Citation
Women in state legislatures

Material Information

Title:
Women in state legislatures representation and the policy process
Creator:
Schumacher, Kristin L
Place of Publication:
Denver, Colo.
Publisher:
University of Colorado Denver
Publication Date:
Language:
English
Physical Description:
xiv, 210 leaves : ; 28 cm

Thesis/Dissertation Information

Degree:
Doctorate ( Doctor of Philosophy)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
School of Public Affairs, CU Denver
Degree Disciplines:
Public Affairs
Committee Chair:
Guy, Mary E.
Committee Members:
Fitzpatrick, Jody L.
Gover, Angela R.
Juenke, Eric G.

Subjects

Subjects / Keywords:
Women legislators -- United States ( lcsh )
Representative government and representation -- United States ( lcsh )
Policy sciences -- United States ( lcsh )
Policy sciences ( fast )
Representative government and representation ( fast )
Women legislators ( fast )
United States ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 188-210).
General Note:
School of Public Affairs
Statement of Responsibility:
by Kristin L. Schumacher.

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:
779973380 ( OCLC )
ocn779973380
Classification:
LD1193.P86 2011D S337 ( lcc )

Full Text
WOMEN IN STATE LEGISLATURES:
REPRESENTATION AND THE POLICY PROCESS
by
Kristin L. Schumacher
B.S., Nebraska Wesleyan University, 2002
M.S.W., The University of Texas, 2004
A thesis submitted to the
University of Colorado Denver
in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Public Affairs
2011


2011 by Kristin L. Schumacher
All rights reserved.


This thesis for the Doctor of Philosophy
degree by
Kristin L. Schumacher
has been approved
by
, Eric G. Juenke

Date


Schumacher, Kristin L. (Ph.D., Public Affairs)
Women in State Legislatures: Representation and the Policy Process
Thesis directed by Professor Mary Guy
ABSTRACT
Women have not achieved equality in regards to political representation within the
US. This exclusion weakens the quality of the political debate and destabilizes the
legitimacy of our democracy. However, inroads have been made in building a more
representative democracy at both the state and federal level in the US. This study
extends the literature regarding gender and representation, exploring the association
between descriptive representation, legislative factors, civil society contexts and the
substantive representation of women.
The quantitative research design employed in this study was adapted from a model
proposed by Beckwith and Cowell-Meyers, rooted in the theory of representation as
developed by Hanna Pitkin. The model analyzes substantive representation as a
function of descriptive representation, legislative factors, and civil society contexts.
The dependent variable, substantive representation, is measured at three different
stages of the policy process: bill sponsorship, roll-call voting, and policy outputs.
Data were collected for all 50 states for two time periods and across 25 states for two
additional time periods. This resulted in a comprehensive and original panel dataset
measuring substantive representation of women at three different stages of the policy
process at four different time periods, n = 150. The data were analyzed using fixed
effects analysis.
The results reveal that the theoretical model did not significantly predict the variance
of the substantive representation of women at any stage of the policy process. These
insignificant results are significant in that this model is the first of its kind to test the
hypotheses using comparative state-level data. Insignificant results provide room for
analysis investigating alternative measurement of key variables, inclusion of
alternative hypotheses, as well as consideration of alternative causal models.
Future research should challenge key assumptions commonly found in the literature.
This includes the assumption that women unilaterally act for women. The new
question should be who acts for women? In addition, scholars must not make a prion


assumptions of substantive representation. Rather, the construct must be analyzed
respective of geographical, temporal, and ideological variation of women. Future
research should challenge the classic definition of substantive representation, as well
as the common tools for measuring substantive representation.
This abstract accurately represents the content of the candidate's thesis. I recommend
its publication.
Signed
Mdry E. Guy


DEDICATION
To Matthew, my love.


ACKNOWLEDGEMENTS
The journey towards my PhD has been marked and advanced by kind words. I would
first like to acknowledge and thank my committee for their kind words, as well as
their hard work on my behalf. The comments and suggestions offered by all have
been utilized with deep appreciation, but it was the kind words of encouragement and
support that spurred me forward in my efforts. Dr. Mary Guy has been a constant
champion offering the right advice and encouragement, while allowing me to move
through my process at my own pace. This has been deeply appreciated. Dr. Jody
Fitzpatrick, Dr. Angela Gover, and Dr. Eric Juenke all offered kind words both early
and late in my process that reinforced my resolve and helped me move forward. Im
deeply grateful for my committees support which allowed me room to craft a
dissertation suited to my own unique interests at my own pace.
I must also acknowledge and thank Dr. Peter deLeon, role model and friend. The
kind words offered by Dr. deLeon in a memo following my first semester in the
program offered a much needed boost in confidence and morale. Without these kind
words, I would not have completed my doctoral degree. These kind words, often
repeated by Dr. deLeon, became a goal that Ive continuously worked towards
throughout my doctoral education. Thank you, Grand Poo-Bah!
It is also necessary to acknowledge and thank my friends and peers, Dr. Katrina
Miller-Stevens and Dr. Saba Siddiki. Their support, evident in both kind words and
deeds, was invaluable in good times and bad. And finally, I would like to
acknowledge the emotional work my partner, Matthew, invested in my process,
grounding me and boosting me along at regular intervals. Thank you, my love!


TABLE OF CONTENTS
Figures.........................................................xiii
Tables..........................................................xiv
CHAPTER
1. INTRODUCTION...................................................1
Overview.....................................................1
Problem Statement............................................4
Purpose.................................................... 6
Methodology..................................................7
Significance.................................................8
Organization of Study...................................... 9
2. REVIEW OF LITERATURE..........................................11
Theoretical Framework: The Concept of Representation........11
Representation, Broadly..................................11
Competing Arguments.................................... 15
Conceptual Framework........................................19
The Model................................................19
Representation of Women..................................21
Descriptive Representation...............................31
Legislative Factors......................................36
IX


Societal Contexts
43
Critical Analysis..............................................64
Theoretical Contributions...................................64
Unit of Analysis.......................................... 65
Sampling & Research Design..................................67
Summary........................................................69
3. METHODOLOGY.................................................... 7]
Research Purpose, Question & Hypotheses........................71
Research Design................................................73
Unit & Period of Analysis...................................74
Variables......................................................75
Substantive Representation: The Dependent Variables.........75
Independent Variables..................................... 77
Data Collection................................................82
Dependent Variables.........................................82
Independent Variables.......................................86
Panel Data Analysis.......................................... 88
Panel Data Explained........................................88
Types of Panel Data Analysis.............................. 92
Panel Data Diagnostics......................................99
Statistical Analysis..........................................101
x


Model One: Policy Priorities................................ 102
Model Two: Policy Preferences...............................108
Model Three: Policy Outputs.................................110
Limitations.....................................................1 11
Reliability.................................................111
Validity....................................................113
Summary.........................................................115
4. RESULTS...........................................................117
Descriptive Statistics..........................................117
Model One: Policy Priorities....................................125
Model Two: Policy Preferences...................................129
Model Three: Policy Outputs.....................................132
Summary.........................................................134
5. DISCUSSION........................................................137
Hypothesis One: Descriptive Representation......................137
Critical Analysis...........................................139
Hypothesis Two: Political Party Influence.......................143
Critical Analysis...........................................145
Hypothesis Three: Womens Caucus................................148
Critical Analysis...........................................150
Hypothesis Four: Citizen Ideology...............................153
xi


Critical Analysis..........................................154
Hypothesis Five: Gender Opportunity Structure............... 156
Critical Analysis..........................................157
Implications..................................................160
Model Misspecification & Type II Error.....................160
Correct Model Specification & True Negative................168
Future Research.............................................. 176
Conclusion....................................................179
APPENDIX............................................................181
A. Womens Interest Legislation Coding Framework.............181
B. LexisNexis Search Terms...................................183
C. State Sampling and Data Collection........................185
D. Fixed Effects Regression Assumptions......................187
REFERENCES..........................................................188
xii


FIGURES
Figure
2.1. BECKWITH & COWELL-MEYERS THEORETICAL MODEL.......20
2.2. DESCRIPTIVE REPRESENTATION OF WOMEN AT THE STATE AND
FEDERAL LEVEL, 1975-2011 .........................67
2.3 THEORETICAL MODEL.................................70
3.1 THEORETICAL MODEL.................................73
3.2 STATE DATA COLLECTION.............................85
3.3 EXAMPLE OF A PANEL DATASET...................... 89
3.4 UNOBSERVED HETEROGENEITY..........................91
3.5 BETWEEN AND WITHIN EFFECTS........................97
4.1 FREQUENCY FOR WOMENS CAUCUS VARIABLE............121
xiii


TABLES
Table
3.1 Operationalization of Dependent Variables................................77
3.2 Operationalization of Independent Variables..............................82
3.3 Comparison of Fixed Effects and Random Effects Models....................99
4.1 Descriptive Statistics for the Dependent Variables......................118
4.2 Descriptive Statistics for the Independent Variables....................120
4.3 Intercorrelations Between Dependent and Independent Variables..........122
4.4 Fixed Effects Results for Model One: Policy Priorities..................127
4.5 Fixed Effects Results for Model Two: Policy Preferences.................130
4.6 Fixed Effects Results for Model Three: Policy Outputs...................133
4.7 Summary Results of Hypotheses Tests.....................................135
5.1 Summary Results of Hypotheses One.......................................138
5.2 Summary Results of Hypothesis Two.......................................145
5.3 Summary Results of Hypothesis Three.....................................150
5.4 Summary Results of Hypothesis Four......................................154
5.5 Summary Results of Hypothesis Five......................................157
xiv


CHAPTER ONE:
INTRODUCTION
We need not suppose that when power resides in an exclusive class, that class will
knowingly and deliberately sacrifice the other classes to themselves: it suffices that,
in the absence of its natural defenders, the interest of the excluded is always in
danger of being overlooked: and, when looked at, is seen with very different eyes
from those of the persons whom it directly concerns.
John Stuart Mill
There never will be complete equality> until women themselves help to make laws and
elect lawmakers.
Susan B. Anthonv
Overview
Women have not achieved equality in regards to political representation
within the United States. This exclusion weakens the quality of the political debate
and destabilizes the legitimacy of our democracy (Mansbridge, 1999). Inroads have
been made in building a more representative democracy at both the state and federal
level in the US, but parity is still a distant spot on the horizon. For example, the
number of women holding office within the United States has steadily increased at
the state and federal level since the 1970s. In 1975, only 9% of those holding office
in the lower chamber of state legislatures were women, but today nearly 25% of the
seats are held by women. This number is lower at the federal level where, currently,
16.5% of the seats of the House of Representatives are held by women.
Comparatively, the US ranks far lower in female political representation than any


other industrialized country (CAWP, 201 la).
Women have advocated for political representation in the US for over a
century, arguing that political representation will result in public policy that reflects
womens rights and interests (Sanbonmatsu, 2003; Swers, 2002a; Thomas, 1994).
From the late 1960s, scholars have focused on analyzing the impact of female
political representation within the US and internationally (Hawkesworth, 1994), and,
since then, the study of female representation in political bodies has produced a
significant collection of empirical and theoretical literature (Lovenduski, 2005b;
Phillips, 1995; Sapiro, 1981).
Research within this vein is concerned with investigating the interaction of
gender, the state, and policy, where gender is .a way of referring to the social
organization of the relationship between the sexes and also as a primary way of
signifying power (Scott, 1986, p. 1053). Introduction of gender into the analysis of
politics and public policy highlights an awareness of inequality based on biological
differences perpetuated through policies created by the state (Mazur. 2002).
Theory and research concerning gender, politics and policy often encompass
questions relating to representation. Representation is a measure of democracy, and
the US was founded upon the values of popular representation and governance
(deLeon, 1997, p. 1). The exploration of gender, representation and democracy is
important considering the relevance of the constructs within US history and within
the academy. As stated by Reingold (2000):
i


Womens presence in such positions of power confirms and upholds cherished
democratic ideals of equal and full participation of all citizens. It affirms the
openness and thus the legitimacy of our democratic system by demonstrating
that access to positions of power and influence in public affairs are available
to all, no matter what their gender, race, class, national origin, or other
attributes, (p.32)
Representation of women and normative concepts of democracy were first
united by the 2nd wave of the US feminist movement when activists began to advocate
for gender equality of women within political life. At the same time, normative
arguments began to appear claiming that gender equality in political institutions is a
gauge of justice and democracy. Since then, large international and transnational
organizations such as the United Nations, European Union, Summit of Americas, and
the Association of Southeast Asian Nations have declared that an increase in the
political representation of women would lead to a truly representative democracy and
greater economic and social progress (Htun, 2004).
As discussed by Phillips (1995), equal representation leads to a shift from the
politics of ideas to that of politics of presence. The politics of presence places
a diverse, representative group of individuals within the policymaking process where
deliberative discourse can lead to advocacy for a wide range of interests and goals.
Research questions relating to representation are commonly based upon the
work of Hanna Pitkin (1967). Pitkins text, The Concept of Representation, has
informed research concerning female representation and public policy, and this
3


framework has become foundational in research exploring gender and representation
(Schwindt-Bayer & Mishler, 2005).
In her classic text, Pitkin established four types of representation: descriptive,
formal, substantive, and symbolic representation. The causal link between
descriptive and substantive representation is the most commonly researched link
among the four types of representation (Schwindt-Bayer & Mishler, 2005), and it is
the focus of this study. Descriptive representation is being typical, or resembling
(p. 90), and is based on demographic characteristics. Descriptive representation is
secured when the representative group is an approximation of the population at large.
Substantive representation is the statement of the goals and interests of a group
represented by the individual. Pitkin defines substantive representation .. .as an
acting for others, an activity on behalf of, in the interest of, as the agent of. someone
else (p.l 13). Gerrity, Osborn and Morehous Mendez (2007) claim
...if descriptive representation translates into substantive representation,
womens current descriptive underrepresentation in legislatures in the United
States can result in a lack of substantive policy that address the unique
concerns of women, such as womens health, child care, and the workplace.
(p.179)
Thus, it is important to understand the context in which descriptive representation
translates to substantive representation as a matter of democracy and equality .
Problem Statement
Research has not yet established the specific mechanisms that result in a
translation of descriptive representation to substantive representation of women
4


(Beckwith & Cowell-Meyers, 2007). Compelling studies point to the influence of a
lone individual, the policy entrepreneur (S. Carroll, 1984; Chaney. 2006; Childs &
Krook, 2005). Others argue that a critical mass of women is necessary to enact
change (Bratton, 2005). Based upon the work of Rosabeth Moss Kanter (1977a,
1977b), the debate within the literature focuses on the point at which a group of
women is large enough to create change. Opinions and research vary on what
percentage is necessary to bring about substantive representation, with the critical
mass ranging from 15% to 30% (Beckwith & Cowell-Meyers, 2007; Childs & Krook,
2005).
Others argue that it is overly simplistic to simply review the percentage of
woman in a representative body without considering the institutional and societal
contexts that may play a large role in the substantive representation of women (Childs
& Krook, 2005; Lovenduski, 2005a; Weldon, 2002). For example, Beckwith and
Cowell-Meyers (2007) argue that in comparative research both parliamentary factors
and the civil society contexts impact substantive representation, in addition to
descriptive representation. Historically, the dominant group has not consisted of
women or other minority populations yet policy advancements have been made for
groups other than the majority party. Therefore, it can be theorized that certain
elements within political institutions and civil society also influence substantive
representation of women (Beckwith & Cowell-Meyers, 2007; Weldon, 2002).
5


Additional research is necessary to understand the mechanisms that result in
the substantive representation of women. Despite the extensive body of research,
scholars have yet to comprehensively explore the relationship between descriptive
and substantive with specific consideration to institutional and societal contexts. This
study fills a gap in the research by analyzing a comprehensive model of descriptive
and substantive representation across all 50 states for two time periods, and across 25
states for an additional two time periods.
Purpose
The research question guiding this study is as follows:
What is the impact of descriptive representation of women, legislative factors,
and societal contexts on the substantive representation of women at three
different stages of the policy process, within the US states?
First, this study addresses the following question by exploring how descriptive
representation, legislative factors, and societal contexts impact substantive
representation. Second, this study examines how this impact varies at different stages
of the policy process. The hypotheses are as follows:
Hi: State legislatures that have more female legislators will have higher rates
of substantive representation of women.
FD: State legislatures that have a greater Democratic majority will have higher
rates of substantive representation of women.
H3: State legislatures that have a womens caucus will have higher rates of
substantive representation of women.
H4: States that have a more liberal citizen ideology will have higher rates of
substantive representation of women.
6


ff: States that have a more liberal gender opportunity structure will have
higher rates of substantive representation of women.
Detailed information relating to the research question and hypotheses is included in
Chapters Two and Three.
Methodology
This study focused on all 50 US states for two periods of time, 1995 and 2005,
and 25 states for two additional periods of time 1975 and 1985. Analysis for the
years 1975 and 1985 were limited to 25 states due to the availability of data. The
research design utilized panel data analysis in order to explore how each of the
independent variables contributes to the variance in each of the dependent variables.
The unit of analysis within this research design is the general session within the lower
chamber of the state legislature.
Data collection resulted in an original dataset with three dependent variables
and five independent variables. The three quantitative models were based on each of
the three dependent variables, measuring the unique impact of the combination of
predictor variables on each dependent variable. The data were analyzed using panel
data analysis to determine if descriptive representation, political party influence,
womens caucus, citizen ideology and gender opportunity structure predict the
variance in substantive representation of women.
7


Significance
This research design addresses three significant gaps within the literature
regarding the unit of analysis, the theoretical model, and sampling. The research
question and hypotheses supplement current research by investigating similar
questions with new data and research methods. As a result, the significance of this
study is largely empirical in nature.
Research on gender and representation often focuses on the individual as the
unit of analysis, closely reviewing how individual traits and characteristics impact
substantive representation. Those that study gender and policy have begun to note
that the context in which policy is made also matters (Cammisa & Reingold, 2004).
The substantive representation of women is subject to a variety of influences such as
the legislative body as a whole, or external political factors. In order to examine
these influences one must also analyze relationships at the macro level. Research at
the aggregate level is not as commonly studied as micro-level analysis focusing on
the individual (Bratton. 2005; Cammisa & Reingold, 2004; Reingold, 2008).
Application of the theoretical model posed by Beckwith and Cowell-Meyers (2007)
addresses these gaps in the literature, requiring a macro-level unit of the analysis as
well as the inclusion of variables regarding political and social contexts.
Current analysis of gender, policy and representation within the US is limited.
It wasnt until 1992, or the Year of the Woman, when enough women filled the seats
of Congress to justify analysis of representation of women (Swers, 2002). At the
8


state level women have filled representative positions in greater numbers. However,
analysis typically focuses on a small sample of six to twelve states. This research
design contributes to the literature by extending analysis to all 50 states. Pursuing
this research design required the development of a new dataset, including a
comprehensive coding of womens interest legislation across all 50 states for two
periods of time and 25 states for an additional two time periods. A similar dataset
does not exist and offers the opportunity to explore uncharted territory in regards to
the research question and hypotheses, as well as a myriad of other research questions
that have yet to be answered. These unanswered questions are directly related to
gender, policy and representation and are a matter of concern for gender equality and
democracy.
Organization of Study
This study is organized in a standard fashion. Chapter One provides an
introduction to the topic and an overview of the study. Chapter Two offers a review
of relevant literature, which has been organized according to the dependent and
independent variables. Chapter Three reviews the methodology including a recap of
the guiding research question and hypotheses, operationalization of variables and data
collection procedures for each variable, and a review of panel data analysis,
regression diagnostics, and the limitations of the research design. Chapter Four
describes the results of the descriptive analysis, as well as the panel data analysis.
9


Chapter Five concludes with a thorough discussion of the statistical results, including
theoretical implications and a future research agenda.
10


CHAPTER TWO:
REVIEW OF LITERATURE
This chapter provides a summary of the literature regarding gender,
representation and the policy process. The first section reviews Pitkins theoretical
framework along with common critiques of the framework. The second section
outlines the conceptual framework, summarizing research and literature regarding
each of the independent and dependent variables. Each sub-section will end with the
related hypothesis. The third section offers a critical analysis of existing literature,
highlighting the gaps in the research and the ways in which the current research
design endeavors to address these issues.
Theoretical Framework: The Concept of Representation
Representation, Broadly
Hanna Pitkins work offers a theoretical framework for political
representation. Current and historical research concerning gender, representation, and
public policy relies heavily on this theoretical foundation (Beckwith & Cowell-
Meyers, 2007; Lovenduski, 2005b; Schwindt-Bayer & Mishler, 2005; Simon
Rosenthal, 1995). In her classic text, Pitkin established four types of representation:
formal, symbolic, descriptive, and substantive representation. Formal representation
refers to the processes of authorization and accountability in which an individual
stands for another. Specifically, formal representation is not concerned with the
11


action of representing, or even the qualifications or characteristics of a suitable
representative. Instead, formal representation is the process in which one becomes a
representative. In the United States, this is typically achieved through the election
process. Authorization of a representative occurs when they are elected to office, and
a representative is accountable to constituents due to the fact that they are required to
seek re-election. Formal representation does not concern itself with what occurs after
the fact; to be elected is to represent (Pitkin, 1967).
The second type is symbolic representation. Pitkin (1967) explains that
symbolic representation is the extent to which a representative stands as a symbol of
the constituency. Popular examples include the British monarchy, the Catholic Pope,
or transformational heads of state. While a representative may be accepted as a
leader and may act for the constituency, it does not mean that the constituency will
believe that the representative is a symbol of the nation. Symbolic representation is
ultimately concerned with how the representative is perceived and evaluated
(Schwindt-Bayer & Mishler, 2005), and the feelings this evokes from constituents
(Dovi, 2007). The ultimate question is: Do women feel like they are represented
(Dovi, 2007)?
Descriptive representation is 'being typical, or resembling or "standing for
and is based on socio-demographic characteristics (Pitkin, 1967, p. 90, p. 60).
Descriptive representation is realized when the representative group is a reflection of
the population at large. As Pitkin explains, descriptive representation does not
12


concern the actions or motivations of the representative, but simply their socio-
demographic characteristics (Pitkin, 1967). A persons DNA ensures that they have a
unique genetic makeup offering hundreds, if not thousands, of characteristics that one
may be represented by. In addition, one may also be represented according to social
location spatial, temporal, and economic. This results in an array of socio-
demographic characteristics from which one can be descriptively represented.
The study of representation and the policy process focuses almost exclusively
on politicized socio-demographic variables. For example, research and theory
regarding representation does not focus on a group of tall individuals or short
individuals, or on whether one is right-handed, or left. Descriptive representation
focuses on physical ability, race and ethnicity, sexuality, and gender, to name a few.
These are socio-demographic characteristics that, when grouped together, typically
result in a substantial difference in social and economic position. They are
considered political because differences in social and economic position result in
differing interests and needs with and between groups (Sapiro, 1981).
Thus far, a review of Pitkins types of representation has illustrated how an
individual becomes a representative through the process of authorization and
accountability, such as formal representation. Symbolic and descriptive
representation refers to types of representation that occur passively through the ways
in which a representative is a symbol of the nation, or through the ways in which a
representative reflects the composition of constituents based on socio-demographic
13


characteristics. The final category, substantive representation, focuses on the activity
of representing.
Substantive representation is the statement of the goals and interests of any
given group represented by the individual. Pitkin defines substantive representation
.. .as an acting for others, an activity on behalf of, in the interest of, as the agent of,
someone else (p. 113). This includes the constituency as a whole or sub-groups
divided by race, class, gender, or sexuality, to name a few. An example of the
substantive representation of women is health policy requiring health insurance
companies to cover mammograms. Another clear example is legislation defining
womens rights to breastfeed their children in public.
The causal link between descriptive and substantive representation is the most
commonly researched link amongst the four types of representation (Schwindt-Bayer
& Mishler. 2005). For the purposes of this research design, the focus remains
exclusively on the link between descriptive and substantive representation. Described
succinctly, descriptive representation is standing for women where substantive
representation is acting for women. Reingold (1992) asks the pertinent question,
Are these women in public office merely women who represent, or are they also
women who represent women? (p. 509). This is directly related to the purpose of
this study and the stated research question. The next section. Competing Arguments,
details the literature critiquing the study of descriptive and substantive representation.
14


Competing Arguments
Research concerning the link between descriptive and substantive
representation has been criticized on two main points. First, there is an assumption
that there is a direct and positive correlation between descriptive and substantive
representation. Consequently, it is assumed that an increase in descriptive
representation will result in a subsequent increase in substantive representation
(Beckwith & Cowell-Meyers, 2007; Reingold, 2008). This limiting assumption may
camouflage other dynamics at play, resulting in erroneous conclusions.
As many have noted, an increase in the number of women in office doesnt
always directly relate to an increase in the number of bills and acts related to
womens interest legislation (Childs & Krook, 2005; Mackay, 2004). Research
suggests that a smaller minority, or even a token few, may be more effective in
pursuing a pro-women agenda (Carrol, 1984, Chaney, 2006, Childs & Krook, 2005).
In fact, some wonder if, by increasing the number of women in office, women no
longer feel obligated to act for women. Or, alternatively, it is speculated that the
increase in representation may result in an altered environment less supportive of a
women-friendly agenda, or even an environment that motivates men and women to
actively work against womens interest legislation (Beckwith & Cowell-Meyers,
2007; Bratton, 2002; Childs & Krook, 2007). Lastly, the intense focus on the
relationship between descriptive and substantive representation often results in a
failure to investigate how the increase in diversity changes the behavior of all
15


representatives, not just women or other minority groups (Bratton, 2002; Cammisa &
Reingold, 2004). As Weldon (2002) notes, it is time for scholars to move beyond the
body count.
As declared by Weldon (2002) and reiterated by others, it is time for scholars
to move beyond the a simple model exploring the results of an increase in descriptive
representation (Childs & Krook, 2005; Lovenduski, 2005a). Cammisa and Reingold
(2004) state Few studies have considered the potential impact of contextual forces
other than critical mass on the behavior and attitudes of female state legislators" (p.
198). Cammisa and Reingold continue stating that scholars should expand the
literature by integrating the concept of the political opportunity structure such as
developed by McAdams (1982). as opposed to current scholarship that simply
incorporates political and environmental variables as controls.
This sentiment is echoed by Beckwith and Cowell-Meyers (2007). Beckwith
and Cowell-Meyers (2007) note that current research continues to produce conflicting
results regarding the relationship between descriptive and substantive representation
of women. Based on existing literature, the authors argue that there are significant
parliamentary factors and civil society contexts that shape substantive representation
in addition to, and including, descriptive representation. As stated by the authors,
What numbers, frameworks, conditions, and contexts govern the ability of women
legislators to make a difference? (p. 554). This question is important because it
16


moves beyond descriptive representation to include the political context of the policy
process.
The second critique regards the definition and operationalization of
substantive representation. Recall that substantive representation is defined as acting
for the group as a representative. Within the literature regarding gender, this is often
synonymous with feminism and the womens movement. However, as many point
out, women as a group are not homogeneous (Bratton, 2002), and the study of
substantive representation must be conducted carefully to avoid essentialism.
Essentialism is eloquently defined by Mansbridge (2003):
Essentialism includes assuming a single or essential trait, or nature, that binds
every member of a descriptive group together, giving them common interests
that, in the most extreme versions of the idea, transcends the interests that
divide them. Such assumptions lead not only to refusing to recognize major
lines of cleavage in a group, but also to assimilating minority or subordinate
interests in those of the dominant group without ever recognizing their
existence, (p. 637)
As such, it is possible to understand how the representation of some groups of women
may come with negative tradeoffs for other groups of women (Dovi, 2007; Young,
1999), or even how womens interest legislation may continue to bestow benefits on
sub-groups of women at the expense of others (Dovi, 2007). One should not
erroneously assume that all women will uniformly advocate for a standardized slate
of womens interest legislation (Ferguson, 1984; Mansbridge, 2003). In fact, women
come to office with a wide variety of political goals and interests, not all of which
17


nicely dovetail with the typical notion of feminism and the womens rights
movement.
Despite this, research clearly demonstrates that women do have divergent
interests from men (Mansbridge, 1999; Phillips, 1995; Sapiro, 1981). Because
women have historically been excluded from the policy process, womens policy
priorities, largely ignored, are relatively new to the policy agenda (Swers. 2002b). As
stated by Mansbridge (2005), Descriptive representation by gender improves
substantive representation for women in every polity for which we have a measure
(p. 622). Consequently, it has been and continues to be relevant to examine how the
descriptive representation of women translates to substantive representation.
The pool of literature describing, analyzing, and advocating for the descriptive
and substantive representation of women is quite large and has certainly been
advanced since the first studies of its kind in the late 1960s (e.g. Werner, 1966.
1968). The arguments summarized in this section do not limit the utility of
descriptive and substantive representation. In fact, these arguments introduce cogent
points that should be integrated into the literature. Additional research is necessary to
explore the relationship between descriptive and substantive representation,
integrating specific legislative factors and societal contexts that may result in the
substantive representation. The following section, Conceptual Framework, reviews
the literature as it relates to an adaption of Beckwith and Cowell-Meyers (2007)
model, briefly reviewed in the first chapter.
18


Conceptual Framework
The Conceptual Framework section explores the model used in this research
design. The first section explores the original model and the authors' arguments for
expanding research on descriptive and substantive representation. The following
three sections, Representation of Women, Legislative Factors and Societal Contexts,
will give the reader an overview of the research exploring the constructs, including
use of the constructs as dependent and independent variables, common definitions
and operationalization of these variables, and the related hypotheses. The section will
conclude with a brief synthesis of the literature and an introduction to the final section
of the literature review.
The Model
In 2007, Beckwith and Cowell-Meyers published a persuasive article
critiquing the limitations of current research. Similar to the arguments posed above,
the scholars were concerned with the simplistic focus on descriptive representation.
The authors argued that the political context is the necessary condition for
translating sheer numbers of women into women-friendly public policy" (p. 557).
The authors then divided the political context into parliamentary and civil society
contexts. The parliamentary context is concerned with party control of government.
Beckwith and Cowell-Meyers argue that liberal or leftwing political agendas typically
coincide with advancing womens issues. Civil society context encompasses the
19


strength of the womens movement and public opinion concerning womens rights.
Figure 2.1 offers a depiction of Beckwith and Cowell-Meyers model.
Descriptive Representation
Parliamentary Factors
Left-wing majority -----------
Strength of opposition
Civil Society Contexts
Active feminist movement
Weak or no opposing movement
Support in public opinion
Women Friendly Public Policy
Figure 2.1. BECKWITH AND COWELL-MEYERS THEORETICAL MODEL.
Beckwith and Cowell-Meyers model is designed to be comparative in nature,
thus explanatory variables included within their model are appropriate for exploring
representation across many countries. However, this does not eliminate the utility of
the model within the United States. In fact, Beckwith and Cowell-Meyers model
provides a sound foundation in which to develop a similar model based on US state
legislatures. The revised model will be a unique contribution to the literature
focusing on female political representation within US state legislatures. The
following three sections provide further review of Beckwith and Cowell-Meyers
model offering the state of current research and application of said model within
study of US state policymaking.
20


The Representation of Women
Substantive Representation
Recall that substantive representation is the statement of the goals and
interests of a particular group, in this case women. In this study substantive
representation is the dependent variable, and is defined as acting for women in
legislative bodies through the expression of the goals and interests of women (Mazur,
2002). As mentioned earlier, within the academic literature the definition of
feminist or women-focused is a widely contested topic, and is often a point of
heated debate. This is evident when reviewing the range of definitions for feminist or
women-focused public policy (Beckwith & Cowell-Meyers, 2007; Mazur, 2002).
Common definitions include policies that mitigate or ameliorate gender-based
discrimination (Bratton, 2002); policies focused on women for either biological or
socially constructed reasons (Lovenduski, 2001); and, policies that target women
based on traditional roles as caregivers or other gendered divisions of labor (Swers,
2002a). Concrete examples of policies within the US that meet many of these
definitions include reproductive policy, policies concerning violence against women,
and family leave policies, to name a few.
Based on the wide range of definitions, use of the term feminist and women-
focused must be used carefully and deliberatively to avoid an essentialist view of
womens interests. Kathleen Bratton (2002) names this type of policy womens
interest legislation stating, womens interest legislation is defined as legislation that
21


would decrease discrimination or counter the effects of discrimination or would
improve the social, economic, or political status of women (p. 135). This iteration
offers a definition specifically focused on gender equality while broad enough to
encompass the varying needs of an intersectional group of women as found in the US.
Due to this, substantive representation will be used synonymously with womens
interest legislation, as defined by Bratton.
This definition of substantive representation was selected because it is found
to be comprehensive and non-essentializing, thus representative of a diverse group of
w omen with a varied array of policy interests. At this point, as previously discussed,
it is necessary to note that women as a group are not homogenous, but, in fact, have
varying interests and priorities related to socio-demographic characteristics other than
gender. This varies according to race and ethnicity, sexuality, political and spiritual
affiliation, and economic status, to name a few. However, it is also commonly agreed
upon that, for the purpose of analysis, women as a group have a set of priorities that
are different from those of men or other groups based on socio-demographic
characteristics (Schwindt-Bayer & Mishler, 2005). These priorities have often been
neglected and ignored, and the US lags behind most other industrialized nations in
providing policy benefits that work to eliminate discrimination or advance womens
policy priorities.
Bratton (2002) divides substantive representation into two sub-constructs
related to the legislative process: policy priorities and policy preferences. This
22


typology is useful and will be utilized in this study. Policy priorities, also referred to
as agenda setting, typically includes actions such as floor remarks (Shogan, 2001;
Walsh, 2002), bill sponsorship (Reingold, 2000; Swers, 2002a; Thomas, 1991; Vega
& Firestone, 1995), and committee membership (Swers. 2000; Tamerius, 1995).
Policy priorities. Bill sponsorship is the first example of the expression of
policy priorities. Wolbrecht (2002) states that bill sponsorship provides a useful
indication of members interest in, commitment to, and preferences over womens
rights... members (co)sponsor legislation with policy outcomes they prefer to the
status quo (p. 177). Bratton and Haynie's (1999) study of women s bill sponsorship
offers an excellent example of analysis of female representatives and the policy
process. The authors conducted a quantitative, longitudinal analysis of agenda
setting, operationalized as bill sponsorship. This was done for three periods in time:
1969, 1979, and 1989 for five different state legislatures chosen for their range in
ideology and levels of representation of both women and racial minorities. Using
negative binomial regression analysis the authors determined that both race and
gender have a distinct and significant impact on policy priorities, p < .01.
Action within committees also falls within the sub-construct policy priorities.
Scholars have determined that womens representation on committees has an impact
on the functioning of the committee as well as the outputs of the committee
(Rosenthal, 1997; Swers. 2002b). For example, it has been determined that women
lead committees with a leadership style focused on consensus and participation,
23


whereas men lead using a command and direct style (Jewell & Whicker, 1994). In
addition, Kathlene (1994) determined that men and women use different
communication patterns in committees. Specifically, men use aggressive
communication through higher rates of interruption along with talking for longer
periods of time. Most importantly, it has been determined that the presence of
women, or lack thereof, impacts passage of legislation of direct concern to women
(Berkman & O'Connor, 1993).
Lastly, floor debates are also a useful operationalization for policy priorities.
In an analysis of transcribed floor debates in the House of Representatives, Walsh
(2002) used analysis of variance to determine that w omen are significantly different
in the amount of time they verbally represented women and other underrepresented
constituencies, such as children or people living with HIV and AIDS, in the 104th
Congress. In addition, Sw'ers (2002) analysis of floor remarks in the 103rd and 104th
Congress found that Democratic women led the w'ay in representing women.
Bill sponsorship, committee representation, and floor remarks, actively help
set the policy agenda, thus substantively representing women, but each action has
strengths and weaknesses. As stated by Swers (2002)
. .depending on legislators goals, their bill sponsorship patterns will reflect
varying levels of commitment toward womens issues. An analysis of bill
sponsorship is therefore a good first step toward determining whieh members
are working to bring womens issues to the national agenda (p. 34).
24


As a result, among the potential measures of policy priorities, bill sponsorship is the
most appropriate operationalization of policy priorities, and the first dependent
variable discussed within this research design.
Policy preferences. For the purposes of this research design, policy
preferences refers to roll-call voting, which, given the accessibility of data, is the
most commonly researched behavior in the literature (Reingold, 2008). A great deal
of literature exists concerning roll-call votes, and researchers have determined that
there is a significant difference between men and women and voting behavior
(Barrett, 1995; Day, 1994; Diamond, 1977; Dodson & Carroll, 1991; Dolan, 1998;
Epstein, Niemi, & Powell, 2005; Swers, 2002a).
For example, Dolan (1998) conducted an Ordinary Least Squares regression
on roll-call data at the federal level for the 103rd Congress and the 48 female elected
officials. Dolan found that women are more likely to vote yes on womens issues
than men. Dolans results are typical for other analyses that focus on representation
using roll-call data, though most studies typically focus on one policy area such as
reproductive policy, as opposed to a spectrum of womens issues (e.g. Norton, 1999).
While roll-call voting is a common and appropriate operationalization of
policy preferences, there are limitations to its utility. It has been determined that roll-
call voting alone is a weak operationalization of substantive representation of women
because it occurs after the policy agenda has been set (Hero & Tolbert, 1995;
Lovenduski, 1998). Also, roll-call votes limit the representative to voicing preference
25


on a predetermined set of choices with a simple yea or nay (Bratton & Haynie, 1999).
Yet, when studied in conjunction with other expressions of substantive representation
throughout the policy process, such as bill sponsorship, one may develop a richer
understanding of how descriptive representation, legislative factors, and societal
contexts translate to substantive representation throughout the policy process.
Michele Swers (2002a) analyzed both bill sponsorship and roll-call voting
behavior in the 103rd and 104lh House of Representatives. In her analysis of bill
sponsorship, Swers determined that women were more likely than men to sponsor
bills related to womens issues, and at times this likelihood exceeded 50 percent. In
addition, Swers also determined that women were more likely to vote for womens
interest legislation. Specifically, Swers found that while Democratic men and
Democratic women were as likely to cast a progressive vote on womens interest
legislation, Republican women do defect from conservative ideology to vote liberally
on womens interest legislation. However, this voting behavior ebbed once the
Republican Party gained the majority. Swers theorized that the decrease in votes
resulted from an increase of party pressure to toe the party line and maintain a
majority on roll call votes.
Swers (2002) analysis offers the most current and comprehensive view of the
impact of descriptive representation on policy priorities and policy preferences.
Despite this, Swers does not analyze the impact of descriptive representation on
policy outputs, which, arguably, is the goal of both the womens movement and of
26


scholars analyzing representation of women. In addition, Swers' analysis, like
Dolan's, is limited to the federal level. While fruitful, a similar analysis at the state
level would be beneficial and may have the potential to reveal new and exciting
results, due to the fact that women have consistently had greater representation at the
state level (Thomas, 1994). The next section. Policy Outputs, discusses the literature
focusing on women's interest legislation emerging from one or both legislative
chambers and/or becoming law.
Policy outputs. Policy outputs refers to the number of women-interest bills
passed by a chamber or enacted into law. Research at the aggregate level exploring
the relationship between descriptive representation and aggregate policy outcomes is
not as commonly studied as micro-level analysis focusing on the individual (Bratton,
2005; Cammisa & Reingold, 2004; Reingold, 2008). In addition, quantitative and
qualitative research produces mixed results concerning the relationship between
descriptive representation of women and aggregate policy outcomes (Reingold,
2008). Quantitative research designs often provide statistically significant evidence
of support for the relationship between descriptive representation and policy outputs
(Crowley, 2004; Reingold & Schneider, 2001); yet, results are not always conclusive
(Thomas, 1994; Tolbert & Steuemagel. 2001). In contrast, qualitative research
consistently provides evidence that there is a relationship between descriptive
representation of women and policy outputs (Dodson. 2006; Hawkesworth, Casey,
27


Jenkins, & Kleeman, 2001). Due to these conflicting results, further analysis is
necessary. Reingold (2008) states:
Clearly, more research on aggregate policy outcomes is needed not only
because of the mixed results produced thus far but also because this type of
research has been neglected for far too long. We cannot simply assume that
being different and acting differently assures that women w ill, in fact, make a
difference, (p. 132)
Thus, consideration of policy outputs is of benefit to the field and would also fill a
gap in the literature.
In conclusion, this research design will encompass three dependent variables:
policy priorities, policy preferences, and policy outputs. These variables will be used
to measure substantive representation. The selection of these three variables results
in a comprehensive analysis meant to encompass key stages within the policy
process. The relevance of analyzing substantive representation across different stages
of the policy process will be reviewed in the next section.
The Policy Process
Harold Lasswell (1956) introduced the concept of the stages of the policy
process as a tool to understand interactions between actors, forces, and institutions, as
well as a tool to bring science to the study of public policy in the "real world (Jann
& Wegrich, 2007). Many scholars contributed to this stream of research with their
own version of the policy process, such as Anderson (1975), Jenkins (1978), May and
Wildavsky (1979), and Brewer and deLeon (1983). Also referred to as the stages
heuristic, it was the main focus of the policy sciences for many decades, into the late
28


1980s (deLeon, 1999). Currently, it is the most widely applied framework of the
policy process (Jann & Wegrich, 2007).
An empirical and theoretical focus on the stages heuristic is a means for
categorizing policy actions as they vary from stage to stage (deLeon, 1999, p. 26).
An advantage of using the stages heuristic is that it provides an intuitive and practical
means in which to study policy by categorizing policy actions from stage to stage. In
addition, the stages heuristic serves as a means for breaking down a complicated
cycle into more easily synthesized pieces.
Critics of the stages heuristic claim that use of the model results in research
focusing on one stage of the process, painting a picture that fails to tell the entire
story. In addition, amongst other critiques, the stages heuristic is lambasted for its
lack of predictive capabilities (Sabatier, 1999). However, as proponents counter, the
stages heuristic was never meant to be a predictive theory, but .. .a device to help
disaggregate an otherwise seamless web of public policy transactions... (deLeon,
1999, p. 24). The utility of the stages heuristic is described by deLeon as ....a basis
for viewing and categorizing actors and actions in ways that help unravel and
elucidate given policies... (p. 26), where each stage is distinguished by
differentiated actions and purposes (p. 24).
Brewer and deLeons (1983) stages include Initiation, Estimation, Selection,
Implementation, Evaluation, and Termination. The use of the stages heuristic as an
operational tool, categorizing different actions throughout the policy process, is useful
29


within this study. Specifically, this study focuses on substantive representation of
women throughout the policy process. As detailed above, policy priorities, policy
preferences, and policy outputs are all ways in which women can be substantively
represented. They also encompass specific stages of the policy process.
Policy preferences fall within the Initiation or agenda-setting stage of the
policy process. Expression of policy priorities occurs during the Estimation stage
where representatives values are on show through hearings, debates, and roll-call
voting in committees and on the floor. Policy outputs are encompassed by the
Selection stage of the policy process. Selection is the legitimization of a policy
through the formal process. It is important to look at substantive representation
across different stages of the policy process because the elements affecting
representation may vary from stage to stage (Soule & King, 2006).
It is theorized that support for a policy, in this case substantive representation
of women, carries fewer consequences with less stringent rules earlier in the policy
process (King, Cornwall, & Dahlin, 2005). In fact, it has been found that women are
more often substantively represented at earlier stages of the policy process such as
Initiation where problems are defined, the agenda is set, and bills are sponsored
(Bratton, 2002; Tamerius, 1995). However, representation also occurs during
Estimation and Selection, where policy preferences and policy outputs are expressed,
respectively. As the process moves beyond Initiation and bill sponsorship the debate
intensifies, the public may become aware of the proposal, and individual
30


accountability increases (Soule & King, 2006). This indicates that the substantive
representation of women at later stages of the policy process will require more
resources and political support. By designating activities related to the substantive
representation of women by stage, one can demonstrate at what stages of the process
women are substantively represented and where and when each independent variable
is most influential.
The next section, Descriptive Representation, moves into a discussion of the
first of five independent variables. It is followed with a discussion of legislative
factors and societal contexts.
Descriptive Representation
Descriptive representation is the primary independent variable of interest in
this study. Descriptive representation is based upon Pitkins (1967) political theory
where it is defined as standing for (p. 60). Application and analysis of descriptive
representation often focuses on the concept of critical mass (Beckwith & Cowell-
Meyers, 2007; Chaney, 2006: Childs & Krook, 2005; Lovenduski, 2001). The idea of
critical mass is based upon the work of Rosabeth Moss Kanter (1977a, 1977b). where
critical mass is also often referred to as the tipping point (Thomas, 1994; Weldon,
2002).
The portion of Ranters work that has translated to the study of descriptive
and substantive representation of women focuses on the power of a minority group to
create change. Kanter defined groups based on the percentage of the total. For
31


example, a uniform group is composed of an entirely homogenous population. In the
instance where the population isnt homogeneous, but, in fact, contains a small
minority group, the group is considered skewed. Skewed organizations have a
minority population that is less than 15% of the total. In a skewed group the
dominant population will control the dynamics, culture, and process of the group.
The tokens, or the underrepresented population, will have difficulty creating
alliances and advancing their interests (Kanter, 1977b, p. 966). In addition, a skewed
group runs the chance of tokenizing one individual as symbolic and representative of
the entire group, as opposed to their status as one individual member of the group.
As the group proportions move toward equal distribution, it becomes tilted.
At this point the minority population is 15% to 35% of the total. The group is
considered balanced when the ratio of majority to minority nears 60:40 or 50:50. At
this point the minority and majority may actually begin grouping into other subgroups
that are not based on these types. As a result, subgroups would no longer be
exclusive to gender but based on some other defining characteristic.
After defining different types of groups based on proportion of minority
populations, Kanter focused on the role of women in skewed organizations where the
percentage of the total limits their ability to perform as anything but a token for the
entire minority group. At the time of publication, Kanter was upfront about the
theoretical nature of the numbers developed to indicate the type of organization -
uniform, tilted, skewed or balanced. Kanter concluded stating that empirical,
32


quantitative analysis was needed in order to determine the exact point at which
groups moved from uniform, to skewed, to tilted, and then to balanced.
Dahlerup (1988) was the first to attempt a quantitative threshold by
researching women in Scandinavian politics. Dahlerup (1988) found that as the
numbers of women increased, the substantive representation of women increased, but
the author also discovered that some women were more effective because of their
token status. Token status is typically defined as 15% or less of the total group
(Kanter, 1977b). It was in this piece that Dahlerup established the threshold of 30%
that has since been claimed by those studying descriptive and substantive
representation as the critical mass necessary for women to coalesce as a group and
enact change.
The connection between a critical mass of women and substantive
representation has since provided mixed statistical results with few proving a direct
causal relationship between a critical mass and substantive representation (Bratton,
2002; Kathlene, 1994; Saint-Germain, 1999; Thomas, 1991, 1994). For example,
Thomass (1991) study of womens bill sponsorship focused on the concept of critical
mass. Thomas primary independent variable was the number of women in both
chambers of state legislatures, which was then divided, according to critical mass
theory, into three groups: less than 10%, 11-20%, and 20% and higher. Twelve states
were selected based on their range of female representation and political culture. The
author used a quantitative survey to collect information from state legislators during
33


the year 1988, and found that women are more likely than men to introduce bills
related to women and children, but Thomas also determined that a critical mass of
30% 35% was often not sufficient enough to impact the policy process. In fact,
Thomas suggested that state legislatures must achieve a balanced distribution by
gender in order to find a significant difference.
Bratton (2002) conducted a similar quantitative study. Brattons longitudinal
analysis was conducted in six states for three different periods in time, 1969, 1979,
and 1989. Bratton found, similar to Thomas (1991, 1994), that women and blacks
sponsor bills related to women and racial minorities in greater numbers than men or
white representatives. In contrast to Thomas, Bratton determined that as time elapsed
and the percentage of women and blacks increased, the overall sponsorship of bills
did not increase, and in fact, sponsorship decreased in some scenarios. Bratton
argued that as minority groups increased, the expression of a groups interests and
goals did not increase in a linear fashion or according to the theory posited by critical
mass theorists.
As demonstrated above, since Dahlerups piece was published, a number of
publications have investigated the merit of critical mass. It has since become evident
that the 30% threshold was arbitrarily adopted and then implemented throughout the
fields of political science and policy analysis. In addition, politicians and
practitioners also adopted 30% as the magic number for representative groups. This
occurred despite the thresholds established by Kanter or the admonitions of Dahlerup
34


herself, who actually theorized that critical acts were more important than the critical
mass (Grey, 2006). Grey states:
...given there is nothing in Kantef s work to indicate that a single figure is all
important, we need to move to the idea that different critical masses may be
needed, depending on the outcome sought. Gaining 15% of the seats in a
political body may allow female politicians to change the political agenda, but
it may take proportions of 40% to have women-friendly policies introduced
(p.494).
Debate within the literature reveals that the concept of critical mass is highly
contested, and scholars vary on their opinions on how to measure descriptive
representation. This is especially relevant within the US where representation of
women at the state or federal level currently averages 24% and 16%, respectively
(CAWP, 2011). It is difficult to test varying levels of critical mass when the
percentage of women in elected office does not near a balanced ratio. Alternatively,
it would be more fruitful to move away from the concept of critical mass and explore
descriptive representation without grouping data into pre-defined groups. This allows
for a focus on the relationship between descriptive and substantive representation at
different stages of the policy process. As a result, the first hypothesis is as follows:
Hi: State legislatures that have more female legislators will have higher rates
of substantive representation of women.
This hypothesis contributes to the literature by opting to forego a measure of
descriptive representation in terms of critical mass. It also contributes to the literature
by testing the hypotheses with a much larger sample of state legislatures at different
stages of the policy process. This may illustrate the need for different levels of
35


descriptive representation at different stages of the process. Lastly, this hypothesis
represents a shift from the individual to the aggregate, to determine if descriptive
representation at the macro level is related to the substantive representation of
women.
The next two sections move beyond descriptive and substantive
representation, to discuss the political contexts that may, either individually or in
combination, result in substantive representation of women. Each section will begin
with a brief discussion of Beckwith & Cowell-Meyers model, followed by a review
of complementary literature based on federal and state level research in the United
States.
Legislative Factors
This section focuses on the legislative factors that may be associated with
substantive representation of women. Beckwith and Cowell-Meyers (2007) state that
there is clear agreement that a leftwing political party is imperative to womens
interest legislation, as is the political position of the leftwing party. The authors
discussion of parliamentary and civil society contexts is limited to variables relevant
in comparative research. As a result, variables used by scholars studying the policy
process within the US justify the analysis of additional or alternate variables that
measure parliamentary and civil society contexts (Gerrity, et al.. 2007; Gottfried &
Reese. 2003; Weldon, 2002). In fact, a number of prominent researchers have also
called for the inclusion of variables related to institutions when studying descriptive
36


and substantive representation (Cammisa & Reingold, 2004; Swers, 2000). A focus
on institutional variables will be termed legislative factors.
Other legislative factors that appear within the literature regarding the US and
state policymaking include a left-wing incumbency, party ideology, electoral systems,
system vetoes, existence of a womens caucus, and the larger policy agenda of the
majority party (Beckwith & Cowell-Meyers, 2007; Mazur, 2002; Swers, 2001;
Weldon, 2002). The next two sections discuss the variables posed by Beckwith and
Cowell-Meyers model as well as other variables that are relevant based on a
thorough review of the literature.
Political Party Influence
Beckwith and Cowell-Meyers (2007) assert that political ideology of the
majority party is instrumental to the substantive representation of women.
Specifically, the authors argue that the left-wing oriented political parties are often
aligned with the goals and interests of women. As stated by Beckwith and Cowell-
Meyers:
The presence of a leftwing party in parliament may not be a sufficient
condition for advancing womens issues but, in general, leftwing parties have
a better record than rightwing parties of initiation and supporting legislation
liberalizing divorce, extending abortion rights, criminalizing violence against
women, expanding employment opportunities, providing womens healthcare
innovations, and advancing social welfare issues, (p. 557)
37


Thus, the authors theorize that a left-wing majority party offers a suitable context for
the substantive representation of women. In order to apply this theory within the US,
one must look at influence within a two-party system.
The idea that political parties are integral to democratic processes has been in
existence since the first half of the 20th century (Aldrich, 1995; Schattschneider,
1942). Since then, party ideology has become a classic variable within public policy
literature (Gabel & Huber, 2000), and it has been found to be a strong predictor of
legislative behavior (Wiggins, Hamm, & Bell, 1992).
A collection of literature within the rational choice arena counters this
argument (Downs, 1957; Mayhew. 1974). The rational choice philosophy argues that
if a majority of the constituency favors a policy, members of both parties will act in
favor of said policy in an effort to seek reelection (Burstein & Linton, 2002). Others
contend that political parties play second fiddle next to a representatives individual
policy preferences (Krehbiel, 1993).
Conversely, Wright and Schaffner (2002, p. 368) state that bundles of issues
(p. 368), such as womens interest legislation, are aligned with specific political
parties for purposes of political gain. The authors describe how the Democratic Party
came to be aligned with civil rights when the New Deal mobilized working class
ethnics. As a counter political move, the Republican Party then appealed to
disenchanted white Southerners that were against the newly adopted civil rights
stance of the Democratic Party. This trend was also observed by Wolbrecht (2002)
38


who stated that in the 1950s and early 1960s the Republican Party was aligned with
the womens rights agenda.
Today, polls and research show that the womens rights movement is
currently aligned with the US Democratic Party. In fact, since 1968, women have
identified with the Democratic Party in greater numbers than men, and since 1982
these differences have been found to be statistically significant even when controlling
for age, race, and education. Lastly, women report a belief that the Democratic Party
has a greater concern for women-related issues (McGlen & O'Connor, 1998).
Research regarding the intersection of gender, representation and political
parties typically revolves around two competing theories. The first focuses on an
oppositional perspective where the substantive representation of women comes at a
cost to the political parties. Here women will act for women, regardless of their party
affiliation and even at the cost of their party. Within this framework, scholars treat
political parties as constraints, and typically include it as a control variable when
modeling substantive representation (Sanbonmatsu, 2008b).
The second viewpoint theorizes that gender and party are in concert, where
representatives act for women through the policy machinations of their party
(Sanbonmatsu, 2008b). This perspective is especially relevant given the
aforementioned alignment of the womens movement, womens rights, and the
Democratic Party. Because the conception of womens rights and the viewpoint of
the Democratic Party are harmonious, one would expect a positive relationship
39


between Democratic Party control and substantive representation. This research
design supports this perspective. The second hypothesis is as follows:
H2: State legislatures that have a greater Democratic majority will have higher
rates of substantive representation of women.
The inclusion of political party within the model offers the opportunity to
advance the field by investigating the relationship between political party control and
substantive representation (Sanbonmatsu, 2008b). The unit of analysis shifts from the
individual level to the macro level with a focus on not just descriptive representation,
but also the institutional and political forces at play within each state. Direct analysis
of the relationship between political party and substantive representation is not as
common within the literature and is a significant contribution to the literature
(Cammisa & Reingold, 2004; Sanbonmatsu, 2008b). The next section discusses the
second variable related to legislative factors.
Womens Caucus
A formal womens caucus is a body within a state legislature that meets
regularly to define policy goals and to create a political strategy in reaching those
goals. The first state level women's legislative caucus formed in Maryland in 1969.
Many early histories of state legislative caucuses claim that the impetus to
organization was a direct result of a male super-majority in state legislatures, as well
as discrimination and harassment many female legislators were experiencing in their
state house (Mahoney, 2011).
40


As stated by Thomas (1991) When a caucus bands together, the result is
political clout a weapon with the potential to overcome skewed groups (p. 973).
As a result, scholars exploring the impact of women in representative positions also
explore the impact of a womens caucus on the substantive representation of women
(Reingold, 1992; Thomas, 1991). The formation of a womens caucus is a clear
indication that the concerns and interests of women are distinct from those of men,
and that gender is a politically relevant category. The womens caucus is also another
tool in which to advance womens interest legislation (Mahoney, 2011). Carroll
(2002) states One of the most important intra-institutional factors [impacting
substantive representation] seems to be whether women legislators met together either
through a formal legislative caucus or as an informal group to discuss legislation that
affects women (p. 18).
Further, establishing a womens caucus serves to provide a means for
coordination of women across chambers, and across parties. As the number of
women grow in state legislatures it becomes increasingly important to organize and
act for women in an effective fashion. Among other things, a womens caucus serves
this purpose (Kanthak & Krause, 2010).
Thomas (1991), using a quantitative survey, looked at 12 different state
legislatures and found that the top five states passing womens interest legislation also
had a women's caucus. Thomas concluded that a womens caucus is a way for
women to achieve clout. In a methodologically different analysis. Reingold (1992)
41


used interviews with state legislators from both Arizona and California. The author's
interviews revealed that in the state with a formal womens caucus (California),
female legislators viewed the caucus enthusiastically and as a way to put forth
women-focused policy, as well as a means of voting as a unified group.
On the other hand, research exists that found the effect of a womens caucus
on substantive representation to be inconclusive or ineffectual (Berkman & O'Connor,
1993; Reingold & Schneider, 2001). For example, Reingold and Schneider (2001)
examined the influence of women within the context of abortion policy. In this study,
women were found to be more effective in either passing or blocking legislation in
their roles as committee members, rather than as a united womens caucus.
The small amount of contradictory research concerning substantive
representation and womens legislative caucuses is a call for more analysis.
Following the lead of Thomas (1991), this research design hypothesizes that the
existence of a formal womens caucus will advance substantive representation of
women. The third hypothesis states:
H3: State legislatures that have a womens caucus will have higher rates of
substantive representation of women.
Analysis of the impact of womens caucuses in state legislatures is quite
limited and few have studied the relationship between womens caucuses and
substantive representation (Mahoney, 2011). This hypothesis seeks to extend the
42


literature by exploring the impact of a womens caucus within state legislatures across
all 50 states for two time periods and for an additional two time periods for 25 states.
The literature also reveals several other legislative factors that may also
impact substantive representation. Consideration of institutional norms, positional
power in committees, and the degree of professionalism are just several recurring
factors mentioned in the literature (Cammisa & Reingold, 2004; Childs & Krook,
2009). Consideration of the Executive branch may also be relevant, such as the
influence of the bureaucracy and the governor on the policy process in state
governments (Crew, 1995). Despite this, research within the field has not commonly
incorporated these variables, and for the sake of creating a parsimonious research
design, they will not be included. The following section. Societal Contexts, discusses
variables outside of government institutions that may also impact substantive
representation of women.
Societal Contexts
Societal contexts also impact substantive representation of women. Weldon
(2002) states that it is problematic to view representation solely as a function of
individual policy preferences (p. 1153). Many policy and administration scholars
argue that policy is a result of the influence of interest groups and of coalitions of
governmental and nongovernmental parties (Sabatier, 1999; Sabatier & Jenkins-
Smith, 1988; Teske, 2004; J. Q. Wilson, 1989), as well as the culture of the state
(Snow, Rochford, Worden, & Benford, 1986). Therefore, it is also necessary to
43


consider extra-institutional sources of substantive representation (Cammisa &
Reingold, 2004; Weldon, 2002). In this section research and theory considering
societal contexts is discussed.
State Citizen Ideology
Beckwith and Cowell-Meyers (2007) state that if the general public holds
positive beliefs concerning womens issues this often results in the creation of a
constructive environment for the substantive representation of women. Public
opinion of womens rights has drastically changed over the past 40 years while
descriptive and substantive representation has also shifted. The relationship between
these constructs is worth exploring.
Leading policy scholars recognize that during the policy process the opinion
of both policy elites and the public is critical to the adoption of public policy
(Kingdon, 1995). Within the policy literature, a number of scholars have investigated
the relationship between public opinion and public policy (Erikson, 1976; Miller &
Stokes, 1963; Page & Shapiro, 1983; Page, Shapiro, & Dempsey, 1987), and most
agree that there is a direct causal link between public opinion and a shift in public
policy (Erikson, Wright, & Mclver, 1993; Hartley & Russett, 1992; Miller & Stokes,
1963; Monroe, 1998; Wlezien, 2004).
In addition, as stated by Page and Shapiro (1983), Responsiveness of
government policy to citizens preferences is a central concern in normative
democratic theory (p. 175). Democratic theory assumes that there is a direct
44


relationship between constituents and elected officials. Elected officials will act
according to the interests and demands of their constituents, and constituents endorse
or reject behavior through elections (Ansolabehere & Jones, 2010).
As a result, there is a general consensus concerning the impact of public
opinion on public policy. The current gaps in the research concerning public opinion
focus on the extent to which other legislative factors and civil society contexts such as
interest groups, social movement organizations, political parties and political elites
impact public policy, in addition to public opinion (Burstein, 2003). If they do have
an impact, research also questions if the impact of public opinion is spurious in
relation to these other variables (Burstein, 1998; Erikson, 1976). Others note that a
model investigating the impact of interest organizations and political parties on public
policy must always include public opinion, due to the belief that it undoubtedly
increases the significance of the model (Burstein, 1998).
Public opinion is difficult to operationalize and measure, but with the advance
of public opinion polls, archival data, and specific statistical techniques, the ability to
capture a shift in public opinion, over time, has become possible (W. D. Berry,
Ringquist, Fording, & Hanson, 1998, 2007a; Lax & Phillips, 2009). In regard to
womens rights, Myra Ferree (1974) states that the question If your party nominated
a woman for President, would you vote for her if she seemed qualified for the job? is
a good measure of public opinion. Ferree makes the case that a yes or no response to
this question is not actually a measure of voter choice, but a measure of the
45


respondents prejudice. This measure has since been used to measure public opinion
concerning womens rights (Burstein, 1985; Costain & Majstorovic, 1994; Schreiber,
1978; Welch & Sigelman, 1982). For example. Costain and Majstorovic (1994) used
the question as a measure of public opinion in their analysis of womens rights
legislation, which focused on both social movements and public opinion.
Unfortunately, this data is limited to the national level, limiting its utility within this
study.
In addition to Ferrees measure, Brace, Arceneaux, Johnson, and Ulig (2004)
constructed a measure of gender attitudes. Using the General Social Survey, the
authors pooled data to construct a cross-sectional measure of state-level gender
attitudes. The measure utilized pooled data for the years 1974-1996, limited to states
in which there was an adequate sample, n=38. Unfortunately, the nature of this
measure does not capture temporal change, assuming public opinion remains stable
over time (W. D. Berry, et al., 2007a; Lax & Phillips, 2009). Due to these limitations,
this measure is also not appropriate for the proposed model, especially in
consideration of public opinion and womens rights, which has changed dramatically
over the past 40 years.
Ultimately, coordinated state-level polls necessary to create state samples
across time regarding women's rights and roles in society do not exist. Alternatively,
there has been a lively debate within the literature concerning the measurement of
public opinion at the state level (W. D. Berry, et al., 2007a; W. D. Berry, Ringquist,
46


Fording, & Hanson, 2007b; Erikson, Wright, & Mclver, 2007; Norrander, 2007). Of
the options available, researchers can pool disaggregated state samples, derived from
national polls, to create a cross-sectional measure (Arceneaux, 2001; Brace, et al.,
2004; Erikson, et al., 1993), or researchers can use a proxy measure of public mood,
or ideology, in order to operationalize public opinion (W. D. Berry, et al., 1998,
2007a; Norrander, 2007).
Berry, Ringquist, Fording and Hanson (1998) constructed just such a measure
of public ideology using interest group roll-call voting scores, congressional election
outcomes, partisan divisions of state legislatures, party of governor and a series of
assumptions concerning political elites. This measure of citizen ideology offers
longitudinal data available from 1960 to 2008 for all 50 states (Fording, 2009).
Scholars have determined that public ideology and public opinion concerning
gender attitudes are strongly and significantly correlated (Arceneaux, 2001; Thomas,
1994). Literature discussing gender and representation commonly acknowledges that
liberal political ideology is directly related to descriptive representation, or the
number of women holding elected office (Arceneaux, 2001; Scola, 2006). As a
result, most research focuses on the causal relationship between political ideology and
descriptive representation (Norrander & Wilcox, 1998; Sanbonmatsu, 2002, 2003;
Scola, 2006). Research incorporating political ideology as a measure of public
opinion typically restricts analysis to domain specific policies, such as abortion policy
(Arceneaux, 2001; Berkman & O'Connor, 1993). There remains a significant gap
47


within the research regarding citizen political ideology and the substantive
representation of women. Scholars have recognized the need to include contextual
factors such as public ideology when analyzing both descriptive and substantive
representation (Cammisa & Reingold, 2004). The fourth hypothesis states:
H4: States that have a more liberal citizen ideology will have higher rates of
substantive representation of women.
Inclusion of this variable within the research design fills a significant gap in
the research. Current analysis has demonstrated that there is a direct link between a
liberal citizen ideology and descriptive representation (Arceneaux, 2001; Scola,
2006). Existing research, outside the bounds of gender and representation, also
document a direct link between citizen ideology and policy outputs (Erikson, et ah,
1993). This relationship is not commonly tested when analyzing substantive
representation (Cammisa & Reingold, 2004; Childs & Krook, 2009). Thus, this
hypothesis seeks to address this gap in the literature by analyzing the relationship
between state citizen ideology and the substantive representation of women.
The Women 's Movement
Beckwith and Cowell-Meyers (2008) determined that the presence of a strong
feminist movement would result in an increase in substantive representation of
women. Through the work of activists, the feminist agenda is publicized in the media
and also aids in the development of an electoral mass in support of the agenda.
Literature exploring the significance of social movements in relation to public policy
48


extends beyond the literature on womens descriptive and substantive representation
(Burstein, 1999; Costain & McFarland, 1998; Giugni, 1998; Giugni, McAdam, &
Tilly, 1999). The following section is divided into two parts. The first explores the
theoretical foundation of the social movement literature, the second focuses
specifically on the womens movement.
Theoretical overview. The study of social movements is grounded in the
theoretical history of pluralism (McFarland, 1998). When discussing public policy,
pluralism is considered a matter of plural points of power within the political system.
Pluralism arose as a counter argument to C. Wright Mills (1956) power elite theory,
which stated that political power within the US is controlled by the top few hundred
political elites (McFarland, 1998). Dahl (1958; 1961) countered this argument stating
that power was dispersed within the American political institution amongst a variety
of interest groups.
Currently, the study of social movements is grounded in post-pluralism theory
(McFarland, 1998), where social movements are classically defined as a set of
opinions and beliefs in a population which represents preferences for changing some
elements of the social structure and/or reward distribution of a society* (McCarthy &
Zald, 1977, pp. 1217-1218), or as collective challenges based on common purposes
and social solidarities, in sustained interaction with elites, opponents, and authorities
(Tarrow, 1998, p. 4). While the policy sciences and political science tend to focus
49


analysis on the nature and power of vying interest groups, the majority of the
literature focusing on mobilization comes from the field of sociology.
Early theory and research regarding collective action within sociology focused
on social psychology of alienated groups that shared a subjective experience of
grievance and deprivation. These shared beliefs were considered a necessary
prerequisite to collective action (McCarthy & Zald, 1977). With the onset of the civil
rights and womens movements, social movement literature exploded onto both the
policy and sociology scene. The first generation of social movement literature
emerged in the early 1970s where the main approach to social movements was a
structural approach dubbed resource mobilization (Mueller, 1992).
Resource mobilization was grounded in political sociology, rational choice
theory, and the work of Mancur Olson (1965). Olsons logic of collective action, as
applied to social movements, argues that individuals weigh costs and benefits before
opting to participate in collective action. Because the outcome of a social movement
is a public good, few will be incentivized to participate. Instead individuals will
become free-riders, reaping the benefits that others sow (McCarthy & Zald, 1977).
This theoretical shift from a social psychological to political-economic model formed
the foundation for the resource mobilization perspective of social movements.
Resource mobilization redefined the central issue of social movement research,
focusing on social movement organizations along with their resources and tactics
(Walder, 2009).
50


Closely aligned to resource mobilization, and sometimes categorized as
component of resource mobilization, is the political opportunity structure (McAdam,
1982; Tarrow, 1988; Tilly, 1978). Where resource mobilization is primarily focused
on the social movement organization, the political opportunity structure model, or
political process model, focuses on the exogenous political and institutional factors
that reside outside the boundaries of the social movement (Meyer & Minkoff, 2004;
Stekelenburg & Klandermans, 2009). Eisinger (1973) first coined the term political
opportunity structure, stating that it was concerned with the ability to gain access to
power and manipulate the political system. A more current definition of the political
opportunity structure comes from Tarrow (1994), consistent but not necessarily
formal or permanent dimensions of the political environment that provide
incentives for people to undertake collective action by affecting their expectations for
success or failure (p. 85).
Since 1973, the political opportunity structure has become foundational in the
study of social movements, and has since been used to incorporate a wide array of
independent variables into research designs. While the strengths of the model lie in
the ability to analyze how the political environment may shape opportunities for
social movements, a weakness of the model includes the sheer breadth of its
definition (Tarrow, 1988). In fact, it has been noted that scholars have used the
model to throw everything but the kitchen sink into collective action research designs
(E. A. Anderson, 2005). Despite this, there is a general consensus that the political
51


opportunity structure is typically comprised of access to formal institutions, allies
within the political system, and power within the system (Tarrow, 1994).
This first generation of research and theory focused on the utilitarian and
instrumental model of the individual, often ignoring structural and cultural
components of social movements (Mueller, 1992). This included ignoring the
relationship between the existing social structure and politics (Walder. 2009). This
has also resulted in a tendency to focus on the who, why and when of social
movements as opposed to social movement success (McCammon, Campbell,
Granberg, & Mowery, 2001). Since the move to a political-economic model of
mobilization, many have attempted to reintegrate the social-psychological elements of
social movements back into research (Klandermas, 1982; Snow, et al., 1986; Walder,
2009).
The second generation of research and theory revisited the notion of
individual grievances, values, and ideology. This perspective, rooted in the social
constructionist community, included consideration of the sociopolitical culture,
identity, emotions, as well as consideration of embedded inequity throughout all
aspects of society (Meyer & Minkoff, 2004; Stekelenburg & Klandermans, 2009).
This perspective is distinct from the structural basis of previous research focusing on
resource mobilization and political opportunity structure, and is often referred to as
cultural framing (Benford & Snow, 1988; Snow, et al., 1986).
52


Cultural framing is defined as the mechanism through which individuals may
understand what happens around them, identify sources of their problems, and devise
methods for addressing their grievances (Noonan, 1995, p. 85). These society-
centered, contextual forces provide opportunity for social movements, outside of
political institutions (McCammon, et al., 2001). Here, the ability of a social
movement to create an enduring message relies upon their ability to draw on the
existing cultural stock for images of what is an injustice (Zald, 1996, p. 266).
Scholars study collective action framing in order to understand the ways in which
social movements mobilize individuals, with particular consideration to values,
meaning, identity and emotions. It also allows for the study of how these cultural and
contextual factors alter policy-making (McCammon, et al., 2001). This is distinct
from the political-sociology perspective that focuses on the rational actor and social
movement organizations.
The women's movement. Social movements have been critical to increasing
substantive representation of women in the US. As stated by Banaszak (2008):
In the United States, social movements represent a major means by which
unrepresented or underrepresented groups gain access to decision making or
achieve social change. Women are among the many groups that have stood
outside politics, needing social movements to acquire change, (p. 79)
Research on womens movements, which in the US is typically found to be
synonymous with the feminist movement, is often at the national level (Costain &
53


Majstorovic, 1994), but studies also exist in a comparative context as well as at the
state level (Meier & McFarlane, 1992; Weldon, 2002).
For example, Weldon (2002) focused her research on social movements
impact on substantive representation, specifically investigating hypotheses similar to
those posed by Beckwith and Cowell-Meyers (2007). The author did this in a cross-
national analysis of feminist activism that focused on violence against women. The
author looked at 36 democratic countries in 1994 in order to determine how feminist
activism resulted in substantive representation in the policy process through national
response. In her analysis, Weldon operationalized the womens movement using
organizational data to compile a count of organizations. The strength and autonomy
of these organizations were determined via content analysis. Weldon found that
womens movements are statistically more significant than the descriptive
representation of women in predicting the variance of substantive representation.
Therefore, Weldon called for future research focusing on multiple sources of
representation outside of the standard analysis of descriptive representation.
Weldons work is exemplary in that it provides evidence that substantive
representation occurs as a result of a variety of different factors outside of legislative
bodies.
Soule and King (2006) studied the affect of social movements on ratification
of the Equal Rights Amendment at three different stages of the policy process across
all fifty states. The authors operationalized social movements according to the
54


resource mobilization model with a measure of National Organization of Women
(NOW) chapters in all 50 states across eleven years (1972-1982). Specifically, the
authors used the WorldCat library database to count the number of newsletters
produced by state and local level NOW chapters across all eleven time periods,
arguing that the ability to publish a newsletter is a sign of organizational strength.
Logistic regression revealed that the strength of the social movement is significant
during the early stages of the policy process, primarily bill introduction. Soule and
Kings analysis is useful for this research design, because it analyzes the impact of
social movements on substantive representation across different stages of the policy
process.
As illustrated above, common ways to operationalize the womens movement
include a variety of methods based on the resource mobilization model of social
movements. For example, many choose to use membership data for social movement
organizations (Meier & McFarlane, 1992). However, this method has typically not
been an option when considering womens interest organizations, as many
researchers have noted that feminist organizations are notorious for choosing not to
collaborate with the academy on membership data (Bratton, 2002; Freeman, 1995).
Others support the use of organizational data, as opposed to raw numbers related to
membership (Barakso, 2004; Schulz, 2008; Soule & King, 2006). This, too, poses
problems in regard to the validity and reliability of data, due to the fact that the
women's movement is comprised of a huge array of organizations, both bureaucratic
55


and institutionalized as well as informal and grass roots. For example, in the 1980s,
post-ERA, many different types of womens organizations sprouted up, such as
community health clinics, rape crisis centers and domestic violence shelters. As such,
measuring a social movement based on membership of a single organization doesnt
fully capture the size and scope of the movement (Staggenborg & Taylor, 2005).
Additionally, during the 1960s and 1970s the feminist movement quickly
split into two factions. One faction was dubbed the womans reform movement and
the second the liberation movement. Where the former was concerned with equality
within the present social structure, the latter focused on radical social change and the
creation of a new paradigm (Ryan, 1989). NOW and many other advocacy
organizations typically fell into the womens reform camp, while womens liberation
groups were more loosely structured, consciousness raising organizations (Barakso,
2004). The liberation group could be further delineated into a number of groups such
as Marxist, socialist, radical and lesbian feminist groups. While these groups
eventually agreed on a common definition of feminism, focusing on equality for all,
the subsets of feminism persisted separate from the womens reform movement
(Ryan, 1989). Because of this, focusing on NOW, or any other single organization,
would result in an incomplete measure of the womens movement. It also would
result in an essentialist operationalization, favoring one group or type of women over
others.
56


Another common way to measure the movement includes feminist self-
identification through polling data. Many maintain that self-identification with a
group or specific ideology is a necessary prerequisite to behavior (Schnittker, Freese,
& Powell, 2003). However, a cursory review of the feminist movement and feminist
research highlights how disparate, and often contradictory, feminist ideology really is
(Lorber, 2001). Also, many women adopt political beliefs consistent with the
feminist agenda, but, due to a negative connotation of the term feminist,
vehemently deny any association as a feminist (Bolzendahl & Myers, 2004;
Schnittker, et al 2003). As a result, using public opinion data concerning feminist
ideology or self-identification is also not a suitable measure of the feminist
movement.
Analysis over time utilizing common measures of the womens movement
such as membership data, organizational counts, protest events, or public opinion
measures of ideology would not capture oscillation in movement activity. Some
claim that we currently exist in a post-feminist society (Hawkesworth, 1994), but
others argue this is simply because the methodology in which one typically measures
social movements has failed to capture relevant elements of the womens movement
(Staggenborg & Taylor, 2005). Due to this, it is necessary to move beyond a
traditional measure of social movements when operationalizing the US womens
movement.
57


Culture & Structure. The review of literature reveals that common measures
of the womens movement fail to capture the entire picture, especially when
measuring change over 40 years. This leaves us with a quandary. How do you
measure the women's movement across forty years of change? A return to the
cultural framing literature, as well as a review of the political opportunity structure
offers a good starting point. Doug McAdam (2000), a leading scholar in the field,
states, 'Social movements are embedded in the cultures within which they act and
should be understood in relation to the broader culture (p. 253). Even Wildavsky
(1994) notes that the concept of self is a socially constructed term that is a direct
product of our social and economic conditions, or, our culture. In fact, many agree
that that cultural change is consequential for individuals and for public policy
(Cuthers & Lockhart, 2000; Wildavsky, 1994).
Conway, Ahem and Steuemagel aptly state (2004), A cultural approach to
public policy emphasizes the impact of factors such as class lifestyle, religion, ethnic
identification and race on womens understanding of the political significance of their
gender, a perception frequently referred to as gender consciousness (p. 2-3). Like a
string of dominoes, cultural change then alters the demands citizens make of
government (Inglehart & Carballo, 1997; Inglehart & Flanagan, 1987; Taylor, 1989).
Thus, it is expected that various cultural changes across different states will result in
varying levels of substantive representation of women.
58


The emergence and growth of the womens movement is thought to be a
direct result of the broad political and economic changes of the late 19lh and 20lh
century. In regards to the womens movement, collective action is most often
attributed to the drastic changes in personal and public life (Chafetz & Dworkin,
1986). In analyzing these broad structural changes in the lives of US women, it is
clearly documented that specific socio-demographic variables impact opinions related
to gender equity (Bolzendahl & Myers, 2004; Reingold & Foust, 1998). As stated by
Banazsak and Plutzer (1993):
...feminist and antifeminist sentiments emerge in response to changes in the
social structure of society, in response to personal changes in work or family
life, and as a consequence of socioeconomic status... Individuals will tend to
adopt political attitudes that endorse or reinforce lifestyle arrangements in
which they find themselves, (p. 31-32).
McCammon et al (2001) studied these contextual factors at the state level in regards
to womens suffrage movements. The authors state
...shifts in political circumstances altered the political calculus on which
decision-makers based their actions, providing a political opportunity for
suffrage. On the other hand, changing gender relations also caused political
decision-makers to alter their views about the proper roles for women in
society, and these changing attitudes about gender- not changing attitudes
about the political viability of a particular stance on suffrage-provided a
gendered opportunity for suffrage success, (p. 51)
Here gender relations is defined as "the social organization of the relationship
between the sexes" (Scott, 1986, p. 1053).
59


Consideration of altered gender relations in the past forty years often include
change in the employment status of women, status as wife and mother, educational
achievement, family size, age, and religiosity, to name a few (Pelak, 1999; Rosenfeld
& Ward, 1991). In fact, McCammon et al (2001) found that a new woman index, a
combined measure of educated, professional women that are politically and civically
active, was statistically significant in predicting suffrage at the state level.
It is important to analyze this shift in womens lives because it altered and
elevated the demands women made of the government and of society regarding their
economic, political and social status (Banaszak, 1996). Despite this, there is a dearth
of empirical research linking gender relations, gender opportunity structures, and
policy outcomes (Banaszak, 1996; Rosenfeld & Ward, 1991; Scott, 1986). While
most research focuses on how various socio-cultural variables impact descriptive
representation of women (Schlozman, Bums, & Verba, 1994, 1999; Verba, Bums, &
Schlozman, 1997), very few have analyzed gender relations and gender opportunity
structures in relation to substantive representation (McCammon. et al., 2001). Yet,
there is agreement that these variables will impact substantive representation.
Gender Opportunity Structure. Women entered the labor force in staggering
numbers following World War II. As women moved into the workforce, their
opinions and world views expanded and they also gained leadership and managerial
skills, both of which have enhanced their ability to become politically active
(Schlozman, et al., 1999). It has been determined that participation in the workforce
60


leads to female political participation (Rosenbluth, Salmond, & Thies, 2006). In fact,
womens labor force participation is important to the representation of women
(Matland & Montgomery, 2003; Salmond, 2006). where an influx of women within
the workforce may demand policies reconciling labor within the private and public
sphere (Henderson & White, 2004). Lastly, Rule (1987) and Norris (1985) both
found that there was a correlation between labor force participation and
representation of women, where representation of women is descriptive. As a result,
it w'ould be of value to include labor force participation rates when considering
gender opportunity structure.
Education is paramount to women obtaining an equivalent social status to
men. Not only does education increase a womans skills and abilities related to the
workforce, but it also serves as a tool to increase knowledge and awareness through
exposure to ideas (Bolzendahl & Myers, 2004). Similar to labor force participation,
results show that educational achievement is directly related to gender consciousness
and gender relations (Bolzendahl & Myers, 2004; Plutzer, 1988; Reingold & Foust,
1998; Rhodebeck, 1996). Scholars have also determined that women's level of
educational achievement is important in predicting womens descriptive
representation (Krook, 2005; McDonagh, 2002; Moore & Shackman, 1996; Weldon,
2002). Again, for these reasons consideration of educational attainment is relevant to
this research design.
61


Lastly, the 1970s and 1980s saw the liberalization of divorce laws across
many states. With a dramatic increase in divorce rates, many women found
themselves adopting nontraditional roles in an effort to support themselves and their
dependents. This included increased participation in the workforce, and, with it,
awareness of issues such as childcare, equal pay, and sexual harassment, to name a
few (Bolzendahl & Myers, 2004). Research has shown that a womens marital status
impacts feminist attimdes and gender relations (Cook, 1989: Klein, 1984). Thus, it is
important to include consideration of divorce rates when operationalizing gender
opportunity structure.
Broad, social structural change was necessary to not only facilitate the
womens movement, but to also transform women into an important component of
the constituency (Costain, 1992; Klein, 1988; Minkoff, 1997). It was this shift in the
social structure that prompted mobilization and policy change. Changes in gender
relations create demand for favorable social policies that support the new' paradigm.
A change in the paradigm creates a new' social base from which the feminist
movement emerged (Rosenfeld & Ward, 1991). By defining the womens movement
as gender opportunity structure, one can assess how powerful structural change in
gender relations impacts substantive representation. By analyzing the gender
opportunity structure, this research design takes another step in addressing a gap in
the literature regarding state-level contextual forces and substantive representation
(Cammisa & Reingold, 2004). The fifth hypothesis is as follows:
62


H5: States that have a more liberal gender opportunity structure will have
higher rates of substantive representation of women.
In conclusion, the literature concerning civil society contexts commonly
includes public opinion, ideology, political culture, social movements, and interest
groups. For the purposes of this research design state citizen ideology has been found
to be a useful and reliable measure for public opinion. In order to measure dynamics
at play concerning the womens movement, gender relations, and the gender
opportunity structure, an index measure will be created measuring labor force
participation, educational attainment and divorce rates. This index measure will be
titled Gender Opportunity Structure.
Inclusion of this variable within the model seeks to address a gap in the
research regarding the relationship between the gender opportunity structure and
policy outcomes (Banaszak, 1996; Rosenfeld & Ward, 1991; Scott, 1986). Existing
research typically focuses on howr various socio-cultural variables impact descriptive
representation of women (Schlozman, et al., 1994, 1999; Verba, et al., 1997), but very
few have analyzed the gender opportunity structure and substantive representation of
women (McCammon, et al., 2001). In order to construct a parsimonious research
design and to also allow the literature to guide the use of specific variables, societal
contexts will be limited to these two variables. The next section provides a critical
analysis of the existing literature.
63


Critical Analysis
Research on women, representation and public policy covers a wide
range of topics at all levels of governance. The following section synthesizes the
major gaps in the research and the ways in which this research design addresses the
gaps. Specifically, discussion regarding the theoretical model, the unit of analysis,
and sampling is reviewed as contributions to the literature. The research question and
hypotheses supplement current research by investigating similar questions with new
data and research methods. As a result, the significance of this study is largely
empirical in nature.
Theoretical Model
This chapter began with an overview of the competing arguments regarding
the representation of women. Analysis of representation focuses on three competing
perspectives. Those arguing in favor of critical mass state that a specific percentage
of women must exist before substantive representation can occur. This level varies
across the literature and according to how substantive representation is defined and
measured. Others argue that as the levels of women increase, their effectiveness
decreases as they lose their token status. Alternatively, some contend that substantive
representation is the work of a policy entrepreneur, where substantive representation
of women will occur due to the efforts of one or a few individuals, regardless of the
level of descriptive representation. Finally, scholars have recently begun to call for a
more comprehensive assessment of substantive representation, signaling a departure
64


from a strict focus on descriptive representation to an analysis of legislative factors
and societal contexts. This line of research is relatively new and unexplored within
the research on substantive representation of women.
The literature review extensively reviewed Beckwith and Cowell-Meyers
(2007) theoretical model and arguments for expanding research beyond the body
count to the legislative and social contexts of the policy process. When analyzing
representation of women it is imperative that one consider the political and social
context where descriptive representation translates to substantive representation
(Cammisa & Reingold, 2004; Swers, 2000). As stated by Reingold (2008), "...we
could benefit from more historical and comparative approaches that might account for
the effects of changes in interest group and social movement formation, public-
opinion, voting behavior, and issue cycles (p. 145). This study addresses this issue
through the development of a model incorporating descriptive representation,
legislative factors, and societal contexts, moving beyond an analysis of how
individuals represent women.
Unit of Analysis
Early research on gender and public policy focused on demographic
characteristics of female representatives. Researchers asked questions relating to the
traits of women (Werner, 1966, 1968) and then moved on to compare men and
women in public office (Cammisa & Reingold, 2004; Diamond, 1977). Most of this
early work focused on the individual as the unit of analysis. This trend continued
65


through the 1980s and 1990"s as scholars amassed an impressive collection of
literature concerning women and individual behavior in elected office {Cammisa &
Reingold. 2004; Swers, 2002b). Yet, since the turn of the century, more and more
scholars are advocating for macro-level analysis, studying aggregate outcomes. In
their review article of gender specific state-level research Cammisa and Reingold
state (2004):
Much less work has made systematic state-level comparison of legislative
gender dynamics, and even less attention has been paid to the actual impact of
women on state legislative outcomes and processes...Moreover, the novelty of
women's entry into state legislatures has meant that questions of what they do
are more easily answered than questions of what impact they have and how
that impact varies across states, (p. 191)
This sentiment is echoed by others advocating for institutional level analysis (e.g.
Reingold, 2008).
Research at the aggregate level exploring the relationship between descriptive
representation and aggregate policy outcomes is not as commonly studied as micro-
level analysis focusing on the individual (Bratton, 2005; Cammisa & Reingold, 2004;
Reingold, 2008). This research design addresses this gap in the literature by studying
descriptive and substantive representation across the policy process in all 50 states for
two time periods and across 25 states for an additional two time periods. By shifting
the unit of analysis from the micro level to the macro level it is possible to investigate
how the variation in descriptive representation, legislative factors, and societal
contexts is associated with the substantive representation of women.
66


Sampling & Research Design
The representation of women within state legislatures has dramatically
changed since the 1970s. With the backdrop of the civil rights movement and the
second wave of feminism, women increased their representation in state legislatures
from less than 6% in 1973 to over 13% in 1983. By 1993 state legislatures, on
average, were composed of approximately 20% of women. In 2011 the tally is at
23.3% (CAWP, 201 la). Figure 2.2 shows the increase of descriptive representation
of women in the lower house of state legislatures and in Congress dating from 1975.
25.0%
20.0%
15.0%
10.0%
us OS rH ro US OS rH ro US OS rH ro us os rH
r*. 00 00 00 00 00 OS OS OS OS OS o O O o O rH
Os os os OS os OS os OS OS Os os OS OS o O o o o c
rH T rH rH Tt tH rH rH rH rH rH rH rH Federal
State
Figure 2.2. DESCRIPTIVE REPRESENTATION OF WOMEN AT THE STATE
AND FEDERAL LEVEL, 1975-2011.
Source (CAWP, 2011)
State legislatures offer researchers a unique look at women in elected office.
As compared to state-level executive positions or representation in Congress, levels
67


of representation in state legislatures are significantly higher. In addition, state
legislatures offer a relatively comparable institutional arrangement, yet the state
political and social context varies dramatically across states (Cammisa & Reingold,
2004). Lastly, issues of concern to women are primarily considered state issues, such
as reproductive justice, womens health, and laws related to violence against women.
This offers fruitful ground for cross-state analysis.
State-level research exploring gender and representation primarily relies on
data from small representative samples of states or single state case studies. While all
three have strengths and weaknesses, no one has yet published a large-N, cross-state,
comparative analysis of descriptive and substantive representation. This is due to the
fact that development of a dataset across states and over time is an arduous process
requiring a great deal of time and resources (Cammisa & Reingold, 2004). This
creates a large gap in the literature. In constructing this type of dataset, one can
explore the entire picture, not just pieces of the puzzle. This may well result in
exposing specific contexts in which substantive representation is most likely to occur.
Pursuing this research design necessitates the development of a new dataset.
This will include comprehensive coding of womens interest legislation at the state
level for four time periods: 2005, 1995, 1985 and 1975. For the earlier periods, data
will be collected for 25 states, and for the two most current time periods data will be
collected for all 50 states. A dataset such as this does not exist and offers the
opportunity to explore uncharted territory in regards to the proposed research
68


question and hypotheses, as well as a myriad of other research questions that have yet
to be answered. These unanswered questions are directly related to gender, policy
and representation and are a matter of concern for gender equality and democracy.
Summary
Beckwith and Cowell-Meyers (2007) model offers a new theoretical
framework in which to explore descriptive and substantive representation of women.
Beckwith and Cowell Meyers constructed a model to be comparative in nature.
Despite this, a thorough review of the literature proves that an adaptation of the
model within US state legislatures would fill a gap within the research exploring
representation of women.
A review of literature determined that an adaptation of the predictor variables
for the US states is necessary based upon the availability of data and specific factors
unique to state-level research. In regards to legislative factors, this includes a
womens caucus variable. Civil society contexts have also been modified to reflect
conditions within the US with the inclusion of state citizen ideology as a measure of
public opinion and the gender opportunity structure in lieu of a measure of the
strength of the womens movement.
Based on the literature within the US concerning the policy process, the
proposed research design varies from Beckwith and Cowell-Meyers (2007) design
concerning the dependent variables. While Beckwith and Cowell-Meyers limited
their dependent variable to outputs, or in this case women-friendly public policy, but
69


it has been documented within the literature that this, alone, is a weak
operationalization. By limiting public policy to outputs one does not capture the
complexity of the policy process nor does it offer a comprehensive measure of
legislators behavior. Thus, the design has been extended concerning the dependent
variable, substantive representation. This includes three measures of substantive
representation meant to capture representation at three different stages of the policy
process. Figure 2.3 offers a revised model.
Descriptive Representation
Civil Society Contexts
State Citizen Ideology
State Gender Structure
Figure 2.3. REVISED THEORETICAL MODEL.
The theoretical model, variables, unit of analysis, and sampling were guided
by the literature in an effort to complement current research and to fill the gaps in
existing literature. The next chapter, Methodology, explores this research design in
greater length detailing definition and operationalization of variables, sampling, data
collection, analysis, as well as threats to reliability and validity..
Legislative Factors
Political Party Influence
Womens Caucus
SUBSTANTIVE
REPRESENTATION:
Policy Priorities
70


CHAPTER THREE:
METHODOLOGY
Chapter Three offers an in-depth review of the research design. First, a
review of the research purpose, research question and hypotheses lays the foundation
for the data collection and methodology. Next, a detailed account of the
operationalization, measurement and data collection procedures is presented for
purposes of replicability. This is followed with a thorough review of the statistical
analysis, including an explanation of panel data analysis and diagnostic tests, as well
as a description of the specific statistical methods used to analyze the data within this
research design. The chapter concludes with a discussion of the threats to reliability
and validity.
Research Purpose, Question, and Hypotheses
A comprehensive review of the literature demonstrates that the nature of the
relationship between descriptive representation of women and substantive
representation of women has not yet been established (Beckwith & Cowell-Meyers,
2007). Scholars argue that it is time to move beyond study of the individual to the
aggregate, examining not just descriptive representation, but legislative factors, and
societal contexts, as well (Cammisa & Reingold, 2004). This research contributes to
the literature regarding gender and representation by expanding the scope of the
71


analysis to the macro level, which has not yet received much attention within the
literature. This macro level unit of analysis allows for the inclusion of variables
beyond descriptive representation, such as legislative factors and societal contexts.
These variables are common in policy studies and political science literature, but they
are not as common when examining descriptive and substantive representation of
women.
Lastly, this research design contributes to the literature through the creation of
a comprehensive dataset for all 50 states for two periods of time, and 25 states for an
additional two time periods. This far outstrips typical research designs examining
just one state, or a small representative sample of 6 to 12 states. As such, the purpose
of this study is to explore the relationship between descriptive representation,
legislative factors, and societal contexts and substantive representation of women.
The review of literature in the preceding chapter established the foundation for the
following research question:
What is the impact of descriptive representation of women, legislative factors,
and societal contexts on the substantive representation of women at three
different stages of the policy process, within the US states?
This research question, in conjunction with a comprehensive review of literature,
guided the development of the following hypotheses:
Hi: State legislatures that have more female legislators will have higher rates
of substantive representation of women.
H2: State legislatures that have a greater Democratic majority will have higher
rates of substantive representation of women.
72


I-I3: State legislatures that have a womens caucus will have higher rates of
substantive representation of women.
H4: States that have a more liberal citizen ideology will have higher rates of
substantive representation of women.
H5: States that have a more liberal gender opportunity structure will have
higher rates of substantive representation of women.
The next section discusses the relationship between the research question and
hypotheses and the research design.
This research design is rooted in Pitkins (1967) theory of representation and
is adapted from Beckwith and Cowell-Meyers model (2007), presented earlier in
Chapter Two. The design seeks to answer the aforementioned research question and
address the related hypotheses. Figure 3.1 provides an illustration of the
hypothesized relationship between the independent and dependent variables.
Descriptive Representation
Research Design
SUBSTANTIVE
REPRESENTATION:
Legislative Factors
Political Party Influence
Womens Caucus
Policy Priorities
Policy Preferences
Policy Outputs
Civil Society Contexts
Citizen Ideology
Gender Opportunity Structure
Figure 3.1. THEORETICAL MODEL.
73


Unit & Period of Analysis
The unit of analysis within this research design is the general session within
the lower chamber of the state legislature. The lower chamber of state legislatures
offers a unique opportunity to explore the substantive representation of women.
Women are more successful at securing seats in state houses as opposed to Congress.
Within state houses, the lower chamber historically has higher rates of female
representation, as opposed to the upper chamber (CAWP, 201 lc). Additionally,
choosing the state as the level of analysis is ideal because a large number of womens
interests fall under the purview of state policy such as welfare, violence against
women, womens health, and reproductive rights (Cowell-Meyers & Langbein,
2009).
Lastly, cross-state variation allows us to examine varying levels descriptive
representation, legislative factors and civil society contexts. Data were collected for
four time periods 1975, 1985, 1995, and 2005. These dates refer to the general
session within each state during that year. These years were selected to encompass
the vast changes made in social, cultural and political rights of women in the past 40
years. These specific years were also selected to coincide with the general session of
states that meet biannually1. Data were collected for the independent and dependent
1 Currently, all 50 states meet annually except for Montana, Nevada, North Dakota, and Texas. These
states meet every other year, in odd years. Oregon. Arkansas, Kentucky, New Hampshire, and
Washington were the last states to change to an annual meeting, which occurred in 2011. 2009, 2001,
1985, and 1981. respectively. Of these states, only one state met biannually in even years Kentucky
74


variables across all 50 states for two time periods, 1995 and 2005. Data were
collected for an additional two time periods for 25 states, 1985 and 1975. This
resulted in a panel dataset uniquely designed to address the research question and
hypotheses through quantitative analysis. The following section discusses the
variables presented in Figure 3.1.
Variables
Substantive Representation: The Dependent Variables
The dependent variables are related to the construct substantive representation
of w omen. Recall that substantive representation is the statement of the goals and
interests of a group represented by the individual (Pitkin, 1967). In regards to
women, this research design follows the definition established by Bratton (2002) who
defines substantive representation of women as "womens interest legislation.
As detailed in the review of the literature, substantive representation can
encompass a wide range of definitions and viewpoints. Bratton (2002) states that
...womens interest legislation is defined as legislation that would decrease
discrimination or counter the effects of discrimination or would improve the
social, economic, or political status of women. These generally involved three
overlapping categories: measures that addressed the health concerns of
women; measures that addressed the social, educational, and economic status
of women; and measures that addressed the political and personal freedom of
women, (p. 139)
These three categories narrow the scope of womens interest legislation, but still
provide a rather obtuse definition. In order to provide a replicable research design, a
(NCSL, 2010). Data were collected for Kentucky in 1994, instead of 1995, to accommodate the even-
year biennium.
75


range of topics related to womens health, womens social, education and economic
status, and womens political and personal freedom were selected based upon the
traditional policy concerns of a wide group of women as defined by leading
institutions2, as well as both academic and popular texts (Dodson & Carroll. 1991;
Friedan, 1963; Gelb & Palley, 1982, 1996; Klein, 1984; Mazur, 2002; Swers, 2002a).
The final coding framework, available in Appendix A, was meant to be a
comprehensive, but not all encompassing list of womens interest legislation. The
construct of substantive representation is separated into three separate sub-constructs:
policy priorities, policy preferences and policy outputs. The next three sections detail
the operationalization and measurement of womens interest legislation under these
three sub-constructs.
The policy priorities variable is operationalized as bill introductions. Bill
introductions are measured as a percent of the total bills introduced in the lower house
during the general legislative session related to womens interest legislation. A
percentage is used, as opposed to a raw number, to control for the variability in bills
introduced between state legislatures. The policy preferences variable is a measure of
roll-call votes. Within this research design, the policy preferences variable measures
the average aye vote for all women's interest legislation introduced in that session
that reached the final reading and vote. Finally, the policy outputs variable is a
: Institute for Women and Public Policy Research. Center for American Women and Politics, National
Organization for Women, Feminist Majority, Moms Rising. American Association of University
Women
76


measure of the womens interest legislation introduced during that session that passed
the lower house. Similar to policy priorities, policy outputs are measured as the
percent of total in order to control for the variability in total bills sponsored between
state legislatures. Table 3.1 offers a concise overview of the operationalization and
measurement of the dependent variables. Each variable is a continuous variable.
Table 3.1. Operationalization of dependent variables.
Dependent Variables Operationalization Measurement
Policy Priorities Bill Introductions in the lower chamber Number of womens interest legislation introduced during the regular session divided by the total number of bills introduced in the regular session
Policy Preferences Aye votes for all womens interest legislation voted on in the final reading Average aye vote at the final roll-call vote for all womens interest legislation receiving a vote
Policy Outputs Womens interest legislation passing the lower chamber Number of womens interest legislation passed by the lower chamber divided by the total number of womens interest legislation introduced during the regular session in the lower chamber
Independent Variables
The independent variables within this research design include descriptive
representation, political party influence, women's caucus, citizen ideology, and
77


gender opportunity structure. The first independent variable is descriptive
representation. Descriptive representation is operationalized as seats held by women
in the lower chamber of US state legislatures3. This variable is measured as a percent
of the total to account for the number of seats across state legislatures. A one unit
change in this variable represents a one percent increase or decrease in the share of
seats held by women in any given state legislative chamber.
Political party influence is the second independent variable. Political party
influence is operationalized as Democratic Party influence, and is measured as a
percent of seats held by Democrat representatives4. A one unit change in this variable
represents a one percent increase or decrease in the share of seats held by the
Democratic Party in the lower chamber of a state legislature.
The final legislative variable is womens caucus. Womens caucus is
operationalized as the presence of a formal, organized group of female
representatives. Womens caucus is a binary variable with a value of 1 = caucus, 0 =
no caucus.
Two variables are categorized as civil society contexts state citizen ideology
and gender opportunity structure. These variables are lagged for one year to establish
temporal precedence (Stimson, Mackuen, & Erikson, 1995). State citizen ideology is
3 In the case of Nebraska, which has a Unicameral, the number of seats held by women in the single
chamber were counted.
4 Nebraska has a nonpartisan Unicameral. In lieu of removing Nebraska from the analysis, the mean
value of the variable for the respective time period was used in the analysis.
78


measured with the citizen political ideology measure constructed by Berry, Fording,
Ringquist and Hanson (1998). This measure is constructed using interest group roll-
call voting scores, congressional election outcomes, partisan divisions of state
legislatures, party of the governor and a series of assumptions concerning political
elites. Here a one unit increase represents a more liberal citizen ideology.
The final independent variable, gender opportunity structure, is an index
variable constructed using three measures.3 5 The first is the labor force participation
rate of women (civilian non-institutional) ages 16 and older, measured as a percent of
total. The second is educational attainment of women between the ages of 25 and 64.
Educational attainment is measured as a percent of the population holding a
bachelors degree or higher. Divorce rates are the last measure. This measure is the
percent of women 16 and older that were divorced in the survey year.
These measures were combined to create a composite index measuring the
gender opportunity structure. A composite index is a set of indicators, not necessarily
a result of a specific factor, that determine the same outcome (DeVellis. 2003). A
3 The methodology proposed by Soule and King (2006), reviewed in Chapter Two, was the proposed
operationalization of social movement strength within this research design. The authors data
collection procedures utilized the WorldCat library database to generate a count of the number of
newsletters produced by NOW chapters across all 50 states for the years 1972-1982. However,
replication revealed that the results did not produce a valid and reliable measure. First,
operationalization is not valid for the 1995 and 2005 time periods, where published newsletters failed
to be the primary source of communication with NOW members, ceding to electronic, web-based
communication. Thus, a measure of chapters publishing newsletters is not an accurate reflection of the
number of NOW chapters in each state for 1995 and 2005. Second, the measure was not found to be
reliable. The Worldcat database searches thousands of library catalogs. Matching database results
with the results of Soule and King revealed inconsistent counts of newsletters. This search was done
utilizing Worldcat at three academic libraries: University of Colorado Denver, University of
Califomia-Irvine and California State University-Fullerton.
79


composite index has several benefits which include richer operationalization and
measurement of a given variable, a reduction in the risk of measurement error,
reduction in the risk of multicollinearity, a parsimonious research design, and more
efficient analysis and interpretation (Babbie, 2010; Scavo, 2008). There are a number
of common ways in which to develop a composite index. In the case of similar units
of measurement, often the values are simply summed. For instances where the units
of measurement are vastly different, the values are often standardized by converting
to z-scores. The values are then summed (Scavo, 2008).
I developed the composite index duplicating methods used by the Institute for
Women and Policy Research in their development of a womens political
participation composite index (Werschkul & Williams, 2004). The data were
standardized by dividing the observed value for each state, from the national average.
The values were then summed to create a score that represents a continuous variable.
A one unit increase represents a more liberal gender opportunity structure.
A number of tests were performed to ensure that the measure is both reliable
and valid. Internal consistency reliability assesses how well the three items measure a
single phenomenon. This is typically done using Cronbachs alpha. Cronbachs alpha
values for these three items was acceptable with a value of a = .78 (DeVellis, 2003).
Both internal and external validity are also important in order to ensure that
the composite measure behaves the way that the construct it purports to measure
should behave with regard to established measures of other constructs (DeVellis,
80


2003, p. 53). I assessed the validity of the index by examining the correlation
coefficients for the index and the individual measures. The average correlation
between the individual measures and the index was r = .85, indicating an internally
valid index.
Lastly, I assessed external validity by examining the correlation of descriptive
representation and the gender opportunity structure index. The expected relationship
is a strong, positive correlation. The correlation, r = .70, indicates an acceptable
degree of external validity. A thorough discussion of the studys limitations
including threats to reliability and validity are included at the end of this chapter.
In summary. Table 3.2 offers a list of the independent variables, their
operationalization and measurement. The independent variables are continuous
variables with the exception of the womens caucus variable. The womens caucus
variable is a binary variable. In the next section the data collection procedures are
detailed for each dependent and independent variable.
81


Table 3.2. Operationalization of independent variables.
Independent Variables Operationalization Measurement
Legislative Factors
Descriptive Representation Number of seats held by women in lower chamber Percent of the total
Political Party Influence Womens Caucus Seats held by Democratic Party Presence of a formal female group of state representatives Percent of total 1 = caucus, 0 = no caucus
Civil Society Contexts
Citizen Ideology State citizen political ideology (W. D. Berry, et al., 1998) Average ideology scores for major party candidates weighted for specific state districts
Gender Opportunity Structure Composite Index of Labor Force Participation, Educational Attainment & Divorce Rates Sum of the standardized measures
Data Collection
Dependent Variables
A number of research studies code womens interest legislation within state
legislatures (Bratton, 2005; Bratton & Haynie, 1999; Bratton, Haynie, & Reingold,
2006; Cammisa & Reingold, 2004; Reingold & Schneider, 2001; Thomas, 1991). yet
these studies dont typically detail their data collection procedures. Therefore, data
82


collection for this study used original methods that do not seek to duplicate existing
data collection procedures.
Data for the dependent variables were collected from four major sources -
LexisNexis State Capitol (LNSC). state legislative journals, state legislative websites,
and directly from state archives or state law libraries. Bills were coded based upon
substantive content for the legislative sessions of the lower chamber in the year 1975,
1985, 1995, and 2005. I restricted coding to legislation proposing new laws and did
not include resolutions, memorials or budget bills (Bratton, 2002). Bills focused on
women, but in direct contrast to social, economic, or political advancement of
women, were not included. An example of this is a bill that seeks to restrict access to
abortion or a bill limiting insurance coverage for breast cancer treatment.
For the 2005 and 1995 time periods, data were collected using LNSC. This
database is an invaluable resource for state legislative scholars, offering
comprehensive information regarding bills, laws, regulations, and legislative
members for all 50 states, dating back to 1991 (LexisNexis. 2011). LNSC offers a
key word search that examines bills originating from each chamber of the state
legislature. The womens interest legislation identified in Appendix A was adapted to
create nine comprehensive groups of search terms, with up to 15 terms per group.
Two states were randomly selected to test the queries. The search terms were refined
and updated after testing. The final list can be found in Appendix B. After refining
and updating the final key word search list, the data collection procedures were then
83


duplicated for the two test states and applied to the remaining 48 states. Using the key
word search, the search hits for each state were examined to ensure that the nature of
the bill matched the womens interest legislation topic. If the bill did pertain to
womens interest legislation, the date of introduction, the sponsor, the bill summary,
the last stage in the legislation process, and roll call votes, if applicable, were
recorded in a spreadsheet. This was done for all 50 states for two time periods.
Data for the dependent variables for the 1975 and 1985 time periods were
collected from state legislative journals. I developed these data collection procedures
by selecting two states at random and coding their content using information available
within the journal. Adhering to the same list of womens interest legislation and
search terms, I referenced the subject index of each state journal to create a list of
womens interest legislation within each state. Again, data relating to the date of
introduction, the sponsor, the bill summary, the last stage in the legislative process,
and roll call votes, if applicable, were recorded in a spreadsheet. The journals also
provided a table of bills, which included information regarding the bill sponsor, date
of introduction, a brief bill summary, and the tracking of the bills process through the
lower chamber. The initial list of womens interest legislation generated from the
index was then cross-referenced with the table of bills in order to be as
comprehensive as possible. After these procedures were piloted with the two test
states, they were then repeated for an additional 23 states, and duplicated for the two
test states.
84


The legislative journals were secured using both interlibrary loans as well as
physically visiting the University of Iowa Law Library. Some states, such as Iowa
1975, California 1975 or New Mexico 1975, simply did not have a subject index in
the legislative journal. Other states, such as Hawaii and Alaska, were impossible to
secure through interlibrary loan. The University of Iowa retains the most
comprehensive collection of state legislative journals outside of the Library of
Congress, yet it has gaps in its collection. Thus, the selection of the 25 states is a
convenience sample based on the availability of journals through interlibrary loans or
the University of Iowa, as well as the journal contents. Figure 3.2 is a map
illustrating data collection for states within this research design. An ordered list can
be found in Appendix C.
m
SD
NY r-vv
Y-v *
M \
ki
^ \
\
CA
kk MO
y ,
TN
4 Time Periods
2Time Periods
O
Figure 3,2. STATE DATA COLLECTION.
85


States varied dramatically in the type of information available within the
legislative journals and on state legislative websites. In addition, at times LNSC had
gaps in the availability of data for bills, especially in relation to the 1995 time period.
Therefore, additional questions about the nature of a bill or roll call votes were
inevitable. These gaps in the data collection were addressed by directing inquires to
state archives, legislative clerks, or state law libraries in order to secure the final
details for each bill. This resulted in the compilation of 5,900 bills coded as women's
interest legislation across 50 states for two time periods and 25 states for an additional
two time periods. Finally, the data were aggregated by state and year according to the
methods described within the preceding section.
Independent Variables
Data collection regarding the independent variables was considerably
straightforward, as compared to the dependent variables. First, data for descriptive
representation were secured from the Center for American Women and Politics
(CAWP, 201 lc). Data regarding political party influence were available from a
supplement from the biannual edition of the Book of the States (CSG. 2011). The
state citizen ideology variable was retrieved from the website State Citizen &
Government Ideology (Fording, 2009). Data collection for both the women's caucus
variable and the gender opportunity structure variable were collected from several
different sources.
86


Comprehensive data regarding the presence of a womens legislative caucus
do not exist. As a result, several sources provided the data necessary for this study.
The National Conference on State Legislatures published a report offering a starting
point for collecting data on womens legislative caucuses (Oliver, 2005). This
material was supplemented with data from the Center for American Women and
Politics (CAWP, 2010). The data were then validated through triangulation, fact
checking through caucus websites (CWLC, 2010; LWC, 2010; LWLC, 2010; VSA,
2010; Wise, 2010), books and scholarly sources (Jones & Winegarten, 2000; Kanthak
& Krause, 2010; Keyserlmg, 1998; Mahoney, 2011; Reingold & Schneider, 2001;
Schenken, 1995; Smiley, 2007; Witt, Paget, & Matthews, 1995), informal email
communication (Keuther, 2010; Mason, 2010; McCoy, 2010; Skipper, 2010; Smith,
2010), and news media coverage of womens caucuses (AP, 1998; Bass, 2000;
Bemick, 1994; Gregg & Sullivan, 1989; Gross, 1992; Intress, 1996; Kieman, 1991;
Leonard, 1992; Lutey, 2008; Nussbaum, 1981; O'Matz, 1994; Propp, 1996; Stiffler,
1999; Whereatt, 1987).
Lastly, data for the gender opportunity structure were sourced from the annual
Community Population Survey6,7. These data were available through the University 6
6 Based on sampling methods, the Community Population Survey was unable to compute an average
rate for a number of states for the year 1974. Therefore, the average labor force participation rate was
computed per region for 36 states. Region One: Alabama & Mississippi; Region Two: New
Hampshire, Maine, Vermont, & Rhode Island; Region Three: South Carolina & Georgia; Region Four:
Kentucky & Tennessee: Region Five: Iowa, North Dakota, South Dakota, Nebraska, Kansas,
Minnesota & Missouri; Region Six: Washington, Oregon, Alaska, & Hawaii; Region Seven: Montana,
Wyoming, Colorado, New Mexico, Utah, Nevada, Arizona & Idaho; Region Eight: Delaware,
Maryland, Virginia & West Virginia; Region Nine: Wisconsin & Michigan.
87