revenue + Assets 3.078 3.061

PAGE 94

84 As these data demonstrate, the probability of bill passage increases as the presence of political elite support increases, even when co ntrolling fo r mean levels of SMI strength, interest group strength, public opinion and media coverage. Further, this is true for all three models, in all three data sets. Public O pinion The level of public opinion in favor of a bill is positively correlated with bill passage in the combined data set for all three models. Figure IV.4 aids in the visualization of this correlation by showing the probability of bill passage (absolute risk) as the public opinion variable increases. Figure IV. 4 Probability of Recognit ion Bill Passage by Public Opinion for All Bills (n = 370) Note: All other independent variables were set at their mean. Table IV.23 supports these data, showing the relative risk of bill passage as public opinion in favor of a bill progresses from its m inimum to its maximum level (RR min to \# \"+# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# +# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# $()*+*,-,./0)102,--0$+33+450 $)(.,)60)10$7*-,809:,6,)60,60;+<)(0)102,--0 &4<4284#H.94G# C??40?#H.94G# &4<4284#`#C??40?#H.94G#

PAGE 95

85 max), as well as the relative risk as public opinion progresses from it s 20 th percentile value to it s 80 th percentile value (RR p20 to p80). Table IV 2 3 Relative Risk of Bill Passage for Public Opinion for All Bills (n = 370) Note: All other independent variables were set at their mean. As shown above, in the combined data set, public opinion is positively correl ated with bill passage. That is bills with more public support (a higher percentage of public opinion in favor of them) are more likely to pass. Indeed, in all three models the probability of bill passage rises from under 9% to over 21% as public suppo rt for the bill increases. Surprisingly these findings do not hold in the redistribution data set Indeed, within the redistribution data set, public opinion is significantly correlated with bill passage in only one of the models (the revenue model). F urther, this correlation is negative as public support for redistribution bills increases, the probability of bill passage decreases. Figure IV. 5 and Table IV.24 below aid in visualizing this correlation. Model RR min to max RR p20 to p80 Revenue 2.816 1.355 Assets 3.536 1.485 Revenue + Assets 3.297 1.434

PAGE 96

86 Figure IV. 5 Probability of Redistributio n Bill Passage by Public Opinion (n = 159) Note: All other independent variables were set at their mean. Table IV.24 Relative Risk of Redistribution Bill Passage for Public Opinion (n = 159) Note: All other independent variables were set at their mean. In the redistribution data set, the probability of bill passage falls from over 12% to just over 3% in the Revenue Model as public support for the bill increases. This finding is somewhat surprising, and I discuss possible explanati ons in the discussion section. However given the p value (p = 0.094) there is a reasonable chance that a Type I error occurred that this relationship is not real. However, the relationship between public opinion and bi ll passage returns to our expected direction in the recognition data set Here, as shown by Figure IV.6 and Table IV.25 below, the probability of bill passage rises as public support increases in both the Revenue Model and the Revenue + Assets Model (the relationship is not significant in the Assets Model). \# \"+# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# +# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# $()*+*,-,./0)102,--0$+33+450 $)(.,)60)10$7*-,809:,6,)60,60;+<)(0)102,--0 &4<4284#H.94G# Model RR min to max RR p20 to p80 Revenue 0.210 0.519

PAGE 97

87 Figure IV.6 Probability of Recognition Bill Passage by Change in Public Opinion (n = 211) Note: All other independent variables were set at their mean. Table IV. 25 Relative Risk of Recognition Bill Passage for Public Opinion (n = 211) Note: All other independent variables were set at their mean. Thus, in the Revenue Model and the Revenue + Assets Model in the reco gnition data set, the probability of bill passage rises from under 13% to over 34% as public support for the bill increases. As demonstrated above in both the combined and the recognition data sets the probability of bill passage increases as public sup port for the bill increases, even when co ntrolling for mean levels of SMI strength, interest group strength, political elite support and media coverage However, contrary to my expectations, in the redistribution data set public opinion in the revenue mo del is negatively correlated with bill passage \# \"+# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# +# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# $()*+*,-,./0)102,--0$+33+450 $)(.,)60)10$7*-,809:,6,)60,60;+<)(0)102,--0 &4<4284#H.94G# &4<4284#`#C??40?#H.94G# Model RR min to max RR p20 to p80 Revenue 3.472 1.464 Revenue + Assets 3.375 1.434

PAGE 98

88 Interest G roups Though interest groups are not significantly correlated with bill passage in either the redistribution or the recognition data sets, there is a significant negative correlation between int erest group strength and bill passage in the combined data set. Figure IV. 7 and Table IV.26 help to illustrate this relationship. Figure IV.7 Probability of Bill Passage by Interest Groups for All Bills (n = 370) Note: All other independent variables were set at their mean. Table IV.26 Relative Risk of Passage for Interest Groups for All Bills (n = 370) Note: All other independent variables were set at their mean. As shown above, in all three modes in the combined data set, the probability of bill passage falls from over 21% to under 10% as interest group strength increases. \# \"+# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# +# a+L# a+F# a+\# aK# a;# ;# K# +\# +F# $()*+*,-,./0)102,--0$+33+450 =6.5(53.0>()7:0?0,602,--,)630 &4<4284#H.94G# C??40?#H.94G# &4<4284#`#C??40?#H.94G# Model RR min to max RR p20 to p80 Revenue 0.444 0.996 Assets 0.363 0.998 Revenue + Assets 0.380 0.997

PAGE 99

89 G iven the surprising nature of these findings, I investigated this varia ble further by removing 501(c) (5)s ( labor u nions and agricultural organizations ) and 501(c) (6)s (business l eagues and chambers of c ommerce) from the data set I then removed the bills (and thus cases) directly related to labor unions (501(c)(5)s) or busin ess leagues and chambers of commerce (501(c)(6)s). When I reran the analysis under these conditions, the corrected coefficient for the interest group variable remained insignificantly correlated with bill passage in both the redistribution and recognition data sets, but significant and negative in the combined data set As such, removing 501(c)(5)s and 501(c)(6)s did not alter this unexpected finding. If replicated, these results would undermine the assumption that there is a direct causal link between in terest group strength and bill passage. Rather, perhaps interest groups tend to target bills that are not likely to pass (as there is little sense in focusing on bills that are likely to pass). If true we would expect interest group strength to be stron gest relative to bills that are otherwise unlikely to pass. In such instances, we might find that results similar to those above interest group strength negatively correlated with bill passage. To test this proposition, I investigated the relationship bet ween interest group strength and both public opinion and political elites. If interest groups do, in fact, target bills that are difficult to pass, we would expect there to be a negative correlation between interest groups and one or both of these variabl es. That is we'd expect interest group strength to be the strongest when public opinion was against the bill and/or when elite support was low. T he evidence in this regard, is mixed.

PAGE 100

90 A biserial correlation demonstrated that there is a statistically sig nificant positive correlation between interest group strength and public opinion in all three models ( Revenue Model: r (368) = 0.25 p = 0.000; Asset Model: r (368) = 0.2 6, p = 0.000 ) ; Revenue + Asset Model: r (368) = 0.2 6, p = 0. 000) This suggests that int erest groups actually coalesce around bills with higher public support. However, there is a negative (though not statistically significant) correlation between interest group stre ngth and political elite allies ( Revenue Model: r (368) = 0.06 p = 0.284; A sset Model: r (368) = 0.03 p = 0.524 ) ; Revenue + Asset Model: r (368) = 0.04 p = 0. 408 ) Though the lack of significance means we can't conclude there is a real correlation between interest group strength and elite support this finding does indicat e that, within these data, political elite support is generally lower when interest group strength is higher Yet, at this point, it is important to note that the political elite support variable is based on partisan sponsorship with the relative strength of elite support based on the party sponsorship of a bill within the context of the partisan divide during the 109 th Congress. That is because Democrats were the minority party in the 109 th Congress, only Democratic sponsorship equates to low elite supp ort (as opposed to Republican or bipartisan sponsorship) As such, the negative correlation between interest group strength and elite support coul d be interpreted another way: p erhaps interest groups interested in social justice target bills that are more likely to be supported by Democrats. To investigate this possibility I compared the political elite values for the 37 cases with the highest interest group strength values (the top 10%), to the political elite values for the whole sample. In particula r I looked at the percentage of cases with a value of 0, indicating only Democratic support. These results are displayed below.

PAGE 101

91 Table IV.27 Percentage of Cases with Democratic S ponsorship Only (Low Political Elite Support) Note: These values are the same because in all three models (revenue, assets, revenue + assets) the 37 cases with the highest interest group values are th e same cases. As shown above, of the bills with the high est interest group support 70% were sponsored by Democrats only, compared with 50% of the bills as a whole. This finding suggests that within these data, the strongest interest group coalitions te nd ed to target bills sponsored by Democrats alone As such, we can explain the negative relationship between interest group strength and bill passage in these data through the correlation between political elite support and interest group strength in the d ata (though we cannot conclude that the correlation itself is real) Further, as we drill down into what's really going on, it appears that strong interest group coalitions (as measured by aggregate revenue and assets) tended to target bills sponsored onl y by Democrats. Because Democrats were the minority party in the 109 th Congress, this means that the social justice interest group coalitions with the greatest financial resources also lacked substantial elite political support within the 109 th Congress Social M ove ment Industrie s (SMI s) The strength of SMI s is positively correlated with bill passage in all three models in the recognition data set (though it is not significantly correlated with bill passage in Model All Cases (n = 370) Top 10% Interest Group Values (n = 37) Revenue 50.0 70.3 Assets 50.0 70.3 Revenue + Assets 50.0 70.3

PAGE 102

92 either the redistribution or the combined data sets ) Figure IV. 8 below displays this relationship graphically. Figure IV.8 Probability of Recognition Bill Passage by SMIs (n = 211) Note: All other independent variables were set at their mean. As shown in Figure IV. 8 above, the shift in the probab ility of bill passage that occurs as SMI strength increases is relatively large. In the Revenue Model, the probability of bill passage rises from about 17% to over 63% as SMI strength increases. In the Assets Model, increa se in SMI strength shifts the pr obability of bill passage from a low of about 1% to just under 50%. Finally, the Revenue + Assets Model shows a shift in the probability of bill passage from j ust under 11% to over 61% as SMI strength increases. Though this spread is quite large, a look at the relative risk from the 20 th percentile to the 80 th percentile shows that a good deal of this shift occurs at the margins. \# \"+# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# +# a@# a;# a+# \# +# ;# @# F# S# K# T# L# $()*+*,-,./0)102,--0$+33+450 @A=0?0,602,--,)630 &4<4284#H.94G# C??40?#H.94G# &4<4284#`#C??40?#H.94G#

PAGE 103

93 Table IV.28 Relative Risk of Passage for SMI s for Recognition Bills (n = 211) Note: All other independent variables were set at their mean. SMI Impact in Context : Various ReLogit Models To further investigate how the context within which social movements are embedde d influe nces the correlation between SMI strength and bill passage, I calculated probabilities and relative risks under four additional circumstances: SMI s Against the World where all IVs are set at their 20 th percentile, and thus SMIs are operating in a context that is not favorable to bill passage. SMI s Riding a Tide of Support where all IVs are set at their 80 th percentile, and thus SMIs are operating in a contest favorable to bill passage. People vs. Power where public opinion is set at it s 80 t h percentile, while all other IVs are set at their 20 th percentile, and thus public opinion is favorable but all other variables are unfavorable for bill passage. Key Elite Allies where political elites are set at their 80 th percentile, while all IVs are set at their 20 th percentile, and thus political elite support is favorable but all other variables are unfavorable for bill passage. I address each of these in turn below. SMI s Against the World This circumstance sets all IVs at their 20 th percentile. I n doing so, it approximates a situation in which interest group support for the bill is low, the bill is supported by the Model RR min to max RR p20 to p80 Revenu e 4.845 1.090 Assets 180.000 1.964 Revenue + Assets 14.800 1.290

PAGE 104

94 minority party only (Democrats), public opinion is split ( at 0.5) and there is very little media coverage. In other wo rds, this model demonstrates SMI impact on the probability of bill passage when all the other variables are not favorable for bill passage. Figure IV.9 Probability of Recognition Bill Passage for SMIs with IVs at p20 (n = 211) Table IV.29 Relative Risk of Recogniti on Bill Passage for SMI s with IVs at p20 (n = 211) Note: All other independent variables were set at their 20 th percentile. As shown above, in an o verall unfavorable context, SMI strength correlates with a shift in the probability that recognition bills will pass from under 8% to over 51%. \# \"+# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# +# a@# a;# a+# \# +# ;# @# F# S# K# T# L# $()*+*,-,./0)102,--0$+33+450 @A=0?0,602,--,)630 &4<4284#H.94G# C??40?#H.94G# &4<4284#`#C??40?#H.94G# Model RR min to max RR p20 to p80 Revenue 6.994 1.100 Assets 69.298 1.869 Revenue + Assets 14.832 1.272

PAGE 105

95 SMI s Riding a Tide of Support This circumstance is the reverse of the above all IVs are set at their 80 th perc entile. As such, it approximates a situation in which interest group support for the bill is high the bill has strong bipartisan support, public opinion is in favor of the bill and there is a substantial amount of media coverage. In other wo rds, this mo del demonstrates SMI impact on the probability of bill passage when all the other variables are favorable for bill passage. Figure IV.10 Probability of Recognition Bill Passage for SMIs with IVs at p80 (n = 211) Table IV 30 Relat ive Risk of Bill Pass age for SMI s with IVs at p80 (n = 211) Note: All other independent variables were set at their 80 th percentile. \# \"+# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# +# a@# a;# a+# \# +# ;# @# F# S# K# T# L# $()*+*,-,./0)102,--0$+33+450 @A=0?0,602,--,)630 &4<4284#H.94G# C??40?#H.94G# &4<4284#`#C??40?#H.94G# Model RR min to max RR p20 to p80 Revenue 2.643 1.061 Assets 23.594 1.592 Revenue + Assets 4.882 1.183

PAGE 106

96 As shown in Figure IV. 10 and Table I V.30 in an overall favorable context, SMI strength correlates with a shift in the probability that recognition bills will pass from under 33% to over 82%. People vs. Power This circumstance sets public opinion at its 80 th percentile and all other IVs at t heir 20 th percentile approximating a situation in which public opinion is in favor of the bill, but interest group support for the bill is low, the bill is supported by the minority party only (Democrats) and there is very little media coverage. In other wor ds, this model demonstrates SMI impact on the probability of bill passage when the public is in favor of the bill but all the other variables are not favorable fo r bill passage. Figure IV.11 Probability of Recognition Bill Passage for SMIs with Publi c Opinion at p80 and All Other IVs at p20 (n = 211) \# \"+# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# +# a@# a;# a+# \# +# ;# @# F# S# K# T# L# $()*+*,-,./0)102,--0$+33+450 @A=0?0,602,--,)630 &4<4284#H.94G# C??40?#H.94G# &4<4284#`#C??40?#H.94G#

PAGE 107

97 Table IV 31 Relative Risk of Recognition Bill Passage for SMI s with Public Opinion at p80 and All Other IVs at p20 (n = 211) Note: Public opinion was set at its 80 th percentile, with all other independent variables set at their 20 th percentile. As shown above, when public opinion is in favor of a bill, but other variables are not favorable fo r bill passage, SMI strength correlates with a shift in the probability that recognition bills will pass from under 12% to over 57%. SMI s with Key Elite Allies This circumstance sets elite political allies at its 80 th percentile and all other IVs at their 20 th percentile approximating a situation in which there is strong bipartisan support for a bill in the legislature, but interest group support for the bill is low, public opinion is split (at 0.5) and there is very little media coverage. In other wo rds, this model demonstrates SMI impact on the probability of bill passage when there is strong bipartisan support for a bill in the legislature but all the other variables are not favorable for bill passage. Model RR min to max RR p20 to p80 Revenue 5.157 1.095 Assets 54.662 1. 840 Revenue + Assets 10.314 1.245

PAGE 108

98 Figure IV.12 Probability of Recognition Bill Pass age for SMIs with Elite Political Allies at p80 and All Other IVs at p20 (n = 211) Table IV 32 Relative Risk of Recognition Bill Passage for SMI s with Elite Political Allies at p80 and All Other IVs at p20 (n = 211) Note: Elite political allies was set at its 80 th percentile, with all other independent variables set at their 20 th percentile. As shown above, when there is strong bipartisan support for a bill in the legislature, but other variables are not favorable for bill passage, SMI strength correlates with a shift in the probability that recognition bills will pass from under 24% to over 72%. \# \"+# \";# \"@# \"F# \"S# \"K# \"T# \"L# \"Y# +# a@# a;# a+# \# +# ;# @# F# S# K# T# L# $()*+*,-,./0)102,--0$+33+450 @A=0?0,602,--,)630 &4<4284#H.94G# C??40?#H.94G# &4<4284#`#C??40?#H.94G# Model RR min to max RR p20 to p80 R evenue 3.135 1.072 Assets 22.215 1.510 Revenue + Assets 5.485 1.193

PAGE 109

99 Comparison of Circumstances Figure IV.13 below di splays the probability of recognition bill passage under each of the circums tances described above, when the SMI value in the revenue model is at $1 b illion, $0 and $1 b illion. Figure IV.13 Probability of Recognition Bill Passage in the Revenue Model b y SMI Strength in Various Circumstances (n = 211) As shown above, within the recogni tion data set, an increase in SMI revenue is correlated with an increase in the probability of bill passage, but the circumst ance matters. Indeed, when SMI s are riding a tide of support (all other IVs are favorable for bill passage) the probability of bill passage is higher at $1 b illion in SMI revenue (the aggregate revenue of SMOs opposed to a bill exceed the aggregate revenue of SMOs in favor of a bill by $1 b illion) than when under general circumstances (all other IVs are set at their mean) the SMO revenue is at $1 b illion (the aggregate revenue of SMOs in favor of a bill exceed the aggregate revenue of SMOs opposed to a bill by $1 b illion) Both the Assets Model \# \"+# \";# \"@# \"F# \"S# \"K# \"T# ab+#R=GG=.2## b\## b+#R=GG=.2## J424/1G# CZ1=2?0#064#E./G9# &=9=2Z#1#%=94#.5#78--./0# [4.-G4#4/# ^4:#BG=04#CGG=4?#

PAGE 110

100 (s ee Figure IV. 14 ) and the Assets + Revenue Model (see Figure IV. 15 ) have similar results. Figure IV.14 Probability of Recognition Bill Passage in the Assets Model by SMI Strength in Various Circumstances (n = 211) Figure IV.15 Probability of Recogn ition Bill Passage in the Revenue + Assets Model by SMI Strength in Various Circumstances (n = 211) \# \"+# \";# \"@# \"F# \"S# \"K# \"T# ab+#R=GG=.2## b\## b+#R=GG=.2## J424/1G# CZ1=2?0#064#E./G9# &=9=2Z#1#%=94#.5#78--./0# [4.-G4#4/# ^4:#BG=04#CGG=4?# \# \"+# \";# \"@# \"F# \"S# \"K# \"T# ab+#R=GG=.2## b\## b+#R=GG=.2## J424/1G# CZ1=2?0#064#E./G9# &=9=2Z#1#%=94#.5#78--./0# [4.-G4#4/# ^4:#BG=04#CGG=4?#

PAGE 111

101 As shown in f igures IV. 13, IV. 14 and IV. 15 above, SMI strength is positively correlated with the passage of recognition bills, but this correlation is hea vily influenced by other varia bles. Most strikingly, when SMI strength is paired with political elite allies, the probability of bill passage is relatively high even in the face of low interest group support, split public opinion and little media coverag e (the Key Political Elite Allies circumstance). Overview of Results This study was guided by two research questions. First: Within the contemporary United States does the strength of a social movement industry impact whether the U.S. Congress passes a bill that is important to that industry? My results suggest that SMIs do in fact impact bill passage within the U.S. Congress, but only under particular circumstances. Thus, in response to my first question, the answer is a qualified "yes." The most impo rtant qualification is related to my second research question: If a relationship between social movement industry strength and bill passage exists, is that relationship moderated by the focus of the bill on redistribution or recognition? The answer to thi s question is a n un qualified "yes" as SMIs appear to be able to impact the passage of recognition bills, but not the passage of redistribution bills. Further, consistent with the literature, two additional contextual factors amplify (or in their absence s uppress) SMI influence on recognition bill passage: public opinion and political elite allies.

PAGE 112

102 CHAPTER V DISCUSSION These findings have profound implications fo r the two major bodies of literature included in this paper: (1) the social movement literature and (2) the critical theory literature on recognition. I address the implications for each below. Implications for the Social Movement Literature Consistent w ith the social movement literature, my results suggest that social movements can and do matter, but the extent to which they do is highly dependent on the context they are embedded within Specifically, these results indicate that within the contemporary United States, SMI s can influence the passage of bills in the legislature, but only under certain conditions. Most importantly, my results indicate that SMI s are able to influence the passage of recognition bills, but not the passage of redistribution bil ls. Further, their influence on r ecognition bills appears to be quite substantial. There are a few possible explanations for this finding. One poss ibility is that the contemporary SMO s comprising SMIs tend to focus their efforts on recognition claims and at least partially ignore redistribution claims. T he data included in this study are unable to fully assess this position doing so would require content analyses of the actual claims made by SMOs. However, this does not appear to be the case as relative SMI strength is actually larger in the redistribution data set than in the recognition data set. This suggests that strategic focus does not explain t he difference we see between SMI impact on redistribution and recognition bills. Another possibility is that distributional structures are simply more difficult to change than elements of the recognition order. My result s provide some evidence to

PAGE 113

103 support this position as far more recognition bills passed than did redistribution bills (6.3% of redistribution bills passed as opposed to 22.7% of recognition bills). Yet this thesis, that redistribution is more difficult to influence than recognition, has several variants of differing scope. In its strongest variant this thesis might hold that the recognition/ redistribution distinction is an essential characteristic of modern society. In its weakest variant, it is possible that this distinction only holds within the particular context studied (the 109 th Congress of the United States) and that in other contexts within a modern society the distinction would disappear. M ore research is needed to make such determinations 18 If the distinction between recognition and redistribution were to hold within different contexts, there would be substantial evidence that th e distributional structure is distinct from the recognition order in modern society and that this distinction has implications for social movement efforts to achieve social justice. In particular, if the findings in this study were replicated across conte xts, it would appear that movements are far more able to influence the recognition (or status) order, rather than the distributional (or class) structure of contemporary modern capitalist societies. Though more research is needed, a promising preliminar y explanation for this is the general absence of class consciousness in contemporary American society, and the 18 To determine if these findings can be extended beyond the 109 th Congress, this study should be replicated with a different Congress, preferably one controlled by Democrats. Additionally, it would be useful to approximate t his study within other political systems, cultural contexts and historical eras.

PAGE 114

104 relative absence of class analysis in the academy. Within academia, the "cultural turn 19 has largely, though certainly not entirely, displaced M arxist class analysis Jeff Goodwin and Gabriel Hetland (2009) found that words associated with class analysis are virtually absent from the two leading English language journals dedicated to social movements: Mobilization and Social Movement Studies In the 183 articles published in Mobilization between 1996 (its founding) and 2007, the word "capitalism" appeared only once in either titles or abstracts. Additionally, no titles or abstracts include the words "class conflict" or "class struggle." Similar ly, in the 71 articles published in Social Movement Studies from 2002 (its founding) to 2007, the word "capitalism" appears in only one title and three abstracts, and the words "class conflict" and "class struggle" do not appear. Outside the academy, the fall of Soviet style communism and the capitalistic metamorphosis of Chinese communism have left a void within which a neoliberal economic ideology has ascended throughout much of the globe (Harvey, 2007). Perhaps these developments have rendered the dis tributional (class) structure at least partially invisible to scholars and the polis alike. Put another way, neoliberalism may operate as a hegemonic master frame for material distribution within contemporary American society making the distributional (cl ass) structure impervious to challenges by movements and the direction of the general will (public opinion). 19 The cultural turn was a widespread shift in the social sciences and the humanities during the late 1970s and early 1980s, away from a focus on economics and politics and to ward culture. Though its form varied by discipline, in general this shift focused scholars on the role played by "meaning," "identity," "values" and "beliefs." Though many scholars never accepted this shift, culture remains a primary focus within social science and the humanities today (Ritzer, 2007).

PAGE 115

105 Meanwhile, both academics and the general public are engaged in work that destabilizes the recognition order Postmodern scholars have taught us t o "deconstruct" some of our most taken for granted cultural understandings while the left leaning public touts multiculturalism and the right leaning pubic insists on a return to "traditional values". As such, it appears that the recognition order is des tabilized because it lacks a hegemonic master frame, and as result, is open to influence from organized efforts (SMIs) and shifts in the general will (public opinion). Yet, if as my findings suggest, class and status are decoupled from each other within mo dern society, the stability of the distributional structure is not necessary threatened by the relative instability of t he recognition order. As an ideology governing the distributional structure (class) of society, n eoliberal economic ideology can easily absorb changes in the recognition order (such as same sex marriage) because such status based shifts don't challenge its fundamental premise Likewise, master frames within the recognition order should be able to coexist with a variety of distributional ideologies. For example, both "traditional values" and multiculturalism can coexist wit h either socialism or capitalism. Thus these data appear to suggest that social movements within contemporary society are able to impact the recognition (status) orde r because it lacks a hegemonic master frame and is thus open to contestation, but because a pervasive neoliberal master frame dominates the distributional (class) structure, it is not vulnerable to movement influence. In addition to this fundamental insigh t, these data have further implications for the social movement literature. These include further support for resource mobilization theory (RMT), the relative imp ortance of political elites, public opinion and master

PAGE 116

106 frames the role of the media, the pos sibility of the division between SMOs and interest groups and the importance of context. I address each of these below. Support for Resource Mobilization Theory (RMT) Social movements were operationalized in this study by aggregating the financial resour ce base (revenue and assets) of SMOs into SMIs As such this variable assumes that both formal organization and financial resources are essential to movement success. This view, consistent with RMT, is confirmed by the results SMI assets and revenue a re significantly correlated with recognition bill passage, even when controlling for interest groups, political elite support, public opinion and media coverage. This suggests that resources, mobilized through and directed by formal organizations, impact recognition policy directly. Though this effect is augmented by the context that movements are embedded within, this finding supports RMT. Political E lites, P ublic Opinion and Master Frames Consistent with previous research ( Bernstein 1998, 1999; Giugni 2 004; Snow & Soule, 2009 ; Tarrow 1993, 2011 ), my results suggest that both political elite allies and public opinion are important factors in bill passage. Indeed, both were significantly correlated with bill passage in all three data sets ( redistribution recognition and combined). Additionally, like Giugni ( 2004 ) my results indicate that political elites have a larger impact on bill passage than public opinion, though both matter. 20 Thus, in key ways, my results further support the existing social moveme nt literature with regard to political elite allies and public opinion. 20 This is demonstrated by comparing the relative risks for each variable and holding all other variables constant. Across all three models (revenue, assets and revenue + assets) all relative risks are hig her for political elite allies than for public opinion.

PAGE 117

107 Yet, my introduction of the redistribution /recognition distinction adds some nuance to these findings. My results suggest that political elite alliances have a larger effect on redist ribution bills than on recognition bills. Additionally, in a rather surprising finding, in the R evenue M odel public opinion is negatively correlated with the passage of redistribution bills but positively correlated with the passage of recognition bills a nd those in the undifferentiated data set. Th ough more research is need ed to explain these findings, a promising possibility is the relative invisibility of the class structure discussed above W ithout the identification of class interests among the pub lic, public opinion may have become somewhat divorced from distributional politics, in which case the causal link between public opinion and bill passage would be broken. Thus, the negative correlation between public opinion and redistribution bill passag e would indicate that the system follows the logic of neoliberal economics in the face of the desires of the general public, but because the class structure is largely concealed from the collective consciousness, there is no check on this development. Add itionally, in the absence of class consciousness, neoliberal economics may have become the guiding logic for distributional policies thus explaining the stronger influence elites have over the passage of redistribution bills. At this point it is too earl y to confirm such a hypothesis. However, these findings do suggest that future research should further investigate the recognition/ redistribution distinction in the context of the interaction between social movements, political elites and public opinion.

PAGE 118

108 Uncertain Effect of Media Coverage My results suggest that there is no correlation between the amount of media coverage and bill passage. However, my methods were not able to account for the directionality of media coverage whether coverage was for or against the bill. Additionally, by using the New York Times as a proxy for the media generally, I was unable to account for both editorial bias (i.e. partisan leanings within the institution that influence coverage) and any difference in influence by me dia type (i.e. newspapers vs. TV etc.). Fully understanding how media coverage influences movement impact on bill passage will take further study that more adequately takes these factors into account. Distinguishing Between Interest Groups and SMOs Th ere is currently no consensus as to how to appropriately distinguish between SMOs and interest groups, or even whether such a distinction is desirable. In this paper, I chose to use the designation provided by the Internal Revenue Code to make this distin ction. Though this method carries with it a variety of potential problems (see M ethods section above for details), my results suggest that there may be some utility in making such distinctions. As Paul Burstein (1998, 1999) correctly points out, such di stinctions are fraught with the challenge of artificially forcing a continuum into distinct taxonomies. However, as shown in the results above, 501(c)(3)s (what I've designated as SMOs) appear to correlate with bill passage in quite different patterns tha n 501(c)(4)s, 501(c)(5)s and 501(c)(6)s (what I've designated as interest groups). As such, it will likely be useful for researchers to continue to distinguish between the types of advocacy organizations involved in efforts to shape public policy.

PAGE 119

109 This is especially true given that, in the results presented above, interest groups (with or without 501(c)(5)s and (6)s) are negatively correlated with bill passage in the combined data set for all three models ( R evenue A ssets, and R evenue + A ssets) even when controlling for SMI s, political elites, pu blic opinion and media coverage. Though the effect is slight, on the surface this is an unexpected finding. However, it appears that we can explain this finding based on the (statistically insignificant) correla tion between interest group strength and political elite support. More specifically, within these data, the bills with the strongest interest group support (measured by revenue and assets) tended to also be bills with only Democratic sponsorship Thus, t hough there may not be a real correlation between interest group strength and political elite support, within these data strong interest group coalitions did, in fact, tend to target bills supported by Democrats only Within the 109 th Congress, this equat es to a relative lack of political elite support (because Democrats were the minority party). However, this explanation leads to an additional question. Are interest group coalitions with a social justice focus generally more likely to target bills spons ored by Democrats? If so, social justice interest group strength should be positively correlated with bill passage when the Democrats are in power. T hough this would be consistent with some of the existing liter ature ( Meyer and Minkoff, 2004) these data cannot confirm such a hypothesis (because the correlation is insignificant in these data) and alternative hypotheses are apparent For example, perhaps interest groups generally target bills that would otherwise be difficult to pass within the existing p artisan structure Based on the logic that it doesn't make sense to expend resources on bills that are likely to pass, this explanation would suggest that when

PAGE 120

110 Democrats are in power interest group strength would coalesce around bills sponsored by Republ icans only Further research is needed to settle this question. SMI Impact and C ontextual Circumstance As described above these results suggest that SMI strength is positively correlated with the passage of recognition bills (though not the passage of re distribution bills) and that the magnitude of this correlation is rather large. 21 That is as the strength of aggregate SMOs (SMIs) in favor of a bill rise, the probability that the bill will pass rises substantially. Yet, consistent with the literature, my results also demonstrate that context matters a great deal. In two of the three models (Revenue and Revenue + Assets), the probability of bi ll passage is higher with an SMI value of 1 in the Riding a Tide of Support circumstance than the probabi lity of bill passage with an SMI value of 1 in General circumstance. Though less stark, a similar outcome is present in the Assets model the probability of bi ll passage is higher with an SMI value of 0 in the Riding a Tide of Support circumstance than the proba bi lity of bill passage with an SMI value of 1 in General circumstance. Additionally, in both the Revenue and the Revenue + Assets models, the probability of bill passage is higher in the General circumstance with an SMI value of 1 than in the SMI s Agains t the World circumstance with an SMI value of 1. Indeed, when all other variables in the models are generally not favorable for bill passage (an SM I s Against the World circumstance ), even with the strongest SMI support the 21 In the R evenue model, the probability of bill passa ge ranges from 17% to 63% as SMI strength increases (holding all other IVs at their mean) and the relative risk from min to max is 4.845. In the A ssets model, the range in proba bi lity is 1% to 50% with a min to max relative risk of 180. Finally, in the Assets + Revenue model, probability rises from about 11% to over 61% with a min to max relative risk of 14.800.

PAGE 121

111 probability of bill passage doe s not exceed 59% in any of the models. Conversely, when all other variables in the models are generally favorable for bill passage (a Riding a Tide of Support circumstance), strong SMI support pushes the probability of bill passage to almost 90% in the As sets model and over 82% in the other two. These findings suggest that circumstance matters a great deal as the combination of support from political elites, public opinion, interest group support and media coverage influence the probability of bill passage substantially, regardless of SMI strength. Yet SMI strength also influences the probability of bill passage, regardless of the circumstance in each of these circumstances, the probability of bill passage increases as SMI strength increases. In other wor ds, while SMI strength matters in each circumstance, the circumstance itself also matters a great deal. A further comparison of the various circumstances reveals that for all three models in all three data sets, the probability of bill passage is almost, t hough not quite, as high in a Key Elite Allies circumstance as it is in a Riding a Tide of Support circumstan ce. This suggests that when SMI strength is paired with strong bipartisan political elite support, the probability of bill passage is relatively h igh even in the face of low public opinion, a lack of media coverage and low support from interest groups. However, the same cannot be said for a People vs. Power circumstance. Indeed, for two of the three models (Revenue and Revenue + Assets), the proba bility of bill passage is higher in a General circumstance than it is in a People vs. Power circumstance (though the reverse is true in the Assets model). That is while f avorable public opinion aids SMI s in influencing bill passage, it is not sufficient to overcome a lack of key political elite support, low interest group support and low media coverage.

PAGE 122

112 These findings provide further support for the view that social movement impact is highly contextual. Though SMI strength is positively associated with recognition bill passage in all of the circumstances I simulated, there was more variance between circumstances than within them. That is to say, these findings suggest that SMI s influence recognition policy, but that the contextual circumstances in whic h they are embedd ed have a larger effect than SMI s alone. Implications for Critical Theory My findings suggest that, in the context of social movement impact the distinction between recognition and redistribution matters. Specifically SMI strength (meas ured by revenue and assets) is positively correlated with recognition bill passage, but not significantly correlated with redistribution bill passage. This adds some much needed empirical evidence into the contemporary philosophical debate around recognit ion, the implications of which stretch from the practical political level, to political theory, social theory and finally, moral philosophy. In the sections below I address the implications of my findings for the debate between Nancy Fraser and Axel Hon neth (2003) at each of these levels. Practical P olitics (SMI T actics) At the level of practical politics, critical theorists interested in praxis are inclined to ask : H ow do efforts to achieve social justice succeed and why do they fail? Though social mov ement organizations do not exhaust collective efforts at achieving social justice and bill passage does not e xhaust social justice outcomes t his study contributes to answering this question by investiga ting the correlation between SMI strength and bill pa ssage in the 109 th Congress. These results suggest that SMI s were influential in the

PAGE 123

113 passage of recognition bills, but did not impact redistribution bills. Further, recognition bills passed at a significantly higher rate than redistribution bills. Thes e findings appear to be inconsiste nt with Honneth's position that distributional justice will follow from attaining recognition. According to Honneth, material distribution in capitalist societies is regulated by an achievement principle that legitimates the distributional structure of a society by according social esteem to particular locations within the social structure. In other words, people get what they get based on how much value society places on their social position. Thus, Honneth views strugg les over distribution as struggles over the contemporary application of the achievement principle a specific form of recognition ( Fraser & Honneth, 2003 ). From this point of view, because redistribution is a form of recognition, there should be no meanin gful difference between the findings in my recognition and redistribution data sets. Yet my results indicate that there is a significant difference, both in overall outcomes (the proportion of bills passed is substantially higher for recognition bills) an d in the ability of a social movement to realize outcomes (the correlation of SMI strength with bill passage is only significant for recognition bills). This indicates that as Fraser (2003) suggests, and contra Honneth (2003) redistribution is substantia lly different from recognition at the practical political level. If true, social inequity will not be overcome in all its forms by attending only to recognition. Rather, achieving equity will require a multipronged approach including efforts to remedy mi sdistribution and misrecognition. 22 This would indicate that Fraser is 22 In her later work, Fraser adds a t hird political dimension: representation. Representation is discussed in more detail in the literature review section.

PAGE 124

114 correct to suggest that, "In practical politicsthe task is to foster democratic engagement across current divides in order to build a broad based programmatic orientation that integrat es the best of the politics of distribution with the best of the politics of recognition." ( Fraser &Honneth, 2003, p. 27). For Fraser, such an approach should include three elements: Perspectival Dualism (Three dimensional Lens) 23 considering the distribu tive implications of recognition efforts and recognition implication of redistribution In the latter form, adding to this the representational implications of and for redistribution and recognition. Cross redressing utilizing redistribution to address re cognition issues, and recognition to address redistribution issues. Boundary awareness thinking about how policies reaffirm or tear down social boundaries, both in recognition and redistribution 23 Though most of Fraser's work that I draw on refers to a dualist approach, her more recent work, which adds the concept of representati on, makes this a "three dimensional" approach. As such I refer to both here.

PAGE 125

115 Given the results above, this approach 24 appears promising. I f I am correct in my assertion that SMIs inability to change the distribution (class) structure is largely a result of the stability provided by a neoliberal master frame critical theorists must make class visible again and in doing so, open the distrib utional structure to critique (though in this post soviet world, new alternatives may be needed ) But this should not come at the expense of the status ( recognition ) order. Thus, we need a Fraserian approach to r eintroduce effective class critique (redis tribution) without compromising on status (recognition) Political Theory At the level of political theory, critical theorists are interested in the most appropriate institutional arrangements (in part, defined by public policy or bills) to achieve socia l justice. As discussed above, recognition bills passed at significantly higher rates than distribution bills, and SMI s appeared to influence recognition bills but not distribution bills. Consistent with Fraser's (2003 ) position, these results suggest th at successfully combating cultural injustice does not necessarily lead to economic justice. 24 In addition to the melding of the politics of recognition and the politics of re distribution (as well as the politi cs of representation in her lat er work), Fraser makes a dist inction between affirmative and transformative approaches. Affirmative approaches seek to correct unjust outcomes, whereas transformative approaches seek to change the underlying structural arrangements in society that generate injustice. For example, Fra ser sees the liberal welfare state as an affirmative approach to distributional injustice because it redistributes resources while leaving capitalism untouched. A recognition corollary to this is multiculturalism, which seeks to embrace socially construct ed difference rather than challenge it. On the other hand, socialism challenges capitalism and is thus a transformational approach to distributional justice (deconstruction is a recognition corollary). Fraser views transformative approaches as preferable to affirmative approaches, but she acknowledges that they are often politically impractical except under exceptional circumstances. As such, following Andre Gorz she advocates a "nonreformist reform" that takes politically practical steps, which may hav e affirmative aspects but that have transformative potential ( Fraser, 2008 ).

PAGE 126

11 6 That is thoug h these results suggest that SMI s have succeeded in achieving at least limited recognition (in the form of recognition bill passage), much less progr ess is apparent in the realm of equitable distribution. This is consistent with Fraser's (2003) suggestion that success in the recognition of difference has come at the expense of egalitarian redistribution. This process, which Fraser terms displacem ent 25 results in recognition displacing distribution as the focal point of claims for social justice ( Fraser, 2008 ). Here it is important to note that my results do not indicate, as Fraser suggests, that a shift in focus from distribution to recognition h as taken place at the movement level. Indeed, my data i ndicate that the strength of SMI s challenging distributional injustice is greater than those challeng ing cultural injustice (mean SMI strength was greater in the distributional data set than the recog nition data set). However, in terms of movement impact it does appear that much more progress is being made on the recognition front that, perhaps, recognition outcomes are displacing distribution outcomes in the contemporary struggle for social justice. This is consistent with Nancy Fraser's assessment that in contemporary struggles for justice, "group identity supplants class interest a nd cultural recognition displaces socioeconomic redistribution as the remedy for injustice" ( Fraser, 2008 p. 11). Y et, as demonstrated by my results, this displacement seems to be at the institutional level, not the level of the individual. As discussed above, public opinion is negatively correlated with bill passage in the distribution data set, but it is positively 25 In addition to displacement, Fraser notes that efforts to recognize difference can have the unintended effect of reification. That is by socially recognizing differences, we r einforce difference. Under some circumstances, this can lead to further division along largely socially constructed cleavages.

PAGE 127

117 correlated with bill passage in the recognition data set. That is the political system appears to be responding to the wishes of the people in terms of recognition, but it is implementing policies (and thus institutional structures) that run contrary to the will of the people in terms of distribution. These results suggest a potentially chilling reality fo r critical theorists. While SMI s seem to be gaining some ground in terms of recognition, it appears that this is not the case in terms of re distributio n recognition bills pass at much higher rates than redistribution bills. E ven more disturbing, neither organized efforts to achieve justice (SMI s) nor the general will (public opinion) appear to have any effect on the rules (bills) that structure distribu tion within our political institutions. As discussed above, this may be linked to the presence of a semi invisible master frame that guides the distributional structure in modern society (neoliberalism) and the lack of such a hegemonic ideology within th e status order. If this account is accurate, neoliberalism will need to be dislodged as the de facto distributional logic within modern society before movements can have a substantial impact on the class structure Further, this will need to be done whil e simultaneously acknowledging the need for recognition within the status order. Such an achievement w ill require a Fraserian approach that accounts for both redistribution and recognition as inter related, but distinct dimensions of social justice Socia l Theory Social theory involves empirically informed accounts of society. Among the many aspects of society social theorists address is the stratification order of society. Both

PAGE 128

118 of the critical theorists utilized by this study (Honneth and Fraser) have d eveloped such social theories based on different accounts of historically specific macro social shifts. Honneth (2003) views recognition as the unitary core of social stratification and the various forms of social oppression. However, for Honneth, recogn ition is a historically specific concept. Prior to modernity, recognition was undifferentiated, and one's recognized status defined social strata (e.g. in medieval Euro pe, being a "noble" defined one s social esteem as well as one's material well b eing). But with modernity, recognition was un coupled into love, law and achievement. For Honneth, distribution in modern society is associated with one of these three forms of recognition: achievement. That is material resources are distributed based on how s ociety recognizes the value of particular achievements. Within this model of modern society, the various strata observable within a given society amount to various forms of social recognition. So for Honneth's social theory (2003) there is no meaningful distinction between distribution and recognition because distribution is a form of recognition (achievement). If this were true, we would expect there to be no difference bet ween the ways the formal social rules (or laws) governing distribution and recog nition are established. In other words, if Honneth's social theory were correct, we would not expect to see statistically significant differences between the recognition and the re distribution dataset. 26 Yet, as 26 A possible response to this, though not one I think Honneth would make, would be to claim that the difference observed is bet ween two types of recognition: (1) legal recognition (recognition) and (2) recognition of achievement (distribution). However, such a move would make the Honneth/Fraser debate (2003) largely a semantic argument over what to call these two different concept s, and as such, the utility of a dual model would remain.

PAGE 129

119 described above, there are differences betw een th e data sets both in bill passage rates and in the relative influence of the variables on bill passage. This suggests that there is an empirical difference between recognition and redistribution in modern society. Such a finding is consistent with the quasi Weberian perspective Fraser has developed. In this model, class and status are analytically distinct, but inter related components of social stratification. Following Weber, Fraser (2003, 2008) views class as an economic distinction defined by one's relationship to the market 27 and status as a cultural distinction defined by the amount and type of social esteem associated with particular categories of people. Though class and status are inter related, from this perspective, they are also analyt ically distinct. Further, like Honneth, Fraser's social theory (2003, 2008) envisions a historical decoupling, but this time of class from status. Following Weber, Fraser's definition of c lass requires capitalism and it s legally regulated markets that f acilitate the rational pursuit of profit. With the rise of capitalism, distribution is partially decoupled from status. Material distribution comes from relations in the market place (either the capital market or the labor market) at least semi independe ntly of social esteem. So, while the two remain inter related ( a privileged class often receives social esteem and social esteem lubricates market relationships) they are analytically distinct. My results can't confirm 27 Unlike Marx, Weber views class as a market relationship. Marx defines class on a two pole relationship to the ownership of the means of production Those who own income producing property are capi talists, those who don't are proletarians forced to sell all that they do own (their labor). Weber also takes the ownership of property into account, using the ownership of property as a primary distinction. But he goes further, making additional distincti ons within this division based on the ways in which individuals interact with the market (the capital market for those with property and the labor market for those witho ut property) (Turner, 2001 ).

PAGE 130

120 this quasi Weberian model of moder n capitalist society that Fraser supports. However, in demonstrating that there are empirical differences between bills related to cultural recognition (status) and bills related to economic distribution (class), they do support it. Yet Fraser s social th eory, following Weber, identifies a third dimension of social stratification the political. According to Fraser, the political dimension consists of the decision rules, which guide politics, as well as the jurisdictional boundaries that define who are inc luded in the polis. Defining who is included is a matter of citizenship, while the relevant decision rules include elements of our political system, including the decision rules that encourage a two party system ( Fraser, 2008 ). This study did not take on the political dimension directly, but it does provide limited evidence that the political is a third dimension of social stratification. My political elite variable represents the partisan nature of support for each bill (from minority party support only, to majority party support only, to bipartisan support). This variable is positively correlated with bill passage in all three data sets suggesting that partisan divisions matter within the American political system. In other words, membership in the ma jority party holds more power than membership in a minority party, which in turn holds more power than membership in one of the "third parties" that are essentially excluded from the political process by the institutionalized decision rules of the system. Further, the correlation between the political elite variable and bill passage persists when controlling for public opinion, suggesting that the political elite matter independently of the feelings of the populace. Thus, these data indicate that the polit ical dimension matters both in the sense of party membership and ascendance to leadership

PAGE 131

121 positions within the parties. In other words, modern society is further stratified by political power. Thus, my results are consistent with Fraser's quasi Weberian model of modern capitalism in which social stratification tracks three independent, yet inter related dimensions: distribution (class), recognition (status) and representation (party). Moral Philosophy These results have substantial implications for cri tical theorists interested in constructing a moral philosophy. While all social theorists seek to describe society and its various elements, and many go one step further by attempting to explain how society works, critical theorists take a final step by e ngaging in a critique of society. So critical social theorists add a normative component to their theorizing, utilizing social theory to identify the socially structured sources of stratification they identify as oppression (or not) through a moral philos ophy. Thus, social theory for a critical theorist involves devising empirically grounded explanations of both social stratification and the institutional mechanisms that maintain the existing stratification coupled with a moral philosophy that normativel y judges some forms of stratification as justified, and others as not. While the justifications embedded within moral philosophies are not an entirely empirical matter, they do rely on as sumptions about the empirical nature of society. Honneth's (2003) moral philosophy views social recognition as important because it is thought to be necessary to achieve self realization. That is to become fully realized human subjects, individuals require social recognition. In its modern form this means being recog nized by the state as a legal citizen (law), being loved by others (love) and being recognized as a valuable member of society based on achievements (achievement). From this perspective, distribution is simply a form of recognition grounded in

PAGE 132

122 achievement Thus, the denial of recognition is the cor e of social oppression because it denies an individual the ability to achieve self realization, and distributional issues are simply a subset of such denials. The necessity of recognition to self realization i s in and of itself, an empirical question, but because my data cannot address this question, I set it aside. What my data can address, however, is Honneth's empirical assumption that redistribution is a subset of recognition. As discussed above, my resu lts suggest that the recognition/ redistribution distinction is an empirical reality at the level of practical politics, political theory and social theory. If true, Honneth's assertion that redistribution is a reflection of recognition doesn't hold and a s a result, the morality he constructs is unable to fully account for economic equity. Though these results do not invalidate Honneth's moral philosophy, they do make it more difficult for him to argue that addressing recognition will necessarily lead to distributional justice. Thus, to adopt Honneth's theory of recognition is to accept a version of social justice that allows for massive economic inequity. In contrast, Fraser's (2003, 2008) participatory parity normative standard is able to accommodate bo th recognition and redistribution as independent, yet inter related dimensions of social justice (as well as representation and any other dimensions yet to be identified). Rather than focus on how individuals become fully actualized subjects, as Honneth does, Fraser turns her attention to participation in society. Arguing that institutional arrangements limit some people's ability to fully participate as peers in society, Fraser identifies parity of participation as the evaluative standard for critique a rrangements that limit participation are unjustified, and those that are neutral or enhance it are justified. Drawing on her social theory, she thus identifies three categories of

PAGE 133

123 institutional barriers to social participation: economic ( redistribution ), cultural (recognition) and political (representation). While my data cannot justify participatory parity as a normative standard, they do support the empirical assumptions Fraser makes. Additionally, they suggest that Fraser's position is able to incorp orate larger portions of contemporary capitalist society than does Honneth's. As such, these data add an empirical element to the debate between Honneth and Fraser. While portions of the moral philosophical components of this debate are impervious to emp irical data, these data do suggest that Fraser's philosophy is built on a social theory that more closely aligns with contemporary society than does Honneth's. Overview of Discussion My results suggest that SMI s are able to substantially influence the pass age of recognition bills, but not the passage of redistribution bills. They also suggest that contextual factors matter a great deal in the probability of bill passage (especially support from political elites), but that SMI s retain their influence on rec ognition bill passage even in the most dire circumstances. This suggests that, following Fraser's quasi Weberian social theory recognition/ status and distribution/class have been at least partially decoupled from each other in modern capitalist society. Further, it appears that the class structure is largely immune to challenge because a neoliberal master frame guides distribution within modern society while the recognition order is open to influence because it lacks a hegemonic conceptual frame for sta tus This has profound implications. At the level of practical politics, grassroots advocates may need to consider working with political elites on distributional matters as SMI s don't appear to influence the rules (bills or l aws) that govern distribution

PAGE 134

124 Additionally, political theory has to acknowledge that institutions addressing recognition do not necessarily address distribution (and vice versa). Finally, moral philosophy has to either choose between distributional and cultural justice or followin g Fraser, wrestle with ways to integrate them into normative foundation that can accommodate both recognition and redistribution

PAGE 135

125 CHAPTER VI CONCLUSION I began this paper with the question: D o social movements impact social justice? To answer this question I drew on social movement theory to identify the important variables in this question. Specifically, Resource Mobilization Theory (RMT) suggests that socia l movement strength largely equates to the resources utilized by formal organizations (Social Movement Organizations) From this I operationalized social movement strength as the total revenue and assets for a Social Movement Industry (the aggregate of S MOs in a particular issue area), less total revenue and assets for it s corresponding countermovement. Then, borrowing from recent mediated models in social movement theory, I added public opinion, political elites and media coverage to my models I then a dded a second theoretical layer to this investigation by introducing Nancy Fraser's distinction between recognition and redistribution. From this perspective, economic class (distribution) has decoupled from social status (recognition) in modern capitalis t society. Though redistribution and recognition remain inter related, they are distinct social phenomenon. Borrowing this concept, I divided my data into two distinct data sets: one built around primarily redistribution bills and one built around primar ily recognition bills. With this set up, I then ran ra re event logistic regressions (rel ogit s ) on each. The results of this anal ysis suggest that SMI s do indeed influence recognition based policy but do not influence redistribution based policy. In oth er words, these data suggest that social movements influence the institutional rules that govern the recognition order

PAGE 136

126 (status) but not the rules that govern the distributional structure (class) Thus, in answer to the question that initiated this paper, social movements appear to be able to impact the cultural (recognition), but not the economic (distribution) elements of social justice. Finally, I proposed a possible explanation for this finding : neoliberalism acts as a semi invisible master frame gu iding distribution in modern society, and thus renders the class structure virtually immune to redistributional efforts In contrast, no equivalent hegemonic master frame exists to guide recognition within modern society, and thus the status order is open to the influence of organized efforts (SMIs) aimed at attaining recognition. Though this is a promising account, additional research is necessary to confirm or refute it. Study Limitations Though this study substantially adds to both the social movement and critical theory literatures, it suffers from some limitations. First, this study is limited by the context within which it is embedded The outcomes I measure (bill passage) take place within a particular instit ution (the federal legislature) embedde d within a particular political system ( the American federal political system ) with unique structural attributes. Additionally, particular individuals within this structure make the decisions that ultimately lead to a bill's passage (or failure to pass). This human context is further complicated because all data within this study are embedded within the cultural context of the United States between 2004 and 2006. As such, with these data alone, it is not possible to know t he extent to which my findings c an be generalized beyond this context. An additional limitation arises with the difficulties involved in operationalizing social movement strength. There are many ways to measure movement strength, each of

PAGE 137

127 which has limitations associated with it. The p rimary limitations associated with my organizational measure (SMI revenue and assets) are : (1) an inability to account for the elements of social movements that extend beyond formal organizations 28 (2) an inability to account for non monetary organizationa l resources, and (3) the differences hidden by the relative nature of this measure 29 Yet a ll alternative measure s of social movements contain their own limitations. Social movements are complex arrangements of organizations, individuals and ideas that ma ke comprehensive measurement exceptionally difficult. Likewise, there are logistic difficulties associated with the h ighly complex and multifaceted nature of many bills B ecause I needed to distinguish between redistribution and recognition bills, I did n ot include many of th e most complex bills (which also tend to be the largest bills) in the analysis. This approach strengthened the analytical rigor of the study, but it also resulted in some of the more important bills being excluded (because 28 For example, the "Million Man March" was a well publiciz ed gathering of hundreds of thousands of civil rights activists on the National Mall in Washington, D.C., on October 16, 1995. This event was organized by a variety of civil rights SMOs, including local chapters of the National Association for the Advance ment of Colored People (NAACP). However, many of the attendees were not formal members of these organizations. Thus, while my approach to operationalizing social movements would capture an important component of the movement that put on this event (the f ormal organizations involved), it entirely misses other important components (namely, people active in the movement but not associated with any formal organizations). 29 To account for the fact that some SMOs within an SMI favor a bill, while others oppose it I calculated SMI strength in favor of a bill by subtracting those likely to be against from those likely to be for. Thus, if the value (revenue, assets or revenue + assets) of all SMOs for a bill were $10 billion, and the value of all SMOs against the same bill were $10 billion, the SMI value for this bill would be 0 (10 10 = 0). But if there were no SMOs at all related to the bill, the SMI value would also be 0 (0 0 = 0). Yet these are clearly two very different situations. In the first, $20 billion is mobilized around a bill, though half for and half against. In the second, $0 is mobilized around a bill, none for and none against.

PAGE 138

128 their multi faceted nuance made it impossible to classify them as either recognition or redistribution ). When applying a recognition/redistribution lens to actual bills, analysts will always face this dilemma. Measures of public opinion also contained some inherent limitations Within this study, I had to rely on nine separate sources since no single data source included opinion data on all of my variables. As a result, I cannot account for differences in general approaches between research institutions. Differenc es in scales further compound this limitation, as different scales were used even within the same institution While I was able to standardize these by converting each scale into a percentage in favor, differences in scales likely impacted this effort. Po tentially even more important, this approach does not allow me to measure the intensity of opinion. That is while I can tell roughly how many people favor a bill, I cannot tell how strongly they favor it. Researchers can overcome t hese limitations by us ing a single source for opinion data, and scales that measure intensity of opinion However, this solution would most likely severely limit the variety of bills included in the study as no single source/scale covers all bills or the concepts that underpi n them Despite these limitations, this study provides valuable insigh t into the way social movements, interest groups, public opinion, political elites and the media interact to influence public policy. Additionally, it provides some much needed empiric al evidence for the contemporary debate over recognition in critical theory. There are however, limits to what empirical work can contribute to the moral philosophical aspects of critical theory. These results support particular accounts of practical po litics, political theory and social theory and in this sense, support the empirical assumptions made by Fraser, while

PAGE 139

129 partially refuting those made by Honneth. However, su ch empirical data cannot, on it s own, support or refute their respective moral phil osophies This task falls to continued discourse around the appropriate normative foundation for critical theory. Directions for Future Research This study has presented exciting findings that further our understanding of both movement impact and the empi rical underpinnings of critical theory. Yet, as discussed above, we need more research to generalize these findings beyond the context of federal bills in the 109 th Congress A logical next step would be to replicate this study with a different Congress preferably one controlled by Democrats This would have the added benefit of helping to determine whether social justice is substantially associated with partisan positions, or not. 30 Further, i f the findings hold under these circumstances, a next step c ould be to approximate this study within a different political system, cultural context and/or historical moment. Such studies would go a long way in determining the extent to which the se findings can be generalized to modern society or represent particul ar characteristics of the social, political cultural and economic context in which this study takes place. Another fruitful line of research entails the pursuit of the redistribution /recognition distinction in movement influence on policy using different measure s for social movement strength. This may include o ther monetary measures 30 Within these data, it is unclear whether the negative cor relation between interest group strength and bill passage is an artifact of these particular data (and the fact that, within these data, powerful interest group coalitions tended to target bills sponsored by the minority party, Democrats) or is a sign that social justice is a partisan issue (i.e., Democrats support bills which further social justice at higher rates than Republicans). Replicating this study with a Congress controlled by Democrats would go a long way in answering this question.

PAGE 140

130 associated with formal organizations (e.g. operating budgets etc.). However, money is not the only organizational resource employed by SMOs. S ome powerful organizations have very little money but an excess of other resources (i.e., charismatic leadership, knowledge, social legitimacy, etc.). As such, studies may continue to use SMI s but measure their strength through other means, such as membership numbers. Other approaches might attempt to measure movements independently of formal organizations, using events or social networks. Additionally, the literature would benefit from extensions of the redistribution /recognition distinction beyond movement outcomes and into areas su ch as mobilization or framing. Finally, the social movement literature would benefit from further study of the role the media plays in mediating movement effects Though this study has made preliminary contributions in this regard, m y media variable suffe rs from an inability to determine directionality. In using raw article counts, I was able to gauge how the total amount of coverage correlates to bill passage, but not how the type of coverage (for or against) matters. Additionally, because I use d a sing le media outlet (the New York Times), I risk ed over or underestimating total media coverage based on the biases associated with this source. Future research could improve upon this measure by including more media sources in their analysis, and by accounti ng for the content of the media (i.e. directionality, tone etc.). The critical theory literature would benefit from investigations into the implications of these findings for the nor mative foundations of a critical moral philosophy. For example, if these findings hold and Fraser's redistribution / recognition distinction is a fundamental characteristic of modern society, can Honneth modify his

PAGE 141

131 approach to include distributional elements of self realization? And if movements are not effective at impacting d istribution policy, can Fraser rethink her practical political stance to come up with a viable means to achieve distributional justice? Many other potential philosophical questions arise from these findings, but it will take another work to address them.

PAGE 142

132 REFERENCES Amenta, E., Carruthers, B. G., & Zylan, Y. (1992). A Hero for the Aged? The Townsend Movement, the Political Mediation Model, and U.S. Old Age Policy, 1934 1950. The American Journal of Sociology 98 (2), 308 339. Amenta, E. (2006). When M ovements Matter: The Townsend Plan and the Rise of Social Security (1St ed.). Princeton University Press. Amenta, E., & Caren, N. (2007). The Legislative, Organizational, and Beneficiary Consequences of State Oriented Social Movements. The Blackwell Compa nion to Social Movements (pp. 461 488). Malden, MA: Blackwell Publishing. Amenta, E., Caren, N., Chiarello, E., & Su, Y. (2010). The Political Consequences of Social Movements. Annual Review of Sociology 36 287 307.doi:10.1146/annurev soc 070308 120029 Amenta E., Caren, N., & Olasky, S. J. (2005). Age for Leisure? Political Mediation and the Impact of the Pension Movement on U.S. Old Age Policy. American Sociological Review 70 (3), 516 538. Andrews, K. T., & Edwards, B. (2004). Advocacy Organizations in the U.S. Political Process. Annual Review of Sociology 30 479 506. Agresti, A. (2007). An Introduction to Categorical Data Analysis (2nd ed.). Wiley Interscience. Benford, R. D., & Snow, D. A. (2000). Framing Processes and Social Movements: An Overv iew and Assessment. Annual Review of Sociology 26 611 639. Berstein, M. (2003). Nothing Ventured, Nothing Gained? Conceptualizing Social Movement "Success" in the Lesbian and Gay Movement. Sociological Perspectives 46 (3), 353 379. Bohman, J. (2012). C ritical Theory. In E. N. Zalta (Ed.), The Stanford Encyclopedia of Philosophy (Spring 2012.). Retrieved from http://plato.stanford.edu/archives/spr2012/entries/critical theo ry/ Buechler, S. M. (2011). Understanding Social Movements: Theories from the Classical Era to the Present Paradigm Publishers. Burstein, P. (1998). Bringing the Public Back In: Should Sociologists Consider the Impact of Public Opinion on Public Policy? Social Forces 77 (1), 27 62. doi:10.1093 /sf/77.1.27

PAGE 143

133 Burstein, P. (1999). Social Movements and Public Policy. In M. Giugni, D. McAdam, & C. Tilly (Eds.), How Social Movements Matter Minneapolis, MN: University of Minnesota Press. Burstein, P., & Linton, A. (2002). The Impact of Political Parties, Interest Groups, and Social Movement Organizations on Public Policy: Some Recent Evidence and Theoretical Concerns. Social Forces 81 (2), 381 408. Burstein, P. (2006). Why Estimates of the Impact of Public Opin ion on Public Policy are Too High: Empirical and Theoretical Implications. Social Forces 84 (4), 2273 2289. doi:10.1353/sof.2006.0083 Burstein, P. (2010). Public Opinion, Public Policy, and Democracy. In K. T. Leicht & J. C. Jenkins (Eds.), Handbook of Po litics (pp. 63 79). New York, NY: Springer New York. Carroll, W. K., & Ratner, R. S. (1999). Media Strategies and Political Projects: A Comparative Study of Social Movements. The Canadian Journal of Sociology / Cahiers canadiens de sociologie 24 (1), 1 3 4. Cook, F. L., Tyler, T. R., Goetz, E. G., Gordon, M. T., Protess, D. Leff, D. R., & Molotch, H. L. (1983). Media and Agenda Setting: Effects on the Public, Interest Group Leaders, Policy Makers, and Policy. Public Opinion Quarterly 47 (1), 16 35. doi:1 0.1086/268764 Costain, A. N., & Majstorovic, S. (1994). Congress, Social Movements and Public Opinion: Multiple Origins of Women's Rights Legislation. Political Research Quarterly 47 (1), 111 135. doi:10.1177/106591299404700106 Cress, D. M., & Snow, D. A (1996). Mobilization at the Margins: Resources, Benefactors, and the Viability of Homeless Social Movement Organizations. American Sociological Review 61 (6), 1089.doi:10.2307/2096310 Cress, D. M., & Snow, D. A. (2000). The Outcomes of Homeless Mobiliza tion: The Influence of Organization, Disruption, Political Mediation, and Framing. The American Journal of Sociology 105 (4), 1063 1104. deLeon, P. (1999). The stages approach to the policy process: What has it done? Where is it going? In P. Sabatier (Ed. ), Theories of the Policy Process (1st ed.). Boulder, CO: Westview Press. Earl, J. (2000). Methods, movements, and outcomes, Research in Social Movements, Conflicts and Change (22), 3 25. doi:10.1016/S0163 786X(00)80033 6 Edwards, B., & McCarthy, J. D. ( 2007). Resources and Social Movement Mobilization. In D. A. Snow, S. A. Soule, & H. Kriesi (Eds.), The Blackwell Companion to Social

PAGE 144

134 Movements (pp. 116 152). Blackwell Publishing Ltd. Retrieved from http://onlinelibrary.wiley.com/doi/10.1002/9780470999103. ch6/summary Fraser, N., & Honneth, A. (2003). Redistribution or recognition?: A Political Philosophical Exchange Brooklyn, NY: Verso. Fraser, N. (2008). Adding Insult to Injury: Nancy Fraser Debates Her Critics Verso. Gamson, W. A. (1990). The Strat egy of Social Protest (2 Sub.). Wadsworth Pub Co. Gamson, W. A. (2007). Bystanders, Public Opinion, and the Media. In D. A. Snow, S. A. Soule, & H. Kriesi (Eds.), The Blackwell Companion to Social Movements (pp. 242 261). Blackwell Publishing Ltd. Retriev ed from http://onlinelibrary.wiley.com/doi/10.1002/9780470999103.ch11/summary Gamson, W. A., & Wolfsfeld, G. (1993). Movements and Media as Interacting Systems. Annals of the American Academy of Political and Social Science 528 114 125. Giugni, M. (199 8 ). Was It Worth the Effort? The Outcomes and Consequences of Social Movements. Annual Review of Sociology 24 371 393. Giugni, M. (2004). Social Protest and Policy Change: Ecology, Antinuclear, and Peace Movements in Comparative Perspective Rowman & Li ttlefield Publishers. Giugni, M. (2007). Useless Protest? A Time Series Analysis of the Policy Outcomes of Ecology, Antinuclear, and Peace Movements in the United States, 1977 1995. Mobilization 12 (1), 53 77. Giugni, M. (2008). Political, Biographical, and Cultural Consequences of Social Movements. Sociology Compass 2 (5), 1582 1600. doi:10.1111/j.1751 9020.2008.00152.x Giugni, M., & Passy, F. (1998). Social Movements and Policy Change: Direct, Mediated, or Joint Effect American Sociological Associatio n Section on Collective Behavior and Social Movements Working Paper Series, 1(4). Goldstone, J. A. (1980). The Weakness of Organization: A New Look at Gamson's The Strategy of Social Protest. American Journal of Sociology 85 (5), 1017 1042. Goodwin, J., & Hetland, G. (2009). The Strange Disappearance of Capitalism from Social Movement Studies. Presented at the Comparative and Historical Sociology Section of the American Sociological Association 2009 Mini Conference, Berkeley, CA. Gorz, A. (1967). Strate gy for labor: A radical proposal Beacon Press. Hair, J. F. Jr., Anderson, R. E., Tatham, R. L. & Black, W. C. (1995). Multivariate Data

PAGE 145

135 Analysis, 3rd edn. New York: Macmillan. Hartley, T., & Russett, B. (1992). Public Opinion and the Common Defense: Wh o Governs Military Spending in the United States? The American Political Science Review 86 (4), 905.doi:10.2307/1964343 Harvey, D. (2007 ). A Brief History of Neoliberalism Oxford University Press. Hill, K. Q., & Hinton Anderson, A. (1995). Pathways of R epresentation: A Causal Analysis of Public Opinion Policy Linkages. American Journal of Political Science 39 (4), 924.doi:10.2307/2111663 Johnson, E. (2008). Social Movement Size, Organizational Diversity and the Making of Federal Law. Social Forces 86 (3 ), 967 993. Johnson, E. W., Agnone, J., & McCarthy, J. D. (2000). The Contingent Impact of U.S. Environmental Movement Organizations and Activities on Agenda Setting and Law Passage. Social Forces 88 (5), 2267 2292. Kelly, W. R., & Snyder, D. (1980). Rac ial Violence and Socioeconomic Changes Among Blacks in the United States. Social Forces 58 (3), 739 760. doi:10.1093/sf/58.3.739 Khawaja, M. (1994). Resource Mobilization, Hardship, and Popular Collective Action in the West Bank. Social Forces 73 (1), 191 220. doi:10.1093/sf/73.1.191 King, G., & Zeng, L. (2001a). Explaining Rare Events in International Relations. International Organization 55 (03), 693 715. doi:10.1162/00208180152507597 King, G., & Zeng, L. (2001b). Logistic Regression in Rare Events Dat a. Political Analysis 9 (2), 137 163. Kriesi, H. (1995). New social movements in Western Europe: a comparative analysis Psychology Press. Lipset, S. M., & Marks, G. (2001). It didn't happen here: why socialism failed in the United States W. W. Norton & Company. Lunder, E. (2007). Tax Exempt Organizations: Political Activity Restrictions and Disclosure Requirements. Congressional Research Service Reports Taxation Tax Exempt Organizations: Political Activity Restrictions and Disclosure Requirements, S eptember 11, 2007 [i]. Luders, J. E. (2010). The Civil Rights Movement and the Logic of Social Change (1st ed.). Cambridge University Press.

PAGE 146

136 McAdam, D., Tarrow, S., & Tilly, C. (2001). Dynamics of Contention (Cambridge Studies in Contentious Politics) Cambridge University Press. McCammon, H. J. (2009). Beyond Frame Resonance: The Argumentative Structure and Persuasive Capacity of Twentieth Century U.S. Women's Jury Rights Frames. Mobilization 14 (1), 45 65. McCarthy, J. D., & Zald, M. N. (1973). The T rend of Social Movements in America: Professionalization and Resource Mobilization. Retrieved from http://deepblue.lib.umich.edu/handle/2027.42/50939 McCarthy, J. D., & Zald, M. N. (1977). Resource Mobilization and Social Movements: A Partial Theory. The American Journal of Sociology 82 (6), 1212 1241. McCarthy J. D., & Zald, M. N. (2001). The Enduring Vitality of the Resource Mobilization Theory of Social Movements. In J. H. Turner (Ed.), Handbook of Sociological Theory (pp. 533 565). Springer US. May er, L. H. (2010). Grasping Smoke: Enforcing the Ban on Political Activity by Charities. SSRN eLibrary Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=991129 Menar d, S. (1995). Applied Logistic Regression Analysis: Sage University Series on Quantitative Applications in the Social Sciences. Thousand Oaks, CA: Sage. Meyer, D. (2003). Social Movements and Public Policy: Eggs, Chicken and Theory. Center for the Study of Democracy, UC Irvine. Meyer, D. S. (2004). Protest and Political Opportunities. Annual Review of Sociology 30 125 145. Meyer, D. S., & Minkoff, D. C. (2004). Conceptualizing Political Opportunity. Social Forces 82 (4), 1457 1492. doi:10.1353/sof.20 04.0082 Michels, R. (1915). Political Parties: A Sociological Study of the Oligarchical Tendencies of Modern Democracy Hearst's International Library Company. Mirowsky II. J. & Ross, C.E. (1981). Protest Group Success: The Impact of Group Characteristi cs, Social Control, and Context. Sociological Focus 14 (3), 177 192. National Center for Charitable Statistics. (2006). Guide to Using NCCS Data Retrieved from http://nccsdataweb.urban.org/kbfiles/468/NCCS data guide 2006c.pdf Nemes, S., Jonasson, J., G enell, A., & Steineck, G. (2009). Bias in odds ratios by logistic regression modeling and sample size. BMC Medical Research Methodology 9 (1), 56.doi:10.1186/1471 2288 9 56

PAGE 147

137 Olzak, S., & Soule, S. A. (2009).Cross Cutting Influences of Environmental Protest and Legislation. Social Forces 88 (1), 201 225. doi:10.1353/sof.0.0236 Page, B. I., & Shapiro, R. Y. (1983). Effects of Public Opinion on Policy. The American Political Science Review 77 (1), 175.doi:10.2307/1956018 Peduzzi, P., Concato, J., Kemper, E., Holford, T. R., & Feinstein, A. R. (1996). A simulation study of the number of events per variable in logistic regression analysis. Journal of Clinical Epidemiology 49 (12), 1373 1379. doi:10.1016/S0895 4356(96)00236 3 Piven, F., & Cloward, R. (1978). Po or People's Movements: Why They Succeed, How They Fail Vintage. Ritzer, G. (Ed.). (2007). The Blackwell Encyclopedia of Sociology (1st ed.). Wiley Blackwell. Schumaker, P. D. (1975). Policy Responsiveness to Protest Group Demands. The Journal of Politics 37 (02), 488 521. doi:10.2307/2129004 Rohlinger, D. A. (2002). Framing the Abortion Debate: Organizational Resources, Media Strategies, and Movement Countermovement Dynamics. The Sociological Quarterly 43 (4), 479 507. Sabatier, P. A., & Weible, C. M. ( 2007). The Advocacy Coalition Framework: Innovations and Clarifications. Theories of the Policy Process (Second., pp. 189 220). Boulder, CO: Verso. Sabatier, P. A. (Ed.). (2007). Theories of the Policy Process (Second.). Boulder, CO: Westview Press. Schu maker, P. D. (1975). Policy Responsiveness to Protest Group Demands. The Journal of Politics 37 (02), 488 521. doi:10.2307/2129004 Schwartz, M. A. (2000). Continuity Strategies among Political Challengers: The Case of Social Credit. American Review of Can adian Studies 30 (4), 455 477. doi:10.1080/02722010009481063 Shorter, E., & Tilly, C. (1971). The Shape of Strikes in France, 1830 1960. Comparative Studies in Society and History 13 (01), 60 86. doi:10.1017/S0010417500006113 Snow, D. A. (2004). Social M ovements as Challenges to Authority: Resistance to an Emerging Conceptual Hegemony. Research in Social Movements, Conflicts and Change (Vol. 25, pp. 3 25). Bingley: Emerald (MCB UP ).

PAGE 148

138 Snow D. A., & Soule, S. A. (2009). A Primer on Social Movements W. W Norton & Company. Snow, D. A., Soule, S. A., & Kriesi, H. (Eds.). (2007). The Blackwell Companion to Social Movements Malden, MA: Blackwell Publishing. Snow, D. A., Soule, S. A., & Kriesi, H. (2008). The Blackwell Companion to Social Movements John Wiley & Sons. Snyder, D., & Kelly, W. R. (1976). Industrial Violence in Italy, 1878 1903. American Journal of Sociology 82 (1), 131 162. Soule, S. A., & Olzak, S. (2004). When Do Movements Matter? The Politics of Contingency and the Equal Rights Amendmen t. American Sociological Review 69 (4), 473 497. Staggenborg, S. (1988).The Consequences of Professionalization and Formalization in the Pro Choice Movement. American Sociological Review 53 (4), 585.doi:10.2307/2095851 Steedly, H. R., & Foley, J. W. (197 9). The success of protest groups: Multivariate analyses. Social Science Research 8 (1), 1 15. doi:10.1016/0049 089X(79)90011 5 Tarrow S. (1993). Social Protest and Policy Reform May 1968 and the Loid'Orientation in France. Comparative Political Studies 25 (4), 579 607. doi: 10.1177/0010414093025004006 Tarrow, S. G. (2011). Power in Movement: Social Movements and Contentious Politics Cambridge University Press. Tilly, C., & Tarrow, S. (2006 ). Contentious Politics Paradigm Publishers. Turina, F. (2009) Social movements and public policy: An examination of the California timber wars (Dissertation). University of Colorado at Denver, Denver CO. Turner, J. H. (Ed.). (2001). Handbook of Sociological Theory Springer US. Uba, K. (2009). The Contextual Dep endence of Movement Outcomes: A Simplified Meta Analysis. Mobilization 14 (4), 433 448. Wilson, R. K., & Young, C. D. (1997). Cosponsorship in the U. S. Congress. Legislative Studies Quarterly 22 (1), 25 43. doi:10.2307/440289 Zald, M. (1992). Looking Ba ckward to Look Forward: Reflections on the Past and Future of the Resource Mobilization Research Program. In A. D. Morris & C. M. Mueller (Eds.), Frontiers in Social Movement Theory (pp. 326 348).