OUT AT WORK: A DEMOGRAPHIC AND POLICY ANALYSIS OF LESBIAN, GAY, AND BISEXUAL DISCRIMINATION IN THE LABOR MARKET

A Dissertation Submitted to the Temple University Graduate Board

In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY

by Colin Hammar May 2021

Examining Committee Members:

Judith A. Levine, Advisory Chair, Department of Sociology James D. Bachmeier, Department of Sociology Robert L. Kaufman, Department of Sociology Douglas A. Webber, External Member, Department of Economics

© Copyright 2021

by

Colin Hammar

All Rights Reserved

ii ABSTRACT

My dissertation explores the demographics and composition of sexual minority populations in the United States, their labor market experiences, and public policy, respectively. Using a novel method of Cross-Survey Multiple Imputation (CSMI), I create a unique dataset which allows me to examine the demographic profile of lesbian, gay, and bisexual (LGB) populations at the national and state levels. I then measure the prevalence of discrimination experienced by these groups in the labor market through regression analyses and decompositions of wages. Finally, I examine the effectiveness of sexual orientation nondiscrimination policies at the state level.

My analyses show that LGB people make up just over four percent of the national population, a sizeable minority though smaller than popular and historical estimates. I show that LGB people tend to be younger, more racially and ethnically diverse, and slightly more educated than the heterosexual majority. However, LGB people are also more likely to be unemployed, more likely to be living below the poverty line, and less likely to have health insurance than heterosexual people. I find that lesbian women and gay men earn a wage premium over similarly situated heterosexual women and men while bisexual men and women experience a significant wage penalty relative to heterosexual men and women. After cataloguing and analyzing all state-level sexual orientation nondiscrimination policies for textual themes, I test for policy effectiveness.

My analyses suggest that while policies raise the wages of all workers, the specific effects of policies on LGB workers’ wages are inconsistent, suggesting other factors play a role in shaping wage differentials.

iii

For Daniel

iv ACKNOWLEDGMENTS

This dissertation was a long time coming and would not have been completed

without the support and guidance of innumerable people. I have had the privilege of

extraordinary teachers throughout my life, too many to name them all. Ann Rider at

Indiana State University was the first scholar and mentor to inspire me to reach further in

my love of learning. James Pennell, Timothy Maher, and Kevin Whitacre at the

University of Indianapolis introduced me to sociology and taught me the value of

research. Tracy Marschall at UIndy was a source of encouragement and friendship. At

Temple I had the privilege of studying with a number of brilliant and generous scholars

including Dustin Kidd, Tom Waidzunas, Kim Goyette, and Michelle Byng, each of

whose courses changed my thinking forever. More thanks than I can offer is deserved by

my dissertation committee. In addition to being exceptional scholars, they are each

exceptional people. Bob, Jim, and Doug each encouraged me throughout this process and helped make this dissertation what it is. It is hard to put into words the gratitude I feel towards my chair, Judith Levine. Since my very first semester at Temple, Judith has inspired me as a teacher, scholar, mentor, colleague, and friend. I hope to someday return

her kindness in more than just baked goods.

Surviving graduate school is impossible without the love and support of friends

and family. I am lucky to have both in abundance. Thanks to my graduate student

colleagues at Temple who went on this journey with me. When I arrived at Temple, AJ

Young was assigned as my mentor, a pairing which will be infamous in Temple

Sociology history. I will forever be AJ’s manatee. I am grateful to him and Stephen

v Dickinson for suffering through early drafts of this dissertation and offering their

feedback. My mom, sisters, nieces, and nephew have been a source of joy and escape throughout the years. I would not have survived without my best Judies: Jordan, Alex,

Stephen, and Ben. Love you, ladies.

Finally, none of this would have been possible without my husband, Daniel. Few

are lucky enough to have someone as kind, compassionate, and encouraging in their

corner. He came along through the highs and lows of graduate school and never once

complained. Without you, I never would have finished this. You are my life and my

world.

vi TABLE OF CONTENTS

Page

ABSTRACT ...... iii

DEDICATION ...... iv

ACKNOWLEDGEMENTS ...... v

LIST OF TABLES ...... xii

LIST OF FIGURES ...... xiv

CHAPTER

1. INTRODUCTION ...... 1

Introduction ...... 1

Research Questions ...... 4

Chapter 2—Demographics of Sexual Minority Populations ...... 5

Chapter 3—Labor Market Discrimination Against Sexual Minorities ...... 8

Chapter 4—Measuring the Effects of State Nondiscrimination Policies on Wages of Sexual Minorities ...... 10

My Contributions ...... 12

2. DEMOGRAPHICS OF SEXUAL MINORITY POPULATIONS ...... 16

Introduction ...... 16

My Contributions ...... 19

Sexual Orientation and Gender Identity ...... 21

Literature Review...... 23

vii Defining Sexual Orientation ...... 23

Attraction ...... 24

Behavior ...... 26

Identity ...... 28

Alignment of Dimensions ...... 29

Existing Data ...... 31

Sampling and Survey Design Challenges ...... 32

Pooled Sample Estimates ...... 34

Household-Level Data ...... 37

Overview of Literature ...... 39

Method ...... 40

Cross-Survey Multiple Imputation ...... 41

Data ...... 43

National Health Interview Survey (NHIS) ...... 43

American Community Survey (ACS) ...... 45

Imputation Procedure ...... 46

Harmonized Variables ...... 50

Prediction Model ...... 52

Results ...... 57

National Profile ...... 57

Demographic Characteristics ...... 59

Economic Characteristics ...... 63

State-Level Profiles ...... 64

viii Discussion ...... 74

Limitations ...... 77

Conclusion ...... 78

3. LABOR MARKET DISCRIMINATION AGAINST SEXUAL MINORITIES ...... 80

Introduction ...... 80

My Contributions ...... 82

Literature Review...... 83

Empirical Approaches to the Study of Discrimination ...... 83

Sexual Minority Discrimination ...... 85

History of Sexual Minority Discrimination ...... 86

Empirical Evidence of Discrimination ...... 88

Hypotheses ...... 94

Data ...... 95

Methods...... 98

Dependent Variable ...... 99

Independent Variables ...... 99

Regression Models ...... 101

Wage Decompositions ...... 106

Results ...... 108

Regression Models ...... 109

Bisexual Men ...... 110

Gay Men...... 116

Bisexual Women ...... 118 ix Lesbian Women ...... 124

Decompositions...... 125

Discussion ...... 132

Conclusion ...... 137

4. MEASURING THE EFFECTS OF STATE NONDISCRIMINATION POLICIES ON WAGES OF SEXUAL MINORITIES ...... 139

Introduction ...... 139

My Contributions ...... 140

History of Sexual Orientation Nondiscrimination Policies ...... 142

Federal Policy ...... 143

State Policy ...... 150

Municipal Policy ...... 158

Measuring Policy Effectiveness ...... 161

Hypotheses ...... 164

Data ...... 166

Methods...... 167

Variables ...... 168

Regression Models ...... 174

Results ...... 176

State Policies ...... 183

Regression Models ...... 189

Gay Men...... 189

Bisexual Men ...... 196

Lesbian Women ...... 197 x Bisexual Women ...... 201

Discussion ...... 202

Limitations ...... 206

Conclusion ...... 207

5. CONCLUSION ...... 209

Introduction ...... 209

Limitations ...... 213

My Contributions ...... 214

Future Research ...... 217

Conclusion ...... 219

REFERENCES CITED ...... 221

APPENDICES

A. STATE SAMPLE SIZES ...... 242

B. STATE-LEVEL DEMOGRAPHIC AND ECONOMIC CHARACTERISTICS ...... 245

C. FULL-STATE DEMOGRAPHIC AND ECONOMIC CHARACTERISTICS ...... 313

D. COMPARISONS OF CSMI STATE ESTIMATES TO EXISTING DATA ...... 319

E. DESCRIPTIVE STATISTICS BY SEX AND SEXUAL IDENTITY ...... 321

F. MEDIAN EARNINGS BY SEX AND OCCUPATIONAL CATEGORY ...... 329

xi LIST OF TABLES

Table Page 2-1. Sexual Orientation (Identity) and Sex of Sex Partners in Last Year (Behavior) by Year and Sex of Respondent ...... 36

2-2. Same-Sex Households by Relationship Status ...... 38

2-3. NHIS Sample Sizes ...... 44

2-4. NHIS Sexual Orientation Sample Sizes ...... 45

2-5. ACS Sample Sizes ...... 46

2-6. CSMI Model Variables ...... 47

2-7. Weighted Multilogit Prediction Model of Sexual Orientation ...... 54

2-8. Population Estimates by Sex and Sexual Identity ...... 59

2-9. Demographic Characteristics by Sexual Identity and Sex ...... 60

2-10. Economic Characteristics by Sex and Sexual Identity ...... 65

2-11. Percent of State Population by Sexual Identity ...... 68

3-1. Descriptive Statistics by Sexual Identity ...... 103

3-2. Mean Wages for Currently Employed by Sexual Orientation ...... 109

3-3. Regression of Log Wages by Sexual Identity for Male Workers ...... 111

3-4. Regression of Log Wages by Sexual Identity for Female Workers ...... 119

3-5. Wage Decomposition for Bisexual and Heterosexual Men ...... 127

3-6. Wage Decomposition for Bisexual and Heterosexual Women ...... 130

4-1. Descriptive Statistics of Model Variables ...... 170

4-2. Median Raw Wages by Sex, Sexual Identity, and State (in Dollars) ...... 180

4-3. Policy Themes ...... 184

xii 4-4. Regression of Log Wages by Sexual Identity for Male Workers ...... 191

4-5. Regression of Log Wages by Sexual Identity for Female Workers ...... 198

A-1. State Sample Sizes ...... 242

B-1. Mean Age by State and Sexual Identity ...... 245

B-2. Percent White, Non-Hispanic by State and Sexual Identity ...... 250

B-3. Percent Black by State and Sexual Identity ...... 255

B-4. Percent Other Race by State and Sexual Identity ...... 260

B-5. Percent Hispanic by State and Sexual Identity ...... 265

B-6. Percent Married by State and Sexual Identity ...... 270

B-7. Percent with Children by State and Sexual Identity ...... 275

B-8. Percent Having at Least a BA by State and Sexual Identity ...... 280

B-9. Percent Not in Labor Force by State and Sexual Identity ...... 285

B-10. Percent Unemployed by State and Sexual Identity ...... 290

B-11. Percent Fulltime, Year-Round Worker by State and Sexual Identity ...... 295

B-12. Median Earnings for Fulltime, Year-Round Workers by State and Sexual Identity ...... 300

B-13. Percent Below Poverty Line by State and Sexual Identity ...... 303

B-14. Percent with Insurance by State and Sexual Identity ...... 308

C-1. Demographic Characteristics by State ...... 313

C-2. Economic Characteristics by State...... 316

D-1. State-by-State Comparisons of CSMI Estimates to Existing Data ...... 319

E-1. Descriptive Statistics by Sex and Sexual Identity ...... 321

F-1. Median Earnings by Sex and Occupational Category ...... 329

xiii LIST OF FIGURES

Figure Page

4-1. State Policies by Type (2014-2018) ...... 157

4-2. Mean Hourly Wages by State for Men ...... 178

4-3. Mean Hourly Wages by State for Women ...... 179

xiv CHAPTER 1

INTRODUCTION

Introduction

It was a crisp and overcast Saturday in October when my friend AJ and I emerged

from our hotel in lower Manhattan. We walked on rain-soaked sidewalks past the Stock

Exchange and the grave of Alexander Hamilton, who would rap two shows uptown later that day, before we got to our subway station. After a thirty-minute train ride and another walk through the brisk autumn air, we arrived at the City University of New York. We had come to present at a conference organized by the Center for LGBTQ Studies. The conference’s theme was “After Marriage.”

Fifteen months prior, the United States Supreme Court ruled that same-sex marriage bans were unconstitutional. A culmination of decades of activism and advocacy, the Court’s opinion represented a sea change for lesbian, gay, bisexual, transgender, and queer (LGBTQ) rights in the United States. And yet, much work was still undone. The issue of marriage equality had loomed large in LGBTQ movements, leading one scholar to note, “The overwhelming attention paid to same-sex marriage [had] eclipsed other civil rights goals and perhaps even slowed progress in attaining them” (Mucciaroni

2016:26). The organizers of the conference had called us together to grapple with this question: Now that marriage equality, the issue that had consumed so much attention and resources, was a reality, what comes after? “From policing to mass incarceration to poverty to employment to housing to education to immigration and deportation to health

1 care to families living outside the marital model, the list of unfinished work is long”

(Jones and Yarbrough 2016:2).

Among the unfinished work for LGBTQ activists was the movement to enact

employment nondiscrimination policies. Activists have been concerned with the issue of

workplace inequality dating back to the first LGBTQ political organizations in the mid-

twentieth century (Stein 2012). From the expulsion of LGBTQ people out of the military

and the civil service, to discrimination against people with AIDS, to the persistence of

incidents of harassment and even violence, the workplace has been a key site in the

struggle for LGBTQ equality. According to a review of numerous studies on the topic, as

many as sixty-eight percent of LGBTQ people report experiencing discrimination at work

(Badgett et al. 2007).

Fifteen years into the twenty-first century, federal employment protections

appeared stalled in the wake of the successful movement for marriage equality.

Legislation in various forms had been introduced frequently over the preceding decades but never found its way into federal law. Part of the struggle was that most Americans

were just unaware of the problem. According to one recent poll, only twenty-three percent of respondents were aware that the United States did not have federal employment protections for lesbian, gay, and bisexual (LGB) workers (Reuters/Ipsos

2019). While a patchwork of state and municipal policies was enacted since the 1970s, the lack of federal policy left many LGBTQ people vulnerable to workplace discrimination.

The issue of workplace inequality raises several questions: Just how widespread is discrimination? How does it affect the economic standing of sexual minorities? And

2 which policies are best suited to addressing it? While these questions are seemingly

straightforward, several obstacles make answering them difficult. To begin with, we actually know very little about who is LGBTQ in the United States. To a large extent, this results from the fact that we rarely directly ask people about their sexual orientation in

survey research. Very few surveys include questions about sexual orientation and none of

the large federal surveys that are responsible for policy and resource allocation (e.g.,

Decennial Census, American Community Survey, Current Population Survey) collect this data directly. Therefore, it is challenging to even begin to answer these basic questions with any certainty. Without knowing who is LGBTQ, we cannot assess the effects of anti-LGBTQ discrimination.

The very act of counting LGBTQ people poses several conceptual and political hurdles. Some theorists posit that queerness resists quantification, that because LGBTQ identities are historical, fluid, and contextual, we cannot classify people into discrete categories of sexual orientation (Ghaziani and Brim 2019). And yet, an identity-based movement for LGBTQ civil rights emerged from the political understanding of individuals “being” LGBTQ (Stychin 2005). Part of the challenge posed by these perhaps competing schools of thought is that sexual orientation is itself a multifaceted concept incorporating individuals’ attractions, behaviors, and identities. When someone speaks about LGBTQ people, it can be unclear exactly about whom they are referring.

Wherever one lands in this theoretical debate, it is undeniable that public policy, as currently envisioned, yields to an empirical preference. Knowing how many people

“are” LGBTQ has critical implications for the distribution of rights and resources. Data on the prevalence and distribution of LGBTQ people “play key roles in arguments about

3 public policies involving the extension, protection, or prohibition of certain rights”

(Laumann et all. 1994:286). Being able to demonstrate how widespread discrimination is

and how many people are impacted by it goes a long way in the process of advocating for

policy interventions. Indeed, “if compulsory heterosexuality ‘others’ queer populations, then counting them may undermine this ‘otherness’ by demonstrating the legitimate needs of the LGBTQ+ population” (Doan 2019:121).

Given this reality, it is important to provide empirical measures of sexual orientation discrimination, its prevalence, effects, and potential remedies. This dissertation seeks to provide this evidence through an analysis of the demographics of sexual minority populations, an examination of their labor market experiences, and an analysis of the effectiveness of nondiscrimination public policies.

Research Questions

This dissertation is organized into three distinct, but interrelated essays. The essays explore the demographics and composition of sexual minority populations in the

United States, their labor market experiences, and public policy, respectively. While each essay contains its own literature review, methods, results, discussion, and conclusion, subsequent essays build on the findings of the prior analyses.

I address these questions in turn:

• What is the demographic profile of the lesbian, gay, and bisexual populations

(LGB1) in the United States? Who gets defined as LGB? What are the overall size

1 As I explain below, conceptual differences between sexual orientation and gender identity as well as data limitations prevent me from considering gender identity in these analyses. 4 of LGB populations? How do LGB people as a group compare to the heterosexual

majority? What demographic differences exist between LGB people?

• How do the economic outcomes of LGB people compare to their heterosexual

counterparts? Do differences in individual productivity characteristics, family

structure, or labor market position account for any differences between LGB and

heterosexual workers? Is there evidence of labor market discrimination against

LGB people?

• What is the current landscape of sexual orientation nondiscrimination public

policy? Do existing sexual orientation nondiscrimination policies adequately

protect LGB people from labor market discrimination? How does the content of

existing policies affect outcomes?

Chapter 2—Demographics of Sexual Minority Populations

In this essay I construct demographic profiles of the lesbian, gay, and bisexual

(LGB) populations. A popular statistic posits that ten percent of the population is

LGB(TQ). The statistic was initially disseminated as part of a political project: to demonstrate the size and potential electoral import of sexual minorities. The political origins of this statistic and the source of its derivation call its accuracy into question (see

Laumann et al. 1994, Eriksen 1999, Voeller 1990). While the ten percent estimate was effective in an era when proving the existence of LGB people was paramount, the policy needs of sexual minority communities today require more accurate estimates of the size and geographic distribution of the population.

Estimating the size of LGB populations poses several challenges. As noted above, sexual orientation is a multifaceted concept, and few surveys attempt to collect data on it.

5 It is also not obvious how one should conceptualize sexual orientation. Sexual orientation exists along several dimensions including attraction, behavior, and identity. One’s definition of sexual orientation should be contingent on the types of questions they seek to answer. Therefore, estimates of LGB populations can vary depending on the context in which the estimates are produced.

I use a novel strategy of Cross-Survey Multiple Imputation (CSMI) to develop a large-sample dataset capable of producing demographic profiles of LGB populations

(Rendall et al. 2013). CSMI uses data present in one survey (the donor survey), typically with a smaller sample size, to impute information not collected or measured in a second survey (the recipient survey), typically with a larger sample size. Using the National

Health Interview Survey (NHIS), I impute sexual orientation into the American

Community Survey (ACS). While the NHIS is nationally representative and contains sexual orientation information, its smaller sample size precludes state-level comparisons.

The ACS, while lacking sexual orientation information, has a large enough sample for state-by-state comparisons and has a robust set of economic indicators which I use extensively in Chapters Three and Four. CSMI allows me to exploit the benefits of both surveys.

My analyses focus exclusively on sexual minorities and do not include estimates of the transgender population for two reasons. First, sexual orientation and gender identity are distinct, though related concepts, and require different measurement strategies. Second, neither the NHIS nor the ACS collects adequate data to properly identify transgender respondents. Therefore, transgender people are present, but unidentifiable in my data. This is unfortunate given the similar, and perhaps more

6 pressing, need for accurate data on transgender Americans. I also recognize the political

implications of this demographic erasure in my work. If I were able to overcome it, I

certainly would endeavor to. As it stands, I join the call of other researchers and activists

demanding the inclusion of gender identity measures in survey research. For a discussion

of transgender measurement strategies and estimates of the transgender population, see

Doan 2019.

After imputing sexual orientation into the ACS using CSMI, I describe the

demographic profiles of LGB populations at the national and state level. I compare multiple sexual identity categories across several demographic and economic characteristics. The estimates that I produce further our sociological and demographic understandings in the areas of sexuality and inequality. Cross-group comparisons allow me to investigate differences and inequalities among and between sexual identity groups.

These comparisons speak directly to contemporary social and cultural movements for

(and against) LGBTQ equality; they inform continuing public policy debates; and they

give us a more complete picture of the sexual composition of the population. I also

demonstrate that CSMI is a useful method for developing datasets with sexual orientation

information.

My estimates suggest that LGB people make up 4.07 percent of the U.S.

population, well below the popular ten percent estimate. Bisexual women make up the

largest share of the LGB population, followed by gay men, then lesbian women, and

finally bisexual men. Overall, women are more likely to identify as LGB than are men.

LGB people are, on average, younger, more racially and ethnically diverse, and slightly

more educated than heterosexual people. However, LGB people are also more likely to

7 be unemployed, more likely to be living below the poverty line, and less likely to have

health insurance than heterosexual people.

Chapter 3—Labor Market Discrimination Against Sexual Minorities

In this essay I examine the labor market experiences of sexual minorities.

Specifically, I measure the extent to which discrimination affects the economic outcomes of LGB people relative to their heterosexual counterparts. There is a long history of explicit and implicit discrimination against LGB people in the labor market, in both policy and interpersonal interactions in the workplace. As noted above, large percentages of LGB people report having experienced discrimination in the workplace.

Prior research has shown that discrimination leads to reduced economic outcomes for some LGB people, particularly in the form of lower average wages. Studies have consistently shown a wage penalty for sexual minority men and a wage premium for sexual minority women. In past research bisexual men and women have often been grouped with gay and lesbian men and women, respectively, often for practical purposes

(e.g., small sample sizes, lack of data). Recent analyses suggest bisexual men and women face unique and significant labor market disadvantages suggesting their outcomes should be analyzed separately (Mize 2016).

Using the dataset developed in Chapter Two, I first construct nested regression models to examine the differences in wages between LGB and heterosexual people. The benefit of this dataset and its large sample size is that I can perform analyses separately for all groups by sexual identity and sex. I test several hypotheses: whether there is a wage penalty for gay men, bisexual men, and bisexual women compared to their heterosexual counterparts, and whether lesbian women experience a wage premium

8 relative to heterosexual women. I control for a number of productivity differences such as

education, work experience, family structure, and occupation.

My analyses suggest that both lesbian women and gay men earn a wage premium

over similarly situated heterosexual women and men, even when controlling for

differences in individual productivity, family structure, and labor market characteristics.

The finding of a wage premium for gay men is relatively novel and perhaps suggests

changes in the social and political climate have had effects on economic outcomes. I find

that bisexual men and women experience a significant wage penalty relative to

heterosexual men and women, though the significant age gap between the two

populations is an important factor, likely contributing to compounding differences in

education, work experience, and family structure. The magnitude of the wage penalty

suggests that bisexual men and women are particularly disadvantaged in the labor market.

Following the regression analyses, I then perform wage decompositions to estimate the impact of discrimination on any observed wage disparities. Wage decompositions allow me to divide an observed wage difference between two groups into two components: that which is explainable by differences in productivity characteristics and a residual, or unexplained component, which is interpreted as a measure of discrimination. I find that over sixteen percent of the gap between bisexual and

heterosexual men’s wages cannot be attributed to differences in productivity

characteristics like education, age, and work experience, suggesting discrimination is

negatively impacting bisexual men’s wages. For bisexual women, nearly eighteen percent

of the gap between their and heterosexual women’s wages cannot be attributed to

9 differences in productivity characteristics, similarly suggesting discrimination is negatively impacting bisexual women’s wages.

It is important to note that, while I do not find evidence of a wage penalty for gay men and lesbian women, this does not mean that these groups face no discrimination in the labor market. It is possible that discrimination against LGB workers prior to their entry into the labor market, perhaps in education, affects individuals’ labor market decisions. This discrimination could have negative impacts on the physical and mental health and wellbeing of LGB workers or drive them into lower paying occupations or industries. It is also possible that discrimination leads some LGB people to exit the labor market all together, leading to inequalities in labor force participation. My findings highlight the need for additional research into the multiple possible effects of workplace discrimination for LGB people.

Chapter 4—Measuring the Effects of State Nondiscrimination Policies on Wages of

Sexual Minorities

In the final essay, I turn my attention to the issue of public policy. At the time of my data (2014-2018), there were no federal protections against discrimination based on sexual orientation. Under federal law, it was perfectly legal to fire, not hire, fail to promote, or pay lower wages to a person just because they were LGB2. The failure to

enact federal policies led several states and dozens of municipalities to fill the void with

nondiscrimination policies of their own. I seek to understand how these policies, where

2 This was true for other sexual identity groups (e.g., pansexual, asexual) and, to an extent, also true for transgender people. Some federal courts had previously found discrimination based on gender identity was covered under Title VII, creating a patchwork of federal policies. Also, several states which have sexual orientation nondiscrimination policies exclude gender identity in their protections. 10 enacted, affect the labor market experiences of sexual minorities. I am interested in both

the symbolic and practical effects of such policies. How are these laws written? What

does the text of policies and statutes tell us about the state’s relationship to sexual

minorities? I am also interested in the effectiveness of such policies in reducing

workplace discrimination against sexual minorities. Do nondiscrimination polices

actually affect the labor market experiences of LGB workers?

I begin by cataloguing the existing state policies. I find that thirty-four states had

some form of sexual orientation nondiscrimination policy: either through an Executive

Order or a legislative policy. I then perform a content analysis of the text of state-level policies, developing a coding scheme to identify themes and patterns in the statutes.

Several themes emerge in the analysis of legislative texts: defining the boundaries of

“sexual orientation,” a discriminators perception of sexual orientation, associations of sexual orientation with crime or pedophilia, sexual orientation as a “preference,” and disclaimers against conferring social approval through statute.

Following the textual analysis, I then test a number of hypotheses to estimate the effectiveness of state policies. Again, using the dataset built in Chapter Two, I expand upon the models from Chapter Three to assess whether LGB workers’ wages are more comparable to heterosexual workers’ wages in states with nondiscrimination polices. I also test for differences in policy type and length of time since enactment. I control for differences in state-level economic outcomes and states’ social and political climates. The

results show inconsistent effects of nondiscrimination policies based on sexual identity.

The presence of both Executive Orders and legislative policies is associated with overall

higher wages for all workers, holding productivity characteristics constant, even when

11 controlling for differences in state-level economic outcomes and states’ social and political climates. Nondiscrimination policies appear to affect the wages of all workers consistently, regardless of sexual identity, suggesting that living in a state with labor protections based on sexual orientation provides a benefit to all workers. Gay men and lesbian women maintain wage premiums over their heterosexual counterparts across all models and policies, though the premium is narrowest in states with Executive Orders, suggesting those policies actually benefit heterosexual workers more so than gay and lesbian workers. A similar pattern was observed for bisexual men but not bisexual women. Bisexual men’s wage penalty relative to heterosexual men is widest in states with Executive Orders whereas bisexual women’s wage penalty relative to heterosexual women is actually widest in states with legislative policies.

These inconsistent findings suggest that other factors may be playing a larger role in shaping the distribution of wages. Perhaps other pro-labor policies or social and political conditions affect wages in ways not measured here. It is also possible that sexual orientation nondiscrimination policies have effects beyond the distribution of wages.

Policies could affect other dimensions of the labor market such as labor force participation rates, which would have an indirect, but ultimately unmeasurable, impact on wage differentials. The symbolic value of the policies, which suggest a safe and welcoming labor market, might outweigh the practical effects of policies on outcomes such as wages.

My Contributions

To my knowledge, no one has used CSMI to study sexual orientation. This dissertation joins a small body of sociological research, mostly in migration studies,

12 which uses the method (Capps et al. 2018). My research provides another case study

supporting the efficacy of the method in studying small and hard to sample populations.

Using CSMI allows me to create a state-of-the-art dataset for demographic, labor market, and policy analyses that overcomes many of the limitations of past research: small sample sizes, biased definitions of sexual orientation, and the inability to separate bisexual populations from gay and lesbian populations. The population estimates that I produce further our sociological and demographic understandings in the areas of sexuality and inequality. Cross-group comparisons allow me to investigate differences and inequalities among and between sexual identity groups. These comparisons speak

directly to contemporary social and cultural movements for (and against) LGBT equality,

inform continuing public policy debates, and give us a more complete picture of the

sexual composition of the population. This work also contributes to ongoing debates over

the inclusion of sexual orientation and gender identity (SOGI) measures in survey

research.

Because the sexual orientation measure in my analyses is based on identity (rather than behavior or attraction), I am able to examine separately the experiences of self- identified gay men, lesbian women, bisexual men, and bisexual women. Many analyses combine gay and bisexual men or lesbian and bisexual women into single categories, typically because of small sample sizes or sexual orientation measures that do not allow the analyst to distinguish between the two groups. Worthen (2013) highlights the importance of examining each group independently, given the multiple dimensions of sexuality and gender which differently shape the lived experiences of each group. The

13 identity measure in my data coupled with the large sample size allows me to examine

each group separately at the national and state levels.

The wage analyses I perform complement and extend the existing sociological

and economic literature on labor market discrimination. These analyses help provide a

more complete picture of labor market inequality broadly and for LGB people

specifically. They provide evidence about the prevalence of inequality and discrimination

experienced by sexual minorities. My data are also more contemporary, examining these

phenomena, at least partially, in the post-marriage equality era. My findings both

corroborate and expand upon the findings of past studies, suggesting that changing social

and political climates have had an effect on the labor market experiences of some sexual

minorities. The use of the CSMI dataset demonstrates the utility of this method for

analyzing economic outcomes.

The demographic and economic analyses also have clear policy implications.

While a small body of literature has examined the effectiveness of existing policies,

existing literature has relied on data with significant limitations. The CSMI dataset allows

for a more holistic examination of policy effectiveness and cross-state comparisons.

Because I am able to distinguish between categories of sexual identity, I examine how policies may differentially impact gay men, lesbian women, bisexual men, and bisexual women. To my knowledge this is the first such study to do so. I also catalogue and analyze the text of every state-level policy, distinguishing between policy types and content. I code policies for themes which become elements in my quantitative analyses, creating a unique dataset of state-level variables. I extend existing policy research by observing states’ social and political climates as factors in my analyses. Textual analyses

14 also allow me to explore the possible symbolic effects of policies, beyond the overt

intention of their purpose (banning discrimination).

Overall, the goal of this dissertation is to render sexual minority populations more

visible in data, to understand the effects of discrimination, and to examine potential

policy remedies. As the social, political, and policy landscape surrounding sexual

orientation continues to shift as we enter the third decade of the twenty-first century, it is imperative we have a clear and accurate picture of sexual minority communities and their policy needs. The sociological frameworks and methodologies I employ allow me to bring this picture into clearer focus. My hope is that this work will contribute to our understandings of LGB populations and inform policy debates going forward.

15 CHAPTER 2

DEMOGRAPHICS OF SEXUAL MINORITY POPULATIONS

Introduction

In March 1975, New York Congresswoman Bella Abzug stood before the press to

announce the introduction of a new piece of legislation. The Civil Rights Amendment of

1975 would have amended the Civil Rights Act of 1964 to include protections based on sexual orientation. Declaring that ten percent of the population was lesbian, gay, or bisexual (LGB), Abzug said the bill would “guarantee that all individuals, regardless of differences, are entitled to share in the fruits of society” (quoted in Faderman 2015:262).

In stating that ten percent of the population was LGB, Abzug sought to bolster support for her bill by suggesting that LGB people were a sizeable minority needing legislative protection. Calling upon the size of the minority was a legitimation strategy for both the bill and the community it sought to protect. But from where did this figure come?

The question of how many people are lesbian, gay, bisexual, and transgender

(LGBT) has captured the attention of activists, academics, and policy makers since at least the middle of the twentieth century. As nascent LGBT movements for civil rights emerged onto the national scene in the post-Stonewall United States, activists engaged the idea of “coming out” as a political strategy. Earlier generations of homophile activists

(the preferred term of many LGBT activist groups in the 1950s and 60s) typically understood coming out as a process of personal identity development and entrance into

LGBT life. Following the Stonewall Uprisings of June and July 1969, a new generation of activists, inspired by the feminist philosophy that “the personal is political,” saw

16 coming out as not only about personal identity but as a challenge to an oppressive society

that often failed to acknowledge their very existence. Gay Liberation activists embraced

“the notion that coming out and pursuing gay and lesbian visibility held the key to [gay

and lesbian] freedom” (Vaid 1995:57). While some continued to view coming out as a strategy for personal self-fulfillment, others recognized the political implications of coming out.

Bruce Voeller was an LGBT rights activists and founder of the National Gay Task

Force (now the National LGBTQ Task Force). A biologist and associate professor at the

Rockefeller Institute, Voeller “saw an opportunity to make a contribution to this new

movement by bringing… scholarly fields to it” (Voeller 1990:33). Drawing from the

most widely known studies on sexuality at the time, Voeller interpreted statistics from

sexologist Alfred Kinsey’s studies on male and female sexuality to estimate the size of

LGB populations (Kinsey et al. 1948, 1953). Kinsey et al. (1948) developed a seven-

point rating scale to describe an individual’s (homo)sexual history. Ranging from zero,

exclusively heterosexual, to six, exclusively homosexual, the scale was one of the first

attempts to demonstrate the diversity of human sexual experience, expanding

conceptualizations of sexuality beyond binary categories of heterosexual and homosexual

(636-641). Noting that thirteen percent of men and seven percent of women fell on the

“homosexual” side of the Kinsey scale, Voeller projected that this must mean that, on

average, ten percent of the overall population was gay and lesbian. Activists quickly

began to propagate this number to demonstrate the potential political strength of the

LGBT rights movement (Ericksen 1999:160, Laumann et al. 1994:289n7, Voeller

1990:34). The methodological and scientific validity of the statistic aside:

17 Convincing the population that LGBT people exist was an important factor in the decision of early LGBT advocates to promote the idea that 10 percent of the population was gay. That figure was large enough to “matter” and convince an American public skeptical about the very existence of LGBT people…. (Gates 2012:712)

The reliance on the Kinsey-derived statistic allowed activists to make concrete demands for rights and political representation, as evidenced by Abzug’s use of it when announcing her legislation.

While having clear political utility, the veracity of the statistic has been frequently called into question in the decades since its emergence. Laumann et al. (1994) refer to it as the “Myth of 10 Percent” (287). Others have critiqued the validity of the underlying data from the Kinsey studies (Cochran et al. 1953) and Voeller’s interpretation of it

(Ericksen 1999:160). More recent studies have found significantly lower estimates of

LGBT populations based on various data sources and collection approaches (Gates 2014).

Though the twenty-first century has seen significant social and political advancement for

LGBT people, particularly in the United States, estimates of LGBT populations continue to be important for assessing policy needs and for targeting funding and public policy interventions (Velte 2020). As Gates (2012) suggests, “The utility and accuracy of LGBT population estimates is now more salient in assessing and understanding the needs of the

LGBT community and evaluating programs designed to meet those needs” (712).

Producing accurate estimates of LGBT populations has significant demographic, economic, and public policy implications.

Despite this need for accurate demographic data, to this day large-sample, scientifically accurate measures of sexual minority populations are still rarely collected, and those data which are collected have significant limitations. In 2017, dozens of federal

18 lawmakers requested that the Census Bureau add questions related to sexual orientation and gender identity (SOGI) to the 2020 Census and the American Community Survey

(ACS), a major, annual survey of the U.S. population. Though this request came from lawmakers, the Director of the United States Census Bureau announced that the agency would not add SOGI measure to its surveys as “there was no federal data need” for such information (Thompson 2017). And yet, the Census Bureau Director’s claim notwithstanding, the lack of available data does impact policymaking. Velte (2020) refers to the failure to collect SOGI data in U.S. government surveys as the “Identity

Undercount” and suggests “policymakers ignore the needs of the LGBT community… because the Identity Undercount renders these needs invisible” (73). This invisibility makes it exceedingly difficult to identify and address the policy needs of LGB people and communities.

My Contributions

In this chapter, I develop national- and state-level demographic profiles of the lesbian, gay, and bisexual (LGB) populations in the United States. By developing these demographic profiles, I aim to render these populations visible. To produce these profiles, I use a strategy of Cross-Survey Multiple Imputation (CSMI) to create a large- sample dataset. CSMI uses data present in one survey (the donor survey), typically with a smaller sample size, to impute information not collected or measured in a second survey

(the recipient survey), typically with a larger sample size. Using the National Health

Interview Survey (NHIS), I impute sexual orientation into the American Community

Survey (ACS). While the NHIS is nationally representative and contains sexual orientation information, its small sample size precludes state-level comparisons. The

19 ACS, while lacking sexual orientation information, has a large enough sample for state-

by-state comparisons and has a robust set of economic indicators which I use extensively

in Chapters Three and Four below. CSMI allows me to exploit the benefits of both

surveys.

After imputing sexual orientation into the ACS using CSMI, I describe the

demographic profiles of LGB populations at the national and state level. I compare

multiple sexual identity categories across several demographic and economic

characteristics. The estimates that I produce further our sociological and demographic

understandings in the areas of sexuality and inequality. Cross-group comparisons allow

me to investigate differences and inequalities among and between sexual identity groups.

These comparisons speak directly to contemporary social and cultural movements for

(and against) LGBT equality; they inform continuing public policy debates; and they give

us a more complete picture of the sexual composition of the population. I then use the

created dataset to measure economic discrimination (Chapter Three) and to measure the

effectiveness of state policies (Chapter Four).

CSMI has not seen widespread use in the social sciences. To the extent that is has

been used, it has primarily been in migration research to study unauthorized immigrants

in the United States (see for example, Capps et al. 2018). This chapter further illustrates the benefits of the methodology and offers a case study of its application to the demographic and sociological analysis of sexual orientation. My analyses suggest that the method may have further applications beyond SOGI and may benefit scholars of other topics in which data availability is limited. This chapter, along with the existing literature

20 in other areas such as migration, provides evidence for the efficacy of CSMI in

sociological research of small or difficult to sample populations.

Additionally, this chapter further demonstrates the need for the inclusion of SOGI

measures in federal, state, and local surveys. As Velte (2020) notes, the failure to collect

SOGI data creates real, material harms for LGBT people, including “the denial of

statutory civil rights protections, the dilution of political power, and the creation of

maintenance of poverty in the LGBT community [sic]” (74). While the ten percent

estimate may have been politically efficacious at a certain time, it offered only a flawed

estimate of the overall proportion of LGBT people in the population. Accurate estimates

of poverty, income, educational attainment, housing insecurity are needed to craft social

policies that can assist LGBT people and counter discrimination. As Gates (2014) notes,

it is “less about the prevalence estimate and more about the ability to compare and

contrast characteristics of LGB/T individuals with their non-LGB/T counterparts” (10). I

hope these analyses contribute to our understandings of these differences and the

alleviation of some of these harms.

Sexual Orientation and Gender Identity

Within social movements, academia, and popular culture, sexual and gender identity groups are often examined together as a single, minority population. This is illustrated through the proliferation of terminology like “queer,” reclaimed from its usage as a vicious slur and used now as an umbrella term for sexual and gender minorities of numerous, and perhaps countless, variations, as well as abbreviations such as LGBT and

LGBTQ+ (for lesbian, gay, bisexual, transgender, queer/questioning, and beyond). While the concepts of gender and sexuality are certainly intertwined (see Richardson 2007),

21 there is a conceptual distinction to be made between the two, especially as it relates to the

empirical analysis of identity categories. Though gender and sexuality are often fluid and dynamic, they remain, for many, quite distinct in their definition and lived experience.

As such, in this chapter I draw a distinction between sexual orientation and gender identity and focus my attention on the former. The demographic profiles I construct examine sexual minority populations: lesbian, gay, and bisexual (LGB) people. Because of the conceptual distinction between sexual orientation and gender identity, I do not explicitly examine transgender populations. Furthermore, neither the NHIS nor the ACS collects data on trans-identified respondents making their explicit identification in my analyses impossible. Trans respondents are present but invisible in Census Bureau data, highlighting further the need to collect SOGI data.

I acknowledge that the decision to focus exclusively on sexual orientation is perhaps politically fraught, given the history of marginalization of trans identities within sexuality studies and by cisgender scholars, like myself. I do feel, however that the decision is warranted for purely practical purposes of measurement and analysis. My intent is not to reify any political or cultural barriers that might exist between sexual and gender minorities. While beyond the scope of this dissertation, the experiences of gender

minorities, and trans-identified individuals specifically, deserve continued investigation.

As this chapter highlights the need to collect sexual orientation data in national surveys, it

is equally important that we advocate for the inclusion of gender identity measures as

well.

22 Literature Review

Defining Sexual Orientation

Developing profiles of LGB populations requires a clear definition of the concept

of sexual orientation. According to the American Psychological Association (APA),

sexual orientation is “an enduring pattern of emotional, romantic, and/or sexual

attractions to men, women, or both sexes. Sexual Orientation refers to a person’s sense of

identity based on those attractions, related behaviors, and membership in a community of

others who share those attractions” (American Psychological Association 2008). The

diversity of dimensions within this definition arises from the interplay between

essentialist conceptualizations, in which one’s sexual self is understood as an innate characteristic of one’s “natural” self, and a constructivist view, in which one’s sexual self is an amalgamation of social and environmental factors (Laumann et al. 1994:284-5). An essentialist conceptualization of sexual orientation suggests there are innate indicators of sexuality that could easily be classified and quantified. Attempts to biologically define sexual orientation, typically through chromosomes, genetics, or hormones, have been frequently discredited and abandoned (D’Emilio 2002:154-64).

Rather than simply a product of an innate or natural self, many theorists and

researchers have come to view sexual orientation as far more complex than the

cataloguing of biological markers. As a social construct, sexual orientation is believed to

be influenced by numerous social and cultural factors (Weeks 1996:42). This has led to

the conceptualization of sexual orientation as, rather than biological, a social identity.

However, “the construct of sexual [identity] lacks a single definition agreed on by

researchers and society” (Gates and Sell 2007:336). Existing research has tended to

23 conceptualize sexual orientation along three dimensions: attraction, behavior, and

identity.

Attraction

Attraction consists of the psychological affinity one feels toward a potential

romantic/sexual partner based on the relationship between that partner’s gender and one’s

own. It is defined as “an intense, physiological, uncontrollable erotic or sexual desire for

males, females, or both sexes” (Savin-Williams 2009:8). An attraction-based definition of

sexual orientation would assign a category label (e.g., gay, lesbian, bisexual, straight) to

someone based on the gender(s) they find romantically/sexually attractive, regardless of

whether that attraction aligns with other dimensions of a person’s sexual orientation (i.e.,

behavior and identity). A person who acknowledges being attracted to persons of the

same gender might be labeled as LGB, depending on the degree and variation of

attraction, even if this attraction never manifests in a self-identification as LGB or even

same-sex sexual experiences.3

Because attraction is rooted in internal desires, it is difficult to conceptualize and

measure. Some survey measures have attempted collect self-reported data on sexual

attraction. For example, the National Survey of Family Growth (NSFG) has surveyed

respondents about their sexual attractions since 2002. Respondents are asked, “People are

different in their sexual attraction to other people. Which best describes your feelings?”

3 The sexualities literature consistently refers to “same sex” activity (e.g., attraction, sexual behavior) when identifying sexual minorities. It is perhaps more appropriate to discuss sexual minorities in terms of “same gender” activity when biological distinctions are not the primary focus. Often, our attractions are more rooted in gender than sex, especially prior to the disclosure of bodies. Similarly, same-gender sexual activity can occur between people of diverse biological sexes. However, to be consistent with the literature, I use “same sex” where otherwise “same gender” might be a more accurate reflection of the phenomena being discussed. 24 Response options include being mostly or only attracted to males/females and being

equally attracted to males and females (National Center for Health Statistics).

Respondents who indicate they are only or mostly attracted to those of the same sex could be construed as being lesbian or gay while those indicating equal attraction to both sexes could be construed as bisexual. However, individuals indicating such attractions may not identify themselves based on these attractions.

Given that individuals might not interpret their attractions as reflective of their

“true” sexual orientation, some research has sought alternatives to self-reported attraction data. As Waidzunas and Epstein (2015) document, numerous psychological (and pseudo- psychological) studies have attempted to measure sexual attraction through the physiological responses of, especially male, bodies. Using phallometric testing, a means

of measuring male genital response to psychological, typically visual, stimuli, researchers

have purported to discover the “true” sexual orientation of subjects. An individual’s

physical, genital response to stimuli is viewed as definitive evidence of sexual attraction.

These studies suggest “the phallometric test appears to yield authoritative claims about

embodied male sexual desire that may override what individuals themselves profess”

(188). The use of physiological evidence of sexual attraction remains controversial and

has been mostly embraced by conservative psychology and practitioners of conversion

therapies which are now banned in numerous states (Waidzunas 2015).

Whether through self-reported responses to survey questions or physiological

responses, attraction has occasionally been used as an indicator of sexual orientation in

research. However, attraction tends to be overlooked by social scientists in favor of other

dimensions of sexual orientation. According to Savin-Williams (2009), researchers often

25 use attraction measures for practical or political reasons—such as when surveying

adolescents, where questions about sexual behavior might be seen as unacceptable—

rather than scientific reasons, and their use is more common in international research

(10). Additionally, how respondents interpret the concept of attraction itself may affect

the validity of their responses. Cognitive testing of sexual attraction questions found

“some participants were simply unable to fully articulate a conceptualization of sexual

attraction, and their descriptions tended to be relatively ambiguous” (Miller 2001). While

attraction may be a meaningful concept worthy of further exploration, as an indicator of

sexual orientation it is perhaps limited in its efficacy.

Behavior

Behavior-based definitions of sexual orientation are determined by the gender of

those persons with whom one has physical, sexual relationships. It is defined as

“involvement in (minimally) genital contact with males, females, or both sexes” (Savin-

Williams 2009:8). Individuals acknowledging physical, sexual relationships with partners

of the same sex would be categorized as LGB. Additional data on the number and

duration of such relationships is needed to distinguish between lesbian/gay and bisexual

categories. Furthermore, the distinctions between these categories (lesbian/gay and

bisexual), requires interpretation from the researcher. How many sexual encounters

between persons of the same sex are required to establish a specific sexual orientation?

Does an individual who has a single same-sex sexual experience have a sexual orientation equivalent to someone who has had multiple, enduring same-sex experiences throughout their lifetime? Much like attraction, behavior can vary depending on one’s situation, social location, or life stage.

26 Several surveys ask questions about the sex of recent and past sexual partners.

The General Social Survey (GSS) asks respondents whether their sexual partners have

been “exclusively male, both male and female, or exclusively female” (Smith et al.

2019). Behavior-based definitions of sexual orientation are common in biological, public

health, and epidemiological studies. Kinsey et al. (1948) used behavioral data as a key

indicator for constructing their infamous seven-point scale. While the scale allows for a

range of sexual orientations, from exclusive heterosexual to exclusive homosexual, it has

been critiqued for, among other reasons, its “potentially inaccurate grouping of attitudinal

and behavioral aspects of sexuality, its focus on enacted sexual behaviors and sexual

fantasies without considering other dimensions of sexual orientation” (Wolff et al.

2017:510).

There are instances where behavioral measures are logically and conceptually

appropriate. In public health research, for example, sexual orientation categories defined

by one’s sexual history are common. Men who have sex with men (MSM) and women who have sex with women (WSW) are common conceptualizations of sexual orientation, especially in epidemiology. These categories explicitly rely on behavioral dimensions of

sexual orientation as indicators of risk for certain health outcomes, such as HIV exposure.

MSM/WSW are logical conceptualizations of sexual orientation when the outcome under

study is dependent on behavior in ways it is not dependent on attraction or identity.

However, like attraction-based measures, Young and Meyer (2005) note that MSM/WSW

are “problematic because they obscure [the] social dimensions of sexuality [and]

undermine the self-labeling of lesbian, gay, and bisexual people” (1144).

27 Identity

Sexual identity is the personal and social meanings one attaches to one’s own

sexuality. It is the “personally selected, socially and historically bound labels attached to

the perceptions and meanings individuals have about their sexuality” (Savin-Williams

2009:8). Most social and psychological studies ask for respondent’s self-reported sexual identity based on a list of pre-determined, mutually exclusive categories. In the GSS, for example, respondents are asked, “Which of the following best describes you? Gay, lesbian, or homosexual; bisexual; or heterosexual” (Smith et al. 2019).

Individuals rely on several factors in determining their sexual identity, and these factors may be fluid and contingent on one’s locations, both physical and social. As Gates and Sell (2007) contend, “identities are very situation-dependent. That is, a person may identify as gay in one setting, bisexual in another, and heterosexual in another” (242).

Fixed and mutually exclusive response categories cannot capture this fluidity and therefore only capture a snapshot of one’s sexual orientation, their sexual identity as described in a single moment in time. It is also important to consider what factors an individual might rely upon when assigning themselves an identity label. When asked to describe themselves as lesbian/gay, bisexual, or straight, a respondent may think to other dimensions of sexual orientation to construct their response:

It is important to note that although individuals may conceptualize their identity within the framework of who they have sex with or who they are attracted to, behavior and attraction in and of themselves do not constitute identity. It is the meaning—specifically the interpretations that the individuals assign those behaviors and experiences—that defines how they ultimately conceptualize their identity. (Miller and Ryan 2011:2)

Sexual identity should therefore be multidimensional in its conceptualization.

28 Additionally, identity-based measures require respondents to understand

both the question and the response categories and see the correlations with their

own identity. Some have found that respondents can be confused by clinical terms

like “heterosexual” and “homosexual” and that sexuality terminology is

understood differently based on characteristics like socio-economic status (Miller and Ryan 2011).

Alignment of Dimensions

Together, these three dimensions—attraction, behavior, and identity—represent the dominant approaches to defining and measuring sexual orientation in existing research. It is often assumed there should be alignment between these three dimensions within individual cases. Research has shown, however, that there is often inconsistency between reports of attraction, behavior, and identity (Badgett 2001, Laumann et al. 1994).

As Gates and Sell (2007) acknowledge, “It would be convenient if measures of sexual attraction, sexual behavior, and sexual orientation identity (or even coupling status) identified the same population. Unfortunately, they do not” (Gates and Sell 2007:237).

Being attracted to someone of the same gender does not necessitate having sexual experiences with such a person nor does it require identifying with a certain sexuality label (see for example, Ward 2015). Similarly, the sexual self-identification one offers in the context of a survey or an interview may not necessarily align with the identification they hold in other contexts.

The diversity and variability of sexual orientation dimensions and the fluidity of individual experiences can seem insurmountable when measuring sexual orientation, but researchers must decide upon the definition that will guide their individual inquiries.

29 Gates (2012) “advocate[s] that when we use these terms [lesbian, gay, and bisexual], we think more critically about providing explicit clarity about whom we are including in any particular definition” (710). It is perhaps most prudent to separate the various dimensions of sexual orientation into discrete measures or indices depending on the specific outcome of analysis. The context of a specific study provides the best rationale for favoring any single dimension or, where possible, multiple dimensions of sexual orientation.

For this chapter’s demographic analysis and the subsequent economic and policy analyses that follow, I will use sexual identity measures. In terms of construct validity, sexual identity measures are most likely to capture the information relevant for these types of analyses. “Because identities are a conception of self, sexual identity is a more tangible and knowable construct, and, theoretically, less problematic for respondents to report” (Ridolfo et al. 2012:115). Like other social and demographic characteristics, self- reported sexual identity is a measure of personal identification which impacts numerous aspects of one’s lived experience. How one identifies themselves publicly, including on a survey, has the potential to affect their material and social reality.

Identity is also particularly relevant for the study of economic outcomes and public policy, the subjects of Chapters Three and Four. Badgett (2007) suggests, “For labor market interactions, self-identity might best capture a characteristic that could cause differential treatment by employers or fellow employees, since identity might influence labor market decisions and openness about one’s sexuality in the workplace” (21). Self- identifying as a sexual minority makes one “available” for discrimination in a way that attraction and behavior might not. It is more likely for one’s identity to be disclosed in the workplace than one’s attractions or behaviors. For an employer to discriminate

30 because of an employee’s sexual orientation, the employer must be aware or, at the very least, infer that an employee is LGB. Unless an employee openly acknowledges attractions or behaviors that do not comport with the presumed standard sexuality

(heterosexual/straight), it is unlikely they would be marked for discrimination in the same ways as someone who identifies as openly LGB. I explore labor market discrimination in

Chapter Three.

Finally, sexual orientation public policy is typically enacted with sexual identity labels as the means of classification. Where the law explicitly protects individuals from discrimination based on sexual orientation, it must define who is covered. Where such laws have been enacted, the definitions of sexual orientation typically rely on sexual identity labels. Individuals are protected because of their being LGB. This ontological formulation raises a number of questions for, as Stychin (2005) posits, “to be suggests there is coherence and finality to identity—it is what one is and is knowable and recognizable to oneself and to others” (90). When one’s identity is incoherent to themselves or others, the law may not be malleable enough to protect them. “When the law defines identity categories, it risks reinforcing regressive and narrow ideas about sexual identities” (Boso 2015:575). I examine sexual orientation and public policy, specifically labor market nondiscrimination policies, in Chapter Four.

Existing Data

Data on sexual orientation have been collected in several surveys, including some small-sample federal surveys of the U.S. population. However, limitations in survey design prevent their use in these analyses. Below I describe these existing data sources and highlight their limitations, emphasizing sampling and survey design challenges, the

31 analytic challenges of pooled sample estimates, and the narrow conclusions drawn from household-level data. These limitations illustrate the need for more robust collection of sexual orientation data and, in the interim, support the use of CSMI.

Sampling and Survey Design Challenges

Whichever dimension(s) a researcher decides to use, collection of sexual orientation data requires overcoming several methodological hurdles. Special attention must be paid to issues of sampling, stigma, selection bias, and the broader politics of sexual orientation in contemporary society. Sampling sexual minorities presents one of the greatest challenges to sexual orientation research, particularly when attempting statistical analyses. Because sexual minorities are likely a small proportion of the overall population, probability sampling methods have typically been cost and time prohibitive for most research. Most early studies and nearly all qualitative studies of sexual minority communities rely on convenience samples or other non-probability sampling procedures.

As when these strategies are employed in other contexts, non-probability samples of sexual minority communities are inherently biased with significant variations dependent on the site of sampling (Weinberg 1970). While the findings of this research are typically valid, it requires a great deal of triangulation and replication because the findings are specific to the context from which the sample was drawn (Gates and Sell 2007).

The persistence of stigma against sexual minorities can impact respondents’ willingness to be forthcoming about their sexual orientation. Respondents may be reluctant to disclose a minority sexual identity, though this reluctance is likely correlated with other characteristics such as age, region, religion, and others (Gates and Sell 2007).

This reluctance may also be influenced by the source of the data collection. Government

32 surveys may be more affected by the threat of stigma than private or university-based research. Gates (2011) suggests that “feelings of confidentiality and anonymity increase the likelihood that respondents will be more accurate in reporting sensitive information”

(2). Stigma may also lead to misreporting in household-level data. LGB couple households might withhold the nature of their relationship by reporting more socially benign household structures. They might describe their relationship as roommates or unrelated adults to avoid disclosing their sexual identities (Gates and Sell 2007). The method of data collection may also affect a respondent’s willingness to accurately disclose their sexual orientation. Face-to-face interviews are more likely to produce undercounts of LGB identity than computer or internet-based surveys (Gates 2011).

Stigma is itself a variable condition and underreporting likely varies across time and place as the social acceptance of sexual minorities has evolved.

As with the non-probability sampling procedures discussed above, existing survey research on sexual orientation may also be prone to selection bias. Because of the threat of stigma and respondents’ possible reluctance to disclose their sexual orientation, those who do disclose are likely not representative of the population of sexual minorities more broadly. It could be that those who do disclose their sexual orientation freely are systematically different from those who are more reluctant to disclose. “It is therefore generally assumed that samples contain the subjects most open about their sexual status or living arrangements” (Gates and Sell 2007:242). As with susceptibility to stigma, a disposition towards disclosure is unlikely to be a randomly distributed characteristic and may be correlated with other variables of interest (Badgett 2007, Berg and Lien 2002).

The possible direction of these effects should be examined and discussed in all research

33 on sexual orientation. Additionally, Berg and Lien (2002) posit that “some heterosexuals

might strategically misreport [respond as LGB], perhaps to support homosexuals by

making them appear more numerous” (397). However, they provide no theoretical basis

or empirical evidence that such misreporting would be prevalent. There is certainly no

contemporary evidence of such misreporting.

Pooled Sample Estimates

Despite the research design difficulties surrounding sexual orientation, some

existing data of varied quality have been used to analyze sexual minority populations.

However, the limitations of these data call in to question the validity and reliability of the

resulting estimates. None of the large-sample national surveys administered by the United

State Census Bureau (Decennial Census, American Community Survey, Current

Population Survey) ask respondents directly about their sexual orientation. Several

topical surveys do ask about sexual orientation but are constrained by small sample sizes,

particularly when making state or local comparisons. A common approach has been to combine multiple years of frequently administered surveys to increase sample sizes.

When using this approach, it is important to consider “changes in social climate” over time that may affect the interpretation of results (Gates 2013:72). Given the dramatic societal shifts in attitudes regarding sexual orientation in the past twenty years, this is a serious consideration that should be adequately addressed with any pooled samples.

The General Social Survey (GSS), administered by the National Opinion

Research Center (NORC), has collected data about the sex of a respondent’s sexual partners since 1988 (Smith et al. 2019). Numerous studies have used the GSS sexual behavior questions to assign respondents to sexual minority categories if they disclose

34 same-sex sexual experiences (Badgett 2001, Martell 2013a, Berg and Lien 2002,

Cushing-Daniel and Yeung 2009). The GSS has also collected sexual identity data since

2008 (Smith et al. 2019). Because the GSS only surveys a moderate sample during each

two-year cycle, the number of LGB respondents is quite low in any given year. Table 2-1

compares sexual behavior and sexual identity data from the 2014, 2016, and 2018 GSS.

LGB respondents average 5.42 percent over these years. Given the small sample sizes

within categories of LGB identity, it is difficult to make inferences about these groups.

The issue is further exacerbated by missing data among LGB respondents on other

measures. Respondents to the GSS do not consistently answer both the sexual identity

and sexual behavior questions leading to varied sample sizes for the questions.

It is common for several years of GSS data to be pooled for analysis. Berg and

Lien (2002) use GSS data from 1991 through 1996. As the GSS did not collect sexual

identity data at that time, the authors use sexual behavior data as an indicator of sexual

orientation. A respondent is coded as “homosexual” if they report having had a same-sex

sexual partner in the past five years (397). This behavior definition is one of the broadest,

even among those using the GSS. After restricting their analysis to full-time workers with no missing data (n=2,887), the authors find roughly four percent (n=116) coded as

lesbian or gay. Cushing-Daniels and Yeung (2009) use a pooled sample from 1988 to

2006. They construct a similar measure using sexual behavior, comparing the reported

sex of sexual partners from the past five years and the past twelve months. They find a

larger sample report having same-sex partners in the past twelve months than the past

five years (168). Their sample (n=15,426), limited to working-age respondents (eighteen

to sixty-four), is roughly three percent (n=452) LGB.

35 Table 2-1. Sexual Orientation (Identity) and Sex of Sex Partners in Last Year (Behavior) by Year and Sex of Respondent 2014 2016 2018 Male Respondents Identity N = 1,058 N = 802 N = 645 Gay 2.65% 2.99% 2.02% Bisexual 1.89% 2.24% 0.93% Straight 94.14% 93.39% 94.57% No Answer 1.04% 1.12% 1.86% Don't Know 0.28% 0.25% 0.62%

Behavior N = 861 N = 636 N = 509 Exclusively Male 4.18% 3.46% 2.55% Both Male and Female 1.05% 1.42% 1.18% Exclusively Female 92.68% 94.03% 94.89% No Answer 2.09% 1.10% 1.38% Don't Know 0.00% 0.00% 0.00%

Female Respondents Identity N = 1,292 N = 970 N = 761 Lesbian 1.32% 2.27% 2.37% Bisexual 3.48% 3.92% 5.91% Straight 92.80% 91.96% 89.49% No Answer 2.17% 1.55% 1.97% Don't Know 0.23% 0.31% 0.26%

Behavior N = 950 N = 725 N = 538 Exclusively Female 2.32% 3.86% 4.09% Both Male and Female 1.16% 2.34% 2.79% Exclusively Male 94.21% 92.69% 92.01% No Answer 2.11% 1.10% 1.12% Don't Know 0.21% 0.00% 0.00% Source: General Social Survey (2014-2018)

36 Martell (2013a), looking exclusively at gay men, uses GSS data from 2008

through 2012 and compares four different definitions of sexual orientation: whether the

respondent reported having exclusively same-sex sexual partners in the past year,

whether the respondent reported having exclusively same-sex sexual partners in the past

five years, whether the respondent reported at least one same-sex sexual partner since age

eighteen, or whether the respondent reported at least half their sexual partners since age

eighteen were same-sex. Because the GSS also collected data on respondent’s self-

reported sexual identity, Martell is able to compare the constructed measures to the

identity measure. Behavioral reports of exclusive same-sex sexual partners within the past five years are most consistent with self-reported sexual identity, however, the sample is reduced to a mere sixteen cases of gay identified respondents.

Household-Level Data

The lack of large-sample surveys, especially those collecting individual-level responses about sexual orientation, has led to alternative measures for inferring sexual orientation within some existing surveys. Many federal surveys collect data on households and their occupants. By comparing the gender and reported relationships among members of a single household, researchers have been able to analyze households purportedly containing a same-sex couple. Household-level data uses behavioral proxies—reporting a same-sex domestic relationship—to define sexual orientation. Since the 1990 Census, the United States Census Bureau has collected data on same-sex couple households. Estimates of the number of same-sex couple households have increased steadily since then. Table 2-2 compares household-level data from four federal surveys: the 2000 and 2010 Decennial Censuses, the 2018 ACS, and the 2019 Current Population

37 Survey’s (CPS) Annual Social and Economic Supplement (ASEC). While these data give

us a glimpse into some sexual minority communities, it is limited to only those sexual

minorities currently occupying a shared household with a same-sex spouse or partner and who disclose such on the survey. Unfortunately, there is no mechanism for identifying sexual minority cases among the non-partnered, those who do not share a household with their partner, those sexual minority respondents in a domestic relationship with a different-sex partner (i.e., certain bisexual respondents), or those who do not describe their relationship as the survey proscribes. Additional definition and measurement considerations limit the usefulness of these data.

Table 2-2. Same-Sex Households by Relationship Status Unmarried Married All Same-Sex Partners Couples Households Percent of all Percent of all Percent of all Survey Total Total Total Households Households Households 2000 314,052 0.30% 44,338 0.04% 358,390 0.34% Censusa

2010 514,735 0.44% 131,729 0.11% 646,464 0.55% Censusa

2018 592,561 0.49% 402,859 0.33% 995,420 0.82% ACS

2019 468,232 0.36% 543,124 0.42% 1,011,357 0.79% CPSb aCensus Bureau Preferred Estimates bMarch 2019 Annual Social and Economic Supplement Source: U.S. Census Bureau

One substantive barrier to interpreting household-level data is the mismatch

between how same-sex couples define their relationship and limited survey response

38 options. Prior to the Supreme Court’s 2013 decision striking down the Defense of

Marriage Act (DOMA), the Census Bureau was prohibited from collecting data on same- sex married couples, whether or not their marriage was legally recognized in the state in which they resided. The Census Bureau limited same-sex couples, married or otherwise, to the category of “unmarried partners.” Many couples likely did/do not identify with this clinical term and may have, for personal or political reasons, indicated a different household arrangement (Hogan et al. n.d.:4).

The Census Bureau has used different strategies for dealing with same-sex couples who designate themselves as something other than unmarried partners. Same-sex couple households are identified using three pieces of data: the sex of the householder, the sex of a second member of the household, and the relationship between these two household members. In the 1990 Census, the Census Bureau recoded any same-sex couples that indicated they were married on the relationship variable by changing one of the spouse’s sex to make them a heterosexual couple. This was similarly done with same- sex couples who indicated they were married on the CPS prior to 2010. In the 2000 and

2010 Censuses, same-sex couples who indicated they were married had their relationship recoded to unmarried partners. Similar changes to same-sex couple households were made in the ACS from 2005 to 2012 and the CPS from 2010 to 2013. Since 2013, same- sex married couples have been counted in the ACS and CPS, and they will be counted in the 2020 Decennial Census.

Overview of Literature

The estimate that ten percent of the population is LGBT has permeated American consciousness. It has been consistently called upon despite its lack of scientific veracity.

39 As Voeller (1990) himself stated, “the concept that 10% of the population is gay has become a generally accepted ‘fact.’ While some reminding always seems necessary, the

10% figure is regularly utilized by scholars, by the press, and in government statistics”

(36). More recently, American perceptions of sexual orientation have skewed towards even larger estimates. Polling by Gallup has found Americans estimate that, on average, nearly a quarter (twenty-three percent) of the population is LGBT (Newport 2015).

Current best estimates of LGB populations fall well below the ten percent that activists have long heralded. Gates (2014), aggregating findings from multiple surveys, found that estimates of the United States LGB population range from 1.70 to 5.80 percent or an average of 3.50 percent. The proliferation of the ten percent figure and the recent perception of LGBT communities as much larger than the data suggest are testaments to the work of activists who, over the past fifty years, have raised the American consciousness to the existence of LGBT people.

While the public consciousness has been raised, public policy makers continue to need scientifically accurate measures of sexual orientation. To address the serious issues facing LGB people, more data about the specific characteristics of this population are needed. Given the limitations of the existing data described above, there has been a concerted movement to expand the collection of sexual orientation data and SOGI data more broadly. While these debates continue and new survey measures are developed and deployed, the immediate need for data persists. This chapter attempts to fill that gap.

Method

To build a dataset with a large enough sample to construct national and state-level profiles of sexual minority populations, I use a strategy of Cross-Survey Multiple

40 Imputation (CSMI). To create my CSMI dataset, I build a model for predicting sexual orientation using variables common to the NHIS and ACS. In this section, I describe the technical requirements of CSMI, I describe the donor (NHIS) and recipient (ACS) surveys, and I evaluate my imputation model.

Cross-Survey Multiple Imputation

Traditional multiple imputation (MI) was developed as a strategy to address missing data within a single survey. MI uses the distribution of observed data to “fill in” missing values with predicted values. According to Rubin (1987), MI “replaces each missing or deficient value with two or more acceptable values representing a distribution of possibilities” (2). Because we cannot perfectly predict the true value of a missing response, generating multiple possible values for missing data accounts for uncertainty in the prediction process. The multiple possible values for missing data are then used to create multiple “complete” data sets. Analyses are performed using the multiple complete data sets and final estimates are produced by pooling the results. CSMI uses these same basic principles.

Developed by Rendall et al. (2013), CSMI combines two surveys to impute information that is not available in one, the recipient survey, but is present in the other, the donor survey. The information to be imputed into the recipient survey is missing for all cases because the measures were not included during the data collection. CSMI is used “to impute values from one survey to a second survey in which that variable is not present by design—that is, no question was asked and no other form of assessment was undertaken in the second survey” (Rendall et al. 2013:485). Using the donor survey, an imputation model is estimated by regressing the variable of interest on a set of predictors

41 present in both surveys. The prediction model is then used to “fill in” the missing data in

the recipient survey. As with traditional MI, multiple possible values are estimated for

each case because the prediction model is not perfectly predictive. The multiple

completed data sets are then analyzed using the conventions of within-survey MI.

CSMI best practices include two main conditions: the surveys must be sampled

from the same universe and all analyzed variables must be jointly observed. The same

universe condition requires that the samples of the donor and recipient surveys be from

the same underlying population. The exclusion of variables not jointly observed requires

that all variables used in the imputation model or subsequent analysis models be present

in both surveys. Rendall et al. (2013) “argue that a cross-survey imputation study should

be designed such that the analysis model can be estimated with one of the surveys alone”

(494). When these conditions are met, CSMI offers benefits over other imputation

methods. CSMI leads to efficiency gains over within-survey MI, the “missing at random” assumption of within-survey MI is easily met, and the missingness in the recipient survey is inherently monotone (Rendall et al. 2013:488). According to Van Hook et al. (2015),

CSMI was found to increase statistical power, allowing for analyses of subgroups and smaller geographies (352).

Rendall et al. (2013) acknowledge that CSMI has not seen widespread use, noting they are “not aware… of any successful implementations in sociology of [CSMI]” (517).

The method has since been used by population analysts and demographers studying immigration (Capps et al. 2018). I expand on this burgeoning literature by using CSMI in the study of sexual orientation. I impute sexual orientation into the American Community

Survey (recipient survey) using the National Health Interview Survey (donor survey).

42 Data

National Health Interview Survey (NHIS)

Because the NHIS includes a measure of sexual orientation, I use it as the donor survey in the CSMI. The NHIS is an interview-based survey of the U.S. population fielded annually since 1957. A project of the National Center for Health Statistics

(NCHS), a part of the Centers for Disease Control and Prevention (CDC) and administered by the United States Census Bureau, “the main objective of the NHIS is to monitor the health of the United States population through the collection and analysis of data on a broad range of health topics” (CDC). The NHIS samples the non- institutionalized population from U.S. households. Overall, each year the sample consists of roughly 35,000 households containing 87,500 individuals (CDC). A randomly selected adult (sample adult) from each household is chosen each year to answer additional questions. Survey adults represent a random sample of the non-institutionalized population aged 18 and older. I use the NHIS for survey years 2014-2018. Table 2-3 describes the sample sizes for the included survey years as well as the sample adults. To use a more complete donor survey, I perform a within-survey multiple imputation to eliminate missing data on several NHIS variables.

The NHIS added a measure of sexual orientation beginning in 2013. The NHIS defines sexual orientation as self-reported sexual identity. Sample adults were asked, “Do you think of yourself as: lesbian or gay; straight, that is, not lesbian or gay; bisexual; something else; or don’t know?” (Miller and Ryan 2011:6). Table 2-4 summarizes responses to the sexual orientation question for the included survey years. Respondents indicating “something else” received a follow up question seeking clarification.

43 Table 2-3. NHIS Sample Sizes4 Year Overall Sample Sample Adults 2014 112,053 36,697 2015 103,789 33,672 2016 97,169 33,028 2017 78,132 26,742 2018 72,831 25,417 Source: IPUMS NHIS

According to Miller and Ryan (2011), who performed cognitive interviews (a

method of evaluating question validity and data quality) about the new NHIS sexual

orientation measure, many transgender respondents chose the “something else” response

option when not presented with a “transgender” response to the sexual orientation

question. While sexual orientation is considered a distinct concept from gender identity,

the two concepts continue to mingle in the lived experiences of actual people.

Additionally, many respondents who identified as asexual, pansexual, or used another

identity label such as queer chose the “something else” response as intended (20). Due to small sample sizes, their follow-up responses are not included in the public data.

Unfortunately, absent the clarifying follow up responses, it is not prudent to include the

“something else” responses in the following imputation of sexual orientation. It is therefore likely that the imputation model excludes some sexual minority respondents and potentially results in an undercount of LGB populations. Future research should seek access to this restricted data to improve the completeness of the NHIS donor survey.

4 The decline in response rates led to a redesign of the NHIS questionnaire in 2019. 44 Table 2-4. NHIS Sexual Orientation Sample Sizes Sexual Orientation 2014 2015 2016 2017 2018 Lesbian or Gay 578 569 526 476 456 Straight, Not Lesbian or Gay 34,606 31,430 30,952 24,826 23,740 Bisexual 282 276 324 333 296 Something Else 88 121 124 131 116 I Don't Know 155 232 256 221 240 No Answer 225 166 194 160 165 Refused 763 878 652 595 404 Source: IPUMS NHIS

American Community Survey (ACS)

The ACS will serve as the recipient survey in the CSMI. While it does not contain

measures of sexual orientation, its large sample size allows for cross-state comparisons.

Additionally, the ACS contains robust economic and labor market data that I will use in

Chapters Three and Four. A nationally representative, annual survey of U.S. households, the ACS samples roughly one percent of the population each year and provides between- census estimates of the U.S. population, including demographic characteristics. I use the

2018 five-year file which includes data collected between 2014 and 2018. Table 2-5 summarizes sample sizes from the 2018 five-year file by survey year. While the five-year

file is less current, its large sample size allows for greater precision compared to the one-

year file, and it is ideal for examining small populations, such as sexual minorities (U.S.

Bureau of the Census 2020a). Because the underlying population of the ACS is also U.S.

households, it is drawn from the same universe as the NHIS. The ACS also has numerous

variables in common with the NHIS, a requirement of CSMI. Its numerous economic and

demographic variables make it suitable for the analyses here and in Chapter Three. The

ACS is also important to use in these analyses because of its policy relevance. ACS data 45 influence legislation, public policy, and “Census and ACS data determine the annual distribution of more than $675 billion of federal funding” (Velte 2020:80). This makes the ACS particularly relevant for the policy analyses in Chapter Four.

Table 2-5. ACS Sample Sizes Year Overall Sample Age 18+ 2014 3,132,610 2,469,680 2015 3,147,005 2,490,616 2016 3,156,487 2,503,750 2017 3,190,040 2,530,726 2018 3,214,539 2,562,591 Source: IPUMS USA

Imputation Procedure

I retrieve the ACS data from IPUMS USA (Ruggles et al. 2021) and the NHIS data from IPUMS Health Surveys (Blewett et al. 2019). Prior to pooling the two surveys,

I prepare each dataset individually for the imputation process. I restrict the NHIS to only sample adults (n=155,556), to whom the sexual orientation question was asked. I restrict the ACS to non-institutionalized adults aged eighteen or older (n=11,808,153). I then harmonize the variables to ensure identical coding across the two surveys. Table 2-6 describes the variables common to both surveys.

46 Table 2-6. CSMI Model Variables Variable Variable Description Harmonized Coding Demographic Variables age Age at survey. age_sq Age at survey squared. sex Self-reported sex. 0 Male; 1 Female 1 White; 2 Black/African American; 3 American Indiana/Alaska Native; 4 race Self-reported race. Chinese; 5 Filipino; 6 Asian Indian; 7 Other Asian; 8 Other/Multi-Racial Self-reported Hispanic hisp 0 Not Hispanic; 1 Hispanic ethnicity. 0 Less than HS; 1 High School degree Highest degree earned. Diploma; 2 BA; 3 MA, Professional Degree, or Higher vetstat Veteran status. 0 Not a Veteran; 1 Veteran householder Householder status. 0 Not Householder; 1 Householder ownership Homeownership status. 1 Owns Home; 2 Rents Home; 3 Other 1 Northeast; 2 North Central/Midwest; 3 region Region of residence. South; 4 West Economic Variables Personal earned income in earnings previous year. 0 Not in Labor Market; 1 Unemployed; empstat Employment status. 2 Employed 0 NIU; 1 Agriculture; 2 Mining; 3 Construction; 4 Manufacturing; 5 Wholesale Trade; 6 Retail Trade; 7 Transportation, Warehousing, Utilities; 8 Information; 9 Finance/Insurance; 10 Real Estate/Leasing; 11 Professional, industry Industry Scientific, and Management, and Administrative, and Waste Management Services; 12 Education/Healthcare; 13 Arts; 14 Food Service; 15 Other Service; 16 Public Administration; 17 Armed Services

47 Table 2-6. (continued) Variable Variable Description Harmonized Coding Received interest or intdiv 0 No; 1 Yes dividend payments. Got Social Security gotss 0 No; 1 Yes payments. Got Supplemental Security gotssi 0 No; 1 Yes Insurance payments. Received government gotwelf 0 No; 1 Yes assistance payments. hourswrk Hours worked in past week. 0 Above Poverty Line; 1 Below Poverty poverty Poverty status. Line Family Variables 1 Married, Spouse Present; 2 Married, Spouse Absent; 3 Separated; 4 marst Marital status. Divorced; 5 Widowed; 6 Single, Never Married mom Mom present in household. 0 No; 1 Yes dad Dad present in household. 0 No; 1 Yes Spouse present in spouse 0 No; 1 Yes household Same-sex spouse/partner sspartner 0 No; 1 Yes present in household. Number of own children in nchild household. eldch Age of own eldest child. Number of own children nchlt5 younger than 5. Number of own siblings in nsibs household. Number of persons in famsize family.

48 Table 2-6. (continued) Variable Variable Description Harmonized Coding Migration Variables citizen U.S. citizenship status. 0 Not U.S. Citizen; 1 U.S. Citizen 0 Born Outside U.S.; 1 Born in U.S. usborn Place of birth. Territory; 2 Born in U.S. State or DC Recently immigrated to recent1 0 No; 1 Yes U.S. within last year. Recently immigrated to recent5 0 No; 1 Yes U.S. within past 5 years. Health Insurance hcovany Has health insurance. 0 No; 1 Yes Has private health hiprivate 0 No; 1 Yes insurance. hipubcove Has public health insurance. 0 No; 1 Yes Has private insurance hipworkr 0 No; 1 Yes through employer. himcaide Receives Medicaid 0 No; 1 Yes himilite Receives insurance 0 No; 1 Yes coverage through military.

49 Harmonized Variables

Independent Variables. Thirty-nine independent variables (IVs) common to both the donor and recipient surveys are used in the prediction model. Both surveys contain several demographic variables. Age is measured as a continuous range from eighteen onward. I also include a quadratic transformation of age. Respondent’s sex is recorded as male or female. Neither survey allows for non-binary or transgender identities to be explicitly recorded: non-binary and trans respondents are forced to choose between the binary male/female response options or leave their response blank, in which case the

Census Bureau allocates their response based on several criteria (Ruggles et al. 2021).

Respondent’s self-reported racial identities are recorded in a categorical variable.

Following Zuberi (2001), I include racial data as an indicator of social stratification and not as a means of essentializing racial categories. Similar considerations are given to the inclusion of the indicator of Hispanic identity. Region indicates in which Census Bureau- designated region the respondent resides. Educational attainment is measured as the highest degree earned. An indicator of householder status is also included. Householder is a Census Bureau designation that refers “to the person (or one of the people) in whose name the housing unit is owned or rented (maintained)… [and] is the ‘reference person’ to whom the relationship of all other household members, if any, is recorded” (U.S.

Bureau of the Census 2020b). Indicators of veteran status and home ownership are also included.

Several economic and labor market variables are included in the model. Earnings is a continuous variable of the respondent’s pre-tax personal income in the past year.

Indicators of whether the respondent received income through interest/dividends

50 payments, Social Security, Supplemental Security Income, and government assistance are

included. Employment variables include a measure of labor force status, industry

categories, and hours worked. Poverty status is measured as above or below the federally

designated threshold for the respondent’s family size and income.

Family variables describe the respondent’s living arrangements. This includes

marital status, family size, and information about the respondent’s children and their

ages. A respondent’s “own” child is self-reported and can include biological, adopted,

and stepchildren. Indicators of the presence of parents, siblings, and a spouse in the home

were also included. Because both surveys now collect data on same-sex partners and

spouses, I include an indicator of whether the respondent identified a same-sex

partner/spouse in the household.

Migration variables include whether the respondent was born in the United States, a U.S. territory, or a foreign country. A citizenship indicator is included as well as indicators of whether the respondent is a recent immigrant within the past year or the past five years.

Health insurance variables include an indicator of whether the respondent has any health insurance, whether that insurance is public or private, whether private insurance is provided through an employer, whether coverage is from the military, or whether the respondent is receiving Medicaid.

Dependent Variable. The dependent variable (DV) in the imputation model is sexual identity, a nominal variable with three categories: straight/heterosexual, gay/lesbian, and bisexual. While many studies collapse gay/lesbian and bisexual into a single category to create a dichotomous sexual orientation variable, I choose to preserve

51 as much variation as possible. According to Worthen (2013), “it is imperative that both qualitative and quantitative studies utilize separate constructs to examine prejudices toward lesbians, gays, bisexual men, [and] bisexual women” (716). The erasure of bisexuality has been a frequent political debate within LGBT social movements (Stulberg

2018), the law (Yoshino 2000), and scholarship (Monro et al. 2017). Recent studies suggest that binary measures of sexual orientation fail to capture the unique threats that bisexual people face. (Mize 2016). Mize and Manago (2018) find that “stereotypes of bisexual men and women as dishonest, indecisive, disingenuous, and as more confused and selfish” lead them to be viewed as less competent and less warm (473). Due to small subsample sizes, it is occasionally necessary to use a binary measure of sexual orientation. I always indicate where such a change has been made.

Prediction Model

Once the variables in the two surveys are harmonized, the next step in CSMI is to build a prediction model using the donor survey for the DV using the IVs common to both surveys. As the DV is a non-dichotomous, unordered categorical variable, I use a multinomial logistic regression model to determine probabilities of sexual identity outcomes (Osborne 2015). Multinomial logistic regression computes the log odds of membership in a category of the variable relative to a reference category. Therefore, I model the probability that the ith case is in the jth sexual identity category by the formula:

= log = + 𝜋𝜋𝑖𝑖𝑖𝑖 ′ 𝜂𝜂𝑖𝑖𝑖𝑖 𝛼𝛼𝑗𝑗 𝑥𝑥𝑖𝑖 𝛽𝛽𝑖𝑖 𝜋𝜋𝑖𝑖𝑖𝑖

52 where αj is a constant and βi is a vector of regression coefficients for the predictor

variables in Table 2-6 above (Rodríguez 2007). Table 2-7 presents results from the

prediction model from the donor survey data. Heterosexual is the base category. The

model is significant (p < 0.00005) with a McFadden’s Pseudo R2 = 0.3120. These suggest

the model adequately predicts sexual identity.

Having developed an adequate prediction model using the donor survey, I pool the donor and recipient surveys. I use the mi package in Stata 16 to create the imputed datasets (StataCorp. 2019a). Stata imputes sexual identity values for cases in the recipient survey using the model developed from the complete donor survey. Because the model does not perfectly predict sexual identity (McFadden’s Pseudo R2 = 0.3120), multiple

imputations are created to account for prediction error. Following the suggested best

practice, I create twenty imputed datasets (StataCorp. 2019b:5). The result is twenty

complete datasets with predicted sexual identity values for all cases in the recipient

survey. After the final datasets are imputed, I drop the NHIS cases for analysis and

preserve the complete ACS with the imputed sexual orientation variable. I perform all

analyses on each of the twenty imputations and the results are then pooled to create the

presented estimates All analyses are weighted using the ACS survey weights with

adjustments to account for coverage error.

53 Table 2-7. Weighted Multilogit Prediction Model of Sexual Orientation Lesbian/Gay Bisexual Independent Variable Coefficient Std. Error Coefficient Std. Error Age 0.0975 0.0003 -0.0293 0.0003 2 Age -0.0010 0.0000 -0.0001 0.0000 Citizen 0.1417 0.0046 -0.0088 0.0044 Dad Present -0.3676 0.0029 0.0694 0.0029 Earnings 0.0000 0.0000 0.0000 0.0000 Age of Eldest Child -0.0342 0.0002 -0.0166 0.0002 Family Size -0.0651 0.0010 0.0267 0.0009 Got Social Security 0.1214 0.0027 0.0813 0.0033 Got SSI -0.2790 0.0037 0.0609 0.0039 Got Welfare -0.0001* 0.0060 0.5700 0.0044 Has Insurance 0.0648 0.0030 -0.2948 0.0033 Has Medicaid -0.4313 0.0050 0.3292 0.0062 Has Military Insurance -0.5635 0.0050 0.0129 0.0051 Has Private Insurance 0.0228 0.0027 0.0832 0.0031 Has Public Insurance 0.5396 0.0052 0.0528 0.0066 Work-Provided Insurance 0.0040 0.0023 0.1343 0.0025 Hispanic 0.1995 0.0021 -0.2257 0.0021 Hours Worked -0.0056 0.0001 -0.0042 0.0001 Householder -0.0592 0.0018 -0.0419 0.0016 Got Interest/Dividends 0.0058 0.0018 0.2315 0.0019 Mom Present -0.2952 0.0027 -0.3826 0.0028 Number of Children -0.0067 0.0020 -0.0893 0.0016 Number of Children Under 5 -0.4565 0.0036 -0.4322 0.0023 Number of Siblings -0.0126 0.0019 -0.2234 0.0017 Survey Weight 0.0000 0.0000 0.0000 0.0000 Below Poverty -0.1092 0.0021 0.1986 0.0019 Migrant Within 1 Year 1.1998 0.0147 -1.8267 0.0323 Migrant Within 5 Years -0.9408 0.0082 -0.0540 0.0059 Sex -0.2886 0.0014 0.9059 0.0016 Spouse Present -2.9442 0.0044 -0.2087 0.0023 Has Same-Sex Partner 8.7241 0.0041 4.1722 0.0044 Veteran 0.1620 0.0030 0.1943 0.0040

Degree High School Diploma 0.2289 0.0025 0.0078 0.0023 Bachelor’s Degree 0.1360 0.0042 -0.0052* 0.0041 54 Table 2-7. (continued) Lesbian/Gay Bisexual IV Coefficient Std. Error Coefficient Std. Error Master’s/Professional Degree+ 0.4745 0.0028 0.2074 0.0028

Employment Status Unemployed -0.1729 0.0032 0.3215 0.0028 Employed -0.1175 0.0027 0.1617 0.0027

Industry Agriculture -0.8821 0.0102 -1.3418 0.0128 Mining -0.3206 0.0128 -0.3205 0.0147 Construction -0.4765 0.0050 -0.2610 0.0050 Manufacturing -0.1144 0.0040 0.0258 0.0037 Wholesale Trade -0.4361 0.0069 -0.8476 0.0084 Retail Trade 0.3306 0.0037 0.1958 0.0032 Transportation, Warehousing, Utilities 0.0789 0.0044 -0.2825 0.0051 Information 0.0128 0.0057 -0.0644 0.0057 Finance/Insurance 0.3572 0.0045 -0.2824 0.0049 Real Estate/Leasing 0.0504 0.0061 -0.2063 0.0068 Professional, Scientific, and Management, and Administrative, and Waste Management Services 0.2489 0.0038 0.1096 0.0035 Education/Healthcare 0.1082 0.0036 -0.2491 0.0033 Arts 0.0893 0.0053 0.4377 0.0043 Food Service 0.7473 0.0036 0.2484 0.0032 Other Service 0.1263 0.0046 0.1777 0.0040 Public Administration 0.4232 0.0042 0.1219 0.0045 Armed Services 0.1585 0.0130 -0.4762 0.0181

Marital Status Married, Spouse Absent -0.8775 0.0063 0.6113 0.0062 Separated -1.3352 0.0064 1.0297 0.0043 Divorced -1.4054 0.0044 0.6677 0.0030 Widowed -2.0166 0.0061 0.0728 0.0058 Single, Never Married 0.0875 0.0041 0.7716 0.0027

55 Table 2-7. (continued) Lesbian/Gay Bisexual IV Coefficient Std. Error Coefficient Std. Error Home Ownership Rents Home 0.1442 0.0016 0.3952 0.0017 Other 0.1677 0.0037 0.5541 0.0035

Race Black/African American -0.0656 0.0019 -0.4737 0.0022 American Indiana/Alaska Native -0.0020* 0.0055 -0.4665 0.0062 Chinese -0.1101 0.0063 -1.1335 0.0092 Filipino 0.4959 0.0052 -0.6267 0.0080 Asian Indian -1.0498 0.0109 -0.6462 0.0081 Other Asian -0.4257 0.0061 -0.1881 0.0050 Other/Multi-Racial 0.6189 0.0072 0.8633 0.0064

Region North Central/Midwest -0.0467 0.0022 0.1499 0.0022 South 0.1474 0.0019 0.0324 0.0021 West 0.2319 0.0021 0.3039 0.0021

U.S. Born Born in U.S. Territory 0.7964 0.0071 -0.3566 0.0114 Born in U.S. State or D.C. 0.4654 0.0031 0.2830 0.0033

Constant -6.0093 0.0096 -4.4803 0.0089 Log Likelihood -22,805,365 LR χ2 11,918.49 p-value <.00005 McFadden’s Pseudo R2 0.3120 * p > 0.05

56 Because standard errors vary across the twenty imputations, I calculate final

reported standard errors that account for within and between imputation variation.

Following Rubin (1987), standard errors are the square root of the total variance which is

calculated as:

= + + 2 2 2 2 𝜎𝜎�𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 𝜎𝜎�𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 𝜎𝜎�𝑤𝑤𝑤𝑤𝑤𝑤ℎ𝑖𝑖𝑖𝑖 𝜎𝜎�𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 where 𝑀𝑀

= 𝑀𝑀 2 2 ∑𝑖𝑖=1 𝜎𝜎�𝑖𝑖 𝜎𝜎�𝑤𝑤𝑤𝑤𝑤𝑤ℎ𝑖𝑖𝑖𝑖 and 𝑀𝑀

( ) = 𝑀𝑀 1 2 2 ∑𝑖𝑖=1 𝛽𝛽𝑖𝑖 − 𝛽𝛽̅ 𝜎𝜎�𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 and where M is the number of imputations and ß𝑀𝑀 is− the parameter estimate. Weighted

sample standard errors are calculated using Stata’s svy package (StataCorp 2019a)

through successive difference replication (Fay and Train 1995).

Results

National Profile

LGB people make up 4.07 (se = 0.0060) percent of the adult, non-institutionalized population. Table 2-8 shows weighted population percentages and sample sizes by sexual identity groups. Bisexual women make up the largest share of the LGB population at

33.38 percent (se = 0.1414), followed by gay men at 29.65 percent (se = 0.1143), lesbian women at 23.30 percent (se = 0.1154), and bisexual men make up the smallest share with just 13.67 percent (se = 0.1256). Women are more likely to identify as LGB than men.

57 4.44 percent (se = 0.0122) of women identify as lesbian or bisexual compared to 3.67 percent (se = 0.0130) of men who identify as gay or bisexual.

Demographic Characteristics

Table 2-9 summarizes demographic characteristics by sex and sexual identity. On average LGB people are younger than their heterosexual counterparts, 39.44 years (se =

16.5332) to 47.81 years (se = 18.2179), respectively. Among LGB people, bisexual people are, on average, more than 10 years younger than lesbian and gay people. The heterosexual population is less racially diverse than LGB populations. 64.44 percent (se =

0.0145) of heterosexuals identify as white, non-Hispanic compared to 60.17 percent (se =

0.1607) of the LGB population. 19.46 percent (se = 0.1387) of LGB people identify as

Hispanic compared to 15.67 percent (se = 0.0111) of heterosexual people, with bisexual men and women identifying as Hispanic more frequently than other sexual identity groups. While more than half of heterosexuals are currently married (52.48 percent, se =

0.0160), just over a quarter of LGB people are currently married (26.10 percent, se =

0.1886). Bisexual men and women have the lowest rates of marriage for any sexual identity group, likely due to their being much younger on average. LGB people are raising children (18.68 percent, se = 0.1529) at lower rates than their heterosexual counterparts (38.23 percent, se = 0.0150) while bisexual women are the most likely among LGB people to be raising children and gay men are the least likely.

58

Table 2-8. Population Estimates by Sex and Sexual Identity Women Sexual Identity Percent of Total SE Percent of Women SE Percent of LGB SE N Lesbian Women 0.95% 0.0044 1.83% 0.0085 23.30% 0.1154 91,548 Bisexual Women 1.36% 0.0058 2.61% 0.0109 33.38% 0.1414 76,634 Total LGB Women 2.31% 0.0065 4.44% 0.0122 56.68% 0.1488 168,182 Heterosexual Women 49.66% 0.0148 95.56% 0.0122 6,029,065 Total Women 51.97% 0.0150 6,197,246

Men

59 Sexual Identity Percent of Total SE Percent of Men SE Percent of LGB SE N Gay Men 1.21% 0.0047 2.51% 0.0098 29.65% 0.1143 111,205 Bisexual Men 0.56% 0.0049 1.16% 0.0102 13.67% 0.1256 30,171 Total LGB Men 1.76% 0.0064 3.67% 0.0130 43.32% 0.1488 141,376 Heterosexual Men 46.27% 0.0149 96.33% 0.0130 5,469,532 Total Men 48.03% 0.0150 5,610,907

Total Sexual Identity Percent of Total SE N LGB 4.07% 0.0060 309,557 Heterosexual 95.93% 0.0060 11,498,596

Table 2-9. Demographic Characteristics by Sexual Identity and Sex Total

Demographic Variables Heterosexual LGB Average Age 47.81 39.44 (18.2179) (16.5332)

White, non-Hispanic 64.44% 60.17% (0.0145) (0.1607)

Black 11.91% 11.72% (0.0098) (0.1011)

Other Race 13.22% 18.27% (0.0105) (0.1401)

Hispanic 15.67% 19.46% (0.0111) (0.1387)

Married 52.48% 26.10% (0.0160) (0.1886)

Has Child 38.23% 18.68% (0.0150) (0.1529)

Education, BA+ 29.57% 30.00% (0.0136) (0.1115)

60

Table 2-9. (continued) Women Heterosexual Demographic Variables Women Lesbian Women Bisexual Women Average Age 48.61 44.89 34.56 (18.4864) (17.0089) (15.0227)

White, non-Hispanic 63.97% 61.99% 58.52% (0.0202) (0.2757) (0.3595)

Black 12.61% 14.53% 10.82% (0.0139) (0.1993) (0.2369)

Other Race 13.18% 14.17% 21.22% (0.0144) (0.1748) (0.2689)

Hispanic 15.23% 16.42% 20.65% (0.0151) (0.2052) (0.2452)

Married 50.47% 33.01% 21.43% (0.0217) (0.2749) (0.3097)

Has Child 41.85% 21.33% 27.88% (0.0210) (0.2128) (0.3274)

Education, BA+ 30.14% 36.80% 25.06% (0.0190) (0.2337) (0.1936)

61

Table 2-9. (continued) Men

Demographic Variables Heterosexual Men Gay Men Bisexual Men Average Age 46.95 43.38 33.52 (17.8851) (16.2301) (14.3748)

White, non-Hispanic 64.94% 61.26% 58.74% (0.0208) (0.2171) (0.5708)

Black 11.17% 11.61% 9.40% (0.0139) (0.1746) (0.2861)

Other Race 13.26% 16.34% 22.27% (0.0152) (0.2182) (0.4971)

Hispanic 16.14% 19.43% 21.76% (0.0162) (0.1629) (0.4960)

Married 54.64% 29.20% 19.02% (0.0226) (0.2471) (0.4930)

Has Child 34.35% 8.42% 13.95% (0.0208) (0.1081) (0.4026)

Education, BA+ 28.96% 34.17% 21.45% (0.0198) (0.2392) (0.4512)

62

Overall, LGB people tend to be slightly more highly educated than their heterosexual counterparts with 30.00 (se = 0.1115) percent of LGB people holding at least a bachelor’s degree (BA) compared to 29.57 (se = 0.0136) percent of heterosexual people. This difference between the full LGB population and the heterosexual population, while slight, is statistically significant, though there is considerable variation within LGB populations. Lesbian women are the most highly educated sexual identity group followed by gay men. Bisexual women and men tend to have the lowest educational attainment, compared to both their lesbian/gay and heterosexual counterparts, likely due to the dramatic age difference across the groups.

Economic Characteristics

Table 2-10 summarizes economic characteristics by sex and sexual identity. LGB people are more likely to be in the labor market where they experience higher rates of unemployment than heterosexual people. LGB people are 6.68 percentage points more likely to be in the labor force. This is due in great part to the much higher rates of labor market non-participation by heterosexual women, of whom 40.25 percent (se = 0.0204) are not in the labor force. For those in the labor force, 5.80 percent (se = 0.0755) of LGB people are unemployed compared to 3.61 percent (se = 0.0057) of heterosexual people.

Among LGB people, unemployment is lowest for lesbian women (3.48 percent, se =

0.1059) and highest for bisexual men (8.89 percent, se = 0.2863). The median lesbian full-time, year-round worker—those working at least thirty-five hours per week, fifty to fifty-two weeks a year, per Census labor force definitions—out earns her heterosexual counterpart $45,000 to $41,000 while the median gay man full-time, year-round worker earns less than the median heterosexual man full-time, year-round worker, $48,700 and

63

$50,000, respectively. Bisexual men and women earn considerably less than their

lesbian/gay and heterosexual counterparts, earning a median $37,000 and $33,500

respectively for full-time, year-round workers, which is again likely due considerably to

their age difference. Poverty rates among all categories of LGB people out-pace heterosexual men and women with dramatic rates of poverty for bisexual men and women. Rates of having health insurance follow a similar pattern to most demographic categories. LGB people are on average less likely to have health insurance than heterosexual people, 86.61 percent (se = 0.1069) to 89.34 percent (se = 0.0094), respectively. However, lesbian women and gay men have higher rates than bisexual women and men.

State-Level Profiles

LGB populations vary dramatically from state to state. Across the fifty states and the District of Columbia, the proportion of state populations identifying as LGB ranges from a high of 7.81 percent (se = 0.1698) in the District of Columbia to a low of 2.36 percent (se = 0.0899) in North Dakota. A note of caution: these estimates may underestimate state-by-state variation due to the lack of state information in the publicly available NHIS data. My imputation model relies on regional variations which may mask

larger state-by-state variations. I discuss this issue further below. Table 2-11 summarizes

state population estimates and standard errors by sexual identity. See Appendix A for

state sample sizes.

64

Table 2-10. Economic Characteristics by Sex and Sexual Identity Total Economic Variables Heterosexual LGB Not in Labor Force 34.33% 27.65% (0.0142) (0.1215)

Unemployed 3.61% 5.80% (0.0057) (0.0755)

Full-Time, Year-Round Worker 44.31% 43.54% (0.0149) (0.134)

Median Earnings $45,500 $41,100

Poverty Rate 12.03% 19.62% (0.0101) (0.1258)

Has Insurance 89.34% 86.61% (0.0094) (0.1069)

65

Table 2-10. (continued) Women Heterosexual Lesbian Bisexual Economic Variables Women Women Women Not in Labor Force 40.25% 30.07% 28.59% (0.0204) (0.24) (0.3323)

Unemployed 3.22% 3.48% 7.39% (0.0075) (0.1059) (0.142)

Full-Time, Year-Round Worker 36.62% 46.14% 35.21% (0.0198) (0.2654) (0.2841)

Median Earnings $41,000 $45,000 $33,500

Poverty Rate 13.63% 15.57% 27.89% (0.0146) (0.1883) (0.2535)

Has Insurance 90.71% 91.00% 85.70% (0.0123) (0.151) (0.2081)

66

Table 2-10. (continued) Men Heterosexual Economic Variables Men Gay Men Bisexual Men Not in Labor Force 27.98% 26.81% 23.04% (0.0195) (0.1917) (0.3759)

Unemployed 4.04% 4.41% 8.89% (0.0087) (0.0882) (0.2863)

Full-Time, Year-Round Worker 52.56% 50.65% 44.08% (0.0219) (0.2706) (0.5561)

Median Earnings $50,000 $48,700 $37,000

Poverty Rate 10.31% 12.91% 20.87% (0.0134) (0.1648) (0.382)

Has Insurance 87.87% 86.97% 80.60% (0.0143) (0.1338) (0.4814)

67

Table 2-11. Percent of State Population by Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 46.01% 51.32% 0.75% 0.67% 0.35% 0.91% 2.67% (0.1200) (0.1200) (0.0283) (0.0258) (0.0258) (0.0308) (0.0385) Alaska 49.01% 47.65% 1.02% 0.75% 0.51% 1.06% 3.34% (0.3510) (0.3501) (0.0954) (0.0879) (0.0788) (0.0923) (0.1216) Arizona 46.65% 49.28% 1.22% 0.93% 0.58% 1.34% 4.07% (0.1034) (0.1035) (0.0261) (0.0236) (0.0317) (0.0365) (0.0399) Arkansas 46.71% 50.44% 0.83% 0.68% 0.38% 0.96% 2.85% (0.1533) (0.1537) (0.0342) (0.0322) (0.0369) (0.0437) (0.0502) California 46.47% 48.78% 1.49% 1.02% 0.66% 1.57% 4.75% 68 (0.0443) (0.0441) (0.0146) (0.0129) (0.0135) (0.0155) (0.0182)

Colorado 47.38% 48.39% 1.25% 0.99% 0.60% 1.39% 4.23% (0.1140) (0.1141) (0.0304) (0.0281) (0.0303) (0.0356) (0.0451) Connecticut 46.24% 50.21% 1.02% 0.87% 0.48% 1.19% 3.56% (0.1384) (0.1390) (0.0347) (0.0336) (0.0434) (0.0494) (0.0506) Delaware 45.46% 50.46% 1.15% 1.01% 0.56% 1.36% 4.08% (0.2708) (0.2722) (0.0719) (0.0701) (0.0903) (0.1024) (0.1061) District of Columbia 41.90% 50.29% 2.71% 1.43% 1.01% 2.67% 7.81% (0.3187) (0.3225) (0.1133) (0.0873) (0.1759) (0.1931) (0.1698) Florida 45.76% 50.10% 1.29% 0.90% 0.56% 1.38% 4.14% (0.0580) (0.0582) (0.0167) (0.0147) (0.0169) (0.0199) (0.0228) Georgia 45.30% 50.73% 1.15% 0.95% 0.51% 1.35% 3.96% (0.0847) (0.0852) (0.0223) (0.0211) (0.0310) (0.0348) (0.0324) Hawaii 46.37% 49.58% 1.25% 0.90% 0.56% 1.34% 4.05% (0.2248) (0.2254) (0.0566) (0.0503) (0.0476) (0.0609) (0.0869)

Table 2-11. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 48.21% 49.30% 0.73% 0.59% 0.35% 0.82% 2.49% (0.2121) (0.2120) (0.0449) (0.0418) (0.0410) (0.0489) (0.0647) Illinois 46.30% 49.79% 1.21% 0.86% 0.54% 1.30% 3.91% (0.0739) (0.0740) (0.0201) (0.0178) (0.0208) (0.0242) (0.0282) Indiana 46.48% 49.45% 1.15% 1.01% 0.55% 1.37% 4.07% (0.1022) (0.1022) (0.0305) (0.0286) (0.0273) (0.0333) (0.0400) Iowa 47.53% 49.10% 0.97% 0.81% 0.47% 1.12% 3.37% (0.1509) (0.1509) (0.0393) (0.0376) (0.0425) (0.0480) (0.0528) Kansas 47.42% 49.62% 0.86% 0.71% 0.42% 0.98% 2.97% 69 (0.1591) (0.1593) (0.0394) (0.0379) (0.0435) (0.0493) (0.0523)

Kentucky 46.83% 50.17% 0.83% 0.76% 0.40% 1.01% 3.01% (0.1252) (0.1254) (0.0284) (0.0270) (0.0305) (0.0361) (0.0420) Louisiana 45.73% 50.81% 0.99% 0.84% 0.46% 1.17% 3.46% (0.1265) (0.1268) (0.0336) (0.0326) (0.0354) (0.0407) (0.0453) Maine 46.21% 49.29% 1.13% 1.26% 0.56% 1.56% 4.51% (0.2290) (0.2295) (0.0604) (0.0622) (0.0777) (0.0896) (0.0932) Maryland 45.46% 50.71% 1.06% 0.97% 0.51% 1.29% 3.83% (0.1083) (0.1088) (0.0326) (0.0320) (0.0251) (0.0306) (0.0409) Massachusetts 45.46% 49.57% 1.39% 1.24% 0.68% 1.66% 4.97% (0.0999) (0.1001) (0.0278) (0.0266) (0.0359) (0.0406) (0.0432) Michigan 46.63% 49.69% 1.06% 0.89% 0.50% 1.23% 3.67% (0.0828) (0.0830) (0.0221) (0.0208) (0.0229) (0.0267) (0.0308) Minnesota 47.53% 48.79% 1.05% 0.90% 0.52% 1.21% 3.68% (0.1135) (0.1135) (0.0325) (0.0311) (0.0361) (0.0403) (0.0419)

Table 2-11. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 45.53% 51.39% 0.87% 0.76% 0.40% 1.05% 3.08% (0.1546) (0.1551) (0.0419) (0.0404) (0.0389) (0.0459) (0.0531) Missouri 46.45% 50.16% 0.98% 0.81% 0.46% 1.14% 3.40% (0.1064) (0.1066) (0.0245) (0.0227) (0.0262) (0.0311) (0.0380) Montana 48.45% 48.89% 0.74% 0.67% 0.38% 0.87% 2.66% (0.2674) (0.2678) (0.0620) (0.0608) (0.0555) (0.0658) (0.0829) Nebraska 47.60% 48.90% 0.99% 0.86% 0.47% 1.17% 3.50% (0.1915) (0.1915) (0.0557) (0.0538) (0.0424) (0.0529) (0.0698) Nevada 46.90% 48.03% 1.68% 1.01% 0.73% 1.65% 5.07% 70 (0.1585) (0.1588) (0.0496) (0.0441) (0.0560) (0.0626) (0.0675)

New Hampshire 46.91% 48.69% 1.16% 1.17% 0.64% 1.42% 4.40% (0.2204) (0.2209) (0.0523) (0.0530) (0.0630) (0.0740) (0.0901) New Jersey 46.05% 50.19% 1.12% 0.87% 0.52% 1.25% 3.75% (0.0873) (0.0875) (0.0220) (0.0200) (0.0253) (0.0293) (0.0331) New Mexico 46.51% 49.51% 1.13% 0.98% 0.56% 1.31% 3.99% (0.1911) (0.1916) (0.0465) (0.0438) (0.0553) (0.0645) (0.0733) New York 45.25% 50.04% 1.44% 1.05% 0.63% 1.58% 4.71% (0.0596) (0.0597) (0.0197) (0.0182) (0.0136) (0.0174) (0.0248) North Carolina 45.58% 50.84% 0.99% 0.90% 0.47% 1.22% 3.57% (0.0833) (0.0836) (0.0194) (0.0190) (0.0250) (0.0287) (0.0305) North Dakota 49.61% 48.03% 0.77% 0.48% 0.35% 0.76% 2.36% (0.3016) (0.3020) (0.0653) (0.0589) (0.0645) (0.0749) (0.0899) Ohio 46.26% 49.79% 1.13% 0.97% 0.53% 1.33% 3.95% (0.0758) (0.0757) (0.0209) (0.0191) (0.0253) (0.0283) (0.0298)

Table 2-11. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 46.77% 49.89% 0.96% 0.81% 0.45% 1.12% 3.34% (0.1382) (0.1387) (0.0314) (0.0296) (0.0337) (0.0410) (0.0494) Oregon 46.46% 48.41% 1.42% 1.30% 0.69% 1.72% 5.13% (0.1291) (0.1292) (0.0404) (0.0382) (0.0326) (0.0411) (0.0562) Pennsylvania 46.30% 49.95% 1.07% 0.92% 0.51% 1.26% 3.75% (0.0734) (0.0738) (0.0243) (0.0242) (0.0178) (0.0215) (0.0276) Rhode Island 45.89% 50.01% 1.19% 0.99% 0.55% 1.38% 4.11% (0.2556) (0.2565) (0.0728) (0.0692) (0.0716) (0.0849) (0.0998) South Carolina 45.70% 51.20% 0.89% 0.75% 0.41% 1.05% 3.10% 71 (0.1182) (0.1186) (0.0270) (0.0257) (0.0239) (0.0305) (0.0409)

South Dakota 48.48% 48.83% 0.72% 0.70% 0.37% 0.90% 2.69% (0.2893) (0.2895) (0.0665) (0.0658) (0.0630) (0.0754) (0.0919) Tennessee 46.23% 50.71% 0.88% 0.74% 0.41% 1.03% 3.06% (0.1013) (0.1017) (0.0261) (0.0248) (0.0199) (0.0254) (0.0348) Texas 46.79% 49.56% 1.09% 0.84% 0.50% 1.21% 3.65% (0.0519) (0.0521) (0.0143) (0.0132) (0.0142) (0.0168) (0.0193) Utah 48.01% 48.54% 1.04% 0.78% 0.48% 1.14% 3.45% (0.1598) (0.1599) (0.0388) (0.0352) (0.0384) (0.0458) (0.0574) Vermont 46.57% 48.66% 1.03% 1.49% 0.61% 1.64% 4.77% (0.3231) (0.3237) (0.0785) (0.0896) (0.1054) (0.1229) (0.1363) Virginia 45.97% 50.51% 1.02% 0.84% 0.47% 1.18% 3.51% (0.0905) (0.0908) (0.0214) (0.0203) (0.0219) (0.0269) (0.0333) Washington 46.90% 48.36% 1.40% 1.11% 0.66% 1.57% 4.74% (0.0968) (0.0967) (0.0286) (0.0268) (0.0243) (0.0303) (0.0408)

Table 2-11. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 46.85% 49.52% 1.07% 0.85% 0.49% 1.22% 3.63% (0.1924) (0.1929) (0.0514) (0.0486) (0.0557) (0.0642) (0.0711) Wisconsin 47.44% 49.07% 1.00% 0.85% 0.48% 1.16% 3.49% (0.1089) (0.1088) (0.0283) (0.0266) (0.0332) (0.0377) (0.0391) Wyoming 49.08% 47.86% 0.87% 0.75% 0.47% 0.97% 3.06% (0.3445) (0.3442) (0.0799) (0.0764) (0.0674) (0.0830) (0.1177)

72

Demographic and economic characteristics also vary considerably by state,

though broad differences mostly follow state-by-state differences in overall rates of

characteristics. Full state-by-state demographic and economic characteristics are

presented in Appendix B. While the NCHS Data Suppression Workgroup recommends

suppressing estimates from sample sizes smaller than thirty cases, I withhold any

estimates from samples smaller than fifty cases to account for the additional error

introduced through CSMI. Because state-by-state comparisons of LGB people are likely affected by state-level differences in outcomes (e.g., LGB people likely have higher poverty rates in states with overall higher poverty rates), Appendix C shows characteristics for the full populations of the states.

Hawaii has the most racially diverse LGB population with only 23.86 percent (se

= 0.0156) identifying as white, non-Hispanic—Hawaii is the most diverse state with only just over twenty-three percent of the total state population identifying as white, non-

Hispanic. Vermont has the most racially homogenous LGB population with 92.56 (se =

0.0193) percent identifying as white, non-Hispanic—Vermont is also one of the least

diverse states with more than ninety-four percent of the total state population identifying

as white, non-Hispanic. Marriage rates vary between a high of 36.17 percent (se =

0.0271) of LGB people in Vermont (which in 2000 became the first state to recognize

same-sex civil unions) to a low of 21.29 percent (se = 0.0094) in Louisiana. The

percentage of LGB people raising their own children ranges from a high of 23.01 percent

(se = 0.0136) in Mississippi to a low of 10.41 percent (se = 0.0141) in the District of

Columbia. The District of Columbia has the most highly educated LGB population with

65.61 percent (se = 0.0194) holding at least a BA. Mississippi and Idaho have the lowest

73

educational attainment with just under twenty percent of LGB people in both states holding at least a BA.

Unemployment ranges from a high of 7.29 percent (se = 0.0094) in Mississippi to a low of 3.81 percent (se = 0.0050) in Minnesota, though small sample sizes make conclusive comparisons across all geographies difficult. Median annual wages for full- time, year-round workers vary from a low of $31,600 in Arkansas to a high of $74,300 in the District of Columbia. The highest poverty rate among LGB people is in West Virginia where 28.66 percent (se = 0.0182) of LGB people are living in poverty compared to

16.05 percent (se = 0.0014) of heterosexual West Virginians, the largest poverty gap of any state. New Hampshire boasts the lowest percentage of LGB people living in poverty with a rate of 12.78 percent (se = 0.0138) while the District of Columbia has the narrowest poverty gap with only 1.36 percentage points separating LGB people from heterosexual people.

Discussion

The profiles I produce through CSMI of the NHIS and the ACS are in line with existing estimates of LGB populations. Gates (2014), comparing estimates from multiple surveys, shows a range of 1.70-5.60 percent for estimates of LGBT populations and finds that an average of 3.50 percent identify as LGB. More recently, Newport (2018), using data from Gallup’s tracking poll, estimates that 4.50 percent identify as LGBT. The 4.07 percent estimate I produce of LGB populations is consistent with these existing estimates.

The demographic and economic profiles are also consistent with existing estimates.

My estimates are consistent with Gates’s (2014) finding that women are more likely to identify as LGB, especially among bisexual people. In Gates’s analysis, “women

74

represented a substantial majority (from 68% to 77%) of bisexuals while they were a minority (from 41% to 49%) among those who identified as lesbian or gay” (5). In my analyses, women make up 70.94 percent of bisexuals and only 44.01 percent of lesbian and gay people. The gender discrepancy in identity, especially among bisexuals, may be a result of socio-cultural factors of gender expectations. Ward (2015) examines the ways in which gender influences social perceptions of sexual fluidities, which differently affect the identity potentialities of men and women. While sexual fluidity is given a degree of freedom from identity proscription among women, at least for a time, men “must manage their sexual fluidity within the context of a culture that they know will immediately equate male homosexual behavior with gay subjectivity” (20). These cultural expectations likely steer sexually fluid men towards a gay identity while precluding many from realizing a bisexual identity.

Many of the differences between bisexual men and women on several characteristics may be a result of the dramatic age difference between bisexuals and lesbian/gay and heterosexual people. Bisexual men and women are, on average, more than a decade younger than other sexual identity groups. This likely explains differences in other characteristics that develop during the life course, such as education, marriage, and employment. There are several possible explanations for the higher rates of bisexual identity among the young. One possibility is that younger people are more fluid in both their sexual practices and identity. According to Kaestle (2019), who analyzes Add

Health data, “young adulthood is still a very dynamic time for sexual orientation development” (821). Similarly, Russell et al. (2009) find more than sixty-five percent of sexual minority adolescents identify as bisexual, queer, questioning, or other labels such

75

as pansexual. While young people may be more fluid with their identity labels, I caution

a conclusion that interprets bisexual identity as a phase or passing identity on a trajectory

towards a concrete gay, lesbian, or heterosexual identity. It is possible that a more

welcoming climate for sexual minorities allows for younger people to be more expressive

and experimental with their identities. It is also possible that older bisexuals, especially

those who are partnered, may transition their identity label to fit perceptions of the gender

makeup of their relationships. This strategy could be an attempt to protect against anti-

bisexual animus from both heterosexual and lesbian/gay people (Ochs 1996)

My estimates are also in-line with Gates’s (2014) finding that, across the surveys he compares, LGB people are between sixty to sixty-nine percent white, non-Hispanic. I

find that 60.17 (se = 0.0016) percent of LGB people are white, non-Hispanic. The

surveys compared by Gates (2014) showed inconsistent findings when it comes to

educational attainment with some surveys showing a correlation between higher

educational attainment and greater likelihood of identifying as LGB and other surveys

showing no relationship. There was also variation when restricting the analysis by age

group. My findings show a weak, though significant, relationship between educational

attainment and sexual orientation with a 0.43 percentage point difference between LGB

and heterosexual people holding at least a BA.

State-level estimates are more difficult to compare to existing estimates for

consistency as few data sources are available at the state level. The Williams Institute,

using data from the Gallup Daily tracking survey provides some state estimates, however

these estimates do not disaggregate LGB and transgender respondents making specific

comparisons with my estimates unreliable (Williams Institute 2019). Combined with

76

additional state-by state estimates of the transgender population (Flores et al. 2016), I can make some top-line comparisons. In Appendix D, I subtract Flores et al.’s (2016) state estimates of the transgender population from the Williams Institute’s state estimates of

LGBT populations. I then compare the resulting figures to my CSMI estimates. I find that my state-level estimates are, on average, within 0.14 percentage points of their figures

(see Appendix D).

Limitations

The NHIS does not publicly provide information about a respondent’s state of residence, only Census Bureau-designated regions. The ACS does include state data, which I recoded into regions for the imputation. To then analyze state-level statistics is potentially a violation of the CSMI requirement that variables never jointly observed be excluded. While this is not ideal, I contend that the inclusion of the region variable mitigates some of the bias introduced. Because the ACS has a much larger sample size than the NHIS, further reduction of bias is possible. As Rendall et al. (2013) note, “the adverse effects of any sampling biases in the smaller sample survey may also be mitigated by anchoring estimates from the richer covariance structure of the smaller data source to the more population-representative larger data source” (486). Future analyses should attempt to access the restricted state data from the NHIS to further corroborate the estimates herein.

The data used here covers the period of 2014-2018. Therefore, I am unable to assess the impact of the 2020 pandemic and the resulting economic recession on LGB people. However, the higher rates of poverty and unemployment observed here suggest that the pandemic and recession may be particularly onerous for LGB people, especially

77

bisexuals. The lower rates of health insurance coverage among LGB people suggests an additional layer of risk. Further, for those who do have health insurance, it is not assured they have access to safe and affirming healthcare providers. Future analyses with more recent data should assess the potential differential effects of these phenomena.

Conclusion

Nearly fifty years after LGBT activists like Bruce Voeller first produced the

“Myth of 10 Percent,” the political and social climate for sexual minorities has changed dramatically. No longer do activists struggle to prove the very existence of LGBT people in the United States. Indeed, that endeavor has been so successful that public perceptions seemingly overestimate the size of LGBT populations four to five times. But the goals and priorities of social movements are different from other spheres of the social world.

While the ten percent figure has permeated the culture, contemporary public policy requires accurate, scientific estimates of sexual minority populations. My analyses suggest that today, just over four percent of the population identifies as LGB. While smaller than the ten percent figure, my estimates, which comport with other contemporary estimates, better reflect the reality of sexual identity in the United States.

My estimates also show notable differences among and between sexual identity groups. Differences in patterns of identity illustrate the continued relevance of gender in contemporary society. Higher rates of identification as sexual minorities among women suggests that gender and sexuality continue to elide in “patterned fluidities” (Richardson

2007). Other demographic differences suggest that my estimates can inform numerous areas of sociological interest including race/ethnicity, education, kinship, poverty, and

78

health, among others. My profiles invite further intersectional inquiries into a number of

fields within the discipline.

The results of this chapter suggest that CSMI is a useful tool for generating

estimates of LGB populations, especially when making state-by-state comparisons. The consistency of the imputed dataset with existing survey results suggests that the national and state profiles I construct are reasonable estimates of LGB populations. CSMI allows me to remedy the “Identity Undercount” and bring sexual orientation information into survey data used for public policy. The success of CSMI in the context of sexual orientation, alongside its use in migration studies, suggest that the method may have efficacy for the study of other small or hard to sample populations.

My estimates help produce a more complete picture of the sociological and demographic state of sexuality in the United States. I now move beyond descriptive demographic profiles and apply the data to the study of inequality. While my profiles present a number of stark differences between and among sexual identity groups, the data are also useful for examining the social mechanisms behind these differences. In the next chapter, I use the imputed dataset to analyze the economic outcomes of LGB people in comparison to their heterosexual counterparts to measure the prevalence of labor market discrimination.

79

CHAPTER 3

LABOR MARKET DISCRIMINATION AGAINST SEXUAL MINORITIES

Introduction

A photo atop a New York Times article shows a smiling Charmaine McGuffey,

clad in a bright pink shirt, meeting with a group of supporters, including one wearing

“Charmaine McGuffey for Hamilton County Sheriff” t-shirts (Cramer 2020). McGuffey

had recently won election to that office, a poetic development in a saga that included the

just ousted sheriff having fired her, a former major in the department, for being a lesbian.

While McGuffey’s story has a satisfying conclusion, it also highlights an all-too common

reality for sexual minorities: workplace discrimination. LGB people are often fired, not

hired, denied promotion, and given lower compensation than heterosexual workers.5

Until recently, these practices have been completely legal in many jurisdictions.

Many LGB people tell similar stories of workplace discrimination. Much of our

knowledge of workplace discrimination against LGB people comes from surveys which

assess individuals’ perceptions of workplace outcomes or from qualitative studies in

which workers recount their individual experiences of discrimination. The Williams

Institute, an LGBT think tank, conducted a review of several studies dating back to the

1980s. Based on their review, previous studies report as many as seventeen percent of

LGB respondents claim they were fired or denied a position because of their sexual

5 There is also significant evidence of labor market discrimination against trans-identified workers as well. Indeed, Burns and Krehely (2011) find “a staggering 90 percent of transgender workers report some form of harassment or mistreatment on the job” (1). While discrimination against trans and gender diverse workers warrants attention, my data prevent explicit analysis of gender identity (see Chapter Two). 80

identity; as many as twenty-eight percent claim they had been denied promotion or were

given a poor performance evaluation; as many as forty-one percent reported having

experienced verbal or physical abuse; and as many as nineteen percent report being given

unequal pay or benefits (Badgett et al. 2007). These studies paint a picture of broad and

pervasive discrimination against LGB workers.

Any LGB person entering the labor market is at risk of experiencing sexual

orientation discrimination. But how might we measure the economic impact that such

discrimination has on sexual minorities as populations? That is, what effect does

discrimination have on the economic opportunities and rewards that LGB people receive?

To answer these questions, we can compare the labor market experiences of LGB people

to heterosexual people. By controlling for differences in characteristics that affect our

labor market experiences—like levels of education, years of experience, occupation, and

others—we can assess the degree to which different labor market outcomes are likely a

result of discrimination. Showing that LGB people, even when they have similar

education, experience, and occupations, receive fewer economic rewards and

opportunities suggests discrimination is occurring. And such evidence supports

arguments for public policy interventions, such as passing nondiscrimination policies.

Part of the difficulty with assessing the impact of anti-LGB discrimination in the

labor market is the lack of available data to compare economic outcomes by sexual

orientation. As discussed in Chapter Two, the existing data on sexual orientation has been

limited to small-sample or household-level surveys that preclude certain analyses and

makes conclusions unreliable. Given that none of the large-sample federal surveys typically used for state and federal assessments of economic outcomes collect explicit

81

sexual orientation data, existing research has had to rely on less-than-ideal sources. There is a need for research which can corroborate or expand the findings of these existing studies.

My Contributions

Using the dataset developed in Chapter Two of this dissertation, I analyze the economic experiences of sexual minorities in the labor market using a large-sample,

nationally representative survey with robust economic indicators. My analyses contribute

to and expand upon the existing research on LGB labor market outcomes. I use my novel

dataset to provide more detailed and current assessments of LGB labor market

experiences. My work has clear implications for future research and informs

contemporary policy debates. These analyses help to create a more complete picture of

labor market inequality broadly and for LGB peoples specifically. They provide evidence

of the prevalence of inequality and discrimination experienced by sexual minorities.

In this chapter, I test several hypotheses about the relationship between sexual

orientation and economic outcomes. To begin with, I run multivariate regression analyses

in nested models to determine whether there is evidence of wage disparities between

sexual identity groups and what factors might account for these disparities. I compare

differences in several factors that might affect wages. These factors, called productivity

characteristics, include demographic differences (such as age and education), family

structure (such as marital status and child rearing), and labor market structure (such as

occupation and part-time status). I then perform wage decompositions to estimate the

impact of discrimination on any observed wage disparities. Wage decompositions allow

me to divide an observed wage difference between two groups into two components: that

82

which is explainable by differences in productivity characteristics and a residual, or

unexplained component, which is interpreted as a measure of discrimination. My

analyses complement and extend the existing sociological and economic literature on

labor market discrimination. My use of the CSMI dataset developed in Chapter Two

further demonstrates the method’s utility for studying small or hard to sample

populations.

Literature Review

There is a large literature in sociology and economics on the effects of

discrimination. Much of the early work in this field understandably focused on the impact

discrimination has had on racial minorities and women in the labor market. More

recently, a growing body of research has used the tools developed by race and gender

scholars and applied them to the analysis of other groups, including sexual minorities.

Here I briefly review some empirical approaches to measuring discrimination before

turning to the literature on sexual orientation discrimination.

Empirical Approaches to the Study of Discrimination

The study of discrimination fits within the broader literature on inequality. From a

sociological perspective, discrimination is the “differential or unequal treatment of the

members of some group or category on the basis of their group membership rather than on the basis of their individual qualities” (Levin and Levin 1982:51). Whether the bias is explicit or implicit, its effects can be observed by comparing the individual and aggregate experiences of group members. Discrimination is often evaluated, especially in public policy, using two standards: disparate treatment and disparate impact.

83

Disparate treatment includes any discrimination that results from treating

individual members of a group differently from others because of their membership in

that group. Protections against disparate treatment based on race, color, religion, sex, or

national origin are included in many federal laws, particularly as regards the workplace,

housing, and public accommodations. “The core concept of disparate treatment

discrimination emanates from the constitutional requirement of equal protection under the

law and is codified in [the Civil Rights Act of 1964]” (Blank et al. 2004). Disparate

impact defines as discriminatory any policies that are neutral on their face but result in

aggregate differences based on legally protected statuses.

In the labor market, inequalities can manifest in several ways including access to

jobs and the setting of wages. Economic discrimination results from the valuing of

“personal characteristics of the worker that are unrelated to productivity” (Arrow

1973:3). Becker (1957) established the framework for the economic study of discrimination in his foundational text on the subject. Discrimination is economically irrational in the sense that those who discriminate narrow their choices in a market and hence their ability to maximize benefits. Becker thus examines discrimination from the perspective of an economic actor’s “taste for discrimination,” that is, their willingness “to pay something, either directly or in the form of reduced income, to be associated with some persons instead of others” (14). For employers, this payment could be the cost of reduced productivity, for workers reduction in wages, or for consumers an increase in price.

Much of the economic literature on discrimination focuses on wages. Wages are understood as the price of labor, determined in a marketplace of supply (workers) and

84

demand (employers/firms) (Hicks 1932). In theory, wages would be set in a freely competitive job market at the equilibrium between supply and demand. In a discriminatory market in which an employer or firm values non-productivity characteristics, wages could be affected. Members of a preferred group might experience positive discrimination and therefore higher wages while members of a disfavored group might experience negative discrimination and receive lower wages. The difference in wages between such groups is therefore interpreted as a measure of discrimination. For

Becker (1957) this difference in wages is the “market discrimination coefficient,” modeled as the percentage difference in wages assuming equal productivity (17).

Differences in the setting of wages is a form of “demand side” discrimination as wages are determined by employers. Workers refusing to accept positions in firms with members of other groups would be a form of “supply side” discrimination.

Sexual Minority Discrimination

There is a long history of discrimination against sexual minority employees in both the policies and practices of public and private employers. Discussions of workplace discrimination in the public sector, particularly at the federal level, have traditionally focused on three fields: the civil service, the granting of security clearances, and the military. Each of these areas have been governed at various times by explicitly anti-LGB policies which forced the rejection of sexual minorities or required their ejection from the labor force. A climate of LGB exclusion and expulsion dominated the domains of public service for most of the twentieth century and in some cases into the twenty first. In the private sector, discrimination has included blatant hostility of employers towards LGB workers as well as more subtle expressions of employer bias. However, as Badgett (2006)

85

notes, “not all observers agree that pervasive wage or employment discrimination occurs

against LGB people, illustrating the need for further study by economists and others”

(163). I take up that call here.

History of Sexual Minority Discrimination

A de facto ban on gay, lesbian, and bisexual workers in the Federal Civil Service

dated to its founding in the nineteenth century. Homosexuality constituted “immoral

conduct,” which was a basis for disqualification and dismissal (Kameny 2000:189).

However, it was not until the mid-twentieth century that significant efforts towards rooting out LGB employees were undertaken. In the 1940s and 1950s, at the onset of the

Cold War and McCarthyism, increased scrutiny was paid to undesirable elements within the federal government and homosexuality was singled out among the highest threats. A subcommittee of the U.S. Senate was tasked “to investigate police reports that about

3,500 sex perverts hold federal jobs” (New York Times 1950:6). The final report from the

Senate investigation, titled “Employment of Homosexuals and Other Sex Perverts in

Government,” concluded that “those who engage in acts of homosexuality and other perverted sex activities are unsuitable for employment in the Federal Government” (U.S.

Senate 1950:19). During his first year in office, President Eisenhower issued Executive

Order 10450 which “explicitly listed ‘sexual perversion’ as sufficient and necessary grounds for disbarment from federal jobs” (D’Emilio 1998:44).

The exclusion of LGB people from the civil service, as a matter of formal policy, continued until 1978, and it was another twenty years before President Clinton issued

Executive Order 13087 in 1998, which banned sexual orientation discrimination in federal employment (Infanti 2007:109). Similarly, the ban on LGB people from receiving

86

security clearances, which was also established in the Eisenhower Executive Order, was not repealed until President Clinton issued Executive Order 12968 in 1995 (Kameny

2000:207).

Perhaps the most heated public debates have surrounded the issue of open service by LGB people in the military. For more than a century and a half, the military lacked any formal policies regarding the exclusion of LGB servicemembers. As Bèrubè (1990) notes,

Traditionally the military never officially excluded or discharged homosexuals from its ranks…. But in World War II a dramatic change occurred. As Psychiatrists increased their authority in the armed forces, they developed new screening procedures to discover and disqualify homosexual men, introducing into military policies and procedures the concept of the homosexual as a personality type unfit for military service and combat—a concept that was to determine military policy for decades after the war (2).

Unlike the Executive Orders reforming civil service and security clearance policies,

President Clinton was unable to lift the ban on gays in the military, an issue on which he campaigned in 1992. The Clinton administration’s political failure on this issue resulted in the infamous “Don’t Ask, Don’t Tell” compromise which had the effect of increasing the rates of discharge among LGB servicemen and women (McFeeley 2000:249-50). Full and open service by LGB people was not a reality until 2011, following congressional action.

As illustrated above, there is a long history of discriminatory policies and practices against LGB employees in public service. While the public sector employs only a fraction of the U.S. workforce, federal policies did not just regulate government practices but provided an example for employers in the private sector. As Kameny (2000)

87

notes, federal policy “set the tone for employment policies and practices throughout all of

American business. After all, if the federal government could exclude gays, and did, then private employers not only could, but should. And they did” (190-1). While the federal government has made considerable progress in remedying its past discriminatory practices, it is worth considering whether the positive example of these policies towards

LGB workers will have a similar influence over the private sector going forward that the exclusionary polices had in the past.

A lack of federal labor market nondiscrimination policies has historically left large portions of the population vulnerable to discriminatory practices by private employers. LGB employees have been at risk of being fired, not hired, losing out on promotions, and receiving unequal pay and benefits. LGB people have also frequently been victims of harassment and abuse in the workplace. In many situations these acts of discrimination have been perfectly legal. Absent state or local policies, LGB workers have had little to no recourse to challenge discriminatory practices by their employers.

Not only that, but employers were often free to openly target their LGB employees for such discrimination. Perhaps one of the most notorious examples involved the restaurant chain Cracker Barrel, which announced an explicit policy against employing LGB people before firing at least eleven LGB employees (Raeburn 2004:43). Because the company operated in jurisdictions without nondiscrimination policies for LGB workers, the policy and firings were within their legal rights.

Empirical Evidence of Discrimination

Numerous studies have sought to determine the degree to which LGB people experience discrimination in the labor market. The emphasis on the labor market reflects

88

the importance of economic outcomes on the broader lives of individuals, although studies of housing and public accommodation discrimination have also been undertaken recently (Mallory and Sears 2016a & 2016b). Some researchers have undertaken audit studies to assess the degree to which LGB people are discriminated against in the hiring process. Tilcsik (2011), in what he describes as “the first large-scale study of discrimination against openly gay men in the United States,” sent paired solicitations to more than seventeen hundred job postings in seven states to measure the difference in the response rates between resumes which indicated a gay applicant (through membership in a college LGB rights organization) and those that did not (586). He found that gay applicants were statistically less likely to receive a call back, but with strong regional variations. Mishel (2016) reports similar findings for queer women following an audit study in which paired applications were sent to more than eight hundred employers.

Controlling for other factors, she found women who were indicated as LGB were 32.50 percent less likely to get a call back.

While labor market discrimination can take on various forms and occurs at multiple points in the job process, economists frequently examine differences in earnings as evidence of discrimination. It is assumed that employers’ “taste for discrimination” against LGB people should, at least in the aggregate, lead to lower earnings for equally productive LGB worker relative to heterosexual workers. “If discrimination does commonly occur and results in similarly qualified and productive people being treated different only because of their sexual orientation, an economist might expect to observe differences in wages” (Badgett 1995:729). As a measurement tool, wage data is also more easily accessible. Most surveys, especially large-sample federal surveys, collect

89

data on income and salaries. Assessing other forms of discrimination often requires more complex data collection, such as the experimental audit studies described above.

Badgett (1995), in her groundbreaking study, was the first to provide evidence of wage discrimination against sexual minorities. She examines what she refers to as the

“myth of privilege,” which suggests that LGB people are an “affluent, well-educated, professional elite, occupying positions of power and influence in the workplace” (1).

Using pooled data from the 1989-1991 General Social Survey (GSS), she finds that behaviorally gay/bisexual men (based on questions about sexual partners rather than sexual identity) are disadvantaged compared to heterosexual men and that lesbian/bisexual women earn less than heterosexual women but with inconsistently significant results (Badgett 1995). According to her analysis, gay/bisexual men earn eleven percent to twenty-seven percent less than heterosexual men while lesbian/bisexual women earn twelve percent to thirty percent less than heterosexual women. “Because this economic disadvantage holds after controlling for education and occupation, it appears that equally productive gay people are being treated differently, that is, they are being discriminated against” (Badgett 1995:737). Similar findings are shown in a larger analysis using pooled GSS data from 1989 to 1994 and the 1992 National Health and

Social Life Survey. Here lesbian/bisexual women show a wage advantage over heterosexual women, a finding replicated in several future studies (Badgett 2001).

Lesbian/bisexual women earn roughly eight percent more than heterosexual women while gay/bisexual men earn seventeen percent less than heterosexual men.

Following Badgett (1995, 2001), numerous subsequent studies use similar methods and data. Blandford (2003) uses pooled GSS data from 1989 to 1996 and finds a

90

wage penalty for gay/bisexual men and a premium for lesbian/bisexual women. Berg and

Lien (2002) use pooled data from the 1991-1996 GSS and a behavioral measure of sexual

orientation. They also find an income penalty for gay/bisexual men below heterosexual

men and an income premium for lesbian/bisexual women over heterosexual women.

Cushing-Daniels and Yeung (2009) use pooled GSS data from 1988 to 2006 with a

behavioral definition of sexual orientation. They find the differences based on sexual

orientation can be attributed to differences in selection into full-time work as well as a

measurable decline over time of the overall penalties and premiums found in previous

research (Cushing-Daniels and Yeung 2009:173). Martell (2013a) uses pooled GSS data

from 1994-2008, comparing multiple behavioral definitions of sexual orientation to

examine wage differentials among men only. He finds that the observed gay/bisexual

wage penalty compared to heterosexuals cannot be explained by differences in

characteristics, suggesting discrimination plays a role. Research using the GSS has been

consistent in finding a wage penalty for sexual minority men and a wage premium for

sexual minority women.

Further studies have used household-level data available in the Census and

Current Population Survey (CPS). Heller Clain and Leppel (2001) analyze the 1990

Census and find that men living with same-sex unmarried partners earn less than other men (partnered or not) while women living with same-sex unmarried partners experience a wage premium over other women (partnered or not). Allegretto and Arthur (2001) also use the 1990 Census to examine wage differences between same-sex unmarried partner men, opposite-sex unmarried partner men, and opposite-sex married men. The married men have a large premium over the same-sex unmarried partner men while the opposite-

91

sex unmarried partner men have a significantly smaller, but measurable, premium. The

authors attribute this to a marriage premium. Baumle and Poston, Jr. (2011) use the 2000

Census and find a wage penalty for same-sex unmarried partner men but a premium for

same-sex unmarried partner women compared to their heterosexual counterparts.

Using the 2004 March Supplement of the CPS, Elmslie and Tebaldi (2007)

employ a flawed definition to identify same-sex households. They identify as LGB

anyone aged twenty-five or older who lives in a household with another unmarried,

unrelated adult of the same sex. As the authors note in a later study (Elmslie and Tebaldi

2014), their definition of same-sex households likely includes several “same-sex”

households in which the inhabitants are merely roommates and not romantically/sexually

linked. “By not properly excluding roommates, the results of [the earlier study] may be

unreliable” (Elmslie and Tebaldi 2014:331). The subsequent study by the authors pools

CPS data from 1995 to 2011 to compare two time periods. Looking exclusively at men,

they find that the wage penalty between gay/bisexual men and opposite-sex married men has declined and that gay/bisexual men hold a wage premium over men in opposite-sex unmarried partner households.

Mize (2016) challenges the practice in most of the previous literature of grouping bisexual men and women with gay and lesbian men and women. Following Worthen

(2013), he suggests that “the experiences of sexual minorities vary greatly from group to group, and the stereotypes that lead to labor market inequalities similarly vary between groups” (1136). Authors typically group bisexual men and women with gay and lesbian men and women for methodological purposes, either to boost sample sizes or because sexual orientation is defined through behavioral measures. According to Mize (2016),

92

bisexuals suffer unique and pervasive forms of discrimination, often beyond that

experienced by gay and lesbian men and women, because of the perception that their

sexual orientation is a result of choice, whereas gay and lesbians are increasingly viewed

as “born that way.” Separating the groups in his analyses, Mize (2016) finds that the

wage penalty often attributed to “gay” men in grouped analyses is mostly the result of

bisexual disadvantage. Similarly, he finds that the wage premium found in grouped

analyses of lesbian and bisexual women only extends to lesbian women and that bisexual

women experience a significant wage penalty.

While many studies have shown evidence of a wage difference between LGB

workers and their heterosexual counterparts, the limitations of the data used (see Chapter

Two above) warrants further inquiry into the question. The frequent use of the GSS,

especially prior to the addition of sexual identity measures in 2008, requires the use of

proxy measures constructed with data about a respondent’s sexual history. The small

samples in the GSS have also led researchers to group bisexuals with gay and lesbian

men and women, potentially masking the unique experiences of bisexuals in the labor

market. Small sample sizes have also led many researchers to pool data across wide spans

of years that include significant changes in the social reality of sexual orientation in the

United States. For studies relying on Census Bureau data prior to the Supreme Court’s

2013 overturning of the Defense of Marriage Act (DOMA), which defined marriage as

only between a man and a woman for federal purposes, researchers have only been able

to compare the wages of LGB people in same-sex unmarried partnerships to partnered

(married or otherwise) opposite-sex couples. There could be a significant selection bias because those who choose to partner are likely different from those who are single.

93

Further, there are economic and social benefits of being coupled, including some benefits

that accrue even when a relationship is not legally recognized. Thus, estimates of wage

discrimination based on household-level analyses are likely conservative.

The first two decades of the twenty first century have seen significant changes in

the social climate for sexual minorities with support for the social acceptance of

homosexuality reaching seventy percent (Pew Research Center 2017) and support for

same-sex marriage at sixty-seven percent (McCarthy 2020). Despite this tremendous

growth, stigma against sexual minorities persists. Nearly sixty percent of sexual minority

students report feeling unsafe at school and more than twenty-five percent report being physically harassed (Kosciw et al. 2020). Anti-LGB stigma continues to affect the mental and physical health of sexual minorities (Matsick et al. 2020). As noted above, this stigma extends to the workplace as well (Badgett et al. 2007). Further examination of sexual minority discrimination is warranted.

Hypotheses

Following the custom in the literature, I focus on wages as an indicator of labor market rewards and disadvantage. While discrimination happens in many forms, wages are commonly used to observe and measure inequality and discrimination. Differences in wages, absent differences in productivity, suggest discrimination is occurring. Following

Mize (2016), I test several hypotheses. I posit that the wages of gay men ( ) and

𝑔𝑔𝑔𝑔 bisexual men ( ) will be less than the wages of similarly situated heterosexual𝑤𝑤 men

𝑏𝑏𝑏𝑏 ( ). 𝑤𝑤

ℎ𝑚𝑚 𝑤𝑤 H1: <

𝑔𝑔𝑔𝑔 ℎ𝑚𝑚 H2: 𝑤𝑤 < 𝑤𝑤

𝑤𝑤𝑏𝑏94𝑏𝑏 𝑤𝑤ℎ𝑚𝑚

I posit that the wages of lesbian women ( ) will be greater than the wages of similarly

𝑙𝑙𝑙𝑙 situated heterosexual women ( ) while𝑤𝑤 the wages of bisexual women ( ) will be

ℎ𝑤𝑤 𝑏𝑏𝑏𝑏 less than the wages of similarly𝑤𝑤 situated heterosexual women ( ). 𝑤𝑤

ℎ𝑤𝑤 H3: > 𝑤𝑤

𝑙𝑙𝑙𝑙 ℎ𝑤𝑤 H4: 𝑤𝑤 < 𝑤𝑤

𝑏𝑏𝑏𝑏 ℎ𝑤𝑤 I also posit that these wage differences can𝑤𝑤 be explained𝑤𝑤 by differences in family structure (rates of marriage and parenthood). A large body of literature suggests that marriage results in a wage premium for men (Loh 1996, Killewald and Gough 2013) and women (Waldfogel 1997, Budig and England 2001, Killewald and Gough 2013). There is also a consistent literature that shows a fatherhood wage premium (Glauber 2018,

Hodges and Budig 2010) and a motherhood wage penalty (Waldfogel 1997, Budig and

England 2001, Glauber 2018).

H5: Wage differences between LGB people and

heterosexual people can be partially accounted for by

differences in family structure.

Data

To test these hypotheses, I use the dataset developed in Chapter Two of this dissertation. Through Cross-Survey Multiple Imputation (CSMI), I impute sexual orientation into the recipient American Community Survey (ACS) using the National

Health Interview Survey (NHIS) as the donor survey. The ACS does not collect data on sexual orientation whereas the NHIS includes a measure of sexual identity. By combining the two surveys I can use information contained in the donor survey that is omitted from the recipient survey, here the ACS, by design. The result is a large-sample, nationally

95

representative dataset that includes the imputed information about sexual identity. I

retrieved the ACS data from IPUMS USA (Ruggles et al. 2021) and the NHIS data from

IPUMS Health Surveys (Blewett et al. 2019). I restrict my sample to those currently

employed (N=7,078,805).

Whereas past studies have relied on small-sample surveys and behavioral measures of sexual orientation, I am able to test the above hypotheses using a more robust and informative set of data. While behavioral definitions measure one dimension of sexual orientation, sexual identity measures are more likely to capture the actual risk of direct discrimination in the labor market of sexual minorities. Sexual behaviors are less likely to be openly discussed in the workplace while sexual identity, even when not explicitly acknowledged, makes LGB workers available for direct and indirect discrimination. Further, many who engage in same-sex sexual activities actively reject identification as a sexual minority. Ward (2015) shows the extreme lengths that some engaged in same-sex behaviors will go to bolster their identities as “not gay.”

While some might contend that LGB workers could simply keep their sexual orientation hidden, there is considerable evidence that sexual orientation is read onto

LGB people without explicit disclosure of their sexual identity. As Mize (2016) notes,

“individuals can almost immediately—and accurately—judge someone’s sexual

orientation from observable cues such as attire, mannerisms, and vocal inflection, and

even by simply examining photographs of faces (Johnson et al. 2007; Rule et al. 2008;

Rule et al. 2009)” (1134). Though certainly not perfect indicators of sexual orientation,

affectations such as attire, mannerisms, and vocal inflection are cultural proxies of sexual

96

identity more so than of sexual behaviors. Sexual identity measures also capture those

who are a sexual minority but not behaviorally active.

It is a fair assumption that someone willing to acknowledge their identity as a

sexual minority on a federal survey would also be likely to disclose their identity in the

workplace, though this is not assured. The NHIS and ACS data cannot distinguish between respondents who disclose in both the survey and the workplace and those who disclose in the survey but not in the workplace. Badgett (1995) suggests that those who disclose in a survey but not in their workplace “might still face indirect discrimination if expectations of discrimination reduce productivity” (732). Disclosure in the survey but not in the workplace potentially biases the results of the analysis, though the bias likely

underestimates the degree of discrimination.

If indirect discrimination lowers earnings less than direct discrimination, or if the net effect of nondisclosure is to induce [sexual minority] employees to work harder and to increase productivity, then the sexual orientation coefficient will underestimate the negative effect of direct discrimination on earnings. (Badgett 1995:732)

Badgett (1995) further notes that, given the small size of the population of sexual

minorities relative to the heterosexual majority, the effect of misclassifying those who disclose in the workplace but not on the survey and those who fail to disclose in either venue will be negligible.

In addition to the benefits of a sexual identity measure, the measure in my dataset separates bisexual men and women as distinct from gay men and lesbian women. Most past studies on sexual minority labor market discrimination group gay and bisexual men as a single category and lesbian and bisexual women as another. This is typically done to overcome small sample sizes of the separate categories or because the studies use

97

household-level data which cannot determine the sexual identity of the inhabitants, only

the sex of their cohabitating partner. Someone cohabitating with a same-sex partner might identify as lesbian/gay or bisexual (or heterosexual for that matter). Household-

level data also cannot capture bisexual people residing with a different-sex partner and

they fail to capture those not cohabitating with a same-sex partner, significant

weaknesses overcome in my data.

Methods

No single variable perfectly predicts wages. Several individual and structural

factors affect wages. Individual characteristics like level of education and work

experience affect a worker’s likely wages as do structural aspects such as occupational

category and region of residence. To measure discrimination, all other possible

determinants of wage differences must be ruled out. Once individual and structural

characteristics are controlled, it is assumed that the remaining difference is the result of

discrimination. Wages for different sexual identity groups (i) are modeled as:

= +

𝑖𝑖 𝑖𝑖 𝑖𝑖 𝑖𝑖 “where wi is the natural log of hourly 𝑤𝑤wages,𝑋𝑋 X𝛽𝛽i denotes𝑣𝑣 a matrix of observed productivity

characteristics, βi denotes the vector of regression coefficients, and vi is a random error

2 term assumed to be normally distributed with variance of σ i” (Rodgers 2006:13).

Comparing wage equations for the various sexual identity groups allows for the isolation of the effects of differences in productivity characteristics. Because we typically cannot control for all of the factors affecting the distribution of wages, the error term also includes the effects of any unmeasured factors that might affect wages. It is important to

98

note that unobserved characteristics can lead to over or underestimating discrimination,

which I address in greater detail below.

Dependent Variable

Following the demand-side approach described above, I use wages as the dependent measure. Wages are computed using respondent’s reported total pre-tax salary and wage income as well as their reported usual hours of work per week, both during the previous year. Assuming a fifty-week schedule, wages are calculated as total personal income divided by the product of hours worked times fifty weeks. Because wages are positively skewed, I take the natural log to normalize the distribution, bottom coding the distribution at $1 prior to transformation.

Independent Variables

The sexual orientation measure in my data comes from the NHIS. Since 2013 the

NHIS asks respondents, “Do you think of yourself as: lesbian or gay; straight, that is, not lesbian or gay; bisexual, something else, don’t know?” (Miller and Ryan 2011:6). This is an identity measure. While there is overlap between identity measures and the behavioral measures common in the literature, they are not perfect correlates (see Chapter Two above). I contend that identity measures better reflect workers’ availability for discrimination as identity is more likely to be salient in the workplace. Even where employers are basing their discrimination on a perceived sexual orientation, it is a perceived LGB identity to which they are likely responding.

Demographic variables are included to control for individual productivity characteristics. Education is measured as the highest degree earned: less than high school, a high school diploma (including those with some college or an associate’s degree), a

99

bachelor’s degree, or a graduate/professional degree. Because work experience is not

directly measured, I use a common proxy measure (age-years of schooling-six) as well as

its quadratic transformation. While the experience measure is a function of age, I also

include an indicator of working age status for those aged twenty-five to sixty-four

because age differences between sexual identity groups are especially stark. Following

Zuberi (2001), I include indicators of self-reported racial identity to control for racial discrimination. Because of the unique history of labor market discrimination faced by

Black Americans, I include separate indicators for self-reported white identity, Black identity, and for all other racial identifications. The inclusion of racial categories cannot establish causal links between different racialized outcomes, but rather I attempt to account for the racial inequities and privileges in American labor markets. Similarly, I include an indicator of self-reported Hispanic ethnicity.

I include several variables to account for difference in family structure. Marital status is an indicator of those currently married, including same-sex married couples. I also include measures of the number of the respondent’s own children in the household as well as a separate measure for the number of the respondent’s own children in the household who are under the age of five. A respondent’s “own” child is self-reported and can include biological, adopted, and stepchildren.

Several variables are included to account for broader labor market differences. An indicator of part-time work is included for those who responded that they usually work fewer than thirty-five hours per week. I include an indicator of metropolitan residency as well as Census regions (Northeast, Midwest, South, and West). Because of the inconsistency with specific occupation codes across survey years in the ACS and between

100

the ACS and NHIS, occupations are coded into twenty-six broad Census categories.

Table 3-1 provides descriptive statistics for all variables in my analyses by sexual

identity. See Appendix E for descriptive statistics by sex and sexual identity.

Regression Models

To assess the degree to which differences in demographic, family structure, and

labor market characteristics affect wage differentials, I construct nested OLS regression

models for the wages of the different sexual identity groups. To account for the gender

differences in family structure and labor market characteristics, I run the regression

analyses for men and women separately. In each model, heterosexual workers are the

omitted category.

Model I regresses wages on sexual identity categories to get a base-line

measurement of any potential wage differences. Wages are modeled as:

= + +

𝑗𝑗 1 𝑗𝑗 where is the natural log of the wages𝑌𝑌� of𝛼𝛼 the𝛽𝛽 jth𝑋𝑋 case,𝜀𝜀 α is a constant, is the vector of

𝑗𝑗 𝑗𝑗 indicators𝑌𝑌 of sexual identity of the ith case, is the effect on wages of𝑋𝑋 membership in the

1 sexual identity categories relative to the heterosexual𝛽𝛽 baseline, and ε is the regression disturbance. Coefficients less than one suggest a wage penalty for membership in that group while coefficients greater than one suggest a wage premium.

Model II accounts for differences in demographic characteristics by adding

controls for work experience, a quadratic transformation of work experience, a work-age

indicator, highest degree earned, an indicator of Hispanic identity, and indicators of racial

identity. Here wages are modeled as:

= + + +

𝑌𝑌�𝑖𝑖 𝛼𝛼 𝛽𝛽1𝑋𝑋1011𝑗𝑗 𝛽𝛽2𝑋𝑋2𝑗𝑗 𝜀𝜀

where is the vector of demographic characteristics.

2 2𝑗𝑗 𝛽𝛽Model𝑋𝑋 III accounts for differences in family structure by adding controls for

marital status, number of own children in the household, and the number of own children

under five in the household. Here wages are modeled as:

= + + + +

𝑗𝑗 1 1𝑗𝑗 2 2𝑗𝑗 3 3𝑗𝑗 where is the vector of𝑌𝑌� family𝛼𝛼 𝛽𝛽structure𝑋𝑋 𝛽𝛽characteristics.𝑋𝑋 𝛽𝛽 𝑋𝑋 𝜀𝜀

𝑗𝑗 3𝑗𝑗 𝛽𝛽Model𝑋𝑋 IV adds controls for differences in labor market characteristics including

an indicator of part-time employment, Census Bureau region, metro status, and

occupation category. Here wages are modeled as:

= + + + + +

𝑗𝑗 1 1𝑗𝑗 2 2𝑗𝑗 3 3𝑗𝑗 4 4𝑗𝑗 where is the vector𝑌𝑌� of𝛼𝛼 labor𝛽𝛽 𝑋𝑋 market𝛽𝛽 characteristics.𝑋𝑋 𝛽𝛽 𝑋𝑋 𝛽𝛽 𝑋𝑋 𝜀𝜀

4 4𝑗𝑗 𝛽𝛽 𝑋𝑋

102

Table 3-1. Descriptive Statistics by Sexual Identity Heterosexual LGB Variable Estimate SE Estimate SE Description Total 95.65% (0.0105) 4.35% (0.0459) Percent of Total Population

Wage/Salary Income (Median) $35,900 $29,700 Pre-Tax Income from Prior Year Wages (Median) $17.77 $15.32 Hourly Wages Log of Wages (Mean) 2.73 (1.0716) 2.61 (1.0509) Natural Log of Hourly Wages

Age (Mean) 42.49 (14.0860) 36.81 (13.5419) Age in Years Work Age 83.09% (0.0149) 75.71% (0.1789) Percentage Age 25-64

103 Experience (Mean) 21.72 (14.2630) 15.92 (13.4054) Work Experience in Years

Education Less than High School 8.49% (0.0109) 7.11% (0.0966) No High School Diploma/GED High School 56.77% (0.0192) 57.48% (0.1666) High School Diploma/GED BA 21.94% (0.0161) 21.70% (0.1483) Bachelor's Degree Grad+ 12.80% (0.0129) 13.72% (0.1012) Graduate or Professional Degree

Racial/Ethnic Identity White 74.60% (0.0171) 70.73% (0.1856) White Racial Identity Black 11.54% (0.0126) 10.79% (0.1270) Black Racial Identity Other 13.86% (0.0139) 18.48% (0.1757) All Other Racial Identities Hispanic 16.73% (0.0148) 19.97% (0.1690) Hispanic Ethnicity

Family Structure Number of Children (Mean) 0.81 (1.1208) 0.33 (0.7846) Number of Own Children

Table 3-1. (continued) Heterosexual LGB Variable Estimate SE Estimate SE Description Children Under Five (Mean) 0.16 (0.4535) 0.07 (0.2884) Number of Own Children under Five Married 54.16% (0.0203) 25.99% (0.2024) Percentage Married

Region Northeast 18.04% (0.0148) 17.19% (0.1211) Percentage in Northeast Midwest 21.68% (0.0161) 19.41% (0.1475) Percentage in Midwest South 36.82% (0.0189) 33.96% (0.1779) Percentage in South West 23.46% (0.0168) 29.45% (0.1914) Percentage in West

104

Within Metro Area 81.63% (0.0152) 85.57% (0.1460) Percentage in Metro Area Part-Time 20.10% (0.0157) 25.70% (0.1606) Percentage Part-Time Workers

Percentage in Occupation Occupational Categories Management, Business, Science, and Arts 10.59% (0.0118) 9.59% (0.0797) Business Operations Specialists 2.69% (0.0063) 2.92% (0.0543) Financial Specialists 2.27% (0.0057) 2.01% (0.0417) Computer and Mathematical 3.06% (0.0067) 2.92% (0.0514) Architecture and Engineering 1.57% (0.0048) 1.16% (0.0379) Technicians 0.31% (0.0021) 0.23% (0.0140) Life, Physical, and Social Science 0.89% (0.0036) 1.04% (0.0331) Community and Social Services 1.72% (0.0050) 1.91% (0.0431) Legal 1.14% (0.0042) 1.34% (0.0410)

Table 3-1. (continued) Heterosexual LGB Variable Estimate SE Estimate SE Description Education, Training, and Library 6.03% (0.0093) 5.67% (0.0848) Arts, Design, Entertainment, Sports, and Media 1.97% (0.0054) 2.83% (0.0508) Healthcare Practitioners and Technicians 6.11% (0.0093) 5.51% (0.0762) Healthcare Support 2.36% (0.0059) 2.45% (0.0571) Protective Service 2.12% (0.0056) 2.08% (0.0514) Food Preparation and Serving 5.26% (0.0093) 10.15% (0.1476)

105 Building and Grounds

Cleaning and Maintenance 3.90% (0.0076) 3.77% (0.0711) Personal Care and Service 3.64% (0.0073) 4.49% (0.0764) Sales and Related 10.19% (0.0120) 12.17% (0.1384) Office and Administrative Support 12.69% (0.0130) 13.18% (0.1287) Farming, Fishing, and Forestry 0.69% (0.0032) 0.29% (0.0221) Construction 5.09% (0.0085) 2.67% (0.0559) Extraction 0.15% (0.0015) 0.06% (0.0083) Installation, Maintenance, and Repair 3.16% (0.0068) 1.98% (0.0514) Production 5.92% (0.0092) 4.45% (0.0815) Transportation and Material Moving 6.46% (0.0096) 5.12% (0.0892) Military Specific 0.01% (0.0003) 0.01% (0.0031)

Wage Decompositions

While regression models estimate the magnitude of wage differentials across the defined characteristics, it is also useful to estimate how much of the differential is attributable to group differences in productivity characteristics versus how much is unexplained by these differences. Decomposing the difference in predicted mean values for the groups of interest, here sexual identity categories, allows me to estimate the proportion of the difference attributable to differences in productivity characteristics, the

“explained” portion of the difference, and a residual or “unexplained” portion. Following the method developed by Oaxaca (1973) and Blinder (1973), I decompose any observed wage penalties to determine how much of the difference is attributable to differences in observed characteristics and how much is unexplained. The unexplained component is typically interpreted to be the result of discrimination as well as any unmeasured factors.

The technique has been frequently used to examine the differences in mean wages based on gender (Oaxaca 1973, Blinder 1973, Gunderson 1989) and race (Blinder 1973,

Winsborough and Dickinson 1971, Kaufman 1983).

Wage decomposition is an extension of regression analysis. For two groups, here for example heterosexual (h) and bisexual (b) workers, the wage gap (G) can be expressed as:

= ( ) ( )

ℎ 𝑏𝑏 where E(Y) is the expected value of𝐺𝐺 the 𝐸𝐸log𝑌𝑌 of −wages𝐸𝐸 𝑌𝑌 for the group. Following Jann

(2008), because wages can be modeled as regression equations (see above) we can therefore express G as the difference between the regression equations for the two groups:

106

= ( ) ( ) = ( ) ( ) ′ ℎ 𝑏𝑏 ℎ ℎ 𝑏𝑏 𝑏𝑏 where X is the vector of𝐺𝐺 productivity𝐸𝐸 𝑌𝑌 − 𝐸𝐸 characteristics𝑌𝑌 𝐸𝐸 𝑋𝑋 and𝛽𝛽 − a 𝐸𝐸constant,𝑋𝑋 ′𝛽𝛽 and β contains the

regression coefficients and the intercept.

The terms of the above equation can be rearranged several ways to express the

relationship between the two groups. Different approaches have been developed to

decompose the equation into the explained and unexplained components (Oaxaca 1973,

Blinder 1973, Winsborough and Dickinson 1971; Jones and Kelly 1984; Neumark 1988).

I employ the strategy outlined in Jann (2008) for a two-fold decomposition with a pooled

model that incorporates a “nondiscriminatory coefficient vector that should be used to

determine the contribution of the differences in the predictors” (455). In this model, the

wage gap can be expressed as:

= { ( ) ( )} + { ( ) ( ) + ( ) ( )} ∗ ′ ∗ ∗ ℎ 𝑏𝑏 ℎ ℎ 𝑏𝑏 𝑏𝑏 where β* is𝐺𝐺 the nondiscriminatory𝐸𝐸 𝑋𝑋 − 𝐸𝐸 𝑋𝑋 ′ 𝛽𝛽coefficient𝐸𝐸 𝑋𝑋 vector𝛽𝛽 −from𝛽𝛽 the𝐸𝐸 pooled𝑋𝑋 ′ model𝛽𝛽 − 𝛽𝛽 which includes a group indicator. In this formulation, the first component:

{ ( ) ( )} ∗ ℎ 𝑏𝑏 is the component explained by differences𝐸𝐸 𝑋𝑋 −in 𝐸𝐸productivity𝑋𝑋 ′𝛽𝛽 characteristics. The second

component:

{ ( ) ( ) + ( ) ( )} ′ ∗ ∗ ℎ ℎ 𝑏𝑏 𝑏𝑏 is the residual or unexplained𝐸𝐸 𝑋𝑋 component𝛽𝛽 − 𝛽𝛽 which𝐸𝐸 is𝑋𝑋 understood′ 𝛽𝛽 − 𝛽𝛽 to be a measure of

discrimination. It is important to emphasize that this component also includes any

differences that can be attributed to differences in unobserved characteristics (Jann

2008:455). If there are productivity characteristics that affect the differential allocation of

wages that are not included in the model, their effect would be included in the residual

107

component and therefore could result in an overestimation of the effect of discrimination

(Blank et al. 2004:123). Badgett (2001) suggests that “the usual forces of unobserved

differences—schooling quality, actual labor force experience, or other factors related to

ability—are not likely to vary systematically by sexual orientation” (35). Because of this,

it is reasonable to assume that the residual component is a fair measure of discrimination

based on sexual orientation. Of course, discrimination in these factors may be

endogenous with wage discrimination measured here. It is possible that the

decompositions could either over or underestimate discrimination. Failing to account for

certain factors which lead to wage differences would result in an over estimation of

discrimination. Controlling out discrimination that occurs elsewhere would result in an

underestimation of discrimination.

The regression model used for the decomposition (Model IV) contains several categorical regressors. In a standard decomposition, the choice of omitted category affects the results with different omitted categories producing varied estimates of the unexplained portion. However, the choice of omitted category is arbitrary. Following

Jann (2008), I use deviant contrast transformations of dummy variables to counter this arbitrariness. Such a transformation produces results in which “the results are equal to the simple averages of the results one would get from a series of decompositions in which the categories are used one after another as the base category” (462). This allows me to isolate the unique contribution of each category of the categorical regressors.

Results

Mean raw hourly wages by sexual identity for those currently employed show results consistent with hypotheses two through four but not hypothesis one (Table 3-2).

108

Bisexual men’s hourly wages are an average of $9.23 less than their heterosexual counterparts (H2). The lesbian wage premium is observed in the raw average of hourly wages with lesbian women earning an average of $3.17 more than their heterosexual counterparts (H3). Bisexual women earn an hourly average of $5.48 less than their heterosexual counterparts (H4). Counter to most of the existing literature, we see here a wage premium for gay men. Gay men earn an hourly average of $1.25 more than their heterosexual counterparts (H1).

Table 3-2. Mean Wages for Currently Employed by Sexual Orientation Men Mean Wages n Percent of Men Sexual Identity Heterosexual $27.53 3,585,508 96.30% Gay $28.78 74,136 2.54% Bisexual $18.30 20,236 1.16%

Women Mean Wages n Percent of Women Sexual Identity Heterosexual $21.23 3,291,279 94.93% Lesbian $24.40 59,109 2.13% Bisexual $15.75 48,538 2.94%

Regression Models

Regression models estimate the potential impacts that demographic, family structure, and labor market characteristics have on the observed wage differences. To establish a baseline effect of sexual identity, Model I contains only the sexual identity indicator. Model II adds controls for demographic differences. Because of the unique gendered effects of family structure, Model III adds controls for marital status, number of

109

children, and number of children under five to the previously controlled demographic characteristics. Model IV adds controls for differences in labor market characteristics.

Bisexual Men

Table 3-3 shows the exponentiated results for the models of the natural log of men’s wages. The baseline results for Model I suggest that bisexual men earn nearly thirty percent less than heterosexual men (eb=0.7037, p < 0.001), a wage difference of roughly $4.90 per hour. The wage penalty is reduced through each subsequent model as additional factors are included. The penalty narrows significantly when demographic characteristics are added in Model II. Here bisexual men’s wages are just under thirteen percent less than heterosexual men’s (eb=0.8730, p < 0.001), a wage difference of roughly sixty-nine cents per hour. The reduction in the wage ratio by more than half from

Model I to Model II owes to the significant age differences between bisexual men and heterosexual men. Bisexual men are on average 10.05 years younger than heterosexual men. Because of this, bisexual men are 15.07 percentage points less likely to being working age (twenty-five to sixty-four). Being working age is associated with a more than forty percent increase in wages over those not working age (eb=1.4061, p < 0.001), a wage difference of roughly $2.20 per hour.

110

Table 3-3. Regression of Log Wages by Sexual Identity for Male Workersa Variables Model I Model II Model III Model IV b se b se b se b se Sexual Identity Gay 1.0255 0.0008 0.9901 0.0007 1.0644 0.0007 1.0703 0.0007 Bisexual 0.7037 0.0012 0.8730 0.0011 0.9201 0.0011 0.9448 0.0010

Work Age 1.4061 0.0004 1.3712 0.0004 1.1861 0.0004 Experience 1.0340 0.0000 1.0263 0.0000 1.0215 0.0000 2 Experience 0.9994 0.0000 0.9996 0.0000 0.9997 0.0000

111 Education High School 1.4258 0.0004 1.4253 0.0004 1.2827 0.0004 BA 2.3618 0.0005 2.3010 0.0005 1.7410 0.0005 Grad+ 3.1603 0.0005 3.0008 0.0005 2.1869 0.0006

Racial/Ethnic Identity Black 0.8506 0.0004 0.8771 0.0004 0.9022 0.0004 Other Race 0.9707 0.0003 0.9683 0.0003 0.9409 0.0003 Hispanic 0.8961 0.0003 0.8946 0.0003 0.9129 0.0003

Household Structure Number of Children 1.0214 0.0001 1.0180 0.0001 Children Under Five 1.0097 0.0003 1.0042 0.0003 Married 1.2487 0.0003 1.1834 0.0003

Table 3-3. (continued) Variables Model I Model II Model III Model IV b se b se b se b se Region Midwest 0.9371 0.0004 South 0.9081 0.0003 West 0.9583 0.0003

Metro Status 1.1778 0.0003 Part Time 0.5653 0.0003

112 Occupation

Business Operations Specialists 0.9576 0.0008 Financial Specialists 1.0136 0.0009 Computer and Mathematical 1.2459 0.0006 Architecture and Engineering 1.2449 0.0008 Technicians 1.0974 0.0016 Life, Physical, and Social Science 0.9407 0.0012 Community and Social Services 0.6889 0.0011 Legal 0.8708 0.0011 Education, Training, and Library 0.7880 0.0007 Arts, Design, Entertainment, Sports, and Media 0.5850 0.0008 Healthcare Practitioners and Technicians 1.1835 0.0007 Healthcare Support 0.7354 0.0015 Protective Service 0.9930 0.0007 Food Preparation and Serving 0.6851 0.0006

Table 3-3. (continued) Variables Model I Model II Model III Model IV b se b se b se b se Building and Grounds Cleaning and Maintenance 0.5501 0.0006 Personal Care and Service 0.5081 0.0010 Sales and Related 0.7985 0.0005 Office and Administrative Support 0.8147 0.0005 Farming, Fishing, and Forestry 0.5487 0.0012 Construction 0.6274 0.0005 Extraction 1.0193 0.0022

113 Installation, Maintenance, and Repair 0.8314 0.0006 Production 0.8516 0.0005

Transportation and Material Moving 0.7161 0.0005 Military Specific 0.9342 0.0104 aAll coefficients significant at p < 0.001

This age difference also contributes to notable differences in educational

attainment and, because it is a function of age and education, estimated work experience.

The coefficients for experience and education suggest additional years of experience and

greater educational attainment are associated with higher average ages. Each additional

year of work experience, holding all else constant, is associated with a 3.40 percent

increase in average wages over the geometric mean (eb=1.0340, p < 0.001), though the

coefficient for the quadratic transformation (eb=0.9994, p < 0.001) suggests these returns

diminish over time.

Education has dramatic effects on wages. Having a high school diploma increases

wages by nearly forty three percent over men without no high school diploma

(eb=1.4258, p < 0.001), a wage difference of roughly $2.30 per hour. Men holding a

bachelor’s degree earn more than double (eb=2.3618, p < 0.001) and men with advanced

degrees earn more than triple (eb=3.1603, p < 0.001) the average wages of those without

a high school diploma holding all else constant, though the inclusion of additional factors

in subsequent models mitigates these effects somewhat. Bisexual men, being so much

younger, are 7.08 percentage points less likely than heterosexual men to have at least a

bachelor’s degree.

Non-white racial identities and Hispanic identity, proxies for racial/ethnic

discrimination, are also associated with lower wages. Bisexual men are 6.05 percentage

points more likely to identify as non-white, and 5.03 percentage points more likely identify as Hispanic than heterosexual men.

The inclusion of family structure variables in Model III narrows the penalty further. Here bisexual men’s wages are slightly less than eight percent those of

114

heterosexual men (eb=0.9201, p < 0.001), a wage difference of roughly forty-two cents per hour. Marriage and parenthood are both associated with an average increase in wages, holding all other variables constant. This is consistent with the literature on wage premiums for fatherhood and marriage. For working men, being married is associated with a nearly twenty-five percent wage premium over being unmarried (eb=1.2487, p <

0.001). Family structure differs dramatically between bisexual and heterosexual men.

Bisexual men are 36.01 percentage points less likely to be married than their heterosexual

counterparts, likely a result of their younger age and the recency of same-sex marriage

legalization for those who might marry a same-sex partner. Furthermore, bisexual men have on average 0.50 fewer of their own children living in the household and 0.10 fewer of their own children under the age of five than heterosexual men. Having children in the household is associated with a modest estimated 2.14 percent per-child increase in average wages (eb=1.0214, p < 0.001). The narrowing of the wage penalty here is evidence for hypothesis five.

The wage penalty for bisexual men narrows slightly further in Model IV with the inclusion of labor market differences. With all control variables included, bisexual men earn over 5.50 percent less than heterosexual men (eb=0.9448, p < 0.001), a wage

difference of roughly fifty cents per hour. In terms of labor market geography, bisexual

men are 2.11 percentage points more likely to live in a metropolitan area than are

heterosexual men, which is associated with a nearly eighteen percent average wage

premium over those living outside a metropolitan area (eb=1.1778, p < 0.001). However,

bisexual men are more likely to live in the Midwest (0.14 percentage points) and the

West (7.49 percentage points) which are associated with average lower wages than the

115

Northeast, though heterosexual men are more likely to live in the South (4.99 percentage

points) which has the largest wage penalty relative to the Northeast. Controlling for the

significant differences in rates of part-time employment contributes to the narrowing of the observed wage penalty. Bisexual men are 10.41 percentage points more likely to be a part-time worker than their heterosexual counterparts, and part-time work is, holding all else constant, associated with wages fifty seven percent that of full-time work

(eb=0.5653, p < 0.001). Bisexual men are also 7.79 percentage points less likely to be

employed in the occupation categories associated with the top median earnings for

working men. See Appendix F for wages by occupation.

Gay Men

Table 3-3 also models the differences in wages between employed gay and

heterosexual men. At the baseline, employed gay men show a modest 2.55 percent wage

premium relative to their heterosexual counterparts (eb=1.0255, p < 0.001), a wage difference of roughly forty-two cents per hour. The inclusion of demographic characteristics in Model II eliminates this premium and shows a less than one percent wage penalty for gay men compared to heterosexual men (eb=0.9901, p < 0.001) a wage

difference of roughly five cents per hour. Gay men’s lower average age and therefore

work experience helps explain the appearance of a wage penalty in this model. Gay men

are on average about two years younger than heterosexual men. Gay men’s higher

educational attainment—they are 6.48 percentage points more likely to hold at least a

bachelor’s degree—does not offset the difference in age leading to a deficit of 2.82 years

in average work experience for gay men compared to heterosexual men. Additionally,

gay men are only 3.26 percentage points more likely to identify as non-white than

116

heterosexual men, almost half the difference between bisexual and heterosexual men, and

only 3.02 percentage points more likely to identify as Hispanic. Overall, gay men appear

demographically far more like heterosexual men than do bisexual men.

The slight wage penalty in Model II is more than accounted for with the inclusion

of family structure characteristics in Model III. Controlling for differences in marriage

and child rearing, gay men have an estimated 6.44 percent wage premium over

heterosexual men (eb=1.0644, p < 0.001), a wage difference of roughly thirty-four cents

per hour. Gay men are 27.84 percentage points less likely to be married than heterosexual men. Part of this difference is likely due to the recency of same-same marriage legalization. Same-same marriage was not legalized nationwide until June 2015, overlapping with the survey years in the ACS data. Gay men also have an average of 0.64 fewer of their own children in the household and 0.14 fewer of their own children under the age of five. The modest effects of family structure variables on wages, which would advantage heterosexual men, are likely outweighed by the larger magnitude effects of

higher educational attainment, which favors gay men. The elimination of the penalty

from Model II to Model III supports hypothesis five.

The wage premium seen in Model III expands slightly with the inclusion of labor

market characteristics in Model IV (eb=1.0703, p < 0.001), a wage difference of roughly

sixty-five cents per hour. Gay men are 6.31 percentage points more likely to reside in a

metropolitan area than their heterosexual counterparts which has a positive effect on

wages. Gay men are also slightly more likely to live in the higher wage Northeast and

West regions. Compared to heterosexual men, gay men are less than four percentage

points more likely to be part-time workers, nearly a third of the difference between

117

bisexual and heterosexual men. Gay men are also more likely to work in the occupation

categories associated with the highest median earnings for working men.

Bisexual Women

Table 3-4 shows the exponentiated results for the models of the natural log of

women’s wages. Like their male counterparts, bisexual women see a significant wage

penalty at baseline relative to heterosexual women. Model I shows bisexual women earn

nearly twenty-five percent less than heterosexual women (eb=0.7519, p < 0.0001), a wage

difference of roughly $3.48 per hour. As additional controls are added in subsequent

models, the penalty narrows but is not eliminated, like that of bisexual men above.

Controlling for demographic characteristics, Model II shows a penalty of just under ten

percent (eb=0.9053, p < 0.0001), a wage difference of roughly forty-five cents per hour.

As above, this significant reduction in the wage penalty likely owes to the dramatic

differences in ages between bisexual and heterosexual women. Bisexual women are, on

average, 9.83 years younger than heterosexual women. Because of this, bisexual men are

16.58 percentage points less likely to being working age (twenty-five to sixty-four).

Being working age is associated with a more than forty-one percent increase in wages over those not working age (eb=1.4182, p < 0.001). Along with being younger, bisexual

women are 6.29 percentage points less likely to hold a bachelor’s degree resulting in an

average 9.47-year deficit in work experience. For women, each additional year of work

experience is associated with a 1.91 percent increase in wages (eb=1.0191, p < 0.001), holding all else constant, though the effect diminishes over time for women as with men.

118

Table 3-4. Regression of Log Wages by Sexual Identity for Female Workersa Variables Model I Model II Model III Model IV b se b se b se b se Sexual Identity Gay 1.1214 0.0008 1.0871 0.0008 1.0887 0.0008 1.0694 0.0007 Bisexual 0.7519 0.0007 0.9053 0.0007 0.9115 0.0007 0.9505 0.0006

Work Age 1.4182 0.0004 1.4073 0.0004 1.2013 0.0004 Experience 1.0191 0.0000 1.0200 0.0000 1.0156 0.0000 2 Experience 0.9997 0.0000 0.9997 0.0000 0.9998 0.0000

119 Education High School 1.4929 0.0005 1.4781 0.0005 1.1979 0.0004 BA 2.3272 0.0005 2.2893 0.0005 1.5491 0.0005 Grad+ 3.0398 0.0005 2.9780 0.0005 1.9045 0.0005

Racial/Ethnic Identity Black 0.9692 0.0003 0.9836 0.0003 0.9776 0.0003 Other Race 1.0034 0.0003 1.0049 0.0003 0.9872 0.0003 Hispanic 0.8987 0.0003 0.9062 0.0003 0.9326 0.0003

Household Structure Number of Children 0.9742 0.0001 0.9904 0.0001 Children Under Five 1.0326 0.0003 1.0269 0.0003 Married 1.0584 0.0002 1.0476 0.0002

Table 3-4. (continued) Variables Model I Model II Model III Model IV b se b se b se b se Region Midwest 0.9140 0.0003 South 0.8645 0.0003 West 0.9421 0.0003

Metro Status 1.1517 0.0003 Part Time 0.6477 0.0002

120 Occupation

Business Operations Specialists 0.9476 0.0007 Financial Specialists 0.9764 0.0007 Computer and Mathematical 1.1520 0.0008 Architecture and Engineering 1.1587 0.0015 Technicians 0.9029 0.0030 Life, Physical, and Social Science 0.8170 0.0011 Community and Social Services 0.7317 0.0007 Legal 0.9796 0.0010 Education, Training, and Library 0.6831 0.0005 Arts, Design, Entertainment, Sports, and Media 0.5053 0.0008 Healthcare Practitioners and Technicians 1.1412 0.0005 Healthcare Support 0.6682 0.0006 Protective Service 0.8343 0.0011

Table 3-4. (continued) Variables Model I Model II Model III Model IV b se b se b se b se Food Preparation and Serving 0.5939 0.0006 Building and Grounds Cleaning and Maintenance 0.3863 0.0007 Personal Care and Service 0.3720 0.0005 Sales and Related 0.6272 0.0005 Office and Administrative Support 0.7634 0.0004 Farming, Fishing, and Forestry 0.4281 0.0018 Construction 0.5411 0.0018 Extraction 0.9723 0.0115 121 Installation, Maintenance, and Repair 0.7760 0.0020

Production 0.6182 0.0006 Transportation and Material Moving 0.6219 0.0008 Military Specific 0.8638 0.0189 aAll coefficients significant at p < 0.001

Higher education has similar effects for women as men with holding a bachelor’s

degree being associated with a more than doubling (eb=2.3272, p < 0.001) of average

wages and holding an advanced degree with more than tripling (eb=3.0398, p < 0.001) the

average wages of those with less than a high school diploma. Interestingly, while bisexual women are more likely to identify as non-white than heterosexual women (30.72 percent versus 26.57 percent respectively), they are less likely to identify as Black (10.03 percent to 13.12 percent). Model II suggests a marginal wage advantage (eb=1.0034, p <

0.001) for non-Black racial minority women, though the inclusion of subsequent controls

accounts for this difference. Bisexual women are also 4.56 percentage points more likely

to identify as Hispanic, the racial/ethnic identity most disadvantaged compared to white

women.

The addition of family structure variables in Model III shows a slight narrowing

of the wage penalty for bisexual women. After accounting for marital status and the

presence of their own children in the household, bisexual women now earn just under

nine percent less than heterosexual women (eb=0.91115, p < 0.001), a wage difference of

roughly forty-two cents per hour. Marriage has a much smaller effect on the wages of

working women. Married women see an average wage increase of only 5.84 percent over

unmarried women (eb=1.0584, p < 0.001) whereas married men received a nearly twenty-

five percent increase. Bisexual women are thirty percentage points less likely to be married than heterosexual women. While men saw an average wage increase per additional child in the household, women’s wages are penalized for each additional child

(consistent with the literature on motherhood wage penalties), though oddly having younger children in the home has the opposite effect (though this effect might be

122

mitigated through interaction with marriage and part-time work status or may be the result of higher-income women being less likely to exit the labor market while raising young children (Bonner 2014). For each additional child, women’s wages reduce by 2.28 percent (eb=0.9772, p < 0.001) compared to women with no children, holding all else

constant. Like bisexual men, bisexual women have fewer children in the household than

heterosexual women, with bisexual women reporting an average of 0.35 fewer children

overall and 0.05 fewer children under five. The motherhood wage penalty and the

marriage wage premium appear to balance in preserving the overall wage penalty for

bisexual women.

The wage penalty narrows further after adding labor market characteristics to the

model. Model IV shows bisexual women earning just under five percent less than

heterosexual women (eb=0.9505, p < 0.001) a wage difference of roughly fifty-five cents

per hour. Bisexual women are slightly more likely than heterosexual women to live in a

metropolitan area (1.81 percentage points) which is associated with an increase in wages,

though they are less likely to live in the Northeast (2.66 percentage points) which is the

reference category and associated with the highest average wages. In terms of part-time worker status, bisexual women are 8.90 percentage points more likely to be a part-time worker than heterosexual women. Holding all else constant, women working part time earn more than one-third less than the wages of full-time working women (eb=0.6477, p <

0.001). Bisexual women are also 7.19 percentage points less likely to work in the

occupational categories associated with the highest median earnings for working women

(see Appendix F).

123

Lesbian Women

As seen in Table 3-4, the lesbian wage premium is observed in all models. At the

base line in Model I, lesbian women earn twelve percent more than heterosexual women

(eb=1.1214, p < 0.001), a wage difference of roughly $1.70 per hour. The premium

narrows in Model II with the inclusion of demographic controls (eb=1.0871, p < 0.001), a wage difference of roughly forty-two cents per hour. Lesbian women are the closest in average age to their heterosexual counterparts of any sexual minority group. Lesbian women are on average 1.70 years younger than heterosexual women and have an average

2.18 fewer years of work experience. While having similar levels of educational attainment through a bachelor’s degree, lesbian women are 5.65 percentage points more likely to hold an advanced degree than heterosexual women.

While lesbian women are more likely to identify as Black than any other sexual identity group of either gender, they are fairly consistent with heterosexual women in the percentage identifying overall as non-white or Hispanic, being only 2.20 percentage points and 1.57 percentage points greater, respectively. The relative closeness in rates of racial/ethnic identity and years of work experience suggests that the greater educational attainment of lesbian women contributes highly to the initial wage premium.

The wage premium for lesbian women persists with the inclusion of family structure variables in Model III (eb=1.087, p < 0.001), a wage difference of again roughly forty-two cents per hour. Lesbian women are the most likely of all sexual minority groups to be married with 32.22 percent (se = 0.0028) currently married; however, lesbian women are 18.58 percentage points less likely to be married than heterosexual women. Lesbian women are raising fewer children than heterosexual and bisexual

124

women, averaging 0.3878 of their own children and 0.0673 of their own children under five in their households. These differences likely balance to maintain the wage premium over heterosexual women.

Model IV shows that the lesbian wage premium persists even while considering labor market characteristics with lesbian women earning 6.94 percent more than heterosexual women in the full model (eb=1.0694, p < 0.001), a wage difference of roughly seventy-seven cents per hour. Lesbian women are more likely than their heterosexual counterparts to live in metropolitan areas and in the Northeast and West regions, all associated with higher wages over the alternatives. Lesbian women are also more likely to work in the occupational categories with the highest median earnings for women workers.

Decompositions

The above regression models suggest that, controlling for differences in productivity characteristics, bisexual men and women continue to experience a wage penalty relative to their heterosexual counterparts while gay men and lesbian hold an earnings premium over heterosexual men and women. For the bisexual wage penalties, it is useful now to estimate how much of the penalty can be attributed to differences in these productivity characteristics and how much is unexplained, a measure of the prevalence of discrimination. I perform wage decompositions using the Model IV equations for bisexual men and women compared to heterosexual men and women, respectively. Because gay men and lesbian women have estimated wage premiums in

Model IV, there is no theoretical basis for performing decompositions on their equations.

125

Table 3-5 summarizes the results of the decomposition for bisexual men and

heterosexual men. Recall that with all control variables included, bisexual men earn over

5.50 percent less than heterosexual men (eb=0.9448, p < 0.001). This is reflected in the

0.3514 log dollar difference between the geometric means for the two groups. 83.91

percent of that difference is explained by differences in productivity characteristics

(0.2949). This suggests that if bisexual men had equivalent characteristics (such as age

and years of work experience) as heterosexual men, then we would expect bisexual men’s

wages would be higher than the observed wages in this data. Indeed, if bisexual men

matched heterosexual men on all observed characteristics, we would expect bisexual

men’s wages to be 34 percent higher (e0.2949=1.3429) than observed here. The bottom

portion of Table 3-5 shows the estimated contributions of each observed productivity characteristic to the explained portion. We can see that age and work experience

contribute a substantial amount to the explained portion of the wage differential. The

higher rates of part-time work and lower rates of marriage among bisexual men are also

notable contributors.

126

Table 3-5. Wage Decomposition for Bisexual and Heterosexual Men Decomposition Components Log Diff. Percent Explained 0.2949 83.91% Unexplained 0.0566 16.09% Total Difference 0.3514 Explained by Observed Characteristics Variable Explained SE Work Age 0.0257 0.0001 Experience 0.2046 0.0005 Experience2 -0.1337 0.0004 Less than High School 0.0031 0.0001 High School Diploma 0.0093 0.0001 Bachelor's Degree 0.0045 0.0001 Graduate/Professional Degree 0.0165 0.0001 White 0.0033 <0.0001 Black -0.0011 <0.0001 Other Race 0.0005 <0.0001 Hispanic 0.0023 <0.0001 Non-Hispanic 0.0023 <0.0001 Number of Children 0.0088 0.0001 Number of Children Under Five 0.0004 <0.0001 Married 0.0309 0.0001 Not Married 0.0309 0.0001 Within Metro Area -0.0017 <0.0001 Not within Metro Area -0.0017 <0.0001 Northeast 0.0013 <0.0001 Midwest <0.0001* <0.0001 South -0.0023 <0.0001 West -0.0006 <0.0001 Management, Business, Science, and Arts 0.0083 0.0001 Business Operations Specialists 0.0004 <0.0001 Financial Specialists 0.0011 <0.0001 Computer and Mathematical 0.0018 0.0001 Architecture and Engineering 0.0022 0.0001 Technicians 0.0002 <0.0001 Life, Physical, and Social Science 0.0001 <0.0001 Community and Social Services -0.0005 <0.0001

127

Table 3-5. (continued) Variable Explained SE Legal 0.0001 <0.0001 Education, Training, and Library -0.0003 <0.0001 Arts, Design, Entertainment, Sports, and Media 0.0026 0.0001 Healthcare Practitioners and Technicians 0.0033 0.0001 Healthcare Support 0.0001 <0.0001 Protective Service -0.0001 <0.0001 Food Preparation and Serving 0.0121 0.0001 Building and Grounds Cleaning and Maintenance 0.0032 0.0001 Personal Care and Service 0.0045 0.0001 Sales and Related 0.0005 <0.0001 Office and Administrative Support 0.0003 <0.0001 Farming, Fishing, and Forestry -0.0025 <0.0001 Construction -0.0048 0.0001 Extraction 0.0002 <0.0001 Installation, Maintenance, and Repair <0.0001 <0.0001 Production 0.0001 <0.0001 Transportation and Material Moving -0.0010 <0.0001 Military Specific <0.0001* <0.0001 Part-Time Worker 0.0297 0.0001 Full-Time Worker 0.0297 0.0001 *p > 0.05

128

The unexplained component suggests that, even if bisexual men’s characteristics were equivalent to heterosexual men’s, their wages would still not be equal. We can attribute this unexplained, or residual, portion to discrimination. According to the decomposition model, 16.09 percent of the difference between bisexual and heterosexual’s men’s wages can be attributed to discrimination (0.0566). If bisexual men’s productivity characteristics were to match heterosexual men’s in this model, they would still earn more than five percent less than heterosexual men (e0.0566=1.0582). It is worth noting again that this unexplained component also includes the effects of any unobserved productivity characteristics.

Table 3-6 summarizes the results of the decomposition for bisexual women and heterosexual women. Recall that with all control variables included, bisexual women earn just under five percent less than heterosexual women (eb=0.9505, p < 0.001). This translates to a 0.2850 log dollar difference in the geometric means for the two sexual identity groups. Of this difference, 82.21 percent is explained by differences in productivity characteristics. This suggests that if bisexual women had equivalent productivity characteristics to heterosexual women, we would expect bisexual women’s wages would be higher than observed here. Based on this model, we would expect them to be twenty six percent higher (e0.2343=1.2640).

129

Table 3-6. Wage Decomposition for Bisexual and Heterosexual Women Decomposition Components Log Diff. Percent Explained 0.2343 82.21% Unexplained 0.0507 17.79% Total Difference 0.2850 Explained by Observed Characteristics Variable Explained SE Work Age 0.0304 0.0001 Experience 0.1452 0.0004 Experience2 -0.0722 0.0003 Less than High School 0.0029 0.0001 High School Diploma 0.0073 <0.0001 Bachelor's Degree 0.0031 <0.0001 Graduate/Professional Degree 0.0124 0.0001 White 0.0005 <0.0001 Black -0.0003 <0.0001 Other Race 0.0001 <0.0001 Hispanic 0.0016 <0.0001 Non-Hispanic 0.0016 <0.0001 Number of Children -0.0034 <0.0001 Number of Children Under Five 0.0013 <0.0001 Married 0.0071 <0.0001 Not Married 0.0071 <0.0001 Within Metro Area -0.0013 <0.0001 Not within Metro Area -0.0013 <0.0001 Northeast 0.0020 <0.0001 Midwest <0.0001* <0.0001 South -0.0035 <0.0001 West -0.0010 <0.0001 Management, Business, Science, and Arts 0.0065 0.0001 Business Operations Specialists 0.0009 <0.0001 Financial Specialists 0.0026 <0.0001 Computer and Mathematical 0.0014 <0.0001 Architecture and Engineering <0.0001 <0.0001 Technicians <0.0001 <0.0001 Life, Physical, and Social Science <0.0001 <0.0001 Community and Social Services <0.0001 <0.0001

130

Table 3-6. (continued) Variable Explained SE Legal 0.0003 <0.0001 Education, Training, and Library -0.0020 <0.0001 Arts, Design, Entertainment, Sports, and Media 0.0025 <0.0001 Healthcare Practitioners and Technicians 0.0152 0.0001 Healthcare Support -0.0004 <0.0001 Protective Service -0.0003 <0.0001 Food Preparation and Serving 0.0135 0.0001 Building and Grounds Cleaning and Maintenance 0.0015 0.0001 Personal Care and Service 0.0078 0.0001 Sales and Related 0.0069 0.0001 Office and Administrative Support 0.0008 <0.0001 Farming, Fishing, and Forestry -0.0010 <0.0001 Construction 0.0001 <0.0001 Extraction <0.0001 <0.0001 Installation, Maintenance, and Repair <0.0001 <0.0001 Production -0.0001 <0.0001 Transportation and Material Moving 0.0002 <0.0001 Military Specific <0.0001 <0.0001 Part-Time Worker 0.0193 0.0001 Full-Time Worker 0.0193 0.0001 *p > 0.05

131

The bottom portion of Table 3-6 shows the estimated contributions of each

observed productivity characteristic to the explained portion of the wage differential. We

can see that, as with bisexual men, age and experience contribute substantially to the

explained portion of the difference in wages between bisexual women and heterosexual

women. The higher rates of part-time work among bisexual women contributes

moderately to the penalty. The different rates of marriage between bisexual women and

heterosexual women have only a marginal impact on the explained portion of the

difference.

The unexplained component suggests that, even if bisexual women’s

characteristics were equivalent to heterosexual women’s, their wages would still not be

equal. For bisexual women, 17.79 percent of the difference in their wages compared to

heterosexual women is attributed to discrimination. If bisexual women were to have

equal productivity characteristics to heterosexual women, we would still expect bisexual

women’s wages to be an estimated five percent less than heterosexual women’s wages

(e0.0507=1.0520).

Discussion

The results above suggest that discrimination against sexual minorities persists, though with unique patterns based on specific identities. Bisexual men and women appear to experience significant wage discrimination relative to heterosexual men and women, consistent with hypotheses two and four. The disadvantage observed for gay men when demographically compared to heterosexual men is accounted for completely by differences in family structure, evidence counter to hypothesis one. Lesbian women, however, see a persistent wage premium over heterosexual women, evidence of

132

hypothesis three. Differences in family structure appear to mitigate the effects of demographic differences for gay men, particularly age/work experience, and to benefit lesbian women while having only marginal impacts on bisexual men and women. This is consistent with hypothesis five. All these findings are statistically significant, though this is not surprising given the large overall sample size and the disparate group sample sizes between heterosexuals and sexual minorities.

For gay men and lesbian women, the premium in wages likely owes considerably to higher levels of educational attainment. Both gay men and lesbian women are considerably more likely to have at least a bachelor’s degree than their heterosexual counterparts. This higher education also correlates with increased likelihood of working in higher paid occupational categories for gay men and lesbian women. Bisexual men and women are considerably younger than both their heterosexual and gay/lesbian counterparts. This differences in age and the correlated lower rates of educational attainment and years of work experience among bisexual men and women account for significant portions of the observed wage penalties.

The persistence of the wage penalty for bisexual men and women, even when accounting for differences in productivity characteristics, suggests that bisexual men and women experience unique forms of labor market discrimination (Mize 2016). For both groups, more than fifteen percent of the difference in wages with their heterosexual counterparts cannot be accounted for by differences in productivity characteristics. While acceptance of sexual minorities has risen over the past decade, it is possible this support does not extend equally to all sexual minorities. According to Mize and Manago (2015), views about bisexuality are especially negative and their “analyses suggest that the

133

negative sentiments toward bisexual men and women largely stem from views of

bisexuals as dishonest, confused, and indecisive” (474). These views may influence the

discriminatory actions of employers. It is also possible that the observed wage penalties

are the result of unobserved productivity characteristics, such as work-limiting disability.

As discussed above, there is evidence that persistent stigmas against sexual minorities

impacts mental and physical health, which could impact one’s labor market experiences.

The lesbian wage premium might be reflective of broader patterns of labor market

attachment and family structure. Lesbian women are considerably more likely to be in the

labor market than their heterosexual counterparts. 40.25 percent of heterosexual women

are out of the labor market compared to 30.07 percent of lesbian women. Given the lower

rate of child rearing among lesbian women, it is possible that lesbian women have had

greater labor market participation over their life course, with heterosexual women

possibly having previously left the labor market only to return later. With heterosexual

women being older and more likely to have children than lesbian women, it is possible

that the observed wage premium is a result of these patterns. Unfortunately, my proxy

measure of work experience cannot capture women who drop out of the labor market

after to raise children. Additional studies which account for labor market attachment

might elucidate these trends further.

The finding of a gay male wage premium here is fairly novel. Only one prior study (Carpenter and Eppink 2017) has found a wage premium for gay men. This other study also uses the NHIS, though not in combination with the ACS. Because my Cross-

Survey Multiple Imputation uses the NHIS as the donor, any peculiarities in that data set

could be carried over into these analyses. It is therefore possible that the NHIS data are,

134

rather than capturing a new phenomenon, simply unreliable. However, Carpenter and

Eppink (2017) use different survey years, meaning the errors would have to be systemic in the survey design which is unlikely. Further, those authors reject the notion that the

NHIS data “are incorrect or otherwise idiosyncratic” given they produce otherwise

reasonable estimates on several other measures (435). Future studies using additional data

sources are needed to corroborate these findings.

This study does have some limitations worth examining. While I use a measure of

sexual identity rather than behavior to better reflect workers’ availability for

discrimination, I cannot measure openness in the workplace directly, which likely better

reflects this availability. Additionally, the risks of disclosing one’s sexual identity are

perhaps unevenly distributed in the labor market based on factors like region and labor

market segment. It could be that the LGB people who have experienced the worst forms of discrimination (explicit harassment and verbal or physical abuse) have dropped out of the labor market all together. The observed wage differences can only account for the experiences of those currently working. If those who have experienced this discrimination have left the labor market all together, these data will not reflect their experiences. The results would then be biased toward those who have experienced either lesser degrees of discrimination or who have all together positive labor market experiences. In fact, Schneider (1986), in a qualitative study of working lesbians, found that higher incomes were associated with a reduction in the likelihood of assuming the risk of disclosure. However, more recent studies on workplace disclosure (Griffith and

Hebl 2002, Wax et al. 2018) have not considered income as a factor. While it is likely that, in the thirty-five years since Schneider’s study (1986), changing societal acceptance

135

may have shifted the calculus on risk to disclosure, if higher income sexual minorities are less likely to disclose their identities and be more available for direct discrimination, then the findings here of wage premiums for gay men and lesbian women are biased against a finding of discrimination.

These analyses only capture differences in wages, not in the other costs or forms of discrimination. For example, all sexual minorities are more likely than their heterosexual counterparts to live in metropolitan areas, where costs of living tend to be higher. It is possible these living patterns reflect a belief among sexual minorities that, even if they desired to live outside a metropolitan area, they might not be welcome or safe. This amounts to a de facto tax on residency. This study also does not assess the effect that health and disability inequality play in the lives of sexual minorities. Sexual minorities are, on average, less likely to have health insurance than their heterosexual counterparts. This might reflect the higher rates of part-time work among sexual minorities, who therefore are unlikely receive employer-provided health benefits.

Relatedly, these data pre-date the 2020 COVID-19 pandemic which has likely disproportionately impacted sexual minorities with lower rates of health insurance and more precarious part-time work.

Finally, while the analyses performed here examine the divergent experiences of diverse sexual identity categories, future research should take a broader intersectional approach. It is likely that the returns of productivity characteristics for Black and white lesbians, for example, are quite distinct. Examining the intersections of race, gender, and sexual identity should elucidate the unique experiences of these populations.

136

Conclusion

The results of this chapter suggest that discrimination against sexual minorities is persistent but is experienced differently by specific identity categories. While gay men and lesbians appear to receive a wage premium over their heterosexual counterparts, bisexual men and women appear uniquely disadvantaged in the labor market. These findings help to illustrate the prevalence of inequality and discrimination against sexual minorities. It is clear the bisexual men and women continue to experience discrimination in the labor market, even when controlling for differences in productivity characteristics.

That gay men and lesbian show markedly different labor market experiences suggests that discrimination may be operating in unique and variable ways across sexual minority groups. It is also clear that the dramatic age differences between bisexual men and women compared to all other groups is a considerable factor in their labor market

experiences. Going forward, it is important to consider why younger generations appear

more likely to identify as bisexual and how that might affect their labor market

experiences. The finding of a wage premium for gay men here suggests that expanding

social acceptance of sexual minorities continues to have an effect on the lived

experiences of at least some growing number of LGB people.

Presently, I have not engaged with the topic of public policy. That is, could it be

that the favorable labor market outcomes experienced by gay and lesbian workers reflects

their geographic distribution in jurisdictions with more favorable public policy? For the

years these data were collected, there were no federal labor market protections for sexual

minority workers. However, several states had enacted such policies. It is therefore

possible that, by controlling for state-level differences in public policy, we can account

137

for the wage premiums of gay and lesbian workers. It is also possible that, absent public

policies, the observed wage penalties for bisexual men and women would be even starker. The effectiveness of public policy is an important question to consider given contemporary debates about extending federal civil rights protections to sexual minorities. I turn to these questions in Chapter Four.

138

CHAPTER 4

MEASURING THE EFFECTS OF STATE NONDISCRIMINATION POLICIES

ON WAGES OF SEXUAL MINORITIES

Introduction

Gerald Bostock loved his job as the child welfare services coordinator for the

Juvenile Court of Clayton County, Georgia. “I was in a dream job. I loved working on

behalf of underserved children” (Bostock 2020). In 2013, ten years after beginning his

work for the county, Bostock lost his dream job. “I was fired. I lost my job, and my

medical insurance, while in recovery after my cancer treatment (Bostock 2020). Earlier

that year, inspired by his cancer diagnosis to be more active, Bostock had joined the

Hotlanta Softball League, a local sports organization for lesbian, gay, bisexual,

transgender, and queer (LGBTQ) Georgians. “But in the months that followed, Mr.

Bostock's participation in the gay Softball league and his sexual orientation were openly

criticized by someone with significant influence in the Clayton County court system” and on June 7, 2013, he was fired (Bostock v. Clayton County, Georgia 2018:6-7). At the time, Georgia was one of twenty-eight states which did not outlaw discrimination on the basis of sexual orientation, and it was perfectly legal for an employer to fire an employee because of the employee’s sexual orientation. Because there were no state protections against the discrimination he experienced, Bostock had to turn to the federal courts.

Until June 2020, when the United States Supreme Court issued its landmark opinion in Bostock v. Clayton County, Georgia, there were no universally applicable federal protections against discrimination for sexual minorities. For workers who

139

experienced discrimination, whether they could seek relief from the state was mostly dependent on where they lived. If their state had a policy, an employee could file a complaint with the state agency tasked with adjudicating such matters or file suit in a state court, depending on how the policy was structured. For employees in states without a policy, there was often little to no recourse to be sought. Protections against discrimination were a patchwork of different policies and jurisdictions.

I seek to understand how these policies, where enacted, affect the labor market experiences of sexual minorities. I am interested in both the symbolic and practical effects of such policies. How are these laws written? What does the text of policies and statutes tell us about the state’s relationship to sexual minorities? I am also interested in the effectiveness of such policies in reducing workplace discrimination against sexual minorities. Do nondiscrimination polices affect the labor market experiences of LGB workers? In Chapter Three, I observe that gay men and lesbian women experience a wage premium over heterosexual men and women, respectively, whereas bisexual men and women experience a wage penalty. In this chapter I ask whether sexual orientation nondiscrimination laws affect or account for these differences. Are the effects of such policies experienced similarly by different sexual identity categories?

My Contributions

Gerald Bostock took his case all the way to the Supreme Court, which ruled that the protections against sex discrimination in Title VII of the Civil Rights Act of 1964 extend to discrimination based on sexual identity. Since that ruling, there are broad

140

federal protections for lesbian, gay, and bisexual (LGB) workers.6 While federal policies

are different from state policies in their breadth of coverage and enforcement, examining state policies and their effects on labor market discrimination can provide insight into how this monumental change in public policy may be felt by workers. Despite the change in federal policy, there are still several states that do not have state-level policies. While federal policy effectively supersedes state policy, there are potential benefits to having state policies in place alongside federal policy, including expanding enforcement mechanisms. Examining the symbolic and practical effects of such policies can inform these continued policy debates.

A small body of literature has examined the effectiveness of state nondiscrimination polices for sexual minorities. The existing body of literature has, however, relied on household-level data which, by their nature, omit several categories of sexual minorities: same-sex couples who do not indicate they are married or unmarried partners on the survey; sexual minorities who do not reside with their partner or who do not have a partner; and bisexual men and women in opposite-sex relationships. My data allow me to examine policies’ effects regardless of respondent’s household structure. I am also able to distinguish between categories of sexual identity, examining how policies may differentially impact gay men, lesbian women, bisexual men, and bisexual women.

To my knowledge this is the first such study to do so.

6 The Supreme Court also ruled in Bostock, which consolidated three similar cases, that Title VII protects against discrimination based on gender identity. Because of the nature of my data and the conceptual differences between sexual identity and gender identity, I do not include transgender people who do not also identify as LGB in my analyses. However, it is important to acknowledge the historic nature of the Supreme Court’s opinion for trans rights. In this dissertation, I refer to LGB people when only talking explicitly about sexual minority populations. When discussing broader themes or social movements, I use LGBT or LGBTQ. 141

My study better reflects the reality of workers’ availability for discrimination in

the labor market and provides a more complete picture of the state of sexual orientation

discrimination. As new state and federal policies are debated and implemented, my study

informs debates about the potential effectiveness of such policies. My study also

complements the broader sociological literature on labor market inequality and

discrimination by demonstrating the unique impact these phenomena have on different

sexual identity groups. Whereas previous literature tended to group bisexual men and

women with gay men and lesbian women, respectively, my data allow me to distinguish

between these sexual identity groups and compare their different experiences.

History of Sexual Orientation Nondiscrimination Policies

Throughout the twentieth century and into the twenty first, most public policy that addressed sexual identity sought to limit the rights of sexual minorities. Sodomy bans, which date back to the colonial era and prohibited sex between people of the same gender, were active in every state until Illinois became the first state to repeal their ban in

1962. During various periods and in various forms, sexual minorities were restricted from joining the military, working in the civil service, accessing federal welfare programs, and were even banned from entering the country as immigrants (Canaday 2009). At the ballot

box, anti-LGBTQ initiatives were common and often successful (Stone 2012). The courts

were also frequently hostile to sexual minorities (Gluck Mezey 2007). By the mid-

twentieth century, LGBTQ activists began to advocate and agitate for civil rights. Groups

organized to stage sit-ins and protests. Political organizations formed to influence

politicians and the media. By the latter decades of the twentieth century, some

142

municipalities and states began to enact policies banning discrimination based on sexual

identity.

Federal Policy

While the June to July 1969 Stonewall Riots in New York City represent a defining moment in the progression of a then nascent LGBTQ rights movement in the

United States, the origins of this movement go back nearly two decades prior to those infamous nights. LGBTQ people first started organizing partially in response to the discriminatory policies and practices of the federal government. In an era in which other identity groups began to coalesce into successful movements for civil rights, early

LGBTQ activist groups “grew out of the notion that homosexuals could and should be able to pursue rights in the same fashion as other minority groups in the United States”

(Armstrong 2002:33). The demand for protection from discrimination in the workplace predates Stonewall. Early protests and legal fights focused on the rights of LGBTQ workers, and in 1968 representatives from several LGBTQ organizations (known at the time as homophile organizations) drafted a “Homosexual Bill of Rights” which included nondiscrimination in public and private employment, including the military (Gregory

2001:155-6).

The first introduction of LGBTQ rights legislation in the United States Congress came when Congresswoman Bella Abzug (D) of New York, a champion of civil rights, introduced the Equality Act of 1974. The bill would have added sexual orientation and marital status to the list of protected classes in the Civil Rights Act of 1964 and would have expanded the protections based on sex beyond the employment protections of Title

VII (Feldblum 2000:152-3). This broad legislation not only would have offered

143

employment protections to LGB workers in the public and private sector, but also would have created protections in the areas of housing and public accommodations, among others. The bill was never brought up for a hearing in committee and died during that session of Congress. Abzug re-introduced the legislation in subsequent sessions, separating the sexual orientation protections from the other proposed additions, until she lost her seat in the late 1970s. The mantle was then taken up by Congressman Ed Koch

(D-NY) and others.

It was not until 1980 that a congressional hearing was conducted on civil rights legislation for LGBTQ workers. While hearings brought broader awareness to the proposed legislation, they also inspired greater attention from opposition groups on the

conservative right, and no votes were held on the legislation. Any momentum that might

have been building towards the passage of a civil rights bill for sexual minorities was

stalled in response to the HIV/AIDS epidemic. Activists were forced to turn their

attention to the pressing issues associated with the disease, the failed government response, and the discrimination experienced by people with AIDS (PWAs). Successful campaigns to win protections for PWAs, as well as the passage of the Americans with

Disabilities Act (ADA), reinvigorated activists’ hopes for federal legislation prohibiting anti-LGBTQ discrimination (Feldblum 2000).

The political failure of the Clinton administration to lift the ban on sexual

minorities from serving in the military and the sweeping takeover of the Congress by

conservative Republicans in the 1994 midterm elections once again halted prospects of

passing a broad civil rights bill for sexual minorities. Activists turned their attention

specifically towards enacting employment protections, setting aside protections in

144

housing, and access to financial services. According to Feldblum (2000), because of the nearly two decades of stagnation, “the painful decision was made to excise all portions of the omnibus bill other than the employment section” (178). The result was the

Employment Non-Discrimination Act (ENDA), a bill that only dealt with employment protections.

The new proposed legislation stated as its purpose “to provide a comprehensive

Federal prohibition of employment discrimination on the basis of sexual orientation” (S.

2238). ENDA’s exclusive focus on employment garnered greater support from legislators. First introduced in June of 1994 by Senator Edward (Ted) Kennedy (D-MA) and Representative Gerry Studds (D-MA), the bill ended the 103rd Congress (1993-1995) with 30 cosponsors in the Senate (S.2238) and 137 cosponsors in the House (H.R. 4636).

The bill did not receive a vote during that congress, but it was re-introduced in the 104th

Congress (1995-1997). The first vote on ENDA came in 1996 as part of negotiations for the Defense of Marriage Act (DOMA), the law defining marriage for federal purposes as exclusively the union of a heterosexual couple. According to Feldblum (2000):

The biggest break for the vote on ENDA resulted from the political machinations surrounding DOMA. The Republican leadership wanted to bring DOMA up for a vote, but were not interested in subjecting the bill to potentially embarrassing votes on gun control or health care. Thus, the Republican leadership offered Senator Edward Kennedy, who was leading the fight for ENDA, a simple deal: an up-or-down vote on DOMA, with no extraneous amendments to be offered to DOMA (other than ENDA), in return for an up-or-down vote on ENDA, without amendments to be offered to ENDA. (185)

While DOMA passed overwhelmingly (342 to 67 in the House and 85 to 14 in the

Senate), ENDA failed in the Senate by a vote of 49 to 50 with one likely supporter absent from the vote. Vice President Al Gore had been planning to vote in favor of the bill

145

should a tie occur but did not get the opportunity, a result Feldblum (2000), herself an

architect of the bill, called “heartbreaking” (185). The defeat of ENDA and the passage of

DOMA were major setbacks for LGBTQ activists. ENDA was re-introduced throughout

the 1990s, gaining 173 cosponsors in the House (H.R. 2355) and 36 cosponsors in the

Senate (S. 1284) by the turn of the century.

ENDA continued to be introduced throughout the 2000s except for the 109th

Congress (2005-2007). As a result of successful lobbying by the transgender community,

ENDA was reintroduced in the 110th Congress (2007-2009) with protections based on

gender identity included for the first time. “Members of the transgender community were outraged when ENDA was introduced in 1994 without explicitly providing employment protection [for gender identity]” and they fought diligently for the next decade and a half to have gender identity included in the bill (Feldblum 2000:183). However, the trans-

inclusive bill failed to pass out of committee, and the sponsors re-introduced a version of

ENDA that stripped out protections based on gender identity, a move that divided activists and lawmakers. The trans-exclusionary version of the bill passed the House by a vote of 235 to 184 but died in the Senate under threat of a veto from President George W.

Bush. A trans-inclusive ENDA failed to get out of committee in the 111th (2009-2011)

and 112th (2011-2013) Congresses and was most recently introduced in the 113th

Congress (2013-2015). During that session, ENDA received a historic number of cosponsors: 205 in the House (H.R. 1755) and 56 in the Senate (S. 815). The Senate passed its version 64 to 32, but the bill never received a vote in the House.

In a return to the past, legislation like that introduced by Congresswoman Abzug

in the 1970s was introduced in 2015. The Equality Act of 2015 (S. 1858, H.R. 3185)

146

sought to amend the Civil Rights Act of 1964 to add sexual orientation and gender

identity as protected classes in all aspects of the legislation. The bill sought broadly “to

prohibit discrimination on the basis of sex, gender identity, and sexual orientation” and

would add protections in employment, public accommodations, federal funding, housing,

and access to credit (S. 1858). In addition to the much broader protections included in the

Equality Act as compared to ENDA, it is thought that revising the Civil Rights Act of

1964 and including sexual orientation in Title VII’s employment protections represents a

stronger legislative remedy to discrimination.

While ENDA, in its most recent incarnation, prohibited discrimination by

employers, employment agencies, and unions, it included exemptions for religious

organizations, prohibited voluntary affirmative action plans, and precluded employees

from making disparate impact claims, that is, seeking recourse when a policy is not

discriminatory on its face but is in its outcomes. Further, Jasiunas (2000) suggests that

enacting a stand-alone bill like ENDA would leave it open to differential interpretation

by the courts. The mere act of passing a bill separate from Title VII might signal to the

courts that the bill was not intended to offer the same level of protections. Because the

courts were viewed as hostile to sexual minorities at the time, it was assumed “that the

protection offered by ENDA [would] fall far below that offered under Title VII”

(Jasiunas 2000:1537).

It is worth briefly highlighting some of the arguments made by opponents of

ENDA or other prohibitions on sexual orientation discrimination. Many conservatives

oppose nondiscrimination protections based on sexual orientation because they fear passing such policies would signal approval of homosexuality. Public opinion,

147

throughout the twentieth century, was divided: Americans supported employment rights for sexual minorities but did not support homosexuality generally (Feldblum 2000:154-

5). By protecting LGB workers from discrimination in the workplace, it was argued, the government would be giving legitimacy to immorality. As early as 1980, conservatives were concerned that sexual orientation nondiscrimination protections might lead to same- sex marriage, affirmative action policies for sexual minorities, or other “special rights.”

Most intriguingly, two consistent lines of argument against nondiscrimination policies based on sexual orientation are the contradictory positions that LGB people are an elite class who do not need protection and that enacting protections would lead to a surge in discrimination claims. The view that LGB people are an affluent elite, what

Badgett (1995) calls a “myth of privilege,” has a long history (1). “This materialization helps legitimate sentiments that gays and lesbians are motivated by questionable ends and undeserving of political or legal reward” (Goldberg-Hiller 2008:225). The myth emerges from a confluence of movement activists seeking to bolster their political power and marketing research which improperly generalizes some segments of LGB communities as representative of all sexual minority experiences.

Distorted views of broad LGB affluence persist to this day. Within recent years an article was published in the Philadelphia Inquirer titled “Chasing LGBT Dollars” which suggested “businesses are eager to cater to the community, with its high income and generous spending patterns” (Parmley 2015:E5). This view has led some to suggest that if

LGB people are a wealthy elite then they must not be experiencing discrimination and thus protections are unneeded. Feldblum (2000) documents the ways in which this belief about gay affluence has been interjected into legislative debates dating to the first

148

hearings in the 1980s (161-2, 181). As discussed in Chapter Three, there is significant

evidence for discrimination against sexual minorities, especially bisexual men and

women. Further, an issue brief from the Center for American Progress suggests lesbian women face high levels of poverty, consistent with the trends discussed in Chapter Two

above (Quintana 2009).

And yet an additional argument against employment protections for sexual

minorities is the fear that they will lead to a burdensome increase in claims of

discrimination. This argument contends that, if nondiscrimination protections are

extended to sexual minorities, the systems in place to adjudicate these claims will be

overwhelmed with filings. This argument is predicated on the assumptions that either

LGB people are prone to filing bad faith claims of discrimination (for which there is no

evidence) or that discrimination against LGB people is extremely widespread (for which

there is evidence). However, Rubenstein (2001) compared complaint filings at the state

level and found that, adjusted for relative size of the community, LGB people were filing

discrimination claims at similar rates to complaints filed for race and gender

discrimination.

Despite the failure to pass legislation protecting LGB employees nationally, there

were some limited non-legislative remedies exercised at the executive level prior to the

Supreme Court’s opinion in Bostock. As mentioned in Chapter Three above, President

Bill Clinton issued several Executive Orders which established nondiscrimination policies for the federal government. President George W. Bush for the most part continued those policies, and in 2014 President Barack Obama expanded protections by issuing an Executive Order which banned sexual orientation discrimination among

149

private employers that contract with the federal government (EO 13672). However,

Infanti (2007) points to the “fragility of protections implemented through executive order: Such protections are subject to change or elimination whenever the occupant of the

White House changes” (109). Executive authority is also subject to review by a potentially hostile judiciary. While federal nondiscrimination legislation languished, several states implemented policies to fill the gap and protect LGBTQ workers.

State Policy

Prior to the Bostock decision, because of “the lack of any comprehensive federal structure extending protection to [LGB workers] … the most viable means of legal protection [was] the adoption of state and local laws, and the enforcement of those laws by local agencies and state courts” (Editors of the Harvard Law Review 1990:169). State policies have proliferated since the 1970s, though the trajectory of their enactment has not always been a straightforward endeavor. State-level policies are diverse in their scope and enforcement. States can enact protections via legislation, which typically covers both public and private employers, or through Executive Orders, which only cover public employees (those working for the state government). State-level policies are also vulnerable are also vulnerable to the whims of executives and electorates. In this section I highlight the history of state policies through informative case studies, examining policy types, processes of enactments, and their vulnerabilities.

The first state to offer explicit protections for LGB workers was Pennsylvania, when Governor Milton Shapp (D) issued an Executive Order in 1975 (EO 1975-5,

4/23/1975) announcing his commitment to end sexual orientation discrimination in his administration. The order was later revised (EO 1975-5, 2/11/1976) and amended (EO

150

1975-5, 9/19/1978) to explicitly bar state agencies from discriminating based on sexual

orientation and to empanel a council to hear complaints and report to the governor.

Governor Edward Rendell (D) later reaffirmed the policy and provided for enforcement

via the state’s Bureau of Equal Employment Opportunity (EO 2003-10, 7/28/2003).

In 1977 the District of Columbia passed the Human Rights Act of 1977, making it

the first state to legislatively ban sexual orientation discrimination in public and private

employment (DC Code § 2-1402.11).7 In 1982 Wisconsin became the second state to

enact a similar policy (1981 AB70, Wis. Stat. §§ 111.31 and 111.36(1)(d)). By the end of

the 1980s, only Massachusetts had joined the District of Columbia and Wisconsin in

enacting legislative protections (M.G.L. c. 151B, § 4(1) and (3)). However, throughout

the 1970s and 1980s seven additional states joined Pennsylvania by adding protections for state workers via an Executive Order by their governors: California (EO B-54-79,

4/4/1979), New York (EO 28, 11/18/83), Ohio (EO 83-64, 12/30/1983), New Mexico

(EO 85-15, 4/1/1985), Rhode Island (EO 85-11, 5/30/1985), Washington (EO 85-09,

12/24/1985), Oregon (EO 87-20, 10/15/87). A common trajectory for policy adoption is

for states to first protect public workers via an Executive Order which is followed later by

legislative protections for private-sector workers. Six out of the eight Executive Order states from the 1970s and 1980s have followed this pattern and extended subsequent legislative protections. The exceptions are Pennsylvania and Ohio.

The histories of policy enactment in several states offer important cases studies in the limitations of executive branch policies. As with presidential Executive Orders, these

7 Because of its role as the federal district, large population, and unique legislative powers, I treat the District of Columbia as a state rather than a municipality. 151

policies are vulnerable to changes of the executive office holder. In Ohio, for example,

Governor Dick Celeste (D) signed Executive Order 83-64 (12/30/1983) which established a policy of sexual orientation nondiscrimination within state agencies. An Executive

Order issued by Governor Bob Taft (R) revoked the previous order and established a vague policy of nondiscrimination which did not mention sexual orientation but referenced “many other groups and classifications of persons that could be subject to discrimination but are not expressly protected by state or federal law” (EO 99-18T,

6/23/1999). This order was later revised to replace references to “groups and classifications of persons” with the vaguer still “persons” (EO 99-25T, 8/11/1999). In

2007, an Executive Order issued by Governor Ted Strickland (D) once again put in place public-sector employment protections based on sexual orientation (EO 2007-10S,

5/17/2007). Governor John Kasich (R) reaffirmed the policy established in 2007 but dropped protections based on gender identity (EO 2011-05K, 1/21/2011) only to add them back several years later (EO 2018-12K, 12/19/2018). Over this 30-year period, state workers in Ohio oscillated between periods with and without protection from discrimination.

In 1987, Oregon Governor Neil Goldschmidt (D) issued that state’s first

Executive Order granting employment protections to LGB state workers (EO 87-20,

10/15/87). A conservative political action organization successfully placed a measure on the 1988 general election ballot asking voters to revoke the Executive Order and forbid the future implementation of nondiscrimination policies based on sexual orientation

(Stone 2012:19). After the measure passed on November 8, 1988, Oregon was left without any sexual orientation employment protections until 1992, when a state appellate

152

court ruled the 1988 voter measure was unconstitutional, essentially reinstating the

Executive Order (841 P.2d 646 (1992), 116 Or. App. 258). It was not until 2007 that the

Oregon legislature codified protections into state law and extended protections to private-

sector employees.

In Iowa Governor Tom Vilsack issued a similar Executive Order (EO 7,

9/14/1999). However, the state legislature based a bill repealing the order, which the governor vetoed, only for a state judge to rule the initial order an unconstitutional exercise of executive power. The Iowa legislature eventually added protections to the state’s statutes (Iowa Code § 216.6). As these examples illustrate, just as Executive

Orders are vulnerable to the whims of changing executives, policies are also often vulnerable to the animus of the electorate, interventions by the state’s legislature, review by the judiciary, or a combination thereof.

In the 1990s the legislatures in Connecticut (Conn. Gen. Stat. § 46a-81c), Hawaii

(H.R.S. § 378-2), Nevada (N.R.S. §§ 233.010 and 613.330), New Hampshire (N.H.

R.S.A. §§ 354-A:6 and 354-A:7), and Vermont (21 V.S.A. § 495) added sexual orientation nondiscrimination policies. The governors of Colorado (EO D0035,

12/10/1990), Illinois (EO 1996, 11/8/1996), Maryland (EO 01.01.1995.19, 7/17/1995),

Minnesota (EO 91-4, 1/29/1991), and New Jersey (EO 39-1991, 8/16/1991) all issued

Executive Orders banning discrimination and the state legislatures subsequently passed laws extending those protections to private-sector employers.8 Just as Executive Orders

are vulnerable to the electoral process, as evidenced by the history of Oregon, so too are

8 Colorado (Colo. Rev. Stat. § 24-34-402), Illinois (§§ 775 ILCS 5/1-103(O-1) and (Q), 5/2102(A), (B) and (C)), Maryland (Md. Code Ann., State Gov’t. §§ 20-601 – 20-609), Minnesota (M.S.A. § 363A.08), New Jersey (N.J.S.A. §§ 10:2-1, 10:5-3, 10:5-4, 10:5-6, 10:5-8, and 10:5-12). 153

legislative protections at risk of being overturned by voters. Maine offers a compelling example of this.

In 1995, voters in Maine rejected a ballot initiative which would have precluded the state government from including sexual orientation as a protected class and would have invalidated any existing municipal protections that included sexual orientation (L.D.

310, I.B. 1). Perhaps interpreting this rejection as a desire to include such protections, in

1997 the Maine legislature passed a bill adding sexual orientation as a protected class in the state’s employment law, and the governor signed it (PL 1997, c. 205). Maine has multiple mechanisms by which voters can directly intervene in the legislative process, including a veto by referendum. After the legislature passed sexual orientation protections, Mainers successfully petitioned to have a veto of the law put before voters, and in February 1998 voters approved the veto, eliminating the protections.

In the next legislative session, lawmakers again passed a bill adding sexual orientation protections, but this time they explicitly put the bill before voters to approve via referendum before it could be enacted. In November 2000, voters again rejected adding sexual orientation employment protections to state law. In 2001, Governor Angus

King (I) issued a Civil Service Bulletin, essentially a human resources policy for the state government, affirming equal employment opportunity based on sexual orientation (CSB

13.4B, 5/1/2001). In 2005 the state legislature passed, for the third time, a bill adding employment protections based on sexual orientation which the governor signed (PL 2005, c. 10). As in 1997, a veto campaign was initiated. However, in November 2005, the veto attempt failed, and Maine officially added employment protections for LGB workers (5

M.R.S. §§ 4552, 4553(9-C) and (10), 4571, and 4572).

154

It is worth remembering that the back-and-forth swings of policies have material and psychic effects on those whose lives are subjected to such political debate. For many,

LGBTQ activists and passive citizens alike, these debates provoke a great deal of tension, which, when left unchecked, can boil over. California offers one such example. In 1991, the California legislature passed a bill (AB101) which would have added sexual orientation to that state’s Fair Employment and Housing Act (FEHA), essentially extending Title VII protections within the state (Dickey 1993:2297). However, Governor

Pete Wilson (R) vetoed the measure, an action which “set off weeks of angry

demonstrations in Los Angeles and Sacramento and near-riots in San Francisco, where

protesters set fires and smashed windows at a state office building and caused $250,000

in damage” (Gross 1992). The following year Governor Wilson relented and signed a law

(AB2601) adding sexual orientation protections to the state’s labor code (CA Labor Code

§1102.1). FEHA was eventually amended with the passing and signing of AB1001 in

1999 (Chapter 592, Statutes of 1999).

Throughout the 2000s states continued to add and to revoke employment

protections for LGB workers. Executive Orders were issued in Alaska (AO 195,

3/5/2002)9, Arizona (EO 2003-22, 6/21/2003), Indiana (Governor’s Policy Statement,

4/26/05), Michigan (EO 2003-24, 12/23/2003), Missouri (EO 10-24, 7/9/2010), and

Montana (EO 41-2008, 11/14/2008) to provide state workers with protections. In

Louisiana, Executive Orders have twice been issued only to expire in subsequent

administrations (EO 92-7, 2/17/1992; EO KBB 2004-54, 12/6/2004). Governor Kathleen

Sebelius (D) of Kansas issued an Executive Order (07-24, 8/21/2007) which was

9 AO 195 is an administrative order and may lack the force of law. 155

rescinded by her successor, Governor Sam Brownback (R) (EO 15-01, 2/10/15), only to be reinstated by Governor Laura Kelly (EO 19-02, 1/15/2019). In Kentucky protections for state employees were similarly added (EO 2003-533, 5/29/2003), dropped (EO 2006-

402, 4/11/2006), and then added again (EO 2008-473, 6/2/2008). A similar pattern occurred in Virginia (EO 1-2002, revised 12/16/2005; EO 6-2010, 2/5/2010; EO 1-2014,

1/11/2014) before Virginia passed legislative protections in 2020 (Virginia Code § 15.2-

1500.1). Delaware’s governor issued an Executive Order (EO 83, 12/2000) and the state legislature subsequently approved legislative protections (19 Del. C. § 711). In 2015 Utah also enacted legislative policies protecting LGB workers in the public and private sectors

(Utah Code 34A-5 § 106). Figure 4-1 summarizes policy types by state during the time of my data (2014-2018).

This detailed history and the extended case studies illustrate the inconsistent and evolving process of extending employment protections to LGB workers. Some states follow a clear progression from no protections to protections for public-sector workers

(via Executive Order) to protections for public and private-sector workers (via legislation). Some states have stalled at an Executive Order without ever progressing to protections for public-sector workers. Pennsylvania has had an Executive Order for 45 years without extending protections to private-sector workers. Still other states show that the progress is not always linear. States adopt policies and rescind them and occasionally adopt them again. Executive and legislative policies alike are vulnerable to interventions by voters, the judiciary, or changing political actors. We are left with the possibility that rights anywhere are always at risk of being rescinded by government actors or by the whims of an electorate.

156

157

Figure 4-1. State Policies by Type (2014-2018).

Municipal Policy

The first jurisdictions to enact nondiscrimination policies were in fact

municipalities. Indeed, municipal governments have been the places where the highest

volume of protections have been enacted. In 1972 East Lansing, Michigan, became the

first city to pass sexual orientation protections (Klawitter and Hammer 1999:22). This

was followed by a wave of similar policies adopted in other liberal university towns

across the country (Button, Rienzo, and Wald 2000:271). According to the Human Rights

Campaign (HRC), an LGBTQ lobbying organization, as of January 2018, there were at

least 225 cities and counties with some sort of employment protection for LGBTQ

workers.10 The municipalities identified by the HRC include several in jurisdictions with

no state-level policies.

While municipal protections are more prolific, they are often less forceful than state policies. “State statutes are more politically secure, do not run as great a risk of preemption, and necessarily have a broader reach and more uniform application”

(Jasiunas 2000:1534-5). Municipal policies are more easily repealed and can be superseded or revoked by state laws. Because most cities govern by virtue of powers vested in them by the state, state governments have extraordinary leeway to intervene in municipal policies. A recent illustration of the supremacy of states over municipalities occurred in Arkansas. Arkansas offers no state-wide employment protections based on sexual orientation. In 2014 the city of Fayetteville passed a nondiscrimination ordinance;

however, the voters in that city forced the ordinance to a referendum where it was

10 http://www.hrc.org/resources/cities-and-counties-with-non-discrimination-ordinances-that-include- gender 158

repealed. The state legislature, hostile to LGBTQ rights and vexed by perceived

municipal overreach, passed Act 137 which declares, “A county, municipality, or other

political subdivision of the state shall not adopt or enforce an ordinance, resolution, rule,

or policy that creates a protected classification or prohibits discrimination on a basis not

contained in state law.” Because Arkansas does not include sexual orientation as a

protected class in state law, Act 137 nullifies municipal protections in the state, though

none exist, and precludes municipal governments from enacting them in the future.

As occurred in Fayetteville, Arkansas, prior to state intervention, municipal

policies are often subject to repeal votes by the citizens of the towns or cities or the wider

state population. The first repeal effort that attracted national attention was the campaign

in Dade County, Florida, in 1977. Spurred by the activism of the singer Anita Bryant, the successful repeal became a galvanizing moment, highlighting the ability of conservative political activists to challenge the tide of gay rights (Button, Rienze, and Wald 2000:272).

Into the twenty first century, repeal through referendum has been a common tactic in cities large and small as evidenced in Fayetteville, Arkansas, and Houston, Texas, where in 2015 the Houston Equal Rights Ordinance (HERO) was repealed by voters.

Furthermore, even in municipalities where the local government and citizens approve of employment protections for LGB workers, the wider state population can still interfere. Such was the case in Colorado where state-wide Amendment 2 “eliminated the handful of existing protections based on sexual orientation, but, more importantly, it placed an absolute bar to the ability of gays, lesbians, and bisexuals to seek such protection in the future” (Donovan, Wenzel, and Bowler 2000:174). This ban was challenged in state and federal courts before ultimately being overturned by the United

159

States Supreme Court in the first LGBTQ victory before the highest court (Romer v.

Evans 1996).11

Municipal ordinances typically have little power over federal entities. For

example, in the early 1980s, the City of Philadelphia sought to prohibit the Temple

University School of Law from using its facilities to place students with the Judge

Advocate General Corps (JAG Corps) of the United States military because a Defense

Department directive prohibited service by sexual minorities. The city viewed this as a

violation of the Philadelphia Fair Practices Ordinance, which outlawed discrimination

based on sexual orientation by employment agencies. The university and the United

States government sued the city, winning rulings at the district and appellate courts which enjoined the city from enforcing its nondiscrimination policy on the law school.

Some research has found correlates between the characteristics of a city and its

likelihood of enacting an antidiscrimination ordinance. Klawitter and Hammer (1999)

study the “spatial and temporal diffusion of local government adoption rates of [sexual

orientation] antidiscrimination policies for private employment” (23). They examine

where and when municipal policies are likely to be enacted based on several conditions.

Using multivariate models, the authors find that the presence of state-level policies

decreases the likelihood of local-level policy adoption (though not statistically

significantly), that regional differences affect the diffusion of ordinances, and that larger,

more urban populations show a greater likelihood of enacting protections (Klawitter and

11 The Supreme Court struck down Colorado’s Amendment 2 because it specifically targeted LGB people from exclusion in nondiscrimination policies. Laws like Arkansas’s Act 137, quoted above, are written in such a way as to exclude LGBTQ people without specifically naming them, thus passing constitutional muster. 160

Hammer 1999:32-3). Dorris (1999) finds that “cities with higher proportions of educated citizens are considerably more likely to have the policy” as are cities that are more racially heterogeneous (51). Given the dramatic shifts in policies and public opinions about LGBTQ people in the more than twenty years since these studies were undertaken, it is possible that the patterns identified by Klawitter and Hammer (1999) and Dorris

(1999) are no longer representative of where and when municipal polices get enacted.

The rejection of employment protections in large, racially diverse cities such as Houston and the acceptance of policies in small communities such Eureka Springs, Arkansas, whose council unanimously passed a nondiscrimination ordinance in 2015, might signal such a change.

Over the past half century, sexual orientation nondiscrimination policies have proliferated at various levels of government. While municipal policies are the most numerous, the force of their protections is often limited. State policies, in the form of

Executive Orders and legislation, have been enacted (and revoked) in many places.

Federal policies, which have the greatest reach, have proven more difficult to enact. The history of these policies detailed here highlights the generally precarious nature of nondiscrimination protections. Beyond the presence of policy, it is also important to consider their effectiveness at reducing experiences of discrimination.

Measuring Policy Effectiveness

A small body of literature has attempted to assess whether existing state and municipal policies have been successful at mitigating the effects of discrimination against

LGB workers. Mostly undertaken by economists, these studies tend to employ methodologies which focus on whether policies reduce or eliminate wage gaps between

161

LGB workers and their heterosexual counterparts. However, data limitations lead the conclusions of these studies to be varied and of little generalizability.

Using data from the 1990 Census, Klawitter and Flatt (1998) find no effect of state and municipal policies on the wages of same-sex couples. While they find average wages in geographies with nondiscrimination policies were higher for all couples, multivariate wage regressions found “no evidence that either public or private employment protections significantly improve earnings or household income for men or women in same-sex couples” (673). However, the 1990 Census, the first to allow for the

identification of same-sex couples, has been shown to have serious limitations in its

usefulness for studying same-sex couples. These census data also only allow for

comparing coupled individuals and cannot be used to assess the effects of polices on

single individuals. Furthermore, prior to 1990 only two states had enacted legislative

protections for private sector employees, fewer than ten had issued Executive Orders

protecting state workers, and a comparatively small number of municipalities had

protections in place. The authors note that their findings could be a result of the short

time in which policies had been in place as employment protections may require

additional time to show a measurable effect (Klawitter and Flatt 1998:676).

Gates (2009), using data from the 2000 Census, finds an effect of state policies on the wages of gay men but a lesser effect for lesbian women. He suggests that the observed effect of policy duration and the exclusive effects on the wages of same-sex couples suggests “a causal relationship between sexual orientation antidiscrimination laws and positive effects on the wages of lesbians and gay men” (Gates 2009:11). Like the 1990 Census, the 2000 Census only allows for an analysis of same-sex couples, which

162

limits and likely biases any conclusions. Unlike Klawitter and Flatt (1998), Gates (2009)

only considers state-level policies and does not analyze municipal policies. There is also

no attempt to distinguish between the types of state policies (Executive Order versus

legislation).

Klawitter (2011) also uses data from the 2000 Census but examines interactions

between state and municipal policies. She also finds that state policies have an effect on

the wages of gay men but not on the wages of lesbian women, and she finds no effect

from municipal policies. However, comparing different models does show that including

local characteristics and policies increases the estimated effect of state policies (Klawitter

2011:355). She notes that this presents a challenge for researchers given the difficulties in

maintaining adequate information on municipal policies.

Baumle and Poston, Jr. (2011) perform multilevel analyses using the 2000

Census. They find that “individual-level factors play a far greater role than state-level factors in explaining earnings outcomes” (1022). Examining the cross-level effects of sexual identity and antidiscrimination policies, they find that policies have an effect for partnered gay men but not for partnered lesbian women. Policies narrowed the observed gap between partnered gay men and partnered heterosexual men but did not have a statistically significant effect on the difference between wages of partnered lesbian and heterosexual women.

Martell (2013b) examines only state policies and gay men. He uses the General

Social Survey (GSS) and looks at men’s sexual histories to determine behaviorally gay men. Using wage decomposition equations and difference-in-difference estimation, he finds that gay men’s wage penalty compared to heterosexual men is not significant in

163

states with nondiscrimination policies but is significant in states without nondiscrimination policies. “This suggests that [policies] decrease wage differentials by decreasing discriminatory treatment experienced by behaviorally gay men” (Martell

2013b:166). This study does attempt to account for variations in state policies. While the use of the GSS allows for individual-level analyses, the operationalization of sexuality based on sexual histories and the small sample sizes are limitations.

The dramatic shifts in the political context of LGBTQ rights in the first two decades of the twenty first century would further suggest that additional analyses are required. The increase in state-level policies, the dramatic implications of Supreme Court decisions, and overall changes in public opinion continue to alter the social and political landscape. As policy priorities among LGBTQ activists shift from the successful marriage equality campaign towards other areas, including the passage of federal employment protections, it is important to continue this body of research. I contribute to this literature through my use of a unique dataset which allows for analyses of both partnered and single individuals, has a large enough sample size for between-group comparisons of sexual identity, and has a large enough sample size to perform state-by- state analyses.

Hypotheses

Whereas in Chapter Three I examine national differences in earnings by sexual identity, here I focus on differences by state with an emphasis on the effect of public policy. I examine the role that state-level nondiscrimination policies have in effecting workplace equality as measured by wage distributions. I hypothesize that:

164

H1: LGB workers’ wages will be more comparable to heterosexual

workers’ wages in states with an explicit sexual orientation

nondiscrimination policy than in states without an explicit

nondiscrimination policy.

In addition to the presence of a policy, it may be that the type of policy and the duration since its implementation may have effects on wage differences. Therefore, I hypothesize that:

H2: LGB workers’ wages will be more comparable to heterosexual

workers’ wages in states with legislative policies than in states with

Executive Orders.

H3: LGB workers’ wages will be more comparable to heterosexual

workers’ wages in states where nondiscrimination policies have been in

effect longer than in states where policies have been in effect for shorter

periods.

Several factors other than policy might affect variations in earnings across geographies and it is possible these factors may not be distributed randomly across states or across sexual identity groups within states. To assess the effects of public policy, it is important to account for these other potential factors which might impact state-level differences in earnings. Broader variations in state economies likely influence the earnings of sexual identity groups therein. It is possible that in more prosperous states,

165

economic rewards are distributed more equitably across sexual identity groups. I

hypothesize that:

H4: LGB workers’ wages will be more comparable to heterosexual

workers’ wages in states with better overall economic outcomes than in

states with worse overall economic outcomes.

Another factor that potentially influences wage differences and varies across

geographies is the political and social climate, especially as relates toward LGB equality.

It is possible that states with a more welcoming political and social climate for LGB

people have a more equitable distribution of wages. I hypothesize that:

H5: LGB workers’ wages will be more comparable to heterosexual

workers’ wages in states with more LGB-friendly political and social

climates than in states with more hostile political and social climates.

Data

To test these hypotheses, I use the dataset developed in Chapter Two of this

dissertation. Through Cross-Survey Multiple Imputation (CSMI), I impute sexual

orientation into the recipient American Community Survey (ACS) using the National

Health Interview Survey (NHIS) as the donor survey. The ACS does not collect data on

sexual orientation whereas the NHIS includes a measure of sexual identity. By combining

the two surveys I can use information contained in the donor survey that is omitted from

the recipient survey, here the ACS, by design. The result is a large-sample, nationally representative dataset that includes the imputed information about sexual identity. I 166

retrieved the ACS data from IPUMS USA (Ruggles et al. 2021) and the NHIS data from

IPUMS Health Surveys (Blewett et al. 2019). I restrict my sample to those currently

employed (N=7,078,805).

Data about states’ economic and political/social climate were compiled from several sources. I use gross state products and personal income data from the Department of Commerce, Bureau of Economic Analysis. I also use state unemployment data from the Department of Labor, Bureau of Labor Statistics. For measures of state urbanity and educational attainment, I use data from the Department of Commerce, Bureau of the

Census. I retrieved results of the 2016 presidential election from the Federal Election

Commission. I use measures of state religiosity from the Pew Research Center (Lipka and

Wormald 2016). I use a measure of state support for sexual identity nondiscrimination policies from the Public Religion Research Institute (Public Religion Research Institute

2019). I use measures of citizen and government ideology developed by Berry et al.

(1998, 2010) and retrieved from Fording (2018). I use the text of state statues from the

Legal Information Institute (https://www.law.cornell.edu/statutes).

Methods

To measure the effect of state policies on wage differentials by sexual orientation,

I begin by cataloging and analyzing the various state policies. I examine the text of

legislation and Executive Orders to assess differences in their scope and their political/cultural meaning. Then I perform regression analysis of wages to quantitatively measure the effects of policies on wages for sexual identity groups.

I first collect and analyze the relevant policies for all the states and the District of

Columbia. Statues are identified by a Boolean search of “sexual orientation” in the state

167

codes for each jurisdiction. I retrieve copies of Executive Orders from state executives,

archives, or libraries. Once all the policies were complied, I performed a content analysis of the text of the policies. Following Hsieh and Shannon (2005), who define content analysis as “a research method for the subjective interpretation of the content of text data through the systematic classification process of coding and identifying themes or patterns,” I develop a coding scheme to identify themes and patterns in the statutes

(1278). Codes are developed inductively from the texts and then I use the coding scheme to classify each state by the breadth and strength of its policies.

To assess the degree to which differences in wages are a product of states’ economic, cultural/political, and policy differences, I construct nested OLS regression models for the wages of the different sexual identity categories. Following Klawitter and

Flatt (1998) and Gates (2009), I disaggregate state-level characteristics to individuals. To account for the gender differences in family structure and labor market characteristics, I run the regressions for men and women separately. In each model, heterosexual workers are the omitted category.

Variables

Following the rationale and procedures outlined in Chapter Three above, wages

are the dependent variable in all models. Differences in aggregate wages, when all other

productivity characteristics are held constant, suggest discrimination is occurring. Wages

are computed using respondent’s reported total pre-tax salary and wage income as well as their reported usual hours of work per week, both during the previous year. Assuming a fifty-week schedule, wages are calculated as total personal income divided by the product

of hours worked times fifty weeks. Because wages are positively skewed, I take the

168

natural log to normalize the distribution, bottom coding the distribution at $1 prior to

transformation. State-level analyses build on the models produced in Chapter Three and

include all the individual productivity, family structure, and labor market characteristics

described in Table 3-1 above. These include age, experience, educational attainment,

racial identity, Hispanic ethnicity, number of children, number of children under five,

marital status, metro status, region, fulltime/parttime status, and occupation.

The main independent variable is the presence of a nondiscrimination policy in a

respondent’s state of residence. I differentiate between states with no policy, an

Executive Order, and a legislative policy. States in which a policy was rescinded during

the survey years are coded as having no policy. This only affects Kansas and Louisiana

which had Executive Orders revoked during the survey years. I also include a measure of

the number of years since the policy was enacted. States with no policy are coded as zero

on the policy length measure. Table 4-1 summarizes all variables in the regression models.

169

Table 4-1. Descriptive Statistics of Model Variables Individual-Level Variables Variable Estimate Description Income $35,130.00 Median Personal Income from Salary and Wages Wages $17.62 Median Hourly Wages in US Dollars Log of Wages 2.7235 Mean of Natural Log of Hourly Wages

Sexual Identity Weighted Proportion of Sample by Sexual Identity Heterosexual Men 0.5054 Gay Men 0.0133 Bisexual Men 0.0061 170 Heterosexual Women 0.4511

Lesbian Women 0.0101 Bisexual Women 0.0140

State-Level Variables Variable Estimate Description Policy Type Proportion of States with Policies by Type No Policy 0.3333 Executive Order 0.2157 Legislative Policy 0.4510

Years Since Policy Enacted 12.0784 Average Years Since LGB Nondiscrimination Policy Was Enacted

States’ Economic Outcomes Gross State Product $369,000 States' Average GSP for 2014-2018 in Million Dollars Personal Income $49,470 States' Average Personal Income for 2014-2018

Table 4-1. (continued) Variable Estimate Description Unemployment 0.0921 States' Average Annual U-6 Unemployment Measure for 2014-2018

Political/Cultural Climate Citizen Ideology 49.6100 Mean State Score for Citizenry Liberalism State Government Ideology 40.6900 Mean State Score for Government Liberalism 2016 Presidential Election Result 0.4118 Proportion of States Voting Democratic in 2016 Presidential Election Religiosity 0.5467 Average Proportion of State Identifying as Highly Religious Proportion Urban 0.7250 Average Proportion of Residents Residing in an Urban Area Proportion BA+ 0.3149 Average Proportion of State Having at Least a Bachelor's Degree

171 Support for 0.7045 Average Proportion of State Supporting an LGB Nondiscrimination Policy Nondiscrimination Policy

Proportion LGB 0.0379 Average Proportion of State Identifying as LGB Years Since Same-Sex 4.6471 Average Years Since Same-Sex Marriage Enacted in State Marriage Enacted

Same-Sex Marriage Method Proportion of States by Method of Enacting Same-Sex Marriage US Supreme Court 0.2941 Federal Court 0.2941 State Court 0.0784 Legislative Statute 0.2745 Voter Referendum 0.0588

To account for differences in state-level economic outcomes, I control for several factors. A measure of gross state product is included to control for differences in state productivity. A measure of state-level personal income is included to account for differences in individual returns to productivity across geographies. I include a measure of state unemployment because economic health varies across states. I use the Bureau of

Labor Statistics (BLS) U-6 definition of unemployment. “U-6 is the broadest measure of

labor underutilization. In addition to the total number of unemployed and all people

marginally attached to the labor force, U-6 includes people at work part time for

economic reasons (also called involuntary part-time workers)” (U.S. Bureau of Labor

Statistics 2020). I use this alternative measure because its broad definition potentially better reflects the state of unemployment, especially for sexual minorities who may be marginally attached to the labor force due to discrimination (see Chapter Three above).

Gross state product, personal income, and unemployment are averaged across the survey

years of the 2018 ACS 5-year sample (2014-2018).

I include numerous measures of states’ political and cultural climate. It is likely

that states that ban discrimination against sexual minorities have a more tolerant climate

independent of policy. The mere presence of “gay-friendly legislation is a sign of social

support” that likely extends to the broader politics and culture of a state (Baumle et al.

2009:127). Therefore, only accounting for policy presence could overestimate the effect

of nondiscrimination policies. I use two measures of state liberalism developed by Berry

et al. (1998, 2010): a measure of citizen ideology and state government ideology. The

citizen ideology measure is from the “revised 1960-2016 citizen ideology series,” and is

measured at the district level by comparing incumbents’ and challengers’ ideologies to

172

election results (Berry et al. 1998). The state government ideology measure is based on the party identification and interest-group ratings of legislators (Berry et al. 2010). Both scales range from 0 to 100 with higher scores indicating more liberal ideologies. I use the

2016 scores in my analyses. I also include an indicator of whether the state voted for the

Democratic candidate in the 2016 Presidential Election to measure liberalism.

I use religiosity as “a measure of the degree of religious conservatism and, thus, lack of tolerance towards homosexuality” (Baumle and Poston, Jr. 2011:1015). Baumle and Poston, Jr. (2011) and Walther and Poston, Jr. (2004) include the percent Southern

Baptist and Martell (2013b) includes the percent Catholic in their analyses. I choose a broader measure of religiosity to avoid sectarian specificity. While religiosity is not inherently indicative if anti-LGB attitudes, religiosity has been long associated with antipathy towards pro-LGB policies, especially same-sex marriage (Babst 2009).

To specifically estimate a state’s general attitude toward sexual minorities, I include several controls. Whether or not a state has a sexual orientation nondiscrimination policy, I include a measure of the proportion of that state which supports such a policy in theory. Because of the recency and the widespread attention that the movement for same-sex marriage achieved, I use when (year) and how a state enacted marriage equality as measures of support for sexual minorities.12 For the method of

enactment of marriage equality, I construct an ordinal measure based on proximity to the

12 Following the Supreme Court’s decision in United States v. Windsor, which struck down the Defense of Marriage Act (DOMA) the Census Bureau has retained information on same-sex married couples. Prior to Windsor, same-sex couples in the ACS were edited to appear as same-sex unmarried partners, even when legally married in their state of residence. It is unclear if the Census Bureau edited same-couples who indicated they were married but resided in a state where same-sex marriage was not legal from 2013-2015. As such, the marital status indicator may or may not capture some variation in state climate as regards same-sex marriage. 173

state’s populace with the United States Supreme Court being the most distant (0) and a voter-approved initiative being the most proximate (4). I also include the proportion of the state population that identifies as LGB. Following Baumle and Poston, Jr. (2011), I anticipate that higher proportions of LGB residents “would provide certain economic benefits, such as social support, career networks and increased tolerance” (1015).

As described above, numerous municipalities have enacted nondiscrimination policies. I do not, however, directly control for municipal policies. Unfortunately, no definitive database of municipal policies exists, and my data do not include local geographies. Omitting the effects of municipal policies could potentially overestimate the effects of state-level policies. However, as Gates (2009) notes, “state-level policies likely provide a more consistent standard of application and enforcement than do local policies,” and are likely more effective at mitigating discrimination (1). Still, it is worth attempting to account for some of the potential effects of municipal policies. Dorris

(1999) finds that municipal policies are more likely in more urban and highly educated areas. Therefore, I include controls for the proportion of a state that lives in an urban area, as defined by the Census Bureau, and the proportion of the state holding at least a bachelor’s degree.

Regression Models

Model I is a baseline measure of differences in wages by sexual identity controlling for productivity differences among workers (equivalent to Model IV in

Chapter Three). Wages are modeled as:

= + + +

𝑌𝑌�𝑗𝑗 𝛼𝛼 𝛽𝛽0𝑋𝑋0𝑗𝑗 𝛽𝛽1𝑋𝑋1𝑗𝑗 𝜀𝜀

174

where is the natural log of the wages of the jth case; α is a constant; is the sexual

𝑖𝑖 0 0𝑗𝑗 identity𝑌𝑌 effect; is a vector of individual productivity, family structure,𝛽𝛽 𝑋𝑋 and labor

1 1𝑗𝑗 market characteristics;𝛽𝛽 𝑋𝑋 and ε is the regression disturbance.

Model II adds the measure of a state’s nondiscrimination policy and an interaction

term between the policy measure and the sexual identity measure. Because the policy

could influence heterosexual workers’ wages, the interaction term is necessary to isolate

the unique effect of the policy on sexual minority groups (Gates 2009). I also control for

the length of policy. Here I model wages as:

= + + + + + +

𝑗𝑗 0 0𝑗𝑗 1 1𝑗𝑗 2 2𝑗𝑗 3 0𝑗𝑗 2𝑗𝑗 4 3𝑗𝑗 where is𝑌𝑌� the policy𝛼𝛼 𝛽𝛽 measure,𝑋𝑋 𝛽𝛽 𝑋𝑋 𝛽𝛽 𝑋𝑋 is the𝛽𝛽 𝑋𝑋 interaction∗ 𝑋𝑋 term𝛽𝛽 𝑋𝑋 between𝜀𝜀 sexual

2 2𝑗𝑗 3 0𝑗𝑗 2𝑗𝑗 identity𝛽𝛽 and𝑋𝑋 policy type, and is𝛽𝛽 the𝑋𝑋 effect∗ 𝑋𝑋 of policy duration.

4 3𝑗𝑗 It is possible that there𝛽𝛽 is𝑋𝑋 a correlation between states having a sexual orientation

nondiscrimination policy and differences in overall state-level economic outcomes. The

direction of the effect could go either way: states with better economies may be more

likely to have nondiscrimination policies, or states with nondiscrimination policies may

be more likely to have better economic outcomes. Model III controls for differences in gross state product, state personal income, and state unemployment rates. Wages are modeled as:

= + + + + + + +

𝑗𝑗 0 0𝑗𝑗 1 1𝑗𝑗 2 2𝑗𝑗 3 0𝑗𝑗 2𝑗𝑗 4 3𝑗𝑗 5 4𝑗𝑗 where 𝑌𝑌� is𝛼𝛼 the vector𝛽𝛽 𝑋𝑋 of 𝛽𝛽state𝑋𝑋 -level𝛽𝛽 𝑋𝑋economic𝛽𝛽 𝑋𝑋 outcomes.∗ 𝑋𝑋 𝛽𝛽 𝑋𝑋 𝛽𝛽 𝑋𝑋 𝜀𝜀

5 4𝑗𝑗 𝛽𝛽Model𝑋𝑋 IV includes several controls for states’ cultural and political climates

towards sexual minorities. These include the Berry et al. (1998, 2010) measures of citizen

175

and state government ideology, 2016 Presidential Election results, religiosity, support for

sexual orientation nondiscrimination policies, proportion of state that is LGB, years since

same-sex marriage was enacted, method of same-sex marriage enactment, proportion of

the state that is urban, and proportion of the state holding at least a bachelor’s degree. I

model wages as:

= + + + + + + + +

𝑗𝑗 0 0𝑗𝑗 1 1𝑗𝑗 2 2𝑗𝑗 3 0𝑗𝑗 2𝑗𝑗 4 3𝑗𝑗 5 4𝑗𝑗 6 5𝑗𝑗 where𝑌𝑌� 𝛼𝛼 is𝛽𝛽 the𝑋𝑋 vector𝛽𝛽 of𝑋𝑋 state𝛽𝛽 political𝑋𝑋 𝛽𝛽 and𝑋𝑋 cultural∗ 𝑋𝑋 characteristics.𝛽𝛽 𝑋𝑋 𝛽𝛽 𝑋𝑋 𝛽𝛽 𝑋𝑋 𝜀𝜀

6 5𝑗𝑗 𝛽𝛽 𝑋𝑋 Results

There is considerable variation in overall wages by state. Figure 4-2 shows variation in median raw wages by state for all men. Wages for men range from a lowest state median of $15.98 per hour in Arkansas to a highest state median of $29.32 per hour

for the District of Columbia. Figure 4-3 shows variation in median raw wages by state for

all women. Wages for women range from a lowest state median of $12.79 per hour in

Idaho to a highest state median of $26.24 per hour for the District of Columbia. These differences suggest that geography is an important factor in the labor market experience of workers. When sexual identity is considered, several key observations are noteworthy.

Table 4-2 summarizes median raw wages by sex and sexual identity for all states.

Looking at the median wages for all workers by sex shows there is a persistent wage premium for men over women, consistent with the literature on the gender pay gap

(Miller and Vagins 2018). We also see consistent evidence of a wage penalty for bisexual men and women compared to their heterosexual counterparts across all states, with no controls. For gay and lesbian workers, wages compare inconsistently with their heterosexual counterparts. The difference in median wages between gay and heterosexual 176

men with no controls ranges from a wage penalty of $4.84 per hour for gay men in

Wyoming to a wage premium of $11.31 per hour in the District of Columbia. Median wages for gay men are less than median wages for heterosexual men in forty-five states.

The difference in median wages between lesbian and heterosexual women with no controls ranges from a wage penalty of $1.04 per hour for lesbian women in Montana to a wage premium of $5.18 per hour in the District of Columbia. Median wages for lesbian women are greater than median wages for heterosexual women in forty-four states. These preliminary estimates suggest that differences in geography affect wages, including by sexual identity, and warrants further analysis. In addition to the individual characteristics examined in Chapter Three, it is possible that variations in state policy play a role in these wage differentials.

177

178

Figure 4-2. Mean Hourly Wages by State for Men.

179

Figure 4-3. Mean Hourly Wages by State for Women.

Table 4-2. Median Raw Wages by Sex, Sexual Identity, and State (in Dollars) Men Women State All Heterosexual Gay Bisexual All Heterosexual Lesbian Bisexual Alabama 17.63 17.73 15.39 12.07 13.88 14.00 13.85 10.45 Alaska 21.63 21.97 21.07 *** 18.39 18.50 21.61 13.45 Arizona 17.93 18.00 16.69 12.19 15.57 15.71 16.51 12.09 Arkansas 15.98 16.00 14.34 11.52 13.43 13.56 13.62 10.13 California 19.57 19.74 20.61 13.26 16.72 16.92 18.85 12.47 Colorado 20.49 20.50 18.80 13.97 17.05 17.25 19.42 13.02 Connecticut 23.82 24.00 21.82 15.43 20.00 20.17 21.86 14.22 Delaware 20.11 20.29 19.31 *** 17.76 17.91 19.84 13.07 District of Columbia 29.32 28.27 39.58 24.86 26.24 26.22 31.41 22.33 180 Florida 16.46 16.50 16.89 12.50 14.82 14.86 15.95 11.87 Georgia 17.99 18.00 17.19 12.69 15.34 15.37 16.36 11.38 Hawaii 20.10 20.37 19.62 15.83 17.58 17.75 18.45 14.20 Idaho 16.76 16.89 14.46 11.97 12.79 12.95 13.05 9.78 Illinois 20.95 21.09 20.76 14.43 16.76 16.93 18.07 12.55 Indiana 19.04 19.11 16.60 13.09 14.86 15.00 15.51 10.81 Iowa 18.85 19.04 15.83 13.10 15.69 15.71 15.52 11.40 Kansas 18.64 18.85 16.21 12.67 15.00 15.03 16.21 10.93 Kentucky 17.47 17.50 16.08 11.79 14.51 14.67 14.43 10.57 Louisiana 18.70 18.85 16.74 12.24 13.63 13.75 14.60 10.33 Maine 18.00 18.13 17.44 13.99 15.71 15.87 17.58 11.51 Maryland 23.82 23.94 22.37 15.72 20.60 20.81 23.07 14.88 Massachusetts 24.59 24.72 24.49 16.70 20.49 20.61 24.35 15.51 Michigan 19.06 19.33 16.56 12.51 15.10 15.32 15.54 10.68 Minnesota 21.16 21.16 21.16 14.86 17.96 18.01 19.82 13.43 Mississippi 16.00 16.23 14.03 11.59 13.24 13.33 13.23 10.45

Table 4-2. (continued) Men Women State All Heterosexual Gay Bisexual All Heterosexual Lesbian Bisexual Missouri 18.23 18.33 16.63 12.38 15.00 15.07 15.81 11.37 Montana 16.59 16.81 13.89 11.89 14.00 14.14 13.10 11.02 Nebraska 18.22 18.41 14.99 13.81 15.35 15.37 15.84 11.71 Nevada 18.00 18.13 18.26 13.38 15.87 15.88 16.77 12.76 New Hampshire 21.63 21.93 19.07 15.51 17.84 17.93 18.94 13.60 New Jersey 24.61 24.87 22.77 15.87 20.00 20.10 21.77 13.96 New Mexico 16.00 16.22 15.67 11.77 13.45 13.57 16.33 10.30 New York 21.17 21.17 23.14 14.78 18.85 18.92 21.76 14.40 North Carolina 17.35 17.46 15.87 12.37 15.00 15.18 15.74 10.95 181 North Dakota 19.99 20.00 16.19 14.78 16.11 16.33 16.69 12.89 Ohio 19.21 19.43 16.93 13.34 15.60 15.71 16.28 11.42 Oklahoma 17.41 17.49 15.56 12.24 13.77 13.97 14.64 10.35 Oregon 18.68 18.88 17.92 13.04 15.71 15.87 17.83 11.88 Pennsylvania 20.49 20.49 18.87 13.99 16.50 16.64 18.06 12.49 Rhode Island 21.16 21.16 20.56 15.04 18.23 18.34 20.11 13.48 South Carolina 16.93 16.94 14.74 12.69 14.30 14.41 14.48 10.72 South Dakota 16.94 17.00 14.24 13.44 14.82 14.94 14.46 11.87 Tennessee 16.76 16.91 15.60 12.14 14.35 14.50 15.13 10.79 Texas 17.93 18.00 17.32 12.76 14.85 14.98 16.44 11.19 Utah 20.00 20.00 16.32 12.54 13.98 14.09 16.47 11.33 Vermont 18.49 18.52 17.79 *** 16.73 16.73 19.94 13.34 Virginia 21.52 21.70 20.91 15.10 17.50 17.62 19.97 13.35 Washington 22.84 23.05 21.30 15.30 17.62 17.79 20.29 13.14 West Virginia 17.93 18.00 15.19 12.23 13.77 13.88 14.64 10.85 Wisconsin 20.00 20.00 17.46 14.27 16.00 16.23 16.39 12.06

Table 4-2. (continued) Men Women State All Heterosexual Gay Bisexual All Heterosexual Lesbian Bisexual Wyoming 20.29 20.49 15.65 *** 14.65 14.81 14.28 10.82 *** n < 50

182

State Policies

During the survey years (2014-2018), thirty-four states had some form of policy

banning discrimination based on sexual identity. Of those states, eleven were covered by

Executive Orders issued by the state’s governor. The remaining twenty-three states, including the District of Columbia, had legislative policies. I only include state policies that were in effect during the survey year for each respondent. I also exclude any policies that were repealed during the survey years (Kansas and Louisiana had Executive Orders revoked during the survey years). Even where policies are present, they vary from state to state in terms of their coverage and enforcement. Executive Orders were consistent in their language and scope: the governor directs heads of executive agencies not to discriminate based on sexual orientation in matters of public employment, i.e., employees of the state government. State statutes are much broader, banning discrimination in public and private employment. Most states include exemptions for small businesses, but none exempts a business with more than fifteen employees. All states have some sort of exemption for religious organizations. Several themes emerged in the analysis of legislative texts: boundaries of “sexual orientation,” perception of sexual orientation, associations of sexual orientation with crime or pedophilia, sexual orientation as a preference, and conferral of approval through statute. Table 4-3 summarizes the themes identified in state legislation.

183

Table 4-3. Policy Themes Policy Year 3 Category Covers Crime/ No State Type Enacted Definition Perception Pedophilia Preference Acceptance Alaska EO 2002 Arizona EO 2003 Indiana EO 2001 Kentucky EO 2008 Michigan EO 2003 Missouri EO 2010 Montana EO 1999 North Carolina EO 2016 Ohio EO 2007 184 Pennsylvania EO 1975

Virginia EO 2014 California LEG 1992 X X Colorado LEG 2007 X X Connecticut LEG 1991 X X X Delaware LEG 2009 X District of Columbia LEG 1977 X X X Hawaii LEG 1991 X X X Illinois LEG 2005 X X X Iowa LEG 2007 X X Maine LEG 2005 X X Maryland LEG 2001 X Massachusetts LEG 1989 X X X Minnesota LEG 1993 X X Nevada LEG 1999 X X

Table 4-3. (continued) Policy Year 3 Category Covers Crime/ No State Type Enacted Definition Perception Pedophilia Preference Acceptance New Hampshire LEG 1998 X X X X New Jersey LEG 1992 X X New Mexico LEG 2003 X X New York LEG 2002 X X Oregon LEG 2007 X X Rhode Island LEG 1995 X X X X Utah LEG 2015 X X Vermont LEG 1992 X X Washington LEG 2006 X 185 Wisconsin LEG 1982 X X X

All states use “sexual orientation” as the basis for banning discrimination against

sexual minorities. Sexual orientation is understood to be an individual’s innate orientation

towards sexual or romantic attraction to partners, defined in relation to that partner’s

gender, and is “determined very early in life, [and] is an enduring and essential

psychological reality that transcends choice” (Reiter 1989:146). Sexual orientation is

distinct though often related to one’s identity or behavior. Orientation implies an

ontological status of being some identifiable thing: a heterosexual, homosexual, or

bisexual, for example. More than ninety percent (n=21) of states with legislative protections define sexual orientation using these three categories. The naming of categories establishes the boundaries of who is or is not included in the protections offered by statute. If someone’s sexual orientation does not fit within these named

categories (such as asexuality or pansexuality), or cannot be made to fit, they are

potentially excluded from protection. Indeed, Delaware’s statue is explicit that sexual

orientation “exclusively means heterosexuality, homosexuality, or bisexuality” (DE Title

19, Chapter 7 § 710(26)). New York is the only state that includes an additional category in its definition, asexuality (NY EXC 15-292-27). The boundaries established by statute suggest that, regardless of how one identifies, one’s sexuality defines them as a type of person in the eyes of the state.

Seventy-four percent (n=17) of states with legislative protections include perception of sexual orientation in their nondiscrimination statutes. Protections based on perception further shift the focus away from how a worker identifies themselves. Here discrimination is rooted in the actions and beliefs of the discriminator. Regardless of whether a worker is or is not a certain sexual orientation, if an employer acts in a

186

discriminatory way because of the employer’s interpretation of someone’s disposition,

that worker is being illegally discriminated against. This allows for the possibility that a

heterosexual worker can seek redress for discrimination if an employer perceives them to

be LGB even if they are not or do not identify as such. The inclusion of heterosexual

within the definition of sexual orientation and perception as a basis for protection suggest

these policies may positively affect all workers.

In thirty-nine percent (n=9) of states with legislative protections, the statute

explicitly notes a connection between sexual orientation and crime (n=6) or makes the

pernicious connection to pedophilia specifically (n=3). Rhode Island’s statute, for

example, states that the nondiscrimination policy “does not render lawful any conduct

prohibited by the criminal laws of this state” (RI §28-5-6(16)). While seemingly innocuous, this suggests that there are elements of “sexual orientation,” rather than acts, which may be criminal. This draws on a long history of the criminalization of sexual minorities in the United States. Indeed, a long-held belief that not only are sexual minorities criminal, but specifically prone to pedophilia, makes its way into some state statutes. Massachusetts, for example, states that their policy “shall not include persons whose sexual orientation involves minor children as the sex object” (MA 151B §3(6)). Of course, sexual acts against children are outlawed in criminal statues and would not be covered under protections based on sexual orientation, as defined in all states.

Similarly, seventeen percent (n=4) of states with legislative protections use language of sexual “preference” in their statutes. For example, Connecticut defines sexual orientation as “having a preference for heterosexuality, homosexuality or bisexuality, having a history of such preference or being identified with such preference”

187

(CT §46a-81a). The language of preference has long been used to suggest that sexual

minorities are acting based on choice and therefore should be encouraged to change their

identity. Discussing sexual identity as a preference is a way of voicing disapproval.

Indeed, nine percent of states (n=2) explicitly say in their statutes that, while they are extending nondiscrimination protections to sexual minorities, the state does not approve of them. New Hampshire’s law “does not confer legislative approval of such status, but is intended to assure basic rights afforded under this chapter” (N.H. Rev. Stat. Ann. §354-

A:2(XIV-c)).

These themes and policy differences have practical and symbolic effects. The setting of boundaries and the inclusion of perceptions has the practical effect of constraining or expanding the breadth of coverage, depending on how the statute is written, interpreted, and implemented. Defining sexual orientation as three distinct and exclusive categories limits the scope of a law to only those who fit within the categories, either through their own positioning or the states’ positioning of them. Including perception as an element in the statute moves coverage away from a worker’s self- definition and, perhaps rightfully, places the determination of discrimination within the mind of the discriminator. While these laws were intended to stop discrimination against sexual minorities, these themes suggest the effects of the laws might extend to heterosexual workers as well. When measuring the effect of such policies, it will be important to take these effects into account.

While two thirds of states have policies (n=34), just under seventy percent of the total population lives in states with a policy including 24.43 percent living in states with an Executive Order and 44.85 percent living in states with a legislative policy. LGB

188

people are slightly more likely than their heterosexual counterparts to live in states with policies. Whereas 69.16 percent of all heterosexuals live in states with some policy, 71.42 percent of lesbian and gay people live in states with a policy and 72.89 percent of bisexual people live in states with a policy. These patterns hold for estimates separated by sexual identity, sex, and labor force participation. It could be that LGB people choose to live in states with sexual orientation nondiscrimination policies or that states with higher

LGB populations are more likely to adopt the policies in the first place.

Regression Models

Regression models estimate the potential impacts that policy differences, state- level economic outcomes, and differences in states’ social and political climates have on observed wage differences by sexual identity. Model I is the baseline model that includes all individual-level demographic, family structure, and labor market controls. Model II adds a measure of states’ sexual orientation nondiscrimination policy (no policy,

Executive Order, or legislative policy), an interaction term between policy presence and sexual identity (which measures the net effect of a policy based on sexual identity), and controls for the years since enactment of the policy. Model III adds state-level economic outcomes. Model IV includes measures of states’ social and political climate.

Gay Men

Table 4-4 shows the exponentiated results for the models of the natural log of men’s wages. The baseline results for Model I show that, absent any state-level differences and controlling for individual productivity, family structure, and labor market characteristics, gay men earn on average 7.03 percent (eb=1.0703, p < 0.0001) more than similarly situated heterosexual men, a wage difference of roughly sixty-five cents per

189

hour. In all states, regardless of policy, the average wage premium for gay men over

similarly situated heterosexual men is observed, though subsequent models show

interesting effects of policy, state-level economic outcomes, and social/political climate.

Adding the policy measure in Model II suggests that policies may be having an inconsistent effect for gay men. In states with no policy, gay men’s average wage premium relative to their heterosexual counterparts is reduced from Model I, suggesting policies have some effect. The wage premium for gay men in states with no policy falls to 6.41 percent (eb=1.0641, p < 0.0001), a wage difference of roughly fifty-three cents per

hour. Model II suggests that the presence of nondiscrimination policies increases the

wages of all workers. Holding all productivity characteristics from Model I constant,

heterosexual men in states with Executive Orders earn on average 4.44 percent

(eb=1.0444, p < 0.0001) more than heterosexual men in states with no policy, a wage

increase of roughly thirty-seven cents per hour. The interaction term suggests wages for

gay men increase by a smaller margin in states with Executive Orders. Gay men’s wages

in such states are on average less than two percent (eb=1.0190, p = 0.0139) higher than gay men’s wages in states with no policy, a wage increase of roughly seventeen cents per hour. However, gay men in Executive Order states still out earn their heterosexual counterparts, though the gap is narrowed because of the different effects by sexual identity. Within Executive Order states, gay men earn on average just under four percent

(eb=1.0381) more than similarly situated heterosexual men, a wage difference of roughly

thirty-three cents per hour. States with Executive Orders show the narrowest gap between

gay and heterosexual wages.

190

Table 4-4. Regression of Log Wages by Sexual Identity for Male Workers Variables Model I Model II Model III Model IV b se b se b se b se Sexual Identity Gay 1.0703 0.0007 1.0641 0.0013 1.0649 0.0013 1.0654 0.0013 Bisexual 0.9448 0.0010 0.9536 0.0020 0.9526 0.0020 0.9525 0.0020

Policy Executive Order 1.0444 0.0004 1.0313 0.0004 1.0444 0.0006 Legislative 1.1328 0.0004 1.0540 0.0005 1.0526 0.0010

191 Interactions

Gay * Executive Order 0.9756 0.0020 0.9750 0.0020 0.9729 0.0020 Gay * Legislation 1.0181 0.0016 1.0148 0.0016 1.0136* 0.0016 Bisexual * Executive Order 0.9775* 0.0029 0.9780 0.0029 0.9781 0.0029 Bisexual * Legislation 0.9923* 0.0025 0.9939* 0.0025 0.9942* 0.0025

Policy Length 1.0000* <0.0001 0.9996 <0.0001 1.0009 <0.0001

State Economy Gross State Product 1.0000 <0.0001 1.0000 <0.0001 Personal Income 1.0000 <0.0001 1.0000 <0.0001 Unemployment 1.9816 0.0107 2.7279 0.0205

Social Climate Religiosity 1.1467 0.0039 Support for Nondisc. Policy 1.0937 0.0099

Table 4-4. (continued) Variables Model I Model II Model III Model IV b se b se b se b se 2016 Election 1.0378 0.0012 Citizen Ideol. 0.9973 <0.0001 State Govt. Ideol. 0.9992 <0.0001 Years SSM 1.0112 0.0001

Method SSM Federal Court 0.9841 0.0005 State Court 1.0024 0.0011

192 Legislative Statute 1.0350 0.0010

Voter Referendum 1.0509 0.0011

Prop. LGB 0.4389 0.0375 Prop. BA+ 0.7125 0.0074 Prop. Urban 1.1656 0.0022 *p > 0.05

Legislative policies have a stronger effect on wages for all men than do Executive

Orders. Holding all productivity characteristics from Model I constant, heterosexual men

in states with legislative policies earn 13.28 percent (eb=1.1328, p < 0.0001) more than

heterosexual men in states with no policy, a wage increase of roughly $1.11 per hour. The

interaction term suggests legislative policies have a stronger effect on gay men’s wages

than heterosexual men’s wages. Gay men’s wages in states with a legislative policy are

on average 15.33 percent (eb=1.1533, p = 0.0355) higher than gay men’s wages in states

with no policy, a wage increase of roughly $1.36 per hour. This difference widens gay

men’s wage premium over similarly situated heterosexual men. Within states with

legislative policies, gay men earn on average over eight percent (eb=1.0833) more than

heterosexual men, holding all else constant. The difference equates to a roughly seventy-

nine cent difference in average hourly wages. This is the widest wage gap between gay

and heterosexual men, suggesting legislative policies have the strongest effect on gay

men’s wages. In Model II, the number of years since a policy was enacted has almost no

measurable effect and the coefficient is not statistically significant.

Because the wage increases attributed to policy presence in Model II may actually

be a result of differences in state-level economic outcomes, Model III controls for

differences in gross state product, state average personal income, and state

unemployment rates. The wage premium for gay men in states with no policy is only

modestly higher when controlling for state-level economic outcomes. Here, holding all

else constant, gay men in states with no policies earn on average 6.49 percent (eb=1.0649, p < 0.0001) more than similarly situated heterosexual men, a wage difference of roughly thirty-five cents per hour. Policies continue to increase the wages of all men, though the

193

effect is greatly reduced for legislative policies. Holding all else constant, heterosexual

men in states with an Executive Order earn on average just over three percent (eb=1.0313, p < 0.0001) more than heterosexual men in states with no policy (a wage increase of roughly seventeen cents per hour) while heterosexual men in states with legislative policies earn over five percent (eb=1.0540, p < 0.0001) more than heterosexual men in states with no policy (a wage increase of roughly twenty-nine cents per hour).

Accounting for differences in state-level economic outcomes results in a nearly sixty percent reduction in the effect of legislative policies on heterosexual men’s wages.

The interaction terms remain fairly consistent from Model II to Model III suggesting policies and economic outcomes have a similar effect on gay men’s wages relative to their heterosexual counterparts. In states with Executive Orders, controlling for state-level economic outcomes, gay men earn on average just over half a percent

(eb=1.0055, p = 0.0092) more than similarly situated gay men in states with no policy, a

wage increase of roughly three cents per hour. Gay men in states with Executive Orders

continue to earn on average just under four percent (eb=1.0382), or roughly twenty-one cents per hour, more than heterosexual men, holding all else constant. In states with legislative policies, gay men earn on average just under seven percent (eb=1.0697, p =

0.0461) more than similarly situated gay men in states with no policy, a wage increase of

roughly forty cents per hour. Compared to similarly situated heterosexual men, gay men

in states with legislative policies continue to earn just over eight percent (eb=1.0807)

more, a wage difference of roughly forty-five cents per hour. Even controlling for state-

level economic outcomes, the wage gap between gay men and heterosexual men is widest in states with legislative policies.

194

Because sexual orientation nondiscrimination policies are likely associated with

states’ broader social and political climates, I attempt to control for such differences in

Model IV. These effects appear to be marginal in this analysis. The wage premium for

gay men in states with no policy is only marginally higher when controlling for states’

social and political climates. Here, holding all else constant, gay men in states with no

policies earn on average 6.54 percent (eb=1.0654, p < 0.0001) more than similarly

situated heterosexual men, a wage difference of roughly thirty-four cents per hour.

Policies continue to increase the wages of all men, though the difference in effects by

policy type is narrowed further in this model. Holding all else constant, heterosexual men

in states with an Executive Order earn on average over 4.44 percent (eb=1.0444, p <

0.0001) more than heterosexual men in states with no policy (a wage increase of roughly

twenty-three cents per hour) while heterosexual men in states with legislative policies

earn just over five percent (eb=1.0526, p < 0.0001) more than heterosexual men in states

with no policy (a wage increase of roughly twenty-seven cents per hour).

The interaction terms continue to remain fairly consistent from Model III to

Model IV, though the effect of legislative policies on gay men’s wages is no longer

significant, suggesting legislative policies do not have a unique effect. In states with

Executive Orders, gay men earn on average less than two percent (eb=1.0167, p = 0.0011)

more than similarly situated gay men in states with no policy, a wage increase of roughly

nine cents per hour. Gay men in states with Executive Orders continue to earn on average

slightly less than four percent (eb=1.0366), or roughly twenty cents per hour, more than heterosexual men, holding all else constant. Compared to similarly situated heterosexual

195

men, gay men in states with legislative policies continue to earn just under eight percent

(eb=1.0798) more, a wage difference of roughly forty-three cents per hour.

Bisexual Men

Table 4-4 also shows results for bisexual men. Overall, the effect of policies on

the wage penalty for bisexual men is less clear. As noted above, policies appear to be

associated with higher wages for all men, even when controlling for state-level economic

outcomes and states’ social and political climate. The interaction effects suggest that both

Executive Orders and legislative policies are associated with lesser increases in wages for

bisexual men than heterosexual men, though the effects are inconsistently significant.

Together, these patterns suggest that policies have an indeterminate effect on the wages of bisexual men.

The baseline results for Model I suggest that, absent any state-level difference and

controlling for individual productivity, family structure, and labor market characteristics,

bisexual men earn on average 5.52 percent (eb=0.9448, p < 0.0001) less than similarly situated heterosexual men, a wage penalty of roughly fifty-one cents per hour.

Interestingly, adding the policy measure in Modell II narrows the wage penalty for bisexual men in states with no policy but widens it in states with policies. In states with no policy, bisexual men earn 4.64 percent (eb=0.9536, p = 0.0006) less than heterosexual men (a roughly thirty-nine cent difference in hourly wages), holding all else constant, while the penalty is more than six percent (eb=0.9322) in Executive Order states and

more than five percent (eb=0.9463) in states with legislative policies. However, neither

interaction term is statistically significant.

196

These patterns hold consistently after adding controls for state-level economic

outcomes and states’ social and political climate in Models III and IV. Bisexual men have

the narrowest penalty (just under five percent) relative to similarly situated heterosexual

men in states with no policy but a larger penalty of just over six percent in states with an

Executive Order, and the effect is statistically significant in these subsequent models. The

unique effect of legislative policies on bisexual men’s wages is never statistically

significant.

Lesbian Women

Table 4-5 shows the exponentiated results for the models of the natural log of

women’s wages. The baseline results in Model I show that, absent any state-level

differences and controlling for individual productivity, family structure, and labor market

characteristics, lesbian women earn on average 6.94 percent (eb=1.0694, p < 0.0001) more than similarly situated heterosexual women, a wage difference of roughly seventy- seven cents per hour. As with gay men, lesbian women’s wage premium over similarly situated heterosexual women is observed across all models.

197

Table 4-5. Regression of Log Wages by Sexual Identity for Female Workers Variables Model I Model II Model III Model IV b se b se b se b se Sexual Identity Lesbian 1.0694 0.0007 1.0774 0.0013 1.0781 0.0013 1.0775 0.0013 Bisexual 0.9505 0.0006 0.9636 0.0012 0.9630 0.0012 0.9631 0.0012

Policy Executive Order 1.0420 0.0004 1.0360 0.0004 1.0406 0.0005 Legislative 1.1437 0.0004 1.0699 0.0005 1.0570 0.0009

198 Interactions

Lesbian * Executive Order 0.9802 0.0020 0.9806 0.0020 0.9804 0.0020 Lesbian * Legislation 0.9912* 0.0016 0.9897 0.0016 0.9902* 0.0016 Bisexual * Executive Order 0.9830 0.0017 0.9838 0.0017 0.9833 0.0017 Bisexual * Legislation 0.9810 0.0014 0.9824 0.0014 0.9828 0.0014

Policy Length 1.0013 <0.0001 1.0006 <0.0001 1.0015 <0.0001

State Economy Gross State Product 1.0000 <0.0001 1.0000 <0.0001 Personal Income 1.0000 <0.0001 1.0000 <0.0001 Unemployment 1.8192 0.0099 1.8550 0.0185

Social Climate Religiosity 1.0062* 0.0035 Support for Nondisc. Policy 1.0187* 0.0090

Table 4-5. (continued) Variables Model I Model II Model III Model IV b se b se b se b se 2016 Election 1.0441 0.0011 Citizen Ideol. 0.9991 <0.0001 State Govt. Ideol. 0.9990 <0.0001 Years SSM 1.0074 0.0001

Method SSM Federal Court 0.9859 0.0004 State Court 0.9871 0.0010

199 Legislative Statute 1.0101 0.0009

Voter Referendum 1.0105 0.0010

Prop. LGB 0.5552 0.0336 Prop. BA+ 0.8596 0.0067 Prop. Urban 1.0782 0.0020 *p > 0.05

Adding the policy measure in Model II suggests that policies may be having an

inconsistent effect for lesbian women as well. In states with no policy, lesbian women’s

average wage premium relative to their heterosexual counterparts actually widens from

Model I. The wage premium for lesbian women in states with no policy grows to 7.74

percent (eb=1.0774, p < 0.0001), a wage difference of roughly seventy-five cents per

hour. The wage premium for lesbian women is consistently the widest in states with no

policy.

As was the case with men, Model II suggests that the presence of

nondiscrimination policies increase the wages of all women. Holding all productivity

characteristics from Model I constant, heterosexual women in states with Executive

Orders earn on average 4.17 percent (eb=1.0420, p < 0.0001) more than heterosexual women in states with no policy, a wage increase of roughly forty-one cents per hour. In

states with legislative policies, heterosexual women’s wages increase by 14.33 percent

(eb=1.1437, p < 0.0001), a wage increase of roughly $1.39 per hour. Controlling for state-

level economic outcomes and states’ social and political climate results in a convergence

of the effects of legislative policies with Executive Orders.

The interaction terms suggest wages for lesbian women increase by a smaller

margin in states with Executive Orders than do the wages of similar heterosexual women.

Lesbian women’s wages in such states are on average just over two percent (eb=1.0213, p

= 0.0090) higher than lesbian women’s wages in states with no policy, a wage increase of

roughly twenty-two cents per hour. However, lesbian women in Executive Order states

still out earn their heterosexual counterparts, though the gap is narrowed because of the

different effects by sexual identity. Within Executive Order states, lesbian women earn

200

on average 5.61 percent (eb=1.0561) more than similarly situated heterosexual women, a

wage difference of roughly fifty-seven cents per hour. Across all models, states with

Executive Orders show the narrowest gap between lesbian and heterosexual women’s

wages with a just under six percent premium for lesbian women.

The effects of legislative policies appear consistent with the patterns observed for

gay men, though they are inconsistently significant for lesbian women. In all models,

legislative policies appear to have a slightly stronger effect than Executive Orders,

resulting in a just under seven percent average wage premium for lesbian women over

similarly situated heterosexual women. However, the unique effect of legislative policies

on lesbian women’s wages is only statistically significant in Model III when controlling

for policy presence and state-level economic outcomes. In Model III, lesbian women in

states with legislative policies earn 6.70 percent (eb=1.0670, p = 0.0413) more than heterosexual women (a roughly forty-six cent per hour wage advantage), holding all else constant.

Bisexual Women

Table 4-5 also shows results for bisexual women. Similar to bisexual men, while policies raise the overall wages of bisexual women, as they do for all women, bisexual women see a larger wage penalty in states with a policy than in states without. The interaction effects show that bisexual women receive a smaller proportional increase in their wages in states with policies relative to heterosexual women, exacerbating the already observed wage penalty. These effects are consistent and significant across models. The baseline results for Model I show that, absent any state-level differences and controlling for individual productivity, family structure, and labor market characteristics,

201

bisexual women men earn on average just under five percent (eb=0.9505, p < 0.0001) less

than similarly situated heterosexual women, a wage penalty of roughly fifty-five cents

per hour. This penalty reduces slightly in subsequent models and consistently shows

bisexual women in states with no policies earn just over 3.50 percent (between twenty-

four and thirty-five cents per hour) less than similarly situated heterosexual women.

Executive Orders and legislative policies, even when controlling for differences in state-level economic outcomes and states’ social and political climates, have similar and consistent effects on the wages of bisexual women. Adding the policy measure in Model

II suggests that, holding constant the productivity characteristics in Model I, bisexual women in states with Executive Orders earn 5.28 percent less (eb=0.9472, p = 0.0032) than similarly situated heterosexual women in those states, a wage penalty of roughly fifty-three cents per hour. For bisexual women in states with legislative policies the wage penalty relative to similar heterosexual women is 5.47 percent (eb=0.9453, p = 0.0427), a wage difference of sixty cents per hour. Across Models III and IV, these differences remain fairly consistent, varying by less than two tenths of a percentage point.

Discussion

The results above suggest that policies have inconsistent effects based on sexual identity. The presence of both Executive Orders and legislative policies, even when controlling for differences in state-level economic outcomes and states’ social and political climates, is associated with overall higher wages for all workers, holding productivity characteristics constant. Nondiscrimination policies appear to affect the wages of all workers consistently, regardless of sexual identity, suggesting that living in a state with labor protections based on sexual orientation is generally a benefit to all. It is

202

possible, and indeed likely, that states with a sexual orientation nondiscrimination also have other pro-labor policies such as a higher minimum. These states may also have a more pro-labor social climate, e.g., greater support for labor unions. Future analyses should attempt to control for these differences.

Gay men and lesbian women maintain wage premiums over their heterosexual counterparts across all models and policies, though the premium is narrowest in states with Executive Orders, suggesting those policies actually benefit heterosexual workers more so than gay and lesbian workers. A similar pattern was observed for bisexual men but not bisexual women. Bisexual men’s wage penalty relative to heterosexual men is widest in states with Executive Orders whereas bisexual women’s wage penalty relative to heterosexual women is widest in states with legislative policies. This provides partial evidence for my Hypothesis One.

Gay men’s wage premium compared to similar heterosexual men is highest in states with legislative policies, suggesting such policies positively impact gay men’s wages. For lesbian women, their wage premium over similar heterosexual women is widest in states with no policy. This suggests that policies have a gendered effect. It is possible this gendered effect is a result of lesbian women’s greater labor market attachment compared to heterosexual women (see Chapter Three above). For bisexual men and women, legislative policies appear to have a similar impact as Executive Orders: legislative policies raise bisexual wages, but do not narrow the wage penalty relative to similar heterosexual workers. This also provides partial evidence for my Hypothesis

Two.

203

Overall, the gender difference in policy effects is consistent with the existing literature on sexual orientation nondiscrimination laws. Klawitter (2011), and Baumle and Poston, Jr. (2011), using household-level data which do not allow them to differentiate between lesbian and bisexual women, found nondiscrimination policies had no effect on the earnings of women in same-sex households whereas they did for men.

Klawitter (2011) posits this “may be explained by the lack of earnings deficit relative to married women” (335). This, however, does not explain the similar findings for bisexual women in my analysis, as bisexual women have a consistent wage penalty compared to heterosexual women. Also, my analyses show gay men with a wage premium like lesbian women, but nondiscrimination policies seemingly influence gay men’s wages and not lesbian women’s. Other explanations for these patterns are warranted.

Perhaps the effect of state-level sexual orientation nondiscrimination policies is muted by the effects of federal policies protecting against discrimination on the basis of sex more broadly. Federal policies such as the Civil Rights Act of 1964 and the Equal

Pay Act provide strong federal protections and remedies for sex-based discrimination that likely outweigh state-level protections. Further, it is possible that women workers do not take advantage of state-level policies to the same degree that men do. Baumle et al.

(2019) found that women made up only forty-two percent of complaints filed with the

Equal Employment Opportunity Commission (EEOC) alleging sexual orientation

discrimination (1139). This could be a result of gender differences in awareness of policy

or beliefs about the potential outcomes of filing complaints.

Some perhaps less likely explanations relate to productivity and the direction of

discrimination. It could be that lesbian and bisexual women in states without

204

nondiscrimination policies increase their productivity to counter the effects of

discrimination, thereby masking the effects of policies for workers in states with policies.

For lesbian women, who experience a wage premium independent of policy, it is possible

that discrimination works in their favor. Employers might respond positively to lesbian

women’s defiance of gender stereotypes and greater labor market attachment (see

Chapter Three) with higher wages over heterosexual women. This statistical

discrimination in their favor could dilute the impact of policies.

It is also possible that the most positive effects of sexual orientation

nondiscrimination policies are not related directly with wages. It could be that such

policies indirectly impact worker compensation by creating safer and more welcoming

environments. As Badgett et al. (2007) note:

The laws’ positive effects may not be quantifiable through wage analyses. For example, the laws may make it easier for gays and lesbians to come out at work, improve intra-office dynamics, or help gays and lesbians to achieve a greater sense of dignity. (16)

Indeed, there is some evidence that working in states with sexual orientation nondiscrimination policies reduces experiences of discrimination in the workplace

(Barron and Hebl 2013). Future analyses using the CSMI dataset developed in this dissertation should explore the effects of sexual orientation nondiscrimination policies on labor force participation rates by sexual identity. It could be that experiencing discrimination makes LGB people more likely to drop out of the labor market altogether, whereas policies may make it possible for LGB people to remain in the labor market.

Policy length does not appear to impact wages, providing no evidence of what I

expected in my Hypothesis Three. It could be that, because sexual orientation is always

205

added as an additional protected class to existing nondiscrimination policies, that there is

no lag in passage and implementation. If the executive infrastructure is already

operational, sexual minorities presumably can immediately begin filing claims. States

may invest differently in these offices or may adjudicate cases in different manners.

Future research could examine, on a state-by-state basis, the administrative records of

filed claims to assess the volume and degree to which policies are being enforced. Some

research in this area has been undertaken for individual states (Colvin and Riccucci 2002,

Colvin 2009). State-level economic outcomes and states’ social and political climates

affect the wages of all workers, but do not appear to have a particularly strong effect on

the relationship between nondiscrimination policies and wages. As noted above, it could

be that the effects of sexual orientation nondiscrimination policies observed here also

include the effects of other generally pro-labor policies, such as higher minimum wages

or support for labor unions.

Limitations

The present approach also has some important limitations worth acknowledging.

Analyses using CSMI data require that all variables be included in the original imputation

model. Because state information is not available in the publicly available NHIS data, my

imputation of sexual orientation into the ACS may not fully capture state-level variation.

The ACS and NHIS are both nationally representative surveys and therefore the state

variation in the ACS should reflect similar patterns in the NHIS, unless there is some

systematic bias in one survey or the other. Future analyses should obtain access to the

restricted NHIS data to confirm the findings of the present study.

206

Additionally, because of the nature of the CSMI dataset I use, I am only able to

disaggregate state-level characteristics to the individual level. While this procedure has

been used before in the existing literature (Gates 2009, Klawitter and Flatt 1998), it is

known that disaggregating higher-level data to lower-level cases violates OLS

assumptions and underestimates standard errors (Raudenbush and Bryk 2002). Others,

using household-level data, have successfully performed multilevel analyses (Baumle and

Poston, Jr. 2011, Klawitter 2011). Future analyses using CSMI may be able to take

advantage of the more robust estimates and inferences of hierarchical models.

Conclusion

The persistence of discrimination against LGB people warrants continued efforts to examine possible policy responses. Workplace discrimination has been a key area of focus for LGBTQ activists for more than half a century. Their successful lobbying has seen policies of various forms enacted in more than half the states, and more recently, an opinion by the Supreme Court that existing statutes, namely the sex protections within the

Civil Rights Act of 1964, provide federal protections for LGBTQ workers. As these policies continue to be enacted and enforced, it is important to consider their strengths and places where they might be improved. This chapter contributes to this ongoing discussion by evaluating the effect of state-level policies in wage differences. Policies have clear effects on the wages of gay men, though inconsistent effects for lesbian women, though both these groups continue to maintain wage premiums over their heterosexual counterparts. Where policies clearly fall short is in alleviating the entrenched wage inequality experienced by bisexual men and women. Further policy responses appear necessary to counter these trends.

207

Despite the inconsistent evidence of effectiveness contained here, nondiscrimination policies are still an important symbol of social equality. While the practical effect of such policies on wages appears mixed, sexual orientation nondiscrimination policies are also a symbolic representation of a state’s commitment to fairness in the labor market. It is possible that sexual minority workers who experience discrimination leave the labor market altogether. Nondiscrimination policies could prompt discouraged LGB workers to reenter to labor market. Policies also signal to the broader population, independent of labor market relations, that sexual minorities are equally deserving of dignity and respect. Future research and policy interventions should assist in making sure the practical and symbolic effects of such policies are matched.

208

CHAPTER 5

CONCLUSION

Introduction

My dissertation addressed three topics related to LGB populations in the United

States: the demographics and composition of sexual minority communities in the United

States, their labor market experiences, and public policy. The myth that ten percent of the population is LGB was originally offered in the early 1970s for political purposes. If

LGB people made such a sizeable minority, then politicians would have to take seriously the issues of a community that garnered very little popular support. Half a century later, we are in a dramatically different social and political climate. The social and political needs of LGB populations have similarly evolved. While the existence of LGB people has been successfully proven, the need for targeted social policy and resources requires more accurate estimates of the population’s size and demographic profile.

In Chapter Two, using Cross-Survey Multiple Imputation I estimated that just over four percent of the overall adult, noninstitutionalized population in the United States identifies as LGB. While significantly smaller than the ten percent myth, this estimate likely better reflects the size of LGB populations in the United States today. I found that

LGB people are younger and more racially and ethnically diverse than their heterosexual counterparts. Importantly, I found that bisexual men and women are, on average, more than a decade younger than both their heterosexual and gay and lesbian counterparts.

Bisexual men and women are also more than five percentage points more likely to identify as Hispanic. This likely reflects a number of important factors in the changing

209

composition of the American population: the growing Hispanic population overall and a more fluid attitude towards sexuality among younger generations. I also show that LGB people reside in all states plus the District of Columbia, with a wide range in state estimates of LGB populations.

I concluded that LGB people, on average, experience lesser economic outcomes relative to heterosexuals. LGB people experience higher rates of unemployment, higher rates of poverty, lower median incomes, and lower rates of health insurance coverage.

This is true for lesbian women and gay men, but especially for bisexual women and men, who appear significantly disadvantaged relative to their heterosexual counterparts, though age is certainly a mitigating factor in this conclusion. This comports with recent research which suggests that bisexual people experience unique and pervasive forms of discrimination in the United States (Mize 2016). Overall, despite the smaller top-line number, these estimates paint a picture of a large and diverse population.

Large numbers of LGB people report experiencing discrimination in the workplace (Badgett et al. 2007). In Chapter Three, I estimated the effects of discrimination on the labor market outcomes for sexual minority workers. Through regression models and decompositions of wage differences, I found that, controlling for differences in individual productivity characteristics, family structure, and labor structure, lesbian women and gay men experience a wage premium relative to heterosexual women and men, respectively. For lesbian women, this is consistent with past estimates. For gay men, this is a somewhat novel finding. Past research, using behavioral measures of sexuality from earlier time periods, has found a wage penalty for gay men relative to similarly situated heterosexual men. My finding suggests that either the changing social

210

and political climate surrounding LGB people (e.g., marriage equality) has benefited gay

men’s economic outcomes or that previous research masked gay men’s economic

outcomes by combining them with bisexual men. My estimates also suggest that bisexual

men and women experience significant labor market disadvantages, even when

controlling for their significant age difference from other groups. I found that more than fifteen percent of the difference in wages between bisexual men and women and heterosexual men and women, respectively, cannot be attributed to differences in productivity characteristics like education, age, and work experience, suggesting discrimination is likely negatively impacting bisexual men and women’s wages. While finding a wage premium for gay men and lesbian women might suggest discrimination is not taking place for these populations, it is quite possible that LGB workers experience discrimination in forms other than lower wages. It could be that discrimination negatively affects their mental and physical health, the types of jobs they choose to take, or even whether they participate in the labor market at all. More research is needed in this area to fully understand the effects of discrimination of LGB people in the workplace.

One possible solution to workplace discrimination is enacting sexual orientation nondiscrimination policies at the federal, state, or municipal level. I examined the effectiveness of state policies in Chapter Four. At the time of my data (2014-2018), there were no federal sexual orientation nondiscrimination policies that protected LGB workers. While thirty-four states had enacted policies—eleven via Executive Orders and twenty-three via legislative policies—I examined how these policies were written and what the symbolic implications of their text might be. I catalogued every state policy,

Executive Order and legislation, and performed a content analysis of the texts, identifying

211

patterns and themes which became elements of my quantitative analyses. I found that

nearly every state’s policy defined sexual orientation using a three-category definition: sexual orientation means heterosexuality, homosexuality, and bisexuality. This definition creates boundaries around who is, and can be made to be, visible to the state as a sexual citizen. I also noted the ways in which many states cover perception of sexual orientation, moving the impetus for coverage away from the discriminated and to the mind of the discriminator. I also noted the ways some states encode anti-LGB animus into their

statues through the use of language of preference and criminality.

Having catalogued state policies, I then turned towards a quantitative analysis of

the effectiveness of such policies. Specifically, I sought to measure how the presence of

sexual orientation nondiscrimination policies affected the differences in wages between

LGB workers and their heterosexual counterparts. Expanding upon the regression models

in Chapter Three, I modeled several wage equations, controlling for differences in

policies, duration of policy, state-level economic outcomes, and states’ social and political climates. I found that policies have inconsistent effects. Nondiscrimination policies appear to affect the wages of all workers consistently, regardless of sexual identity, suggesting that living in a state with labor protections based on sexual orientation is generally a benefit to all. The wage premiums I observed for gay men and lesbian women hold across all states, regardless of policy, while bisexual men and

women continue to be disadvantaged relative to their heterosexual counterparts. That

policies appeared to have an effect for all workers suggests that other factors may be

influencing the wage differences, such as the broader labor protections in states with

sexual orientation nondiscrimination policies.

212

Limitations

It is important to note some methodological limitations and opportunities for improving these analyses by accessing restricted data. The measure of sexual orientation which makes up the basis for these analyses is an identity-based measure introduced by the NHIS in 2013. NHIS sample adults were asked, “Do you think of yourself as: lesbian or gay; straight, that is, not lesbian or gay; bisexual; something else; or don’t know?”

(Miller and Ryan 2011:6). Those indicating “something else” are then asked an open- ended follow up. This includes respondents who identify as queer, pansexual, asexual, or others. Due to small sample sizes (during the survey years in my analyses, those answering “Something Else” range from roughly a quarter to half a percent of all NHIS sample adults), the National Center for Health Statistics (NCHS), which collects the

NHIS, restricts access to these follow-up responses. My population estimates represent the portion of the population identifying as LGB but undercount the number of sexual minorities overall. It is also possible that those identifying outside the more commonly known LGB labels experience different forms of discrimination. Therefore, their exclusion from my analyses potentially underestimates the degree to which discrimination is taking place. This would also have implications for policy, especially given the degree to which current public policy specifically names heterosexuality, homosexuality, and bisexuality as the constructs currently covered by sexual orientation nondiscrimination laws. In the future, I would like to access these restricted data for further analyses.

Another limitation deals with state-by-state variation. The NHIS similarly does not release respondent’s state of residency in the publicly available data. CSMI assumes

213

that all variables used in analyses are jointly observed in both the donor and recipient

survey and included in the original imputation model. My use of the ACS’s state data is a

violation of this assumption. Because the ACS is the larger, more robust survey, and

because I do include Census regions in the imputations, some of this bias is reduced.

Further, the bias would only result from systematic differences between the underlying

samples of the two CSMI surveys, which is unlikely, or from systematic differences in

the distribution of populations based on sexual identity, which is possible and indeed

likely. Interpretation of any state estimates in this dissertation should bear this fact in

mind. For this reason as well, I would like to access the restricted NHIS data to

corroborate these findings in the future.

In Chapter Four, I disaggregate state-level variables to individual cases for my

policy analyses. While this procedure has been used before in the existing literature

(Gates 2009, Klawitter and Flatt 1998), it is known that disaggregating higher-level data

to lower-level cases violates OLS assumptions and underestimates standard errors

(Raudenbush and Bryk 2002). A better approach would be to use hierarchical modeling which considers the different levels of the data in my analyses. However, given the nature of my CSMI data and technological restraints, this proved impractical for this project. In

the future, it would be beneficial to find ways to make use of both CSMI data and

hierarchical modeling.

My Contributions

Despite these limitations, my dissertation advances our knowledge of the demographics of sexual minorities in the United States, the inequality experienced by

LGB workers, and the effectiveness of policy interventions. My use of CSMI provides

214

additional evidence for the efficacy of the method in sociological and economic research.

I argue that CSMI is a useful tool for the study of small and difficult to sample populations. I also join the growing call for the inclusion of sexual orientation and gender identity (SOGI) measures in federal surveys to help address the resource and policy needs of LGBTQ populations.

The lack of federal data on sexual minorities in the United States renders these communities invisible and presents significant challenges to the design and implementation of public policy. Through the creation of my CSMI dataset, I attempted to make these communities more visible in the data, to examine their demographic profiles at the national and state level, and to show the ways in which LGB people differ from the larger heterosexual population. These demographic profiles expand our demographic and sociological knowledge in the areas of sexuality and inequality. Cross- group comparisons allowed me to investigate differences and inequalities among and between sexual identity groups. These comparisons speak directly to contemporary social and cultural movements for (and against) LGBTQ equality; they inform continuing public policy debates; and they give us a more complete picture of the sexual composition of the population.

My labor market analyses contribute to and expand upon the existing research on

LGB labor market outcomes. My examination of the unique experiences of lesbian, gay, and bisexual workers extends beyond the previous literature which often combined these groups in their analyses, and it furthers our understandings of the unique disadvantages experienced by bisexual people in the labor market. My finding of a wage premium for gay men suggests that either past research, which often combined gay and bisexual men

215

in their analyses, was overestimating the unique effects of discrimination on gay men or

that changes in the social and political climates surrounding sexual orientation have

affected the labor market experience for at least some sexual minorities. While I do not

find direct evidence for discrimination against lesbian women and gay men through

examination of wages, I do find such evidence for bisexual men and women. As noted above, labor market discrimination against LGB people may take forms other than wage differentials, and it is possible that wage discrimination is endogenous with other sites of discrimination controlled in the model (e.g., education). Further, the use of CSMI data in

these analyses demonstrates the versatility of the method in sociological and economic

research.

Finally, while a small body of literature has examined the effectiveness of state

nondiscrimination polices for sexual minorities, existing literature has relied on

household-level data which, by their nature, omit several categories of sexual minorities:

same-sex couples who do not indicate they are married or unmarried partners on the survey; sexual minorities who do not reside with their partner or who do not have a partner; and bisexual men and women in opposite-sex relationships. My data allowed me to examine policies’ effects regardless of respondents’ household structures. I am also able to distinguish between categories of sexual identity, examining how policies may differentially impact gay men, lesbian women, bisexual men, and bisexual women. To my knowledge this is the first such study to do so. My study better reflects the reality of workers’ availability for discrimination in the labor market and provides a more complete picture of the state of sexual orientation discrimination and public policy. As new state

216

and federal policies are debated and implemented, my study informs discussions about the potential effectiveness of such policies.

Future Research

This dissertation has demonstrated the versatility of CSMI data in the study of sexual orientation. Given the flexibility of the data, there are a number of areas of future research for which I could use these data. I plan to extend the analyses in this dissertation to examine differences in labor force participation. While wage analyses provide one lens for the examination of inequality, discrimination against LGB people likely takes multiple forms. Indeed, labor discrimination may be such that some LGB people opt to withdraw from the labor market all together. My CSMI data lends itself to analyses of labor force participation at the national and state level (acknowledging the limitations discussed above). I plan to similarly examine differences in labor force participation among LGB people to see if bisexual disadvantage extends to this dimension as well.

Another area where CSMI data can be effective at informing contemporary public policy debates is in the area of family research. While same-sex marriage has been legalized, debates around family structure and childrearing continue. Public policy surrounding issues such as foster parenting, adoption, and second parent rights is becoming an increasingly important site for LGB activists. In November 2020, the United

States Supreme Court heard arguments in Fulton v. City of Philadelphia, whose pending opinion will determine the legality of discriminating against LGB people as potential foster parents by social service organizations. Other same-sex couples have seen the state department deny citizenship to their children born abroad to surrogates (Bruce 2019). For these issues and others, my CSMI data can help expand our understanding of LGB

217

families, their composition, and their needs. In my data I can identify different family

structures (e.g., married, cohabitating), demographic profiles of the parents and children

in LGB households, and their distribution across the states.

This dissertation emphasized the empirical value of CSMI data and the

demographics, labor market, and policy landscape for sexual minorities. There are two

areas of theoretical literature with which I would like to bring my work into conversation.

The first is the growing intersection between queer theory and social science

methodology. A number of texts have been written recently on the topic of queer

methods (e.g., Ghaziana and Brim 2019). These conversations seek to bridge the seeming

gap between hypothetico- deductivist social science and a humanities-oriented queer theory that often rejects quantification and empiricism (e.g., Muñoz 1996). For many in the latter camp, sexuality and its meanings are historical, contextual, and fluid and cannot be captured through the collection of “data.” And yet data are a necessary tool for the enactment of policy and the distribution of resources (Velte 2020). I contend that CSMI, which rather than imposing identity certainty to individuals examines a range of statistical possibilities to construct a complete picture of the whole, is, in its own way, a queer methodology. I plan to extend my empirical analyses here by bringing them into conversation with this growing literature on queer methods.

I also plan to bring the qualitative policy analysis I perform in Chapter Four into the literature on sexual citizenship. Sexual citizenship interrogates the ways in which formal and informal mechanisms of belonging are regulated through sexual regimes, namely heterosexuality (Weeks 1998). In many ways, access to rights and benefits (the perks of citizenship) is predicated on one’s sexual identity. Indeed, the formation of state

218

infrastructure has itself been tied up in projects of sexual regulation (Canaday 2009).

Contemporary policy shifts including the end of sodomy bans, the legalization of same- sex marriage, and the integration of sexual minorities into the armed forces has extended the meaning of citizenship to include, at least some, sexual minorities. I plan to extend my analyses of sexual orientation nondiscrimination policies to argue that these policies represent an additional site in which some LGB people are given (conditional) access to citizenship. I plan to examine the ways that the boundaries written into the text of the laws, both in how sexual orientation is defined and in how some statutes draw on anti-

LGB stereotypes, allow some LGB people to become visible to the state while other groups are purposefully excluded.

Finally, I would like to extend the use of CSMI to study trans populations. My analyses in this dissertation exclude consideration of gender identity for two reasons: gender identity is conceptually distinct from sexual orientation and neither the NHIS nor the ACS explicitly collects data on trans identities. Given the proliferation of anti-trans policies targeting trans people in schools, sports, public spaces, and the military, more detailed and current data are needed to counter these trends. I hope that CSMI could aid scholars who are already working to overcome the many challenges to produce accurate estimates of trans populations (Doan 2019). It may be possible to impute gender identity from an existing smaller survey into a larger survey, such as the ACS, similar to my imputations based on sexual orientation.

Conclusion

On June 15, 2020, the United States Supreme Court issued its opinion in Bostock v. Clayton County. The court held that Title VII of the Civil Rights Act of 1964 bans

219

discrimination against LGBTQ Americans in the workplace under its protections against sex discrimination. For the first time, LGBTQ people have legal protections against employment-based discrimination in every state. The Court’s decision represents a significant sea change in the landscape for LGBTQ rights in the United States. As this change in law and policy takes shape, it will be important to examine its effects on the labor market experiences of LGBTQ people. But it will also be important to monitor ways in which these protections can be undermined. There are already concerted efforts to carve out exemptions for certain religious employers and organizations. Additionally,

Title VII’s protections extend only to employment. LGBTQ people can still legally be discriminated against in public accommodations, healthcare, education, adoption, and other arenas. Despite the persistence of such threats to equality, scholars, activists, and policymakers continue to expand our understandings of LGBTQ populations, their experiences, and policy needs. I remain hopeful that the momentum of past victories will propel the movement for LGBTQ equality forward as we enter the third decade of the twenty-first century.

220

REFERENCES CITED

Allegretto, Sylvia A., and Michelle Arthur. 2001. “An Empirical Analysis of

Homosexual/Heterosexual Male Earnings Differentials: Unmarried and

Unequal?” Industrial and Labor Relations Review 54(3):631-646.

American Psychological Association. 2008. Answers to Your Questions: For a Better

Understanding of Sexual Orientation and Homosexuality. Washington, DC.

Armstrong, Elizabeth A. 2002. Forging Gay Identities: Organizing Sexuality in San

Francisco, 1950-1994. Chicago: University of Chicago Press.

Arrow, Kenneth J. 1973. "The Theory of Discrimination." Pp. 3-33 in Discrimination in

Labor Markets., edited by O. Ashenfelter, and A. Ress. Princeton, NJ: Princeton

University Press.

Babst, Gordon A. 2009. "Introduction." Pp. vii-xiv in Moral Argument, Religion, and

Same-Sex Marriage: Advancing the Public Good., edited by G. A. Babst, E. R.

Gill and J. Pierceson. Lanham, MD: Lexington Books.

Badgett, M. V. Lee. 1995. “The Wage Effects of Sexual Orientation Discrimination.”

Industrial and Labor Relations Review 48(4):726-739.

------. 2001. Money, Myths, and Change: The Economic Lives of Lesbians and Gay Men.

Chicago: University of Chicago Press.

------. 2006. "Discrimination Based on Sexual Orientation: A Review of the Literature in

Economics and Beyond." Pp. 161-186 in Handbook on the Economics of

Discrimination., edited by W. M. Rodgers III. Cheltenham, UK: Edward Elgar.

221

------. 2007. "Discrimination Based on Sexual Orientation: A Review of the Literature in

Economics and Beyond." Pp. 19-43 in Sexual Orientation Discrimination: An

International Perspective., edited by M. V. L. Badgett, and J. Frank. New York:

Routledge.

Badgett, M. V. L., Holning Lau, Brad Sears, and Deborah Ho. 2007. Bias in the

Workplace: Consistent Evidence of Sexual Orientation and Gender Identity

Discrimination. Los Angeles, CA: The Williams Institute.

Barron, Laura G., and Michelle Hebl. 2013. "The Force of Law: The Effects of Sexual

Orientation Antidiscrimination Legislation on Interpersonal Discrimination in

Employment." Psychology, Public Policy, and Law 19(2):191-205.

Baumle, Amanda K., M. V. L. Badgett, and Steven Boutcher. 2020. "New Research on

Sexual Orientation and Gender Identity Discrimination: Effect of State Policy on

Charges Filed at the EEOC." Journal of Homosexuality 67(8):1135-1144.

Baumle, Amanda K., D'Lane Compton, and Dudley L. Poston, Jr. 2009. Same-Sex

Partners: The Social Demography of Sexual Orientation. Albany: SUNY Press.

Baumle, Amanda K., and Dudley L. Poston, Jr. 2011. "The Economic Cost of

Homosexuality: Multilevel Analyses." Social Forces 89(3):1005-1031.

Becker, Gary S. 1957. The Economics of Discrimination. Chicago: University of Chicago

Press.

Berg, Nathan, and Donald Lien. 2002. "Measuring the Effect of Sexual Orientation on

Income: Evidence of Discrimination?" Contemporary Economic Policy

20(4):394-414.

222

Berry, William D., Richard C. Fording, Evan J. Ringquist, Russell L. Hanson, and Carl E.

Klarner. 2010. "Measuring Citizen and Government Ideology in the U.S. States:

A Re-Appraisal." State Politics & Policy Quarterly 10(2):117-135

Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson. 1998.

"Measuring Citizen and Government Ideology in the American States, 1960-93."

American Journal of Political Science 42(1):327-348.

Bèrubè, Allan. 1990. Coming Out Under Fire: The History of Gay Men and Women in

World War Two. New York: The Free Press.

Blandford, John M. 2003. “The Nexus of Sexual Orientation and Gender in the

Determination of Earnings.” Industrial and Labor Relations Review 56(4):622-

642.

Blank, Rebecca M., Marilyn Dabady, and Constance F. Citro. 2004. Measuring Racial

Discrimination. Washington, D.C.: National Academies Press.

Blewett, Lynn A., Heather A. Dahlen, Donna Spencer, Julia A. Rivera Drew, and

Elizabeth Lukanen. 2016. "Changes to the Design of the National Health

Interview Survey to Support Enhanced Monitoring of Health Reform Impacts at

the State Level." American Journal of Public Health 106(11):1961-1966.

Blewett, Lynn A., Julia A. Rivera Drew, Miriam L. King, and Kari C.W. Williams.

IPUMS Health Surveys: National Health Interview Survey, Version 6.4 [dataset].

Minneapolis, MN: IPUMS, 2019. https://doi.org/10.18128/D070.V6.4

Blinder, Alan S. 1973. “Wage Discrimination: Reduced Form and Structural Estimates.”

The Journal of Human Resources 8(4):436-455.

223

Bonner, Valerie K. 2014. “Explaining Changes in Women’s Earnings and Employment

from 1970 to 2010: Quantitative Analysis of Discrimination and Labor Force

Hypotheses.” PhD dissertation, Department of Sociology, Temple University.

ProQuest, 3671872.

Boso, Luke A. 2016. "Acting Gay, Acting Straight: Sexual Orientation Stereotyping."

Tennessee Law Review 83(2):575-643.

Bostock v. Clayton County. 2020. 590 US _ (2020) Retrieved June 16, 2020.

(https://www.supremecourt.gov/opinions/19pdf/17-1618_hfci.pdf)

Bostock, Gerald. 2020. "Gerald Bostock on SCOTUS' Landmark LGBTQ Ruling."

https://www.cbsnews.com/news/gerald-bostock-on-scotus-landmark-lgbtq-ruling/

Bruce, Lena K. 2019. “How to Explain to Your Twins Why Only One Can Be American:

The Right to Citizenship of Children Born to Same-Sex Couples Through

Assisted Reproductive Technology.” Fordham Law Review 88(3):999-1031.

Budig, Michelle J., and Paula England. 2001. "The Wage Penalty for Motherhood."

American Sociological Review 66(2):204-225.

Burns, Crosby, and Jeff Krehely. 2011. “Gay and Transgender People Face High Rates of

Workplace Discrimination and Harassment.” Retrieved November 10, 2020.

(https://cdn.americanprogress.org/wp-

content/uploads/issues/2011/06/pdf/workplace_discrimination.pdf)

Button, James W., Barbara A. Rienzo, and Kenneth D. Wald. 2000. “The Politics of Gay

Rights at the Local and State Level.” Pp. 269-289 in The Politics of Gay Rights,

edited by C. A. Rimmerman, K. D. Wald, and C. Wilcox. Chicago: University of

Chicago Press.

224

Canaday, Margot. 2009. The Straight State: Sexuality and Citizenship in Twentieth-

Century America. Princeton, NJ: Princeton University Press.

Capps, Randy, James D. Bachmeier, and Jennifer Van Hook. 2018. "Estimating the

Characteristics of Unauthorized Immigrants using U.S. Census Data: Combined

Sample Multiple Imputation." Annals of the American Academy of Political and

Social Science 677(1):165-179.

Carpenter, Christopher S., and Samuel T. Eppink. 2017. "Does it Get Better? Recent

Estimates of Sexual Orientation and Earnings in the United States." Southern

Economic Journal 84(2):426-441.

Chandra, Anaji, William D. Mosher, Casey Copen, and Caitlainn Sionean. 2011. "Sexual

Behavior, Sexual Attraction, and Sexual Identity in the United States: Data from

the 2006-2008 National Survey of Family Growth." National Health Statistics

Report 36:1-36.

Cochran, William G., Frederick Mosteller, and John W. Tukey. 1953. "Statistical

Problems of the Kinsey Report." Journal of the American Statistical Association

48(264):673-716.

Colvin, Roddrick. 2009. "Adding Sexual Orientation to New York State's Human Rights

Law: Initial Information about Implementation and Effectiveness." Journal of

Homosexuality 56(4):485-498.

Colvin, Roddrick A., and Norma M. Riccucci. 2002. "Employment Nondiscrimination

Policies: Assessing Implementation and Measuring Effectiveness." International

Journal of Public Administration 25(1):95-108.

225

Cramer, Maria. 2020. "Woman Who Says She was Fired for being a Lesbian is Elected

Sheriff." New York Times, November 8,

(https://www.nytimes.com/2020/11/08/us/charmaine-mcguffey-sheriff-

hamilton.html).

Cushing-Daniels, Brendan, and Tsz-Ying Yeung. 2009. "Wage Penalties and Sexual

Orientation: An Update using the General Social Survey." Contemporary

Economic Policy 27(2):164-175.

D'Emilio, John. 1998. Sexual Politics, Sexual Communities: The Making of a

Homosexual Minority in the United States, 1940-1970. Chicago: University of

Chicago Press.

------. 2002. The World Turned: Essays on Gay History, Politics, and Culture. Durham,

NC: Duke University Press.

Dickey, Todd R. 1993. "Reorienting the Workplace: Examining California's New Labor

Code Section 1102.1 and Other Legal Protection Against Employment

Discrimination Based on Sexual Orientation Notes." Southern California Law

Review 66(5):2297-2330.

Doan, Petra L. 2019. "To Count or Not to Count: Queering Measurement and the

Transgender Community." Pp. 121-142 in Imagining Queer Methods., edited by

A. Ghaziani, and M. Brim. New York: New York University Press.

Donovan, Todd, Jim Wenzel, and Shaun Bowler. 2000. “Direct Democracy and Gay

Rights Initiatives after Romer.” Pp. 161-192 in The Politics of Gay Rights, edited

by C. A. Rimmerman, K. D. Wald, and C. Wilcox. Chicago: University of

Chicago Press.

226

Dorris, John B. 1999. “Antidiscrimination Laws in Local Government: A Public Policy

Analysis of Municipal Lesbian and Gay Public Employment Protections.” Pp. 39-

61 in Gays and Lesbians in the Democratic Process: Public Policy, Public

Opinion, and Political Representation, edited by E. D. B. Riggle and B. L.

Tadlock. New York: Columbia University Press.

Editors of the Harvard Law Review. 1990. Sexual Orientation and the Law. Cambridge,

MA: Harvard University Press.

Elmslie, Bruce, and Edinaldo Tebaldi. 2007.”Sexual Orientation and Labor Market

Discrimination.” Journal of Labor Research 28:436-453.

------. 2014. “The Wage Gap against Gay Men: The Leveling of the Playing Field.”

Kyklos 67(3):330-345.

Ericksen, Julia A. 1999. Kiss and Tell: Surveying Sex in the Twentieth Century.

Cambridge, MA: Harvard University Press.

Faderman, Lillian. 2015. The Gay Revolution: The Story of the Struggle. New York:

Simon & Schuster.

Fay, Robert E., and George F. Train. "Aspects of Survey and Model-Based Postcensal

Estimation of Income and Poverty Characteristics for States and Counties." Paper

presented at the American Statistical Association Conference Annual Meeting in

Orlando, Florida on August 13-17, 1995.

(https://www.census.gov/content/dam/Census/library/working-

papers/1995/demo/faytrain95.pdf)

227

Feldblum, Chai R. 2000. “The Federal Gay Rights Bill: From Bella to ENDA.” Pp. 149-

187 in Creating Change: Sexuality, Public Policy, and Civil Rights, edited by J.

D’Emilio, W. B. Turner, and U. Vaid. New York: St. Martin’s Press.

Flores, Andrew R., Jody L. Herman, Gary G. Gates, and Taylor N. T. Brown. 2016. How

Many Adults Identify as Transgender in the United States? Los Angeles, CA: The

Williams Institute.

Fording, Richard C. 2018. "State Ideology Data.", Retrieved October 19, 2020.

(https://rcfording.com/state-ideology-data/)

Gates, Gary J. 2011. How Many People Are Lesbian, Gay, Bisexual, and Transgender.

Los Angeles: UCLA School of Law.

------. 2012. "LGBT Identity: A Demographer's Perspective." Loyola of Los Angeles Law

Review 45(3):693-714.

------. 2013. "Demographics and LGBT Health." Journal of Health and Social Behavior

54(1):72-74.

------. 2014. LGBT Demographics: Comparisons among Population-Based Survey. Los

Angeles: UCLA School of Law.

Gates, Gary J., and Randall Sell. 2007. "Measuring Gay and Lesbian Couples." Pp. 235-

244 in Handbook of Measurement Issues in Family Research., edited by S. L.

Hofferth, and L. M. Casper. Mahwah, NJ: Lawrence Erlbaum Associates.

Ghaziani, Amin, and Matt Brim, eds. 2019. Imagining Queer Methods. New York: New

York University Press.

Glauber, Rebecca. 2018. "Trends in the Motherhood Wage Penalty and Fatherhood Wage

Premium for Low, Middle, and High Earners." Demography 55(5):1663-1680.

228

Gluck Mezey, Susan. 2007. Queers in Court: Gay Rights Law and Public Policy.

Lanham, MD: Rowman & Littlefield.

Goldberg-Hiller, Jonathan. 2008. "Sowing the Dragon’s Teeth: Materialization in

Lesbian and Gay Anti-discrimination Rights." Pp. 219-235 in Handbook of

Employment Discrimination Research: Rights and Realities., edited by L. B.

Nielsen, and R. L. Nelson. New York, NY: Springer New York.

Gregory, Nancy. 2001. “The Gay and Lesbian Movement in the United States.” Pp. 152-

164 in Doing Democracy: The MAP Model for Organizing Social Movements,

edited by B. Moyer, J. McAllister, M. L. Finley, and S. Soifer. Gabriola Island,

British Columbia: New Society Publishers.

Griffith, Kristin H., and Michelle R. Hebl. 2002. "The Disclosure Dilemma for Gay Men

and Lesbians: ‘Coming Out’ at Work." Journal of Applied Psychology

87(6):1191-1199.

Gross, Jane. 1992. "California Governor, in Reversal, Signs a Bill on Gay Rights in

Jobs." New York Times, September 26, 1992.

Gunderson, Morley. 1989. “Male-Female Wage Differentials and Policy Responses.”

Journal of Economic Literature 27(1):46-72.

Heller Clain, Suzanne, and Karen Leppel. 2001. “An Investigation into Sexual

Orientation Discrimination as an Explanation for Wage Differences.” Applied

Economics 33:37-47.

Hicks, John. 1964. The Theory of Wages. New York: St. Martin's Press.

229

Hodges, Melissa J., and Michelle J. Budig. 2010. "Who Gets the Daddy Bonus?

Organizational Hegemonic Masculinity and the Impact of Fatherhood on

Earnings." Gender and Society 24(6):717-745.

Hogan, Howard, Martin O'Connell, and Sarah Feliz. n.d. "Same Sex Households in the

United States Census: Measurement Issues and Substantive Results." Draft report

of the U.S. Bureau of the Census.

Hsieh, Hsiu-Fang, and Sarah E. Shannon. 2005. "Three Approaches to Qualitative

Content Analysis." Qualitative Health Research 15(9):1277-1288.

Infanti, Anthony C. 2007. Everyday Laws for Gays and Lesbians and Those Who Care

about Them. Boulder, CO: Paradigm Publishers.

Jann, Ben. 2008. "The Blinder-Oaxaca Decomposition for Linear Regression Models."

The Stata Journal 8(4):453-479.

Jasiunas, J. Banning. 2000. “Is ENDA the Answer: Can a ‘Separate but Equal’ Federal

Statue Adequately Protect Gays and Lesbians from Employment Discrimination?”

Ohio State Law Journal 61:1529-1557.

Johnson, Kerri L., Simone Gill, Victoria Reichman, and Louis G. Tassinary. 2007.

"Swagger, Sway, and Sexuality: Judging Sexual Orientation from Body Motion

and Morphology." Journal of Personality and Social Psychology 93(3):321-34.

Jones, Angela, and Michael Yarbrough. 2016. “Dear After Marriage Community.” Letter

to conference attendees of the After Marriage Conference at the Center for

LGBTQ Studies, City University of New York, New York City, October 1-2,

2016.

230

Jones, F. L., and Jonathan Kelley. 1984. "Decomposing Differences between Groups: A

Cautionary Note on Measuring Discrimination." Sociological Methods &

Research 12(3):323-343.

Kaestle, Christine E. 2019. "Sexual Orientation Trajectories Based on Sexual Attractions,

Partners, and Identity: A Longitudinal Investigation from Adolescence through

Young Adulthood using a U.S. Representative Sample." Journal of Sex Research

56(7):811-826.

Kameny, Franklin E. 2000. "Government v. Gays: Two Sad Stories with Two Happy

Endings, Civil Service Employment and Security Clearances." Pp. 188-207 in

Creating Change: Sexuality, Public Policy, and Civil Rights., edited by J.

D'Emilio, W.B. Turner and U. Vaid. New York: St. Martin's Press.

Kaufman, Robert L. 1983. “A Structural Decomposition of Black-White Earnings

Differentials.” American Journal of Sociology 89(3):585-611.

Killewald, Alexandra, and Margaret Gough. 2013. "Does Specialization Explain

Marriage Penalties and Premiums?" American Sociological Review 78(3):477-

502.

Kinsey, Alfred C., Wardell B. Pomeroy, and Clyde E. Martin. 1975. Sexual Behavior in

the Human Male. Bloomington, IN: Indiana University Press.

Kinsey, Alfred C., Wardell B. Pomeroy, Clyde E. Martin, and Paul H. Gebhard. 1998.

Sexual Behavior in the Human Female. Bloomington, IN: Indiana University

Press.

231

Klawitter, Marieka. 2011. “Multilevel Analysis of the Effects of Antidiscrimination

Policies on Earnings by Sexual Orientation.” Journal of Policy Analysis and

Management 30(2):334-358.

Klawitter, Marieka M., and Victor Flatt. 1998. “The Effects of State and Local

Antidiscrimination Policies on Earnings for Gays and Lesbians.” Journal of

Policy Analysis and Management 17(4):658-686.

Klawitter, Marieka, and Brian Hammer. 1999. “Spatial and Temporal Diffusion of Local

Antidiscrimination Policies for Sexual Orientation.” Pp. 22-38 in Gays and

Lesbians in the Democratic Process: Public Policy, Public Opinion, and Political

Representation, edited by E. D. B. Riggle and B. L. Todlock. New York:

Columbia University Press.

Kosciw, Joseph G., Caitlin M. Clark, Nhan L. Truong, and Adrian D. Zongrone. 2020.

“The 2019 National School Climate Survey: The Experiences of Lesbian, Gay,

Bisexual, Transgender, and Queer Youth in Our Nation’s Schools.” Report by

GLSEN. Retrieved October 20, 2020. (https://www.glsen.org/research/2019-

national-school-climate-survey)

Laumann, Edward O., John H. Gagnon, Robert T. Michael, and Stuart Michaels. 1994.

The Social Organization of Sexuality: Sexual Practices in the United States.

Chicago: University of Chicago Press.

Levin, Jack, and William C. Levin. 1982. The Functions of Discrimination and

Prejudice. New York: Harper & Row.

232

Lipka, Michael, and Benjamin Wormald. 2016. "How Religious Is Your State?”

Retrieved October 19, 2020. (https://www.pewresearch.org/fact-

tank/2016/02/29/how-religious-is-your-state/)

Loh, Eng S. 1996. "Productivity Differences and the Marriage Wage Premium for White

Males." Journal of Human Resources 31:566-89.

Mallory, Christy, and Brad Sears. 2016. “Evidence of Housing Discrimination Based on

Sexual Orientation and Gender Identity: An Analysis of Complaints Filed with

State Enforcement Agencies, 2008-2014.” Report by the Williams Institute.

Retrieved March 1, 2016. (http://williamsinstitute.law.ucla.edu/wp-

content/uploads/Housing-Discrimination-Complaints-2008-2014.pdf)

Martell, Michael E. 2013a. “Differences Do Not Matter: Exploring the Wage Gap for

Same-Sex Behaving Men.” Eastern Economic Journal 39:45-71.

------. 2013b. "Do ENDAs End Discrimination for Behaviorally Gay Men?" Journal of

Labor Research 34:147-169.

Matsick, Jes L., Britney M. Wardecker, and Flora Oswald. 2020. "Treat Sexual Stigma to

Heal Health Disparities: Improving Sexual Minorities’ Health Outcomes." Policy

Insights from the Behavioral and Brain Sciences 7(2):205-213.

McCarthy, Justin. 2020. "U.S. Support for Same-Sex Marriage Matches Record High."

Gallup, June 1, 2020. Retrieved June 5, 2020.

(https://news.gallup.com/poll/311672/support-sex-marriage-matches-record-

high.aspx)

McFeeley, Tim. 2000. "Getting It Straight: A Review of the 'Gays in the Military'

Debate." Pp. 236-250 in Creating Change: Sexuality, Public Policy, and Civil

233

Rights., edited by J. D'Emilio, W.B. Turner and U. Vaid. New York: St. Martin's

Press.

Miller, Kevin, and Deborah J. Vagins. 2018. The Simple Truth about the Gender Pay

Gap: Fall 2018 Edition. Retrieved November 18, 2020.

(https://www.aauw.org/app/uploads/2020/02/AAUW-2018-SimpleTruth-nsa.pdf)

Miller, Kristen. 2001. Cognitive Testing of the NHANES Sexual Orientation Questions.

National Center for Health Statistics.

Miller, Kristen, and J. M. Ryan. 2011. Design, Development and Testing of the NHIS

Sexual Identity Question. National Center for Health Statistics.

Mishel, Emma. 2016. “Discrimination against Queer Women in the U.S. Workforce: A

Resume Audit Study.” Socius: Sociological Research for a Dynamic World 1:1-

13.

Mize, Trenton D. 2016. "Sexual Orientation in the Labor Market." American Sociological

Review 81(6):1132-1160.

Mize, Trenton D., and Bianca Manago. 2018. "The Stereotype Content of Sexual

Orientation." Social Currents 5(5):458-478.

Monro, Surya, Sally Hines, and Antony Osborne. 2017. "Is Bisexuality Invisible? A

Review of Sexualities Scholarship 1970-2015." The Sociological Review

65(4):663-681.

Mucciaroni, Gary. 2016. "Will Victory Bring Change? A Mature Social Movement Faces

the Future." Pp. 17-41 in After Marriage Equality: The Future of LGBT Rights.,

edited by C. A. Ball. New York: New York University Press.

234

Muñoz, José Esteban. 1996. “Ephemera as Evidence: Introductory Notes to Queer Acts.”

Women & Performance: A Journal of Feminist Theory 8(2):5-16.

National Center for Health Statistics. "2017-2019 NSFG Questionnaires." Retrieved

October 20, 2020.

(https://www.cdc.gov/nchs/nsfg/nsfg_2017_2019_questionnaires.htm)

Neumark, David. 1988. "Employers' Discriminatory Behavior and the Estimation of

Wage Discrimination." The Journal of Human Resources 23(3):279-95.

New York Times. 1950. "Pervert Inquiry Ordered." June 15, A6.

Newport, Frank. 2018. "In U.S., Estimates of LGBT Population Rises to 4.5%." Gallup,

Retrieved August 5, 2020. (https://news.gallup.com/poll/234863/estimate-lgbt-

population-rises.aspx)

Oaxaca, Ronald. 1973. “Male-Female Wage Differentials in Urban Labor Markets.”

International Economic Review 14(3):693-709.

Ochs, Robin. 1996. "Biphobia: It Goes More Than Two Ways." Pp. 217-239 in

Bisexuality: The Psychology and Politics of an Invisible Minority, edited by B. A.

Firestein. Thousand Oaks, CA: SAGE Publications.

Osborne, Jason W. 2015. Best Practices in Logistic Regression. Los Angeles: Sage.

Parmley, Suzette. 2015. "Chasing LGBT Dollars: Businesses are Eager to Cater to the

Community, with its High Income and Generous Spending Patterns."

Philadelphia Inquirer, September 6, E5.

Pew Research Center. 2017. The Partisan Divide on Political Values Grows Even Wider.

Retrieved October 19, 2020. (https://www.pewresearch.org/politics/wp-

content/uploads/sites/4/2017/10/10-05-2017-Political-landscape-release-updt..pdf)

235

Public Religion Research Institute. 2019. "American Values Atlas." Retrieved October

19, 2020. (http://ava.prri.org/)

Quintana, Nico S. 2009. “Poverty in the LGBT Community.” Retrieved March 9, 2014.

(https://cdn.americanprogress.org/wp-

content/uploads/issues/2009/07/pdf/lgbt_poverty.pdf?_ga=2.142564091.8770543

04.1613593373-2131911234.1613403037)

Raeburn, Nicole C. 2004. Changing Corporate America from the Inside Out: Lesbian

and Gay Workplace Rights. Minneapolis: University of Minnesota Press.

Raudenbush, Stephen W., and Anthony S. Bryk. 2002. Hierarchical Linear Models:

Applications and Data Analysis Methods. Thousand Oaks, CA: SAGE

Publications.

Reiter, Laura. 1989. "Sexual Orientation, Sexual Identity, and the Question of Choice."

Clinical Social Work Journal 17(2):138-150.

Rendall, Michael S., Bonnie Ghosh-Dastidar, Margaret M. Weden, Elizabeth H. Baker,

and Zafar Nazarov. 2013. "Multiple Imputation for Combined-Survey Estimation

with Incomplete Regressors in One but Not both Surveys." Sociological Methods

and Research 42(4):483-530.

Reuters/Ipsos. 2019. “Ipsos Poll Conducted for Reuters: Stonewall Anniversary Poll.”

Retrieved January 15, 2021.

(https://static.reuters.com/resources/media/editorial/20190612/StonewallFinalRes

ults.pdf)

Richardson, Diane. 2007. "Patterned Fluidities: (Re)Imagining the Relationship between

Gender and Sexuality." Sociology 41(3):457-474.

236

Ridolfo, Heather, Kristen Miller, and Aaron Maitland. 2012. "Measuring Sexual Identity

using Survey Questionnaires: How Valid are our Measures?" Sexuality Research

and Social Policy 9(2):113-124.

Rodgers, Yana van der Meulen. 2006. “A Primer on Wage Gap Decompositions in the

Analysis of Labor Market Discrimination.” Pp. 11-28 in Handbook on the

Economics of Discrimination, edited by W. M. Rodgers III. Northampton, MA:

Edward Elgar Publishing, Inc.

Rodriguez, Germán. 2007. "Lecture Notes on Generalized Linear Models.", Retrieved

May 15, 2020. (https://data.princeton.edu/wws509/notes/)

Rubenstein, William B. 2001. “Do Gay Rights Laws Matter? An Empirical Assessment.”

Southern California Law Review 75(65):65-120.

Rubin, Donald B. 1987. Multiple Imputation for Nonresponse in Surveys. New York:

John Wiley and Sons.

Ruggles, Steven, Sarah Flood, Sophia Foster, Ronald Goeken, Jose Pacas, Megan

Schouweiler, and Matthew Sobek. 2021 IPUMS USA: Version 11.0 [dataset].

Minneapolis, MN: IPUMS, 2021. https://doi.org/10.18128/D010.V11.0

Rule, Nicholas O., Nalini Ambady, Reginald B. Adams Jr., and C. Neil Macrae. 2008.

"Accuracy and Awareness in the Perception and Categorization of Male Sexual

Orientation." Journal of Personality and Social Psychology 95(5):1019-28.

Rule, Nicholas O., Nalini Ambady, and Katherine C. Hallett. 2009. "Female Sexual

Orientation Is Perceived Accurately, Rapidly, and Automatically from the Face

and its Features." Journal of Experimental Social Psychology 45(6):1245-51.

237

Russell, Stephen T., Thomas J. Clarke, and Justin Clary. 2009. "Are Teens "Post-Gay"?

Contemporary Adolescents' Sexual Identity Labels." Journal of Youth and

Adolescence 38(7):884-90.

Savin-Williams, Ritch C. 2009. "How Many Gays Are There? It Depends." Pp. 5-41 in

Contemporary Perspectives on Lesbian, Gay, and Bisexual Identities., edited by

D. A. Hope. New York: Springer.

Schneider, Beth E. 1986. "Coming Out at Work: Bridging the Private/Public Gap." Work

and Occupations 13(4):463-487.

Smith, Tom W., Michael Davern, Jeremy Freese, and Stephen L. Morgan. 2019. General

Social Surveys, 1972-2018 [machine-readable data file]. Chicago: NORC.

StataCorp. 2019a. Stata Statistical Software: Release 16. College Station, TX: StataCorp

LLC.

StataCorp. 2019b. Stata 16 Multiple Imputation Reference Manual. College Station, TX:

Stata Press.

Stein, Marc. 2012. Rethinking the Gay and Lesbian Movement. New York: Routledge.

Stone, Amy L. 2012. Gay Rights at the Ballot Box. Minneapolis: University of Minnesota

Press.

Stulberg, Lisa M. 2018. LGBTQ Social Movements. Cambridge, UK: Polity.

Stychin, Carl F. 2005. "Being Gay." Government and Opposition 40(1):90-109.

238

Thompson, John H. 2017. "Director's Blog: Planned Subjects for the 2020 Census and the

American Community Survey." United States Census Bureau. www.census.gov

Blogs. Retrieved July 1, 2020.

(https://www.census.gov/newsroom/blogs/director/2017/03/planned_subjects_202

0.html)

Tilcsik, András. 2011. “Pride and Prejudice: Employment Discrimination against Openly

Gay Men in the United States.” American Journal of Sociology 117(2):586-626.

U.S. Bureau of the Census. 2020a. Understanding and Using American Community

Survey Data: What All Data Users Need to Know. Washington DC: United States

Department of Commerce. Retrieved October 19, 2020.

(census.gov/content/dam/Census/library/publications/2020/acs/acs_general_hand

book_2020.pdf)

------. 2020b. “Subject Definitions.” Retrieved July 1, 2020.

(https://www.census.gov/programs-surveys/cps/technical-documentation/subject-

definitions.html)

U.S. Senate. 1950. “Employment of Homosexuals and Other Sex Perverts in

Government.” 81st Congress, Senate Document 241.

Vaid, Urvashi. 1996. Virtual Equality: The Mainstreaming of Gay and Lesbian

Liberation. New York: Anchor Books, Doubleday.

Van Hook, Jennifer, James D. Bachmeier, Donna L. Coffman, and Ofer Harel. 2015.

"Can we Spin Straw into Gold? an Evaluation of Immigrant Legal Status

Imputation Approaches." Demography 52(1):329-354.

Velte, Kyle C. 2020. "Straightwashing the Census." Boston Law Review 61(1):69-127.

239

Voeller, Bruce. 1990. "Some Uses and Abuses of the Kinsey Scale." Pp. 32-38 in

Homosexuality/Heterosexuality: Concepts of Sexual Orientation., edited by D. P.

McWhirter, S.A. Sanders and J. Machover Reinisch. New York: Oxford

University Press.

Waidzunas, Tom. 2015. The Straight Line: How the Fringe Science of Ex-Gay Therapy

Reoriented Sexuality. Minneapolis: University of Minnesota Press.

Waidzunas, Tom, and Steven Epstein. 2015. "'For Men Arousal is Orientation': Bodily

Truthing, Technosexual Scripts, and the Materialization of Sexualities through

Phallometric Tests." Social Studies of Science 45(2):187-213.

Waldfogel, Jane. 1997. "The Effect of Children on Women's Wages." American

Sociological Review 62(2):209-217.

Walther, Carol S., and Dudley L. Poston. 2004. "Patterns of Gay and Lesbian Partnering

in the Larger Metropolitan Areas of the United States." Journal of Sex Research

41(2):201-214.

Ward, Jane. 2015. Not Gay: Sex between Straight White Men. New York: New York

University Press.

Wax, Amy, Kimberlee K. Coletti, and Joseph W. Ogaz. 2018. "The Benefit of Full

Disclosure: A Meta-Analysis of the Implications of Coming Out at Work."

Organizational Psychology Review 8(1):3-30.

Weeks, Jeffrey. 1996. "The Construction of Homosexuality." Pp. 41-63 in Queer

Theory/Sociology, edited by S. Seidman. Cambridge, MA: Blackwell Publishers,

Ltd.

------. 1998. "The Sexual Citizen." Theory, Culture & Society 15(3-4):35-52.

240

Weinberg, Martin S. 1970. "Homosexual Samples: Differences and Similarities." Journal

of Sex Research 6(4):312-325.

Williams Institute. 2019. "Adult LGBT Population in the United States." Retrieved

January 3, 2020. (https://williamsinstitute.law.ucla.edu/wp-

content/uploads/LGBT-Adult-US-Pop-Mar-2019.pdf)

Winsborough, H. H., and Peter J. Dickinson. 1971. “Components of Negro-White Income

Differentials.” Pp. 6-8 in the Proceedings of the Social Statistics Section, Meeting

of the American Statistical Association. Fort Collins, CO.

Wolff, Margaret, Brooke Wells, Christina Ventura-DiPersia, Audrey Renson, and

Christian Grov. 2017. "Measuring Sexual Orientation: A Review and Critique of

U.S. Data Collection Efforts and Implications for Health Policy." Journal of Sex

Research 54(4-5):507-531.

Worthen, Meredith G. F. 2013. "An Argument for Separate Analyses of Attitudes

Toward Lesbian, Gay, Bisexual Men, Bisexual Women, MtF and FtM

Transgender Individuals." Sex Roles 68(11-12):703-723.

Yoshino, Kenji. 2000. "The Epistemic Contract of Bisexual Erasure." Stanford Law

Review 52(2):353-461.

Young, Rebecca M., and Ilan H. Meyer. 2005. "The Trouble with ‘MSM’ and ‘WSW’:

Erasure of the Sexual-Minority Person in Public Health Discourse." American

Journal of Public Health 95(7):1144-1149.

Zuberi, Tukufu. 2001. Thicker than Blood: How Racial Statistics Lie. Minneapolis:

University of Minnesota Press.

241

APPENDIX A

STATE SAMPLE SIZES

Table A-1. Sample Sizes by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 81,713 94,458 1,291 1,167 366 999 3,823 Alaska 10,766 10,596 202 144 79 185 610 Arizona 116,285 126,854 2,733 2,234 806 1,987 7,760

242 Arkansas 50,784 56,933 760 668 243 643 2,314 California 648,740 704,889 18,526 13,139 4,947 12,205 48,818

Colorado 95,823 100,612 2,170 1,896 686 1,658 6,410 Connecticut 61,846 69,179 1,219 1,073 282 730 3,304 Delaware 15,875 18,015 427 369 75 195 1,065 District of Columbia 10,378 13,007 897 446 90 249 1,681 Florida 350,531 399,695 8,592 6,238 1,614 4,188 20,632 Georgia 164,024 189,796 3,406 2,863 782 2,141 9,192 Hawaii 23,846 26,199 642 482 185 448 1,758 Idaho 27,956 29,081 405 334 170 417 1,326 Illinois 220,737 241,357 4,007 2,996 1,234 3,092 11,330 Indiana 115,140 124,595 1,772 1,595 642 1,634 5,643 Iowa 55,969 58,911 689 630 279 680 2,278 Kansas 49,931 53,291 637 580 263 647 2,127 Kentucky 76,964 85,389 1,243 1,180 386 1,013 3,822 Louisiana 73,503 85,684 1,479 1,292 359 992 4,122

Table A-1. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Maine 23,409 25,003 390 433 96 269 1,188 Maryland 100,632 115,425 2,145 2,037 464 1,227 5,873 Massachusetts 117,922 131,739 2,855 2,725 623 1,589 7,791 Michigan 177,459 189,072 2,660 2,274 962 2,376 8,272 Minnesota 98,820 99,527 1,370 1,221 475 1,110 4,176 Mississippi 47,290 56,631 750 696 208 584 2,237 Missouri 106,191 116,782 1,599 1,401 579 1,483 5,062 Montana 18,360 18,591 281 243 106 254 884 Nebraska 33,192 34,900 416 364 159 401 1,340 243 Nevada 49,901 52,651 1,394 886 393 912 3,585

New Hampshire 24,846 25,963 426 455 116 268 1,264 New Jersey 152,738 171,814 2,867 2,369 677 1,689 7,602 New Mexico 32,642 36,487 818 763 222 565 2,368 New York 337,092 380,920 7,834 6,090 1,739 4,517 20,180 North Carolina 168,271 194,216 3,068 3,004 794 2,172 9,039 North Dakota 14,216 13,949 171 118 70 156 515 Ohio 205,853 225,467 3,304 2,880 1,181 3,001 10,366 Oklahoma 62,206 67,242 995 904 331 861 3,091 Oregon 72,398 77,237 1,654 1,646 530 1,350 5,180 Pennsylvania 226,288 245,265 3,632 3,224 1,035 2,581 10,472 Rhode Island 17,965 20,556 408 352 95 250 1,104 South Carolina 82,565 96,067 1,468 1,326 384 1,038 4,216 South Dakota 15,041 15,391 160 160 71 176 568 Tennessee 114,743 128,882 2,005 1,818 565 1,494 5,881

Table A-1. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Texas 439,666 486,349 8,623 7,135 2,143 5,446 23,347 Utah 48,467 50,298 795 651 307 756 2,508 Vermont 11,611 12,264 184 244 50 136 614 Virginia 140,876 160,318 2,765 2,432 666 1,791 7,654 Washington 127,896 136,422 2,937 2,541 888 2,243 8,608 West Virginia 32,330 35,325 533 450 164 417 1,564 Wisconsin 107,502 109,144 1,454 1,244 527 1,272 4,497 Wyoming 10,336 10,629 147 138 63 149 497

244

APPENDIX B

STATE-LEVEL DEMOGRAPHIC AND ECONOMIC CHARACTERISTICS

Table B-1. Mean Age by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 47.43 49.09 43.19 45.13 34.07 35.19 39.76 (17.9634) (18.4968) (16.7755) (17.3898) (15.0802) (15.3602) (16.9103) Alaska 45.05 45.61 44.13 43.60 32.90 34.03 39.05

245 (17.0472) (17.1979) (16.4996) (16.5332) (13.7977) (14.2374) (16.2377)

Arizona 47.45 48.94 43.05 44.75 33.22 34.61 39.27 (18.5985) (18.8471) (16.7009) (17.6576) (14.6590) (15.3929) (16.9430) Arkansas 47.50 49.02 42.61 45.03 33.89 34.63 39.33 (18.1052) (18.6065) (16.6375) (18.0971) (14.9910) (15.1895) (16.9907) California 45.65 47.45 42.61 43.57 32.56 33.46 38.39 (17.7160) (18.3925) (15.8535) (16.8159) (13.5100) (14.2557) (16.0238) Colorado 45.89 47.43 41.65 43.88 32.91 34.11 38.46 (17.3597) (17.9829) (15.6397) (16.4112) (13.7517) (14.6396) (15.9072) Connecticut 47.92 49.97 45.31 46.77 34.57 35.77 41.01 (17.8526) (18.4835) (16.8885) (16.6426) (15.0380) (15.6682) (17.0278) Delaware 48.28 49.84 47.92 48.56 35.79 35.92 42.40 (18.1871) (18.3786) (16.6818) (16.7271) (15.9044) (15.8001) (17.4104) District of Columbia 43.23 44.85 43.02 40.77 33.13 33.48 38.08 (16.9583) (18.2476) (13.6127) (14.7178) (11.8383) (12.3509) (13.9359)

Table B-1. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Florida 49.19 50.81 45.72 46.47 34.67 35.86 41.08 (18.7356) (19.0231) (16.6867) (17.4428) (15.2874) (15.7799) (17.2005) Georgia 45.99 47.40 42.97 43.48 33.59 34.29 38.92 (17.3483) (17.8676) (15.6487) (16.2563) (14.2116) (14.6089) (15.9354) Hawaii 48.60 50.10 45.50 44.71 34.55 34.32 40.10 (18.2625) (18.9364) (16.1751) (17.2892) (14.6219) (14.6183) (16.6259) Idaho 47.21 48.29 41.25 43.59 32.91 33.89 38.20 (18.1508) (18.5526) (16.4294) (16.9269) (14.1149) (15.0066) (16.3890) Illinois 46.60 48.41 43.08 44.31 33.35 34.23 39.07 246 (17.6809) (18.4773) (15.8281) (17.0539) (14.1644) (14.8650) (16.3101)

Indiana 46.94 48.67 43.88 45.43 33.37 34.65 39.75 (17.7820) (18.5146) (16.5293) (16.9889) (14.7867) (15.1843) (16.8018) Iowa 47.65 49.46 43.10 45.92 33.53 34.51 39.59 (18.0875) (18.9366) (17.3580) (18.3653) (15.2035) (15.6705) (17.5640) Kansas 46.88 48.56 41.87 45.79 33.22 33.88 38.95 (18.0597) (18.7654) (16.1339) (18.1637) (14.6627) (15.1639) (16.9465) Kentucky 47.34 49.00 42.98 45.42 34.08 35.24 39.79 (17.7237) (18.2650) (16.5611) (17.4734) (14.9845) (15.3202) (16.8579) Louisiana 46.45 48.00 43.61 44.19 33.85 34.71 39.45 (17.8126) (18.4259) (16.5831) (17.0161) (14.4098) (14.8931) (16.5448) Maine 49.98 51.43 46.39 49.11 35.40 37.20 42.59 (17.8092) (18.2277) (16.6510) (16.1749) (15.4843) (16.2646) (17.1897) Maryland 46.96 48.54 43.48 44.88 34.01 35.23 39.79 (17.5422) (18.0551) (16.1266) (16.1903) (14.5218) (15.0779) (16.2714)

Table B-1. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Massachusetts 47.23 49.14 44.49 46.27 33.80 35.32 40.41 (17.8245) (18.5006) (15.7333) (16.3141) (14.2245) (15.2729) (16.4030) Michigan 47.65 49.39 44.14 45.77 33.70 34.88 40.01 (18.0435) (18.6410) (16.9413) (17.3154) (14.9801) (15.5751) (17.1416) Minnesota 47.14 48.68 43.89 46.52 33.56 34.96 40.13 (17.6956) (18.4052) (16.5096) (17.1667) (14.5371) (15.6241) (16.9961) Mississippi 46.88 48.51 43.09 44.79 33.95 35.14 39.61 (17.9190) (18.4605) (16.9774) (17.1598) (14.7168) (15.0038) (16.7318) Missouri 47.52 49.16 43.55 45.29 33.82 34.98 39.77 247 (17.9750) (18.5890) (16.4271) (17.4098) (14.9259) (15.6204) (16.9307)

Montana 48.75 49.88 45.14 46.14 34.58 35.38 40.65 (18.2785) (18.5556) (18.8452) (17.2063) (15.2835) (15.7502) (17.7430) Nebraska 46.87 48.54 42.27 46.25 33.13 33.87 39.20 (17.9424) (18.7103) (17.4093) (18.3520) (14.7336) (15.4004) (17.5187) Nevada 47.08 48.01 42.64 43.11 34.30 34.41 38.84 (17.7130) (18.0288) (15.6915) (16.1937) (14.2576) (14.2945) (15.7247) New Hampshire 48.73 50.16 45.98 49.03 34.17 36.45 41.96 (17.5895) (18.0173) (16.8255) (17.2933) (14.8664) (15.9208) (17.4646) New Jersey 47.30 49.34 44.85 46.75 33.96 35.51 40.68 (17.6260) (18.3663) (16.2320) (17.3904) (14.4471) (15.4623) (16.9004) New Mexico 47.49 49.04 44.80 45.69 33.76 35.14 40.27 (18.4012) (18.7509) (17.4103) (17.1951) (14.8370) (15.6338) (17.2589) New York 46.88 48.83 43.45 45.41 33.71 34.95 39.72 (17.8920) (18.6600) (15.3964) (16.6083) (13.9863) (14.7998) (16.0647)

Table B-1. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB North Carolina 47.33 48.78 43.46 45.09 33.80 34.95 39.72 (17.6871) (18.2436) (16.5015) (17.0666) (14.5152) (15.1550) (16.6742) North Dakota 45.87 47.80 41.53 41.39 32.01 32.75 37.24 (18.2151) (19.0703) (16.9823) (18.2849) (13.6222) (15.0908) (16.7933) Ohio 47.73 49.52 43.82 45.06 34.36 35.15 39.94 (17.9202) (18.5887) (16.4147) (17.0891) (14.8977) (15.3354) (16.7228) Oklahoma 46.69 48.38 42.20 44.18 32.96 34.17 38.74 (18.0379) (18.6427) (16.7150) (17.0293) (14.1420) (15.0012) (16.5871) Oregon 47.61 49.20 43.60 46.44 33.74 34.73 40.01 248 (18.0203) (18.5589) (16.3052) (16.8445) (14.4135) (15.1628) (16.7055)

Pennsylvania 48.27 50.21 44.91 46.84 34.63 35.71 40.91 (18.0614) (18.7777) (16.6759) (17.4506) (15.1412) (15.7585) (17.1892) Rhode Island 47.48 49.58 44.12 45.44 34.57 35.05 40.10 (17.9853) (18.7806) (16.1847) (16.9103) (14.5402) (15.4879) (16.6649) South Carolina 47.89 49.39 43.22 44.76 34.66 35.12 39.71 (18.0034) (18.3863) (16.9435) (17.2271) (15.1255) (15.3410) (16.8616) South Dakota 47.56 49.08 45.27 46.48 34.68 34.15 40.42 (18.0409) (18.7901) (17.9565) (18.6181) (15.6666) (15.6395) (18.0403) Tennessee 47.28 48.77 43.02 44.83 33.82 34.91 39.49 (17.7629) (18.3497) (16.2168) (17.0427) (14.5794) (15.0474) (16.5042) Texas 44.90 46.37 41.10 42.35 32.63 33.62 37.73 (17.3494) (17.9550) (15.6229) (16.3730) (13.6614) (14.3761) (15.7094) Utah 43.56 44.66 39.63 43.38 31.35 32.15 36.85 (17.5939) (18.1193) (16.1801) (17.3681) (13.4769) (13.9393) (16.1447)

Table B-1. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Vermont 49.20 50.68 47.02 46.40 34.38 36.11 41.45 (17.8799) (18.2523) (16.6324) (16.0204) (15.2634) (15.8458) (16.9198) Virginia 47.07 48.42 43.62 44.47 33.69 34.67 39.48 (17.6117) (18.1471) (16.2730) (16.6807) (14.2942) (14.9271) (16.4011) Washington 46.72 48.16 43.54 45.43 33.47 34.23 39.50 (17.6698) (18.2361) (15.9734) (16.7766) (14.3212) (14.8153) (16.4265) West Virginia 49.13 50.75 45.92 47.11 35.18 36.33 41.53 (18.0353) (18.5712) (17.4658) (17.7650) (15.5902) (16.0108) (17.6141) Wisconsin 47.78 49.35 43.70 45.90 33.95 34.86 39.95 249 (17.8581) (18.5211) (16.6811) (17.3957) (14.9540) (15.6045) (17.0647)

Wyoming 47.31 48.68 43.62 44.03 33.83 34.34 39.25 (17.9148) (18.3777) (18.2315) (19.0026) (15.1302) (15.2548) (17.7329)

Table B-2. Percent White, Non-Hispanic by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 69.78% 67.27% 62.98% 62.10% 67.44% 63.93% 63.67% (0.1653) (0.1572) (2.2804) (2.1356) (3.9127) (2.8332) (1.3649) Alaska 67.36% 63.95% 60.45% 66.26% *** 57.39% 59.66% (0.4640) (0.4772) (5.1679) (6.4073) *** (6.4727) (2.9961) Arizona 60.39% 60.30% 58.32% 58.57% 52.73% 52.35% 55.62% (0.1485) (0.1409) (1.5491) (1.4328) (3.2504) (1.9586) (0.7406) Arkansas 76.42% 75.54% 70.20% 73.30% 71.10% 70.48% 71.15% (0.1922) (0.1896) (2.6095) (3.3596) (5.4606) (3.3956) (1.8118)

250 California 41.47% 40.64% 45.95% 44.63% 35.93% 35.18% 40.70% (0.0631) (0.0604) (0.5675) (0.6469) (1.2449) (0.8103) (0.3875)

Colorado 72.25% 72.18% 70.00% 72.22% 67.90% 67.79% 69.49% (0.1480) (0.1457) (1.4096) (1.8898) (2.5459) (1.8467) (0.9353) Connecticut 71.29% 70.65% 67.61% 68.88% 61.86% 60.90% 64.87% (0.1856) (0.1793) (2.3860) (2.4598) (4.6547) (3.2474) (1.5699) Delaware 66.93% 66.26% 71.68% 69.61% 63.67% 65.25% 67.92% (0.3798) (0.3585) (3.3321) (3.2343) (8.9259) (5.1686) (2.4346) District of Columbia 41.26% 36.91% 58.05% 39.76% *** 45.27% 48.61% (0.4975) (0.4396) (2.0533) (3.2671) *** (5.6220) (2.2224) Florida 58.24% 57.13% 59.67% 58.00% 55.24% 54.84% 57.08% (0.0847) (0.0799) (0.8402) (0.9046) (1.9305) (1.4340) (0.5713) Georgia 57.89% 55.53% 54.87% 49.43% 56.50% 53.43% 53.28% (0.1252) (0.1177) (1.2197) (1.3656) (3.7918) (1.8649) (1.0215) Hawaii 24.38% 21.82% 29.44% 23.40% *** 19.60% 23.86% (0.2900) (0.2608) (2.9643) (2.6549) *** (3.2484) (1.5578)

Table B-2. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 84.72% 85.37% 83.05% 80.14% 80.02% 77.39% 80.06% (0.2208) (0.2219) (2.8723) (3.2224) (4.9904) (4.2981) (2.1832) Illinois 65.39% 64.61% 63.52% 60.59% 58.84% 59.16% 60.78% (0.1040) (0.0994) (1.1662) (1.4629) (2.5390) (1.4203) (0.7323) Indiana 82.65% 82.23% 77.54% 79.74% 77.18% 77.79% 78.12% (0.1146) (0.1123) (1.6701) (1.6359) (2.9202) (1.7968) (0.9702) Iowa 88.37% 89.16% 85.17% 86.67% 84.43% 83.50% 84.87% (0.1385) (0.1334) (2.2072) (2.1041) (3.9398) (2.5967) (1.1725) Kansas 79.36% 80.11% 70.18% 79.81% 73.91% 73.50% 74.11% 251 (0.1863) (0.1787) (2.6077) (2.5871) (5.0294) (3.2292) (1.7826)

Kentucky 87.12% 87.19% 81.87% 83.77% 83.24% 83.58% 83.12% (0.1239) (0.1188) (1.7929) (1.8648) (3.7103) (2.2381) (1.1281) Louisiana 63.33% 60.74% 59.38% 58.52% 59.56% 59.05% 59.08% (0.1818) (0.1687) (2.1178) (1.8954) (4.7153) (2.6593) (1.2338) Maine 94.82% 94.91% 92.12% 95.59% 90.59% 90.58% 92.35% (0.1508) (0.1459) (2.3569) (1.4300) (5.7843) (3.4258) (1.3147) Maryland 55.19% 53.15% 53.50% 52.06% 52.31% 51.66% 52.35% (0.1587) (0.1492) (1.5607) (1.7106) (3.6457) (1.8132) (0.8972) Massachusetts 75.32% 75.18% 74.09% 77.60% 66.94% 69.92% 72.59% (0.1270) (0.1236) (1.1316) (1.3151) (2.6955) (2.0089) (0.9316) Michigan 78.98% 77.31% 72.64% 71.13% 73.86% 72.47% 72.37% (0.0988) (0.0993) (1.3965) (1.6169) (2.2393) (1.7039) (0.7801) Minnesota 83.89% 84.02% 80.95% 82.94% 76.05% 77.76% 79.68% (0.1187) (0.1221) (1.5765) (2.2846) (3.1697) (2.5590) (1.1057)

Table B-2. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 61.52% 58.46% 52.03% 51.36% 59.81% 55.14% 53.94% (0.2296) (0.2124) (3.1956) (3.1391) (6.3351) (3.9073) (2.2118) Missouri 82.81% 81.50% 78.46% 76.58% 78.19% 77.58% 77.68% (0.1184) (0.1167) (1.7696) (1.9025) (2.9871) (1.9258) (0.9490) Montana 89.31% 88.62% 82.75% 81.47% 82.63% 82.26% 82.25% (0.2344) (0.2519) (3.6305) (3.6756) (5.9827) (5.0012) (2.3847) Nebraska 82.66% 83.45% 79.95% 83.89% 78.16% 79.56% 80.55% (0.2112) (0.2057) (2.5000) (3.0174) (6.9125) (3.7294) (2.0576) Nevada 55.55% 53.39% 53.34% 50.68% 47.20% 44.87% 49.16% 252 (0.2257) (0.2241) (1.8197) (2.5319) (3.8365) (2.6392) (1.2187)

New Hampshire 92.09% 92.22% 90.64% 93.86% 89.09% 89.85% 91.00% (0.1745) (0.1691) (2.1425) (1.7428) (4.9340) (3.3291) (1.4996) New Jersey 58.97% 58.48% 55.95% 57.32% 52.27% 51.54% 54.30% (0.1274) (0.1202) (1.3721) (1.2900) (3.0861) (1.7290) (0.8353) New Mexico 42.07% 41.80% 41.83% 42.39% 37.07% 35.86% 39.32% (0.2800) (0.2640) (2.5029) (2.6979) (5.9980) (2.7486) (1.6064) New York 59.23% 57.39% 56.68% 57.57% 51.06% 51.10% 54.25% (0.0865) (0.0823) (0.8798) (0.9743) (2.0080) (1.1251) (0.5990) North Carolina 67.76% 66.30% 65.76% 64.16% 63.67% 63.03% 64.15% (0.1164) (0.1103) (1.2203) (1.3209) (3.1873) (2.3886) (1.0137) North Dakota 87.63% 88.03% 81.43% 82.68% 83.58% 82.31% 82.26% (0.2838) (0.2867) (5.1722) (6.5511) (8.5238) (5.2774) (3.0416) Ohio 82.50% 81.37% 77.48% 77.00% 77.10% 76.16% 76.86% (0.0852) (0.0841) (0.9876) (1.1576) (1.9522) (1.3638) (0.6425)

Table B-2. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 70.48% 70.76% 65.11% 64.66% 60.26% 60.48% 62.79% (0.1853) (0.1808) (2.2347) (2.6414) (4.0290) (2.7696) (1.3019) Oregon 79.44% 80.10% 77.73% 80.48% 72.45% 73.70% 76.36% (0.1520) (0.1479) (1.2850) (1.5040) (2.8781) (2.0844) (0.8458) Pennsylvania 80.65% 79.55% 74.16% 74.78% 75.60% 73.38% 74.24% (0.0880) (0.0827) (1.3561) (1.2792) (2.8537) (1.2713) (0.6882) Rhode Island 75.99% 76.91% 71.55% 73.67% 70.09% 68.01% 70.64% (0.3257) (0.3024) (3.6228) (3.4390) (7.4551) (4.6500) (2.2718) South Carolina 68.12% 65.84% 61.53% 60.05% 65.60% 63.11% 62.24% 253 (0.1676) (0.1557) (2.4265) (1.9700) (4.1744) (2.4149) (1.2288)

South Dakota 86.24% 86.23% 84.27% 83.73% 79.98% 78.54% 81.66% (0.2884) (0.2870) (4.6757) (5.1810) (7.6581) (4.9410) (2.3952) Tennessee 77.71% 76.44% 71.70% 72.82% 73.16% 73.86% 72.89% (0.1246) (0.1213) (1.4523) (2.0077) (3.2374) (1.6104) (0.7649) Texas 46.35% 46.03% 46.02% 47.00% 42.93% 43.71% 45.05% (0.0765) (0.0728) (0.7795) (0.8605) (1.8008) (1.1543) (0.5660) Utah 80.57% 81.13% 76.05% 79.88% 73.70% 74.88% 76.19% (0.1831) (0.1787) (2.2509) (2.0950) (4.0413) (2.4362) (1.2862) Vermont 94.51% 94.17% 92.23% 94.11% *** 91.08% 92.56% (0.2157) (0.2199) (3.2788) (1.9116) *** (4.4430) (1.9317) Virginia 65.73% 64.31% 62.92% 61.72% 62.13% 62.09% 62.24% (0.1284) (0.1213) (1.2732) (1.3584) (2.8429) (1.4554) (0.7903) Washington 72.95% 72.98% 71.60% 75.76% 65.82% 66.28% 70.01% (0.1280) (0.1253) (1.2580) (1.2413) (2.7066) (1.8336) (0.7920)

Table B-2. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 93.52% 93.66% 90.51% 91.91% 90.27% 91.64% 91.18% (0.1403) (0.1321) (1.9332) (1.7849) (3.8178) (2.5774) (1.0931) Wisconsin 85.37% 84.80% 80.99% 81.98% 78.29% 79.29% 80.29% (0.1103) (0.1142) (1.5706) (1.8065) (3.2634) (2.2725) (1.0129) Wyoming 86.42% 87.01% 82.71% 83.45% 79.22% 82.94% 82.39% (0.3544) (0.3320) (6.0463) (4.1308) (10.8788) (5.1751) (2.9704)

254

Table B-3. Percent Black by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 23.54% 26.80% 29.11% 31.43% 21.98% 26.48% 27.86% (0.1524) (0.1470) (2.1725) (1.8056) (3.8947) (2.5208) (1.0840) Alaska 3.34% 2.86% *** *** *** *** *** (0.1810) (0.1676) *** *** *** *** *** Arizona 4.19% 3.88% 4.82% 6.01% *** 3.74% 4.62% (0.0616) (0.0562) (0.7754) (0.7264) *** (0.8180) (0.3904) Arkansas 13.17% 15.10% 16.99% 18.62% *** 14.62% 15.98% (0.1532) (0.1547) (2.1659) (2.6508) *** (2.2553) (1.1876)

255 California 5.45% 5.78% 5.38% 7.18% 4.44% 5.04% 5.53% (0.0290) (0.0282) (0.2872) (0.3248) (0.4324) (0.3400) (0.1439)

Colorado 4.03% 3.65% 3.56% 4.10% *** *** 3.33% (0.0654) (0.0609) (0.6134) (0.7282) *** *** (0.4049) Connecticut 9.29% 10.36% 9.05% 9.26% *** 8.75% 8.89% (0.1194) (0.1179) (1.3448) (1.5763) *** (1.4649) (0.7484) Delaware 19.35% 21.49% *** 19.69% *** *** 17.83% (0.3198) (0.3097) *** (3.1427) *** *** (1.9117) District of Columbia 42.95% 48.07% 24.17% 40.75% *** 33.04% 31.00% (0.5018) (0.4546) (1.9329) (3.6198) *** (4.6416) (1.8305) Florida 13.87% 15.15% 11.94% 15.81% 12.65% 14.44% 13.72% (0.0601) (0.0583) (0.6595) (0.7413) (1.4560) (0.9633) (0.4212) Georgia 28.18% 31.86% 30.21% 39.89% 25.28% 30.18% 31.89% (0.1156) (0.1106) (1.3207) (1.5032) (3.7261) (1.9957) (1.0569) Hawaii 1.67% 1.15% *** *** *** *** *** (0.0862) (0.0686) *** *** *** *** ***

Table B-3. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 0.74% 0.40% *** *** *** *** *** (0.0569) (0.0399) *** *** *** *** *** Illinois 12.18% 14.44% 13.31% 19.26% 12.10% 14.34% 14.80% (0.0722) (0.0738) (0.8983) (1.2094) (1.7272) (1.0021) (0.6351) Indiana 7.88% 9.10% 10.76% 11.65% *** 8.77% 9.88% (0.0815) (0.0837) (1.1307) (1.2372) *** (1.2225) (0.6637) Iowa 3.01% 2.67% *** *** *** *** *** (0.0738) (0.0684) *** *** *** *** *** Kansas 5.34% 5.23% *** *** *** *** 6.66% 256 (0.1037) (0.1000) *** *** *** *** (1.0283)

Kentucky 7.16% 7.49% 9.21% 9.84% *** 7.67% 8.55% (0.0959) (0.0927) (1.5956) (1.3785) *** (1.6514) (0.7748) Louisiana 28.26% 31.89% 30.12% 35.08% 27.30% 30.76% 31.17% (0.1695) (0.1613) (1.8494) (1.9247) (4.1320) (2.6716) (1.1937) Maine 1.10% 0.75% *** *** *** *** *** (0.0692) (0.0568) *** *** *** *** *** Maryland 27.44% 30.62% 26.46% 34.37% 23.02% 26.64% 28.08% (0.1435) (0.1396) (1.4971) (1.8184) (3.6886) (2.0518) (1.0718) Massachusetts 6.78% 7.07% 6.15% 6.33% *** 5.37% 5.93% (0.0746) (0.0723) (0.8195) (0.7500) *** (0.8358) (0.4408) Michigan 11.78% 13.73% 15.79% 18.74% 11.81% 14.06% 15.38% (0.0782) (0.0819) (1.1039) (1.4069) (1.5441) (1.2803) (0.6277) Minnesota 5.03% 5.00% *** *** *** *** 5.64% (0.0729) (0.0725) *** *** *** *** (0.7490)

Table B-3. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 33.77% 37.30% 38.78% 44.43% 31.55% 37.39% 38.76% (0.2217) (0.2078) (3.0791) (3.1070) (5.2789) (3.9488) (1.9912) Missouri 9.83% 11.42% 11.71% 15.05% *** 11.26% 12.10% (0.0932) (0.0957) (1.4396) (1.5257) *** (1.3624) (0.7017) Montana 0.68% *** *** *** *** *** *** (0.0677) *** *** *** *** *** *** Nebraska 4.13% 4.16% *** *** *** *** *** (0.1108) (0.1107) *** *** *** *** *** Nevada 8.09% 8.55% 8.31% 11.70% *** 7.85% 8.62% 257 (0.1251) (0.1266) (1.1520) (1.7351) *** (1.7064) (0.7334)

New Hampshire 1.38% 0.97% *** *** *** *** *** (0.0752) (0.0628) *** *** *** *** *** New Jersey 11.94% 13.61% 11.73% 16.66% 9.78% 12.99% 13.02% (0.0842) (0.0846) (0.9271) (1.3704) (2.0947) (1.4514) (0.6155) New Mexico 2.31% 1.61% *** *** *** *** *** (0.0861) (0.0681) *** *** *** *** *** New York 13.88% 15.96% 12.77% 15.65% 11.01% 12.27% 13.01% (0.0606) (0.0608) (0.5537) (0.7336) (1.0983) (0.8134) (0.3941) North Carolina 19.31% 21.86% 21.26% 25.49% 17.61% 20.00% 21.43% (0.0973) (0.0964) (0.9116) (1.1986) (2.1068) (1.7675) (0.7847) North Dakota 2.83% 1.85% *** *** *** *** *** (0.1548) (0.1197) *** *** *** *** *** Ohio 10.57% 11.96% 12.90% 14.92% 10.65% 11.69% 12.69% (0.0691) (0.0702) (0.8021) (1.0940) (1.9702) (1.1460) (0.5322)

Table B-3. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 6.51% 6.92% 7.62% 8.19% *** *** 6.96% (0.1023) (0.1017) (1.5471) (1.7927) *** *** (0.7636) Oregon 1.87% 1.53% *** *** *** *** 1.70% (0.0516) (0.0458) *** *** *** *** (0.2741) Pennsylvania 9.25% 10.54% 12.24% 13.46% 8.54% 10.58% 11.49% (0.0640) (0.0637) (0.9662) (1.0319) (1.9504) (0.9815) (0.5612) Rhode Island 5.99% 5.51% *** *** *** *** *** (0.1820) (0.1639) *** *** *** *** *** South Carolina 24.06% 27.24% 29.17% 33.38% 22.00% 26.10% 28.20% 258 (0.1531) (0.1465) (2.1308) (1.9613) (3.4833) (2.4742) (1.1378)

South Dakota 1.78% 1.27% *** *** *** *** *** (0.1110) (0.0932) *** *** *** *** *** Tennessee 14.74% 16.53% 17.49% 19.62% 14.12% 15.31% 16.83% (0.1061) (0.1064) (1.1741) (1.6405) (2.2615) (1.6132) (0.7126) Texas 11.01% 12.34% 11.33% 14.58% 9.92% 11.02% 11.78% (0.0485) (0.0482) (0.6532) (0.6373) (1.2328) (0.7518) (0.4068) Utah 1.23% 0.89% *** *** *** *** *** (0.0520) (0.0430) *** *** *** *** *** Vermont 1.02% *** *** *** *** *** *** (0.0944) *** *** *** *** *** *** Virginia 17.49% 19.25% 18.55% 23.97% 15.47% 16.81% 18.86% (0.1035) (0.1008) (1.0664) (1.2104) (2.6953) (1.4973) (0.6755) Washington 3.74% 3.14% 3.97% 3.23% *** *** 3.25% (0.0542) (0.0497) (0.5234) (0.5894) *** *** (0.3202)

Table B-3. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 3.33% 3.03% *** *** *** *** *** (0.1028) (0.0920) *** *** *** *** *** Wisconsin 4.85% 5.74% 7.06% 7.80% *** *** 6.76% (0.0681) (0.0727) (1.3332) (1.2757) *** *** (0.6034) Wyoming 1.04% *** *** *** *** *** *** (0.1013) *** *** *** *** *** ***

259

Table B-4. Percent Other Race by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 4.45% 4.14% 5.32% *** *** 7.79% 6.41% (0.0737) (0.0674) (0.9171) *** *** (1.3035) (0.6297) Alaska 26.27% 29.89% 30.37% 29.37% *** 35.92% 33.29% (0.4367) (0.4592) (4.8298) (5.9764) *** (6.4204) (2.9056) Arizona 15.87% 16.30% 19.23% 18.01% 26.05% 26.03% 22.16% (0.1104) (0.1065) (1.1558) (1.0167) (2.5132) (1.7280) (0.7235) Arkansas 6.52% 6.00% 9.63% *** *** 11.99% 9.87% (0.1125) (0.1023) (1.7554) *** *** (1.8198) (1.0288)

260 California 32.45% 33.15% 32.27% 31.74% 43.43% 42.60% 37.13% (0.0602) (0.0584) (0.5529) (0.6761) (1.1857) (0.8005) (0.4482)

Colorado 9.96% 10.54% 12.92% 12.50% 17.17% 17.27% 14.86% (0.0984) (0.0998) (0.9547) (1.2002) (2.1760) (1.5541) (0.6124) Connecticut 11.43% 10.91% 14.23% 13.13% *** 21.69% 17.54% (0.1345) (0.1225) (1.9979) (1.6627) *** (2.6099) (1.3684) Delaware 8.15% 7.48% *** *** *** *** 9.77% (0.2212) (0.1989) *** *** *** *** (1.3833) District of Columbia 11.45% 11.01% 10.93% *** *** *** 15.12% (0.3291) (0.2947) (1.4291) *** *** *** (1.8488) Florida 7.50% 7.51% 10.02% 8.77% 14.85% 13.69% 11.63% (0.0458) (0.0423) (0.4966) (0.6361) (1.6455) (0.7920) (0.3641) Georgia 8.55% 8.06% 9.39% 6.93% 13.25% 12.43% 10.34% (0.0700) (0.0642) (0.7089) (0.7090) (2.0873) (1.1922) (0.5384) Hawaii 71.70% 74.62% 66.44% 73.01% 74.42% 76.98% 72.49% (0.2990) (0.2744) (2.8881) (2.8343) (5.4735) (3.1192) (1.4237)

Table B-4. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 7.62% 7.66% *** *** *** 15.27% 12.44% (0.1643) (0.1675) *** *** *** (3.6551) (1.7544) Illinois 13.21% 12.54% 14.17% 11.77% 20.69% 18.83% 16.09% (0.0751) (0.0692) (0.7902) (0.8306) (2.6023) (1.2315) (0.5854) Indiana 5.92% 5.57% 7.90% 6.18% 11.68% 10.61% 8.89% (0.0715) (0.0682) (0.9621) (0.8973) (2.1169) (1.4028) (0.5908) Iowa 5.23% 5.11% *** *** *** *** 8.13% (0.0956) (0.0971) *** *** *** *** (0.9430) Kansas 8.32% 8.06% 14.87% *** *** 14.56% 12.99% 261 (0.1270) (0.1235) (2.1746) *** *** (2.4395) (1.3321)

Kentucky 3.77% 3.63% 5.62% *** *** 7.22% 6.24% (0.0708) (0.0663) (1.1918) *** *** (1.4861) (0.5964) Louisiana 5.06% 4.59% 6.10% 4.12% *** 7.64% 6.56% (0.0825) (0.0724) (0.8397) (0.7867) *** (1.0669) (0.5205) Maine 3.13% 3.45% *** *** *** *** 5.65% (0.1214) (0.1196) *** *** *** *** (1.0679) Maryland 12.92% 12.36% 14.76% 10.68% 21.01% 18.37% 15.77% (0.1074) (0.0997) (1.0588) (1.1826) (2.9742) (2.0988) (0.8154) Massachusetts 12.56% 12.27% 13.62% 10.97% 21.20% 19.04% 15.81% (0.0986) (0.0934) (0.9385) (0.8713) (2.1802) (1.6844) (0.7372) Michigan 6.48% 6.37% 8.18% 7.45% 11.59% 10.85% 9.36% (0.0600) (0.0572) (0.7756) (0.7769) (2.1402) (0.8491) (0.4911) Minnesota 8.58% 8.63% 11.56% 8.31% 15.31% 14.29% 12.20% (0.0916) (0.0917) (1.4148) (1.2230) (2.6134) (1.8064) (0.7551)

Table B-4. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 3.13% 3.03% *** *** *** *** 5.51% (0.0816) (0.0768) *** *** *** *** (0.9101) Missouri 5.14% 5.06% 6.19% 5.87% *** 9.19% 7.62% (0.0690) (0.0678) (0.9263) (1.2059) *** (1.3246) (0.6976) Montana 7.99% 9.02% *** *** *** *** 13.94% (0.2060) (0.2238) *** *** *** *** (1.9972) Nebraska 6.42% 6.37% *** *** *** *** 8.96% (0.1396) (0.1355) *** *** *** *** (1.3599) Nevada 22.05% 24.09% 25.30% 26.63% 33.67% 35.17% 30.00% 262 (0.1902) (0.1922) (1.6529) (2.0062) (3.0744) (2.3378) (1.0526)

New Hampshire 4.38% 4.66% *** *** *** *** 5.58% (0.1339) (0.1338) *** *** *** *** (1.2498) New Jersey 17.70% 17.04% 20.30% 16.11% 26.67% 24.63% 21.66% (0.1000) (0.0933) (1.3177) (1.1060) (2.6258) (1.9534) (0.8067) New Mexico 21.79% 21.98% 25.73% 27.10% 33.78% 32.41% 29.39% (0.2325) (0.2283) (2.5430) (2.4878) (4.9364) (3.4824) (1.3987) New York 19.29% 19.32% 21.65% 19.59% 30.80% 30.01% 25.23% (0.0700) (0.0654) (0.7002) (0.7720) (1.9264) (1.0270) (0.4728) North Carolina 8.30% 7.85% 8.56% 7.00% 14.97% 13.14% 10.57% (0.0704) (0.0640) (0.8022) (0.8393) (2.5553) (1.5452) (0.6661) North Dakota 7.48% 8.57% *** *** *** *** 11.82% (0.2270) (0.2513) *** *** *** *** (2.5350) Ohio 4.92% 4.83% 7.13% 5.94% 10.44% 10.29% 8.35% (0.0488) (0.0460) (0.6283) (0.6957) (1.4488) (0.8798) (0.3646)

Table B-4. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 17.53% 17.52% 21.01% 22.89% 29.42% 28.97% 25.28% (0.1551) (0.1510) (1.9056) (2.2417) (3.7282) (2.9698) (1.2090) Oregon 11.66% 11.84% 13.79% 11.97% 19.78% 19.39% 16.02% (0.1201) (0.1225) (1.0686) (1.2159) (2.3724) (1.8283) (0.7836) Pennsylvania 6.78% 6.68% 8.88% 8.39% 12.41% 12.46% 10.44% (0.0551) (0.0519) (0.8857) (0.8858) (1.9463) (1.0518) (0.5034) Rhode Island 10.71% 10.42% *** *** *** *** 15.61% (0.2345) (0.2196) *** *** *** *** (1.8447) South Carolina 4.60% 4.40% 6.41% *** *** 8.39% 6.92% 263 (0.0753) (0.0682) (0.9736) *** *** (1.3812) (0.6006)

South Dakota 9.88% 10.99% *** *** *** *** 14.47% (0.2477) (0.2606) *** *** *** *** (2.0192) Tennessee 4.45% 4.42% 6.45% 5.00% *** 8.34% 7.16% (0.0625) (0.0597) (0.8426) (0.7984) *** (1.3143) (0.5652) Texas 13.05% 12.59% 15.49% 12.95% 20.57% 19.22% 16.84% (0.0514) (0.0493) (0.5388) (0.6704) (1.3927) (0.8852) (0.4121) Utah 11.19% 10.87% 14.95% 12.59% *** 17.73% 15.93% (0.1452) (0.1436) (1.9277) (1.5535) *** (2.3942) (1.1899) Vermont 3.33% 3.99% *** *** *** *** *** (0.1716) (0.1864) *** *** *** *** *** Virginia 11.44% 11.48% 11.72% 10.09% 17.21% 16.48% 13.67% (0.0873) (0.0816) (1.0599) (0.8123) (2.6310) (1.4941) (0.6880) Washington 17.68% 18.49% 19.06% 16.67% 26.44% 26.10% 21.86% (0.1103) (0.1088) (1.0418) (1.0291) (2.5938) (1.5241) (0.6889)

Table B-4. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 2.35% 2.36% *** *** *** *** 3.94% (0.0890) (0.0828) *** *** *** *** (0.7723) Wisconsin 6.40% 6.32% 8.05% 6.66% *** 11.37% 9.42% (0.0781) (0.0767) (1.1651) (1.0391) *** (1.7408) (0.7351) Wyoming 6.26% 6.26% *** *** *** *** 10.81% (0.2593) (0.2399) *** *** *** *** (2.9622)

264

Table B-5. Percent Hispanic by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 3.66% 2.86% 4.91% *** *** *** 4.21% (0.0684) (0.0571) (1.1360) *** *** *** (0.6060) Alaska 5.38% 5.84% *** *** *** *** *** (0.2268) (0.2364) *** *** *** *** *** Arizona 27.20% 26.80% 28.31% 27.72% 32.42% 33.36% 30.41% (0.1340) (0.1280) (1.2662) (1.4209) (3.3804) (1.7212) (0.7273) Arkansas 6.42% 5.43% *** *** *** *** 6.66% (0.1114) (0.0974) *** *** *** *** (0.9677)

265 California 35.59% 34.40% 34.78% 33.88% 44.98% 44.83% 39.33% (0.0619) (0.0579) (0.5825) (0.6167) (1.0707) (0.7184) (0.3571)

Colorado 18.65% 18.21% 20.84% 18.33% 21.26% 21.17% 20.42% (0.1301) (0.1267) (1.2641) (1.5677) (2.5467) (1.8628) (0.9266) Connecticut 13.79% 13.32% 19.43% 17.48% *** 22.39% 20.37% (0.1428) (0.1350) (1.8341) (2.0802) *** (2.7527) (1.4751) Delaware 7.98% 6.89% *** *** *** *** 7.80% (0.2218) (0.1939) *** *** *** *** (1.6938) District of Columbia 10.64% 8.73% 12.48% *** *** *** 12.94% (0.3141) (0.2681) (1.3242) *** *** *** (1.6739) Florida 24.19% 23.69% 24.61% 22.46% 25.15% 23.84% 23.95% (0.0739) (0.0686) (0.7087) (0.8306) (2.0270) (0.9830) (0.4561) Georgia 8.54% 7.08% 9.84% 7.16% 11.21% 9.07% 9.12% (0.0696) (0.0611) (0.6932) (0.8041) (1.7362) (1.0719) (0.5262) Hawaii 7.74% 7.77% 9.37% *** *** *** 10.84% (0.1802) (0.1797) (1.9767) *** *** *** (1.4732)

Table B-5. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 10.62% 9.75% *** *** *** *** 13.19% (0.1912) (0.1846) *** *** *** *** (1.8591) Illinois 15.62% 13.95% 17.42% 14.96% 21.37% 19.13% 17.99% (0.0804) (0.0722) (1.0388) (1.0178) (2.5176) (1.1365) (0.5678) Indiana 5.89% 5.11% 7.68% 4.82% *** 7.04% 6.85% (0.0708) (0.0651) (0.9677) (0.7884) *** (1.1951) (0.6029) Iowa 4.90% 4.27% *** *** *** *** 5.77% (0.0941) (0.0854) *** *** *** *** (0.7624) Kansas 9.94% 9.05% 14.79% *** *** 11.31% 11.64% 266 (0.1375) (0.1285) (2.3472) *** *** (2.4560) (1.1207)

Kentucky 3.11% 2.45% *** *** *** *** 3.75% (0.0643) (0.0556) *** *** *** *** (0.5231) Louisiana 5.05% 3.96% 6.69% *** *** *** 5.27% (0.0815) (0.0681) (0.8492) *** *** *** (0.5674) Maine 1.29% 1.21% *** *** *** *** *** (0.0790) (0.0745) *** *** *** *** *** Maryland 9.21% 7.68% 11.84% 6.64% 13.76% 11.72% 10.74% (0.0936) (0.0820) (1.1711) (0.8263) (1.8914) (1.6244) (0.8206) Massachusetts 9.92% 9.66% 13.16% 10.02% 17.60% 15.10% 13.64% (0.0896) (0.0841) (1.0729) (0.9615) (2.1564) (1.4418) (0.6294) Michigan 4.15% 3.86% 6.00% 5.03% *** 5.48% 5.53% (0.0482) (0.0461) (0.6966) (0.7824) *** (0.9331) (0.4403) Minnesota 4.38% 3.95% 5.69% *** *** *** 5.71% (0.0684) (0.0656) (1.1591) *** *** *** (0.7389)

Table B-5. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 2.68% 2.06% *** *** *** *** 4.03% (0.0785) (0.0616) *** *** *** *** (0.6993) Missouri 3.45% 3.10% 5.09% *** *** 4.40% 4.50% (0.0573) (0.0522) (0.8754) *** *** (1.0298) (0.4527) Montana 2.87% 2.97% *** *** *** *** *** (0.1270) (0.1350) *** *** *** *** *** Nebraska 8.94% 8.02% *** *** *** *** 8.65% (0.1592) (0.1500) *** *** *** *** (1.3104) Nevada 25.01% 24.19% 26.74% 23.74% 32.40% 33.25% 29.09% 267 (0.1963) (0.1945) (1.4804) (2.3509) (3.2429) (2.4133) (1.1623)

New Hampshire 2.90% 2.80% *** *** *** *** *** (0.1084) (0.1054) *** *** *** *** *** New Jersey 18.62% 17.52% 24.44% 19.82% 28.85% 26.69% 24.73% (0.1028) (0.0941) (1.2697) (1.2062) (3.3917) (1.8699) (0.8838) New Mexico 45.06% 44.97% 46.60% 45.10% 48.32% 50.68% 47.83% (0.2835) (0.2654) (2.5441) (2.8142) (6.9477) (2.8798) (1.7040) New York 17.44% 16.99% 23.41% 20.41% 29.04% 27.43% 24.85% (0.0669) (0.0628) (0.7290) (0.7508) (2.0325) (1.0747) (0.5088) North Carolina 7.82% 6.62% 8.11% 6.22% 10.61% 9.23% 8.34% (0.0680) (0.0585) (0.7641) (0.6966) (2.2881) (1.1831) (0.6768) North Dakota 2.97% 2.44% *** *** *** *** *** (0.1478) (0.1429) *** *** *** *** *** Ohio 3.13% 2.82% 4.19% 3.91% *** 4.22% 4.12% (0.0398) (0.0357) (0.5649) (0.5642) *** (0.6141) (0.3141)

Table B-5. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 8.89% 7.59% 11.24% 8.08% *** 10.15% 10.16% (0.1166) (0.1072) (1.3352) (1.5610) *** (2.1147) (1.0482) Oregon 10.82% 9.61% 11.54% 9.90% 13.54% 12.22% 11.62% (0.1176) (0.1101) (0.9084) (1.1549) (2.5235) (1.8079) (0.7148) Pennsylvania 5.84% 5.52% 9.85% 8.15% 9.55% 9.23% 9.18% (0.0516) (0.0476) (0.8612) (0.7441) (1.7246) (0.8800) (0.5513) Rhode Island 13.16% 12.25% *** *** *** *** 16.59% (0.2585) (0.2321) *** *** *** *** (1.7104) South Carolina 5.00% 3.90% 5.70% *** *** *** 5.29% 268 (0.0764) (0.0636) (0.8265) *** *** *** (0.4617)

South Dakota 3.15% 2.44% *** *** *** *** *** (0.1474) (0.1339) *** *** *** *** *** Tennessee 4.51% 3.67% 6.71% 3.77% *** 4.56% 5.19% (0.0621) (0.0532) (0.7483) (0.7565) *** (0.8127) (0.4510) Texas 36.38% 35.15% 37.07% 33.75% 40.08% 38.20% 37.09% (0.0735) (0.0689) (0.6732) (0.7326) (1.6416) (0.8927) (0.4270) Utah 12.47% 11.90% 15.49% 13.00% *** 15.80% 15.08% (0.1528) (0.1480) (1.8350) (1.8878) *** (1.8575) (0.9381) Vermont 1.39% 1.55% *** *** *** *** *** (0.1151) (0.1151) *** *** *** *** *** Virginia 8.32% 7.50% 11.23% 7.15% 11.29% 9.93% 9.82% (0.0753) (0.0675) (0.8404) (0.8900) (2.3098) (1.0629) (0.6092) Washington 10.32% 9.45% 11.96% 8.91% 13.28% 13.13% 11.82% (0.0881) (0.0827) (0.8915) (0.8321) (1.8664) (1.3401) (0.6396)

Table B-5. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 1.13% 1.20% *** *** *** *** *** (0.0599) (0.0591) *** *** *** *** *** Wisconsin 5.42% 4.94% 7.59% *** *** 7.11% 7.23% (0.0714) (0.0679) (1.1506) *** *** (1.1225) (0.6561) Wyoming 8.58% 7.87% *** *** *** *** *** (0.2899) (0.2677) *** *** *** *** ***

269

Table B-6. Percent Married by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 54.74% 49.35% 25.18% 29.98% 19.72% 22.78% 24.85% (0.1775) (0.1662) (1.7064) (1.6142) (3.3660) (2.3263) (1.1932) Alaska 51.63% 52.96% 31.00% *** *** *** 26.62% (0.5015) (0.5108) (5.7362) *** *** *** (3.1762) Arizona 53.18% 49.97% 25.47% 31.70% 17.06% 21.19% 24.30% (0.1497) (0.1432) (0.9927) (1.1822) (1.8599) (1.4978) (0.7271) Arkansas 57.08% 52.03% 24.76% 28.46% 20.85% 22.25% 24.26% (0.2245) (0.2145) (2.2257) (2.3815) (4.4817) (2.8376) (1.2728)

270 California 52.67% 49.35% 30.05% 33.00% 17.70% 20.55% 25.82% (0.0654) (0.0620) (0.5085) (0.6537) (0.8807) (0.5475) (0.3319)

Colorado 55.63% 54.29% 26.86% 35.09% 19.03% 23.24% 26.49% (0.1631) (0.1622) (1.1355) (1.5455) (2.1399) (1.5995) (0.8094) Connecticut 55.01% 50.07% 32.55% 37.43% 18.95% 19.90% 27.65% (0.2018) (0.1927) (1.7015) (1.9927) (4.1370) (2.1582) (1.0731) Delaware 54.95% 49.34% 33.88% 38.07% *** *** 28.68% (0.4079) (0.3781) (3.2615) (3.5089) *** *** (1.8963) District of Columbia 33.96% 28.56% 32.14% 25.74% 12.78% 12.04% 21.59% (0.4718) (0.4027) (1.7615) (2.4009) (4.8069) (3.3874) (1.3945) Florida 52.70% 47.17% 31.60% 31.96% 19.06% 20.90% 26.39% (0.0860) (0.0799) (0.7732) (0.6845) (1.6987) (0.9631) (0.4586) Georgia 54.49% 48.57% 28.33% 30.03% 20.22% 21.80% 25.46% (0.1259) (0.1171) (1.0949) (1.1099) (2.0820) (1.3962) (0.7400) Hawaii 54.36% 53.36% 30.21% 38.25% *** 23.91% 28.46% (0.3265) (0.3155) (2.3547) (2.7936) *** (3.3658) (1.3802)

Table B-6. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 60.13% 58.66% 24.31% 33.08% *** 26.29% 26.44% (0.3030) (0.2999) (2.9381) (3.3424) *** (3.5755) (1.9039) Illinois 54.40% 49.99% 28.57% 30.22% 17.79% 20.69% 24.83% (0.1069) (0.1045) (0.9085) (1.0457) (1.7058) (1.3315) (0.6748) Indiana 55.78% 52.25% 29.69% 34.33% 19.11% 22.24% 26.91% (0.1493) (0.1443) (1.4569) (1.7234) (2.2554) (1.6259) (0.8418) Iowa 58.99% 56.61% 33.83% 37.85% 20.56% 22.21% 29.09% (0.2122) (0.2098) (2.4711) (2.7033) (3.9521) (2.0101) (1.2573) Kansas 58.20% 56.27% 30.04% 37.09% 20.24% 24.91% 28.64% 271 (0.2240) (0.2227) (2.6488) (2.5985) (3.8739) (3.1321) (1.4824)

Kentucky 56.20% 52.30% 25.84% 35.68% 21.29% 23.51% 26.93% (0.1812) (0.1746) (1.6335) (1.7905) (3.1257) (1.9290) (1.0402) Louisiana 50.23% 45.21% 22.87% 25.26% 17.54% 18.57% 21.29% (0.1870) (0.1754) (1.5185) (1.7060) (3.2903) (2.0361) (0.9369) Maine 56.61% 53.43% 33.97% 44.17% *** 22.73% 31.32% (0.3330) (0.3195) (3.0594) (2.9047) *** (3.9216) (2.0684) Maryland 54.48% 48.60% 31.31% 34.62% 19.83% 21.41% 27.27% (0.1596) (0.1510) (1.2599) (1.5090) (3.2783) (1.8809) (1.0014) Massachusetts 53.82% 48.92% 32.61% 41.95% 16.81% 19.75% 28.47% (0.1481) (0.1418) (1.1378) (1.2719) (2.7094) (1.7552) (0.8217) Michigan 53.95% 50.69% 26.62% 31.48% 17.74% 20.18% 24.43% (0.1206) (0.1170) (1.3248) (1.2995) (1.9030) (1.2543) (0.6630) Minnesota 57.63% 55.88% 30.86% 41.79% 20.05% 22.57% 29.27% (0.1615) (0.1607) (1.6896) (1.8077) (2.8470) (1.7883) (0.9616)

Table B-6. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 51.99% 46.13% 23.20% 27.60% *** 21.81% 23.23% (0.2325) (0.2119) (2.0451) (2.3354) *** (2.4795) (1.1987) Missouri 56.26% 52.06% 26.23% 33.74% 19.28% 22.49% 25.83% (0.1552) (0.1504) (1.3336) (1.6859) (2.5334) (1.7316) (1.0085) Montana 56.18% 56.03% 22.34% 32.68% *** 22.98% 24.37% (0.3827) (0.3821) (4.3551) (4.1805) *** (4.9495) (2.1694) Nebraska 58.94% 57.00% 26.51% 36.04% *** 23.79% 27.04% (0.2757) (0.2708) (2.8578) (3.5471) *** (3.7947) (1.6947) Nevada 50.10% 48.44% 26.95% 26.61% 17.56% 20.65% 23.46% 272 (0.2299) (0.2245) (1.6086) (2.1499) (2.9184) (2.1993) (1.0730)

New Hampshire 57.85% 55.31% 33.86% 49.14% *** 23.45% 32.54% (0.3223) (0.3121) (3.0911) (3.1342) *** (3.3391) (1.5877) New Jersey 56.84% 50.81% 33.83% 35.30% 20.48% 21.52% 28.23% (0.1282) (0.1218) (1.1192) (1.2692) (2.3997) (1.3660) (0.6731) New Mexico 50.60% 47.03% 22.96% 31.54% *** 20.39% 23.25% (0.2866) (0.2672) (2.1918) (2.2352) *** (2.4883) (1.2904) New York 52.14% 46.11% 32.60% 32.35% 18.42% 19.15% 26.12% (0.0887) (0.0822) (0.7414) (0.8633) (1.2776) (0.8738) (0.3998) North Carolina 56.05% 50.54% 27.31% 32.03% 20.48% 22.04% 25.82% (0.1230) (0.1145) (1.0406) (1.0562) (2.4539) (1.2653) (0.6712) North Dakota 56.41% 58.99% *** *** *** *** 24.56% (0.4270) (0.4283) *** *** *** *** (3.3579) Ohio 54.35% 50.44% 23.60% 28.26% 18.75% 19.89% 22.84% (0.1109) (0.1073) (0.9737) (1.1732) (1.9983) (1.2355) (0.6454)

Table B-6. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 56.15% 52.57% 30.15% 37.80% 20.95% 24.02% 28.71% (0.2015) (0.1957) (1.8001) (2.1811) (2.9846) (2.1127) (1.0635) Oregon 55.04% 52.56% 30.13% 36.62% 18.28% 20.78% 27.03% (0.1874) (0.1883) (1.3529) (1.5810) (2.5591) (1.8712) (0.8674) Pennsylvania 55.24% 50.76% 25.82% 31.67% 18.72% 20.82% 24.61% (0.1080) (0.1033) (1.0240) (1.1430) (2.0695) (1.4232) (0.6785) Rhode Island 51.99% 46.52% 31.39% 36.05% *** 17.59% 25.96% (0.3781) (0.3522) (2.9696) (3.3966) *** (4.1577) (2.0904) South Carolina 54.59% 48.77% 26.06% 28.34% 19.67% 20.65% 23.93% 273 (0.1764) (0.1637) (1.4898) (1.5498) (3.2019) (1.8244) (0.9456)

South Dakota 57.98% 56.92% *** 40.40% *** *** 28.09% (0.4104) (0.4067) *** (5.4762) *** *** (2.4822) Tennessee 55.84% 51.08% 25.49% 29.96% 20.15% 21.74% 24.59% (0.1489) (0.1428) (1.2410) (1.4136) (2.7570) (1.8197) (0.9277) Texas 55.77% 51.65% 28.51% 30.29% 20.89% 22.85% 25.98% (0.0766) (0.0731) (0.6548) (0.7904) (1.4776) (0.9413) (0.4819) Utah 61.86% 60.94% 34.86% 36.49% 24.26% 28.28% 31.56% (0.2261) (0.2299) (2.1708) (2.5855) (3.7446) (3.2874) (1.6552) Vermont 56.00% 52.70% 39.22% 57.33% *** *** 36.17% (0.4732) (0.4598) (5.3466) (4.7948) *** *** (2.7099) Virginia 56.96% 52.10% 30.17% 30.82% 21.11% 22.75% 26.62% (0.1335) (0.1267) (1.1468) (1.1595) (2.6740) (1.6905) (0.7267) Washington 56.10% 54.25% 33.84% 41.40% 19.40% 22.99% 30.01% (0.1419) (0.1377) (1.1097) (1.3803) (2.5380) (1.3272) (0.7041)

Table B-6. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 55.64% 52.37% 25.12% 33.64% *** 23.55% 25.94% (0.2812) (0.2703) (2.6726) (2.8219) *** (3.4752) (1.6917) Wisconsin 56.58% 54.61% 25.68% 31.76% 19.70% 21.92% 25.08% (0.1530) (0.1566) (1.4540) (1.7912) (2.6514) (1.8042) (1.0179) Wyoming 58.47% 58.48% *** 39.98% *** *** 29.07% (0.5020) (0.4869) *** (5.4697) *** *** (2.7909)

274

Table B-7. Percent with Children by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 30.88% 40.56% 8.19% 22.40% 13.85% 30.40% 20.00% (0.1631) (0.1647) (0.9692) (1.9134) (2.6848) (2.7949) (1.1086) Alaska 34.99% 44.47% *** *** *** 28.21% 18.68% (0.4792) (0.4947) *** *** *** (6.0832) (2.4262) Arizona 32.92% 40.67% 8.16% 21.57% 13.65% 27.43% 18.34% (0.1392) (0.1396) (0.6529) (1.0287) (2.0963) (1.3270) (0.5765) Arkansas 32.86% 40.53% 8.45% 21.96% *** 31.77% 20.53% (0.2113) (0.2092) (1.6090) (2.5840) *** (2.9262) (1.3365)

275 California 37.86% 45.58% 8.30% 20.53% 13.74% 26.98% 17.88% (0.0623) (0.0613) (0.2621) (0.5303) (0.7768) (0.6075) (0.2876)

Colorado 33.63% 39.95% 7.94% 20.01% 12.26% 24.49% 16.82% (0.1556) (0.1611) (0.7897) (1.1694) (2.0566) (1.8251) (0.8028) Connecticut 34.68% 41.54% 6.23% 21.59% *** 27.60% 18.15% (0.1923) (0.1905) (0.9098) (1.8181) *** (2.8070) (1.0029) Delaware 32.61% 39.64% *** 17.29% *** *** 17.09% (0.3759) (0.3727) *** (2.9792) *** *** (2.2006) District of Columbia 19.48% 29.63% *** 15.38% *** *** 10.41% (0.3934) (0.4089) *** (2.3633) *** *** (1.4092) Florida 29.88% 38.53% 6.79% 18.76% 12.25% 25.84% 16.52% (0.0783) (0.0784) (0.3326) (0.7450) (1.5049) (1.1979) (0.5105) Georgia 34.53% 43.97% 7.87% 23.62% 13.98% 30.48% 20.14% (0.1187) (0.1153) (0.6828) (1.1481) (2.1007) (1.5303) (0.6474) Hawaii 36.36% 44.79% 7.66% 22.37% *** 25.28% 17.86% (0.3166) (0.3188) (1.7241) (2.5070) *** (3.7832) (1.4964)

Table B-7. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 34.87% 40.09% *** 21.88% *** 27.64% 18.96% (0.2888) (0.2956) *** (3.7773) *** (4.0684) (1.9746) Illinois 35.09% 42.46% 8.52% 20.61% 13.30% 27.37% 18.12% (0.1022) (0.1017) (0.6358) (1.0976) (1.7121) (1.0774) (0.5162) Indiana 34.14% 41.13% 9.19% 23.24% 14.79% 30.58% 20.62% (0.1411) (0.1411) (1.0234) (1.4094) (2.4655) (1.7148) (0.8150) Iowa 33.30% 37.96% 8.86% 21.88% *** 28.04% 19.11% (0.2012) (0.2051) (1.4415) (2.2738) *** (2.7937) (1.1741) Kansas 34.49% 40.61% 12.72% 17.07% *** 28.53% 19.30% 276 (0.2147) (0.2176) (1.6183) (2.0327) *** (3.4600) (1.4735)

Kentucky 33.75% 41.16% 8.09% 24.03% 16.02% 31.22% 20.97% (0.1726) (0.1701) (1.1822) (1.7044) (3.2110) (2.2163) (0.9472) Louisiana 31.50% 41.57% 7.17% 22.35% 14.05% 29.80% 19.44% (0.1728) (0.1714) (1.0360) (1.8560) (3.0668) (2.1446) (0.9968) Maine 30.69% 34.60% *** 13.46% *** 24.52% 16.01% (0.3037) (0.3045) *** (2.2507) *** (3.8927) (1.6145) Maryland 35.72% 43.21% 11.04% 22.14% 13.79% 28.28% 20.02% (0.1523) (0.1491) (0.8996) (1.2087) (2.3230) (2.2571) (0.8931) Massachusetts 34.45% 40.96% 7.89% 23.55% 11.57% 25.02% 18.00% (0.1397) (0.1375) (0.6836) (1.0626) (2.4378) (1.8677) (0.6427) Michigan 32.78% 40.02% 8.00% 20.05% 13.83% 27.50% 18.22% (0.1123) (0.1149) (0.7081) (1.2312) (1.5869) (1.3631) (0.6494) Minnesota 34.32% 39.02% 8.74% 20.27% 14.32% 25.28% 17.79% (0.1540) (0.1578) (1.0129) (1.3726) (3.4152) (2.1148) (0.9814)

Table B-7. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 31.92% 43.41% 7.89% 27.39% *** 35.19% 23.01% (0.2166) (0.2117) (1.5513) (3.0730) *** (2.8255) (1.3639) Missouri 33.13% 39.75% 9.39% 21.45% 14.43% 28.27% 19.28% (0.1461) (0.1474) (0.9635) (1.8200) (2.4916) (2.0159) (0.8939) Montana 29.49% 34.90% *** *** *** 25.65% 17.73% (0.3419) (0.3562) *** *** *** (4.2245) (1.9274) Nebraska 34.81% 40.76% *** 22.21% *** 27.59% 18.72% (0.2653) (0.2706) *** (3.4665) *** (4.1782) (1.9659) Nevada 32.50% 41.83% 8.79% 20.43% 13.74% 27.93% 18.06% 277 (0.2139) (0.2168) (0.9800) (2.0581) (2.5957) (1.9529) (0.9059)

New Hampshire 33.59% 38.61% *** 20.82% *** 25.74% 17.88% (0.3024) (0.3054) *** (2.7055) *** (4.2074) (1.6442) New Jersey 38.86% 45.43% 10.14% 22.51% 14.62% 28.52% 19.72% (0.1258) (0.1221) (0.6804) (1.3254) (2.1373) (1.5403) (0.6367) New Mexico 31.47% 40.10% 6.80% 22.56% *** 29.01% 18.92% (0.2629) (0.2592) (1.3819) (2.0617) *** (3.1513) (1.3032) New York 34.50% 41.93% 8.71% 22.20% 12.92% 26.94% 18.42% (0.0834) (0.0815) (0.4491) (0.6874) (1.3796) (1.1486) (0.5276) North Carolina 33.06% 40.71% 9.03% 21.83% 14.75% 29.42% 19.95% (0.1160) (0.1130) (0.7267) (0.8931) (2.7460) (1.7515) (0.7381) North Dakota 31.28% 37.97% *** *** *** *** 18.21% (0.3933) (0.4177) *** *** *** *** (2.3724) Ohio 32.86% 40.06% 8.26% 22.55% 14.76% 30.54% 20.11% (0.1042) (0.1045) (0.6605) (1.3140) (1.5506) (1.2860) (0.5156)

Table B-7. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 34.09% 41.03% 10.06% 25.29% 16.56% 31.29% 21.76% (0.1928) (0.1922) (1.3578) (2.3767) (2.9205) (2.8893) (1.1578) Oregon 32.19% 37.52% 8.12% 18.89% 13.53% 24.60% 17.09% (0.1753) (0.1803) (0.9688) (1.4095) (2.2279) (2.2029) (0.9376) Pennsylvania 33.59% 39.94% 8.06% 22.11% 14.35% 27.71% 18.93% (0.1010) (0.1001) (0.6556) (0.9958) (1.9189) (1.5081) (0.6726) Rhode Island 33.04% 40.28% *** 21.94% *** 28.20% 19.54% (0.3534) (0.3483) *** (3.1775) *** (4.1529) (1.6691) South Carolina 30.58% 39.97% 9.52% 19.58% 13.85% 28.56% 18.99% 278 (0.1618) (0.1601) (1.1899) (1.9114) (2.9139) (2.4716) (1.0627)

South Dakota 32.75% 38.19% *** *** *** 29.49% 17.70% (0.3856) (0.4037) *** *** *** (5.7053) (2.9556) Tennessee 32.69% 40.67% 8.46% 22.71% 14.39% 30.17% 20.03% (0.1394) (0.1409) (0.9015) (1.3683) (2.2423) (2.4067) (0.9325) Texas 37.56% 46.68% 9.47% 24.11% 14.79% 30.50% 20.58% (0.0746) (0.0728) (0.4713) (0.6911) (1.2963) (0.9915) (0.4312) Utah 42.22% 48.53% 12.56% 22.83% 16.44% 28.85% 20.83% (0.2277) (0.2269) (1.3800) (2.3472) (3.1313) (2.4240) (1.2386) Vermont 29.86% 34.24% *** *** *** *** 18.94% (0.4295) (0.4366) *** *** *** *** (2.2580) Virginia 34.80% 41.72% 8.11% 20.02% 14.31% 26.47% 17.99% (0.1284) (0.1264) (0.7428) (1.1970) (2.4789) (1.8750) (0.8313) Washington 34.20% 40.43% 7.08% 18.41% 13.69% 25.61% 16.81% (0.1336) (0.1352) (0.6786) (1.0481) (1.6190) (1.4028) (0.6165)

Table B-7. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 31.03% 37.35% 8.37% 19.23% *** 28.87% 18.80% (0.2605) (0.2634) (2.2586) (2.7456) *** (4.3355) (1.9340) Wisconsin 32.83% 38.35% 9.38% 17.16% 15.08% 26.61% 17.79% (0.1448) (0.1513) (0.9256) (1.4919) (2.9511) (2.4519) (1.1164) Wyoming 31.88% 37.59% *** *** *** *** 16.79% (0.4638) (0.4830) *** *** *** *** (2.8865)

279

Table B-8. Percent Having at Least a BA by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 22.39% 23.62% 22.50% 26.98% 15.48% 17.94% 21.14% (0.1474) (0.1409) (1.4994) (2.0858) (3.1077) (1.8874) (1.0432) Alaska 24.08% 29.90% *** 41.11% *** *** 26.69% (0.4238) (0.4628) *** (6.2037) *** *** (2.9758) Arizona 26.80% 26.43% 28.63% 31.07% 17.78% 20.40% 24.95% (0.1322) (0.1265) (1.2929) (1.5250) (2.1520) (1.2560) (0.7756) Arkansas 20.01% 22.11% 20.81% 26.66% *** 17.55% 20.10% (0.1807) (0.1786) (2.3884) (2.8646) *** (2.2617) (1.5052)

280 California 30.48% 30.98% 37.86% 38.39% 22.09% 24.54% 31.37% (0.0593) (0.0558) (0.5180) (0.5652) (1.1136) (0.5489) (0.3260)

Colorado 36.84% 38.41% 39.27% 46.17% 26.98% 33.46% 37.23% (0.1606) (0.1561) (1.7519) (1.5592) (2.8876) (1.8128) (0.9800) Connecticut 36.56% 37.64% 37.39% 41.71% 26.01% 29.66% 34.30% (0.1952) (0.1877) (1.8900) (2.0536) (3.3973) (2.6523) (1.2626) Delaware 28.64% 30.34% 36.34% 42.42% *** 26.46% 32.62% (0.3618) (0.3503) (3.1971) (3.8847) *** (4.8012) (2.2684) District of Columbia 55.94% 55.44% 73.42% 65.12% 57.97% 60.87% 65.61% (0.5084) (0.4507) (1.9458) (3.6280) (8.2336) (5.0471) (1.9386) Florida 27.45% 27.37% 32.21% 32.63% 19.46% 23.07% 27.51% (0.0766) (0.0725) (0.6616) (0.9770) (1.7962) (1.1027) (0.5506) Georgia 27.64% 29.31% 32.73% 35.23% 20.18% 23.69% 28.62% (0.1121) (0.1055) (1.1830) (1.2105) (1.9925) (1.3603) (0.6187) Hawaii 28.88% 32.45% 32.84% 33.45% *** 25.74% 28.71% (0.2994) (0.2961) (2.6549) (3.2109) *** (3.1941) (1.5575)

Table B-8. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 25.04% 24.06% 20.53% 25.69% *** 17.49% 19.99% (0.2656) (0.2545) (2.7002) (3.4720) *** (2.7471) (1.7293) Illinois 31.50% 32.91% 38.56% 37.69% 24.60% 28.74% 33.18% (0.1016) (0.0967) (1.3306) (1.2760) (2.1649) (1.1975) (0.6918) Indiana 23.78% 25.19% 26.90% 28.88% 16.72% 20.12% 23.74% (0.1268) (0.1244) (1.4988) (1.3544) (2.6116) (1.3771) (0.7185) Iowa 25.73% 28.14% 25.84% 31.26% *** 22.42% 24.90% (0.1884) (0.1874) (2.5050) (2.4617) *** (2.1573) (1.2544) Kansas 29.88% 31.50% 27.40% 36.14% 21.20% 24.37% 27.61% 281 (0.2070) (0.2039) (2.2245) (2.7944) (4.2635) (2.2469) (1.3118)

Kentucky 21.27% 23.40% 23.95% 26.17% 14.97% 18.08% 21.32% (0.1488) (0.1474) (1.5282) (1.8608) (3.1194) (1.8541) (0.9413) Louisiana 20.53% 23.44% 25.40% 26.25% 15.49% 19.08% 22.16% (0.1501) (0.1477) (1.5875) (1.8678) (3.1667) (2.0532) (1.0208) Maine 27.56% 31.26% 34.57% 44.84% *** 25.60% 32.49% (0.3006) (0.2991) (3.9735) (3.1391) *** (4.1516) (2.3872) Maryland 36.57% 37.96% 40.16% 45.42% 26.94% 31.68% 36.87% (0.1544) (0.1451) (1.5844) (1.5741) (3.4405) (1.9521) (0.9060) Massachusetts 40.33% 41.72% 48.23% 54.81% 32.09% 39.02% 44.58% (0.1463) (0.1387) (1.4857) (1.3652) (3.0392) (1.8175) (0.9392) Michigan 26.32% 27.30% 28.82% 31.80% 19.00% 21.48% 25.74% (0.1055) (0.1044) (1.2947) (1.5735) (1.9143) (1.3489) (0.7194) Minnesota 32.36% 34.56% 39.18% 43.97% 24.99% 30.05% 35.34% (0.1524) (0.1534) (2.1654) (1.8555) (3.0265) (2.2982) (1.1796)

Table B-8. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 18.45% 22.16% 19.66% 24.00% *** 17.75% 19.39% (0.1813) (0.1780) (2.1226) (2.3077) *** (3.3374) (1.4882) Missouri 25.98% 27.99% 30.82% 33.81% 18.44% 23.55% 27.42% (0.1357) (0.1351) (1.3684) (1.8183) (2.8170) (1.8474) (1.0335) Montana 27.15% 29.70% 24.99% 31.81% *** 22.66% 24.93% (0.3344) (0.3471) (3.7070) (4.2973) *** (4.4672) (2.1038) Nebraska 28.35% 30.92% 30.15% 35.68% *** 24.98% 28.80% (0.2504) (0.2539) (3.0895) (3.6272) *** (3.9713) (1.9399) Nevada 22.20% 22.69% 24.71% 23.82% 14.69% 16.24% 20.32% 282 (0.1884) (0.1877) (1.4919) (2.3800) (2.2699) (2.2374) (1.0587)

New Hampshire 33.55% 35.76% 33.37% 44.99% *** 29.77% 33.81% (0.3051) (0.3049) (3.6258) (3.3871) *** (5.0911) (2.3119) New Jersey 36.98% 37.27% 37.74% 40.92% 26.50% 30.83% 34.62% (0.1250) (0.1178) (1.2655) (1.4226) (2.9152) (1.6377) (0.7728) New Mexico 23.47% 25.48% 29.66% 35.11% *** 20.38% 26.11% (0.2403) (0.2369) (2.4171) (2.6383) *** (3.2858) (1.5845) New York 33.10% 35.41% 44.07% 45.34% 27.80% 33.66% 38.67% (0.0823) (0.0787) (0.7610) (0.8750) (1.7142) (1.2377) (0.4909) North Carolina 27.59% 29.81% 29.70% 37.05% 20.12% 25.04% 28.72% (0.1112) (0.1055) (1.2565) (1.2123) (2.4096) (1.3594) (0.6484) North Dakota 24.65% 30.56% *** *** *** *** 23.39% (0.3691) (0.4011) *** *** *** *** (3.1988) Ohio 25.86% 26.59% 28.45% 32.00% 18.78% 21.66% 25.74% (0.0977) (0.0947) (1.0732) (1.1552) (1.8839) (1.1925) (0.6794)

Table B-8. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 22.69% 24.34% 26.34% 27.46% 16.61% 16.55% 22.01% (0.1723) (0.1707) (2.1694) (2.0687) (3.8206) (2.1146) (1.4741) Oregon 30.10% 31.17% 33.88% 44.04% 21.62% 25.95% 32.14% (0.1733) (0.1711) (1.6074) (1.7870) (3.1119) (1.8518) (1.0171) Pennsylvania 29.00% 29.81% 32.71% 35.65% 23.61% 26.46% 30.10% (0.0973) (0.0944) (1.2097) (1.1716) (2.4509) (1.6186) (0.7094) Rhode Island 30.91% 31.99% 35.20% 41.23% *** 26.12% 32.22% (0.3494) (0.3292) (3.2567) (3.3894) *** (3.4707) (1.9127) South Carolina 25.31% 26.35% 25.02% 28.17% 18.49% 21.25% 23.64% 283 (0.1538) (0.1440) (1.8903) (1.8986) (3.0664) (2.1403) (1.0322)

South Dakota 25.51% 28.50% *** *** *** *** 24.28% (0.3618) (0.3752) *** *** *** *** (3.2637) Tennessee 24.43% 25.69% 26.52% 28.83% 18.63% 21.27% 24.25% (0.1287) (0.1236) (1.3106) (1.4999) (2.5555) (1.5250) (0.8182) Texas 26.55% 27.25% 30.43% 31.80% 19.08% 21.79% 26.30% (0.0677) (0.0647) (0.7944) (0.6927) (1.1060) (0.8567) (0.4283) Utah 30.88% 27.80% 27.23% 33.68% 18.39% 21.09% 25.42% (0.2128) (0.2027) (2.3630) (2.4088) (3.5478) (2.2296) (1.0994) Vermont 32.78% 37.77% 36.99% 51.97% *** 32.99% 39.00% (0.4440) (0.4464) (5.2071) (4.1330) *** (5.8368) (3.0887) Virginia 35.60% 36.23% 38.79% 42.10% 27.84% 32.20% 35.90% (0.1295) (0.1236) (1.2747) (1.4136) (2.5727) (2.0893) (0.8705) Washington 32.87% 32.96% 37.27% 41.50% 23.98% 26.94% 32.99% (0.1338) (0.1305) (1.2289) (1.4484) (2.7228) (1.2649) (0.7917)

Table B-8. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 18.56% 20.43% 21.53% 29.27% *** 17.62% 20.91% (0.2196) (0.2175) (2.8183) (2.5823) *** (3.1592) (1.6708) Wisconsin 26.59% 29.36% 28.65% 35.79% 20.31% 25.01% 28.02% (0.1364) (0.1417) (1.6127) (1.9005) (2.8415) (1.9569) (0.9534) Wyoming 23.67% 26.28% *** *** *** *** 21.96% (0.4268) (0.4368) *** *** *** *** (3.2343)

284

Table B-9. Percent Not in Labor Force by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 33.91% 46.21% 36.01% 38.92% 30.04% 35.50% 35.77% (0.1691) (0.1659) (2.0865) (2.0363) (4.4366) (2.7943) (1.4391) Alaska 25.15% 34.53% 27.91% *** *** 24.59% 26.40% (0.4300) (0.4704) (5.1646) *** *** (4.8558) (2.8789) Arizona 31.92% 44.50% 28.75% 32.82% 23.40% 29.96% 29.32% (0.1408) (0.1428) (1.5207) (1.4346) (2.0904) (1.5290) (0.8259) Arkansas 33.13% 44.84% 34.67% 37.25% 29.39% 33.94% 34.34% (0.2125) (0.2112) (2.8488) (2.6651) (4.0136) (2.7436) (1.6262)

285 California 26.48% 41.16% 25.11% 28.79% 22.56% 29.97% 27.16% (0.0558) (0.0597) (0.4033) (0.6353) (0.9290) (0.5542) (0.2736)

Colorado 23.30% 36.00% 21.79% 24.71% 18.25% 23.39% 22.50% (0.1387) (0.1533) (1.2967) (1.4543) (2.6388) (1.5950) (0.6867) Connecticut 25.02% 36.14% 25.74% 27.10% 21.64% 25.53% 25.46% (0.1763) (0.1874) (1.8265) (2.0662) (3.7929) (3.0185) (1.3122) Delaware 30.59% 39.65% 33.69% 35.08% *** 27.77% 30.96% (0.3713) (0.3693) (3.1795) (3.5740) *** (5.3281) (2.3339) District of Columbia 23.60% 29.78% 15.48% 19.60% *** *** 18.22% (0.4338) (0.4101) (1.5939) (2.8469) *** *** (1.3706) Florida 33.65% 44.62% 29.98% 33.47% 25.33% 29.68% 30.01% (0.0813) (0.0797) (0.7676) (0.8619) (1.6913) (1.1779) (0.5534) Georgia 27.83% 40.31% 26.80% 30.05% 23.65% 30.41% 28.40% (0.1116) (0.1137) (0.9718) (1.3415) (2.1378) (1.6569) (0.6610) Hawaii 28.86% 38.68% 27.38% 30.67% *** 27.56% 27.35% (0.2974) (0.3110) (2.3148) (3.6900) *** (3.1030) (1.6042)

Table B-9. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 28.61% 42.50% 29.60% 30.29% *** 28.01% 28.61% (0.2762) (0.2933) (3.1933) (3.2889) *** (3.3964) (1.8252) Illinois 26.03% 38.01% 24.58% 27.78% 21.16% 25.84% 25.24% (0.0947) (0.1009) (1.0606) (1.3564) (2.0258) (1.5343) (0.7057) Indiana 27.40% 39.52% 28.51% 29.21% 23.34% 28.01% 27.82% (0.1342) (0.1405) (1.9008) (1.5043) (2.5154) (1.7479) (0.9803) Iowa 24.92% 35.46% 26.07% 29.29% 20.72% 22.29% 24.85% (0.1866) (0.2024) (2.4638) (2.8887) (3.7738) (2.7159) (1.5128) Kansas 24.88% 37.32% 26.24% 30.00% 20.88% 25.23% 26.06% 286 (0.1961) (0.2129) (2.6500) (2.8934) (3.5070) (2.8178) (1.4507)

Kentucky 32.83% 44.08% 32.58% 35.25% 27.36% 32.56% 32.55% (0.1728) (0.1717) (2.3157) (1.8115) (3.6671) (2.1940) (1.1251) Louisiana 31.60% 42.68% 32.29% 34.29% 27.35% 32.21% 32.09% (0.1748) (0.1725) (1.7066) (2.1726) (4.6998) (2.4559) (1.2888) Maine 31.48% 39.47% 30.47% 29.49% *** 28.11% 28.68% (0.3071) (0.3117) (3.1277) (3.0237) *** (4.4312) (1.6687) Maryland 24.06% 34.54% 22.70% 24.77% 20.21% 24.57% 23.52% (0.1364) (0.1414) (1.2411) (1.2655) (2.9075) (1.9734) (0.8069) Massachusetts 25.24% 35.28% 23.72% 25.68% 21.67% 24.97% 24.34% (0.1305) (0.1335) (1.2104) (1.1450) (3.2466) (1.9025) (0.8966) Michigan 31.20% 41.79% 31.33% 32.55% 24.95% 28.95% 29.95% (0.1115) (0.1156) (1.4160) (1.4447) (2.1512) (1.5252) (0.7030) Minnesota 23.65% 32.96% 22.79% 24.90% 18.36% 21.72% 22.32% (0.1385) (0.1532) (1.7198) (1.8254) (2.8443) (2.4804) (1.2086)

Table B-9. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 33.91% 44.79% 38.31% 39.24% 29.74% 36.46% 36.79% (0.2209) (0.2123) (2.5247) (2.6039) (4.9551) (3.4472) (1.4928) Missouri 29.09% 39.68% 28.70% 31.15% 22.29% 28.24% 28.26% (0.1425) (0.1458) (1.8552) (1.8257) (2.9093) (1.9870) (1.1723) Montana 29.93% 39.55% 34.32% 31.96% *** 26.69% 29.48% (0.3433) (0.3671) (4.0522) (4.3336) *** (4.6463) (2.1140) Nebraska 22.31% 33.54% 23.56% 27.09% *** 22.76% 23.38% (0.2320) (0.2580) (3.1463) (3.4394) *** (3.3297) (1.7501) Nevada 28.39% 40.44% 25.64% 28.25% 21.09% 26.37% 25.73% 287 (0.2058) (0.2188) (1.6342) (2.2285) (3.1462) (2.0142) (0.9294)

New Hampshire 24.81% 34.78% 26.29% 30.28% *** 24.43% 25.82% (0.2773) (0.2996) (2.9628) (3.1697) *** (4.2873) (1.9800) New Jersey 24.56% 38.04% 24.21% 29.87% 21.78% 28.22% 26.52% (0.1114) (0.1192) (1.0595) (1.4435) (2.4010) (1.7308) (0.7998) New Mexico 34.38% 45.30% 35.61% 34.73% 27.81% 33.36% 33.55% (0.2718) (0.2674) (2.7687) (2.5821) (4.1271) (3.3803) (1.5506) New York 28.34% 39.65% 24.15% 29.79% 25.31% 29.72% 27.44% (0.0787) (0.0808) (0.6899) (0.8805) (1.3482) (1.0699) (0.5115) North Carolina 29.44% 41.15% 29.68% 31.61% 23.44% 28.83% 29.07% (0.1124) (0.1126) (1.2434) (1.2827) (2.2952) (1.3358) (0.6677) North Dakota 22.10% 33.25% 25.62% *** *** *** 22.19% (0.3554) (0.4073) (4.6829) *** *** *** (3.1796) Ohio 28.75% 39.71% 28.42% 29.64% 23.28% 27.37% 27.67% (0.1018) (0.1041) (1.2979) (1.1657) (1.9024) (1.2223) (0.6150)

Table B-9. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 29.28% 42.92% 28.93% 31.17% 23.88% 31.68% 29.72% (0.1843) (0.1941) (1.8669) (2.3764) (3.9587) (2.4755) (1.2355) Oregon 30.33% 41.34% 28.61% 31.04% 23.68% 27.64% 28.23% (0.1751) (0.1810) (1.7421) (1.5977) (2.8378) (1.7989) (0.8821) Pennsylvania 29.00% 40.25% 30.36% 33.93% 25.35% 28.99% 30.09% (0.0969) (0.1004) (1.1349) (1.0967) (2.3980) (1.6378) (0.7221) Rhode Island 27.30% 38.02% 23.93% 27.97% *** 26.77% 25.84% (0.3363) (0.3437) (3.1327) (3.4834) *** (4.0708) (1.9429) South Carolina 31.67% 42.73% 32.11% 34.08% 26.07% 29.89% 31.04% 288 (0.1669) (0.1630) (2.3182) (1.8149) (3.6020) (2.4905) (1.4018)

South Dakota 23.92% 33.84% 30.04% 27.84% *** *** 25.53% (0.3525) (0.3919) (5.1250) (5.4336) *** *** (2.8947) Tennessee 30.41% 42.28% 29.16% 32.18% 24.55% 30.80% 29.84% (0.1383) (0.1399) (1.5423) (1.7118) (2.5876) (1.8367) (1.1075) Texas 24.21% 40.47% 23.29% 28.01% 20.99% 29.91% 26.27% (0.0658) (0.0717) (0.6504) (0.7874) (1.4928) (0.9509) (0.4801) Utah 20.96% 38.95% 22.59% 28.35% 18.34% 25.86% 24.38% (0.1877) (0.2241) (2.2849) (2.4992) (3.9084) (2.5807) (1.4773) Vermont 27.82% 35.40% 28.82% 23.59% *** *** 25.66% (0.4212) (0.4403) (4.6515) (4.5032) *** *** (2.3641) Virginia 25.90% 37.71% 25.19% 26.46% 21.55% 27.01% 25.63% (0.1177) (0.1226) (1.0334) (1.3872) (2.4509) (1.4884) (0.7173) Washington 27.26% 40.21% 26.33% 29.44% 22.18% 27.77% 26.96% (0.1272) (0.1361) (1.1193) (1.3215) (2.1697) (1.5978) (0.8098)

Table B-9. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 38.86% 49.86% 41.89% 42.84% 36.28% 41.53% 41.24% (0.2803) (0.2702) (3.9413) (3.0910) (6.7195) (3.8975) (2.2518) Wisconsin 26.60% 35.62% 26.83% 30.08% 19.93% 22.24% 25.14% (0.1366) (0.1465) (1.4601) (2.0585) (3.1551) (1.9097) (0.8925) Wyoming 24.44% 37.60% *** *** *** *** 26.38% (0.4342) (0.4768) *** *** *** *** (3.0876)

289

Table B-10. Percent Unemployed by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 4.16% 3.41% 4.79% *** *** 8.18% 6.27% (0.0717) (0.0617) (1.0268) *** *** (1.4852) (0.6548) Alaska 6.07% 3.92% *** *** *** *** 6.69% (0.2359) (0.1931) *** *** *** *** (1.4279) Arizona 4.17% 3.31% 4.57% 3.79% 9.34% 7.78% 6.12% (0.0600) (0.0514) (0.5590) (0.5988) (1.4529) (0.8990) (0.3984) Arkansas 3.53% 2.79% *** *** *** *** 5.40% (0.0854) (0.0712) *** *** *** *** (0.7101)

290 California 4.64% 3.79% 4.70% 4.05% 9.69% 8.33% 6.46% (0.0270) (0.0242) (0.2260) (0.2777) (0.6933) (0.3558) (0.1721)

Colorado 3.47% 2.66% 3.93% *** *** 5.58% 4.64% (0.0609) (0.0532) (0.6762) *** *** (0.8216) (0.3973) Connecticut 4.73% 3.69% 5.43% *** *** 8.89% 7.09% (0.0870) (0.0747) (0.9766) *** *** (1.6981) (0.8226) Delaware 3.82% 3.24% *** *** *** *** *** (0.1582) (0.1367) *** *** *** *** *** District of Columbia 5.38% 4.84% *** *** *** *** 5.39% (0.2293) (0.1949) *** *** *** *** (0.8434) Florida 3.95% 3.36% 4.26% 3.48% 8.64% 7.46% 5.75% (0.0340) (0.0293) (0.3441) (0.3487) (1.1464) (0.7173) (0.3090) Georgia 4.10% 3.76% 4.60% 3.81% 8.58% 8.20% 6.15% (0.0500) (0.0447) (0.5599) (0.5239) (1.6795) (0.9140) (0.4273) Hawaii 3.00% 2.26% *** *** *** *** 4.32% (0.1133) (0.0976) *** *** *** *** (0.8015)

Table B-10. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 3.42% 2.32% *** *** *** *** 5.35% (0.1137) (0.0996) *** *** *** *** (1.3280) Illinois 4.76% 3.72% 4.99% 4.46% 10.38% 8.19% 6.68% (0.0468) (0.0395) (0.4919) (0.5504) (1.4599) (0.7549) (0.3590) Indiana 3.69% 3.06% 4.07% *** 8.08% 7.12% 5.30% (0.0572) (0.0514) (0.8365) *** (1.8009) (1.1320) (0.4741) Iowa 2.79% 2.11% *** *** *** *** 4.40% (0.0713) (0.0657) *** *** *** *** (0.7914) Kansas 3.15% 2.52% *** *** *** *** 4.66% 291 (0.0798) (0.0727) *** *** *** *** (0.8636)

Kentucky 3.96% 2.98% *** *** *** 8.12% 6.02% (0.0731) (0.0594) *** *** *** (1.3440) (0.6502) Louisiana 4.77% 3.46% 5.81% *** *** 7.66% 6.69% (0.0827) (0.0656) (1.1386) *** *** (1.4908) (0.8098) Maine 3.41% 2.13% *** *** *** *** *** (0.1222) (0.0945) *** *** *** *** *** Maryland 4.03% 3.46% 4.70% 3.72% *** 7.96% 6.05% (0.0630) (0.0561) (0.6786) (0.6560) *** (1.2655) (0.4801) Massachusetts 3.94% 2.96% 4.29% 2.56% 8.94% 6.60% 5.26% (0.0579) (0.0475) (0.5216) (0.4708) (2.0436) (0.9651) (0.3852) Michigan 4.48% 3.30% 4.99% 3.93% 9.82% 7.98% 6.39% (0.0509) (0.0422) (0.6631) (0.6322) (1.6222) (0.8285) (0.4390) Minnesota 3.16% 2.11% *** *** *** *** 3.81% (0.0575) (0.0479) *** *** *** *** (0.5042)

Table B-10. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 4.86% 4.35% *** *** *** 9.88% 7.29% (0.1016) (0.0881) *** *** *** (1.9202) (0.9370) Missouri 3.66% 2.69% 4.83% *** *** 6.48% 5.37% (0.0594) (0.0501) (0.9231) *** *** (1.1669) (0.5966) Montana 3.06% 2.16% *** *** *** *** *** (0.1320) (0.1127) *** *** *** *** *** Nebraska 2.35% 2.06% *** *** *** *** *** (0.0844) (0.0822) *** *** *** *** *** Nevada 4.81% 3.96% 5.26% *** *** 8.20% 6.63% 292 (0.0990) (0.0881) (0.9262) *** *** (1.2195) (0.5578)

New Hampshire 3.18% 2.28% *** *** *** *** *** (0.1143) (0.0950) *** *** *** *** *** New Jersey 4.36% 3.50% 4.58% 3.68% 8.61% 7.44% 5.88% (0.0534) (0.0449) (0.4985) (0.5139) (1.6859) (0.9127) (0.3736) New Mexico 4.86% 3.32% *** *** *** *** 6.37% (0.1253) (0.1001) *** *** *** *** (0.9919) New York 4.25% 3.21% 4.54% 3.34% 9.49% 7.39% 5.89% (0.0355) (0.0297) (0.3171) (0.3725) (1.1391) (0.6512) (0.2740) North Carolina 3.98% 3.58% 4.88% 3.76% 9.75% 8.23% 6.37% (0.0497) (0.0437) (0.6044) (0.5093) (1.9280) (1.0579) (0.4234) North Dakota 1.87% 1.80% *** *** *** *** *** (0.1152) (0.1195) *** *** *** *** *** Ohio 4.01% 3.07% 4.64% 3.06% 9.17% 7.59% 5.86% (0.0452) (0.0378) (0.6543) (0.4905) (1.5149) (0.9265) (0.4538)

Table B-10. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 3.51% 2.83% *** *** *** 6.44% 5.24% (0.0761) (0.0656) *** *** *** (1.1618) (0.5488) Oregon 4.06% 3.04% 4.96% *** *** 6.98% 5.55% (0.0745) (0.0654) (0.6401) *** *** (1.1431) (0.4352) Pennsylvania 4.13% 2.99% 4.36% 3.37% 9.64% 7.27% 5.81% (0.0431) (0.0357) (0.5722) (0.5344) (1.5178) (0.8555) (0.4009) Rhode Island 4.60% 3.26% *** *** *** *** *** (0.1596) (0.1285) *** *** *** *** *** South Carolina 4.19% 3.46% 4.64% *** *** 8.41% 6.64% 293 (0.0719) (0.0623) (0.9736) *** *** (1.5510) (0.7177)

South Dakota 2.57% 2.04% *** *** *** *** *** (0.1324) (0.1169) *** *** *** *** *** Tennessee 3.79% 3.36% 4.13% 3.36% 8.86% 8.08% 5.91% (0.0581) (0.0519) (0.7096) (0.8076) (1.6339) (0.9253) (0.5320) Texas 3.79% 3.11% 4.10% 3.30% 7.65% 6.87% 5.33% (0.0292) (0.0262) (0.3143) (0.3852) (0.8028) (0.5975) (0.2543) Utah 2.83% 2.18% *** *** *** *** 4.16% (0.0773) (0.0675) *** *** *** *** (0.6532) Vermont 2.96% 2.10% *** *** *** *** *** (0.1617) (0.1353) *** *** *** *** *** Virginia 3.50% 2.90% 3.39% 3.29% *** 6.55% 4.96% (0.0511) (0.0430) (0.6607) (0.5676) *** (0.8113) (0.4093) Washington 3.73% 2.83% 3.86% 3.34% 8.97% 6.84% 5.43% (0.0538) (0.0465) (0.5157) (0.5523) (1.5422) (0.8500) (0.3402)

Table B-10. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 4.36% 2.78% *** *** *** *** 6.20% (0.1165) (0.0925) *** *** *** *** (1.2425) Wisconsin 3.04% 2.23% *** *** *** 5.47% 4.59% (0.0550) (0.0469) *** *** *** (1.0845) (0.5899) Wyoming 3.88% 2.47% *** *** *** *** *** (0.1960) (0.1590) *** *** *** *** ***

294

Table B-11. Percent Fulltime, Year-Round Worker by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 49.19% 33.90% 43.97% 39.41% 40.98% 31.51% 38.21% (0.1793) (0.1579) (2.6357) (2.3331) (3.8709) (2.6523) (1.5091) Alaska 47.55% 38.20% 41.60% 44.03% *** 36.89% 40.65% (0.4972) (0.4917) (6.3982) (5.3832) *** (7.4193) (3.7666) Arizona 48.78% 33.95% 49.05% 43.99% 42.55% 34.78% 42.28% (0.1499) (0.1363) (1.4679) (1.6362) (2.9752) (1.6311) (0.8251) Arkansas 50.10% 35.87% 45.53% 41.75% 42.64% 33.90% 40.32% (0.2247) (0.2080) (2.6763) (2.8308) (4.5538) (3.3070) (1.5613)

295 California 51.40% 33.74% 50.08% 43.96% 42.19% 31.82% 41.61% (0.0636) (0.0579) (0.6031) (0.7195) (1.3606) (0.6531) (0.3608)

Colorado 56.92% 39.16% 54.56% 50.62% 48.95% 39.79% 47.98% (0.1628) (0.1574) (1.4382) (1.7413) (3.7361) (1.7818) (1.0331) Connecticut 53.81% 37.32% 49.98% 46.25% 42.70% 33.95% 42.71% (0.2037) (0.1886) (2.3106) (1.8229) (4.6954) (3.2516) (1.6091) Delaware 50.71% 38.83% 45.20% 44.13% *** 37.44% 42.19% (0.4026) (0.3687) (3.3186) (3.8448) *** (5.1069) (2.4813) District of Columbia 56.98% 49.06% 68.87% 61.60% 55.07% 51.36% 59.77% (0.5091) (0.4539) (2.1572) (3.2214) (9.3242) (4.8234) (1.9412) Florida 47.49% 34.49% 47.98% 44.38% 43.36% 36.41% 42.70% (0.0858) (0.0769) (0.8463) (0.8615) (1.9075) (1.3896) (0.5911) Georgia 53.48% 37.95% 51.90% 48.18% 44.91% 35.59% 44.54% (0.1252) (0.1138) (1.2630) (1.4454) (2.7211) (2.2675) (0.8137) Hawaii 51.96% 40.44% 50.66% 47.68% 46.53% 39.31% 45.67% (0.3314) (0.3086) (3.0178) (3.3563) (6.4371) (3.8476) (1.6103)

Table B-11. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 52.24% 32.12% 46.76% 44.48% 42.75% 32.33% 40.94% (0.3114) (0.2815) (4.3047) (4.1867) (6.8203) (4.1915) (2.6913) Illinois 53.75% 37.45% 53.22% 46.75% 44.67% 36.73% 45.13% (0.1076) (0.0998) (1.1948) (1.2657) (2.5699) (1.3068) (0.5935) Indiana 54.14% 36.68% 49.43% 48.14% 44.45% 35.04% 43.60% (0.1479) (0.1399) (1.7525) (1.7535) (3.1573) (1.9338) (0.9746) Iowa 57.56% 41.24% 49.99% 45.51% 46.41% 38.89% 44.72% (0.2131) (0.2098) (2.6107) (2.9736) (4.6761) (3.4162) (1.6549) Kansas 57.34% 39.67% 53.64% 47.24% 47.45% 38.12% 46.12% 296 (0.2243) (0.2177) (2.9551) (3.1446) (5.4891) (3.4195) (1.8278)

Kentucky 49.09% 35.03% 46.19% 41.86% 40.07% 32.77% 39.75% (0.1841) (0.1660) (2.2548) (2.2839) (3.9567) (2.0305) (1.2924) Louisiana 49.82% 36.48% 45.49% 42.08% 40.66% 33.56% 39.97% (0.1863) (0.1671) (1.6305) (1.8199) (4.5538) (2.5835) (1.1594) Maine 48.86% 36.07% 50.36% 46.14% *** 34.79% 42.85% (0.3308) (0.3085) (3.7241) (3.3025) *** (4.7126) (2.2778) Maryland 56.92% 42.51% 56.26% 52.48% 48.47% 39.94% 48.76% (0.1593) (0.1474) (1.5648) (1.3136) (3.5002) (2.1305) (0.8847) Massachusetts 54.16% 37.99% 54.08% 49.51% 44.62% 37.37% 46.07% (0.1489) (0.1359) (1.3830) (1.2073) (3.1419) (1.8276) (0.9876) Michigan 48.62% 33.24% 45.16% 41.94% 39.90% 31.17% 38.99% (0.1211) (0.1103) (1.5525) (1.6664) (2.4434) (1.4176) (0.7429) Minnesota 56.37% 40.76% 54.74% 47.65% 47.38% 37.87% 46.42% (0.1636) (0.1610) (2.4747) (2.1507) (3.3496) (2.8676) (1.4403)

Table B-11. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 47.99% 35.32% 40.38% 40.75% 39.73% 31.63% 37.39% (0.2318) (0.2052) (2.7087) (2.8014) (5.9015) (3.3180) (1.7893) Missouri 52.18% 38.90% 49.37% 46.64% 44.08% 37.39% 43.99% (0.1578) (0.1470) (2.2902) (1.9956) (3.3471) (2.2536) (1.2185) Montana 49.03% 35.27% 39.78% 42.28% *** 36.81% 39.56% (0.3776) (0.3577) (4.4971) (4.0731) *** (4.7523) (2.4056) Nebraska 60.75% 42.58% 51.99% 50.65% 52.03% 40.72% 47.89% (0.2727) (0.2689) (3.9318) (3.8330) (7.2263) (3.8573) (2.1503) Nevada 50.29% 36.58% 51.51% 46.83% 44.85% 36.25% 44.66% 297 (0.2277) (0.2165) (2.1458) (2.8494) (3.8484) (2.5648) (1.2618)

New Hampshire 56.22% 39.46% 50.86% 48.52% 47.79% 39.93% 46.22% (0.3187) (0.3074) (3.2370) (3.4642) (6.8803) (4.9054) (2.0941) New Jersey 55.65% 37.71% 54.55% 47.39% 46.63% 36.26% 45.72% (0.1289) (0.1185) (1.3494) (1.6024) (2.7204) (1.4388) (0.7257) New Mexico 46.29% 32.83% 42.63% 42.84% 39.65% 31.93% 38.74% (0.2821) (0.2527) (2.3907) (2.8773) (4.7996) (2.9565) (1.5354) New York 51.75% 37.49% 54.22% 47.74% 42.65% 36.33% 45.20% (0.0876) (0.0797) (0.8147) (0.7923) (1.8550) (1.1302) (0.4929) North Carolina 51.81% 36.35% 46.68% 44.79% 44.17% 34.98% 41.89% (0.1235) (0.1111) (1.3481) (1.2596) (2.6762) (2.1024) (0.9679) North Dakota 59.78% 42.18% 52.58% *** *** 39.68% 47.48% (0.4308) (0.4276) (7.0106) *** *** (6.4325) (3.9747) Ohio 52.24% 36.74% 48.59% 47.18% 43.31% 35.43% 43.11% (0.1120) (0.1029) (1.3607) (1.3283) (2.1613) (1.3068) (0.7122)

Table B-11. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 53.72% 36.82% 49.34% 46.89% 46.03% 35.51% 43.66% (0.2028) (0.1897) (2.0821) (2.9988) (4.3328) (2.2102) (1.3362) Oregon 48.36% 32.87% 45.24% 42.26% 39.99% 32.14% 39.39% (0.1877) (0.1718) (1.5201) (1.4945) (3.0810) (1.8179) (0.9430) Pennsylvania 52.17% 36.87% 47.99% 44.61% 42.27% 35.13% 42.08% (0.1063) (0.0990) (1.1765) (1.2098) (2.7639) (1.5668) (0.6575) Rhode Island 51.37% 36.78% 51.80% 48.98% *** 34.66% 44.37% (0.3758) (0.3425) (3.1786) (3.5922) *** (5.5796) (2.3156) South Carolina 50.13% 36.03% 46.22% 43.11% 43.63% 35.23% 41.39% 298 (0.1779) (0.1585) (2.0118) (2.0625) (4.3140) (2.4718) (1.4397)

South Dakota 59.41% 42.86% 48.55% 50.28% *** 42.81% 47.26% (0.4057) (0.4070) (5.6501) (6.2424) *** (6.0124) (3.2768) Tennessee 51.60% 36.55% 49.70% 45.47% 43.58% 35.13% 42.94% (0.1489) (0.1364) (1.4853) (1.7069) (3.1715) (1.8800) (0.9850) Texas 57.32% 38.37% 55.60% 49.55% 48.96% 37.20% 47.17% (0.0760) (0.0709) (0.8751) (0.8921) (1.5177) (0.8939) (0.5198) Utah 58.88% 32.89% 53.18% 47.49% 46.33% 33.15% 44.30% (0.2283) (0.2149) (2.5373) (2.5182) (5.3041) (2.8711) (1.6280) Vermont 52.22% 39.09% 49.82% 52.92% *** *** 45.54% (0.4718) (0.4495) (4.9950) (5.1241) *** *** (3.2766) Virginia 55.68% 39.76% 53.03% 50.97% 47.88% 38.45% 46.93% (0.1339) (0.1236) (1.3514) (1.3611) (2.9952) (1.6168) (0.7911) Washington 52.80% 35.01% 50.63% 45.46% 43.32% 34.15% 42.93% (0.1410) (0.1324) (1.1869) (1.4344) (2.4250) (1.5962) (0.7451)

Table B-11. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 44.09% 31.27% 37.89% 38.36% 34.30% 27.60% 34.04% (0.2816) (0.2493) (3.3196) (3.2496) (6.6480) (4.0409) (1.9245) Wisconsin 54.94% 39.00% 49.44% 44.15% 47.11% 37.54% 43.89% (0.1554) (0.1507) (1.8593) (2.2324) (4.1756) (2.0990) (1.1526) Wyoming 56.00% 37.89% 47.15% 44.00% *** 33.97% 42.42% (0.4955) (0.4834) (6.2231) (6.3766) *** (6.8726) (3.8837)

299

Table B-12. Median Earnings for Fulltime, Year-Round Workers by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama $45,900 $34,900 $40,100 $36,100 $32,900 $28,500 $34,300 Alaska $60,400 $47,800 $60,500 $53,900 *** $36,500 $48,400 Arizona $46,100 $38,900 $41,300 $40,900 $32,700 $32,000 $37,100 Arkansas $41,000 $32,900 $35,400 $34,600 $30,900 $26,900 $31,600 California $51,200 $45,000 $54,100 $51,100 $37,100 $36,000 $45,200 Colorado $52,900 $43,600 $47,200 $48,900 $37,800 $36,300 $42,300 Connecticut $63,400 $52,400 $59,000 $55,400 $45,100 $40,600 $50,300 Delaware $51,500 $43,700 $50,900 $52,800 *** $35,400 $43,700

300 District of Columbia $73,600 $66,600 $97,200 $73,700 $63,100 $59,700 $74,300 Florida $41,900 $36,700 $41,900 $39,800 $32,100 $30,900 $36,300

Georgia $46,200 $38,300 $44,500 $41,300 $33,900 $31,700 $38,200 Hawaii $50,000 $41,900 $48,300 $44,900 $42,100 $36,300 $42,100 Idaho $43,800 $33,600 $38,400 $31,700 $31,900 $27,900 $32,000 Illinois $52,900 $42,300 $53,600 $46,400 $40,300 $35,500 $43,900 Indiana $49,200 $37,000 $41,400 $39,200 $35,900 $30,400 $36,600 Iowa $47,900 $38,100 $40,100 $40,900 $37,300 $31,400 $36,800 Kansas $48,000 $37,700 $40,900 $42,200 $34,100 $30,700 $37,000 Kentucky $45,000 $36,100 $40,800 $36,600 $33,500 $28,400 $35,200 Louisiana $50,000 $33,900 $44,100 $36,800 $34,000 $28,500 $36,000 Maine $47,100 $39,200 $41,500 $43,600 *** $31,900 $37,900 Maryland $61,500 $52,400 $57,900 $59,300 $43,900 $41,800 $51,100 Massachusetts $63,500 $52,900 $63,800 $61,900 $48,800 $43,600 $54,300 Michigan $51,200 $40,000 $44,800 $40,700 $37,500 $31,300 $38,600 Minnesota $52,900 $44,400 $53,200 $50,800 $41,000 $37,600 $45,900

Table B-12. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi $41,700 $31,900 $36,600 $31,900 $32,600 $28,400 $32,000 Missouri $46,100 $37,000 $42,500 $39,200 $34,700 $31,100 $36,700 Montana $44,600 $34,500 $36,900 $32,900 *** $28,200 $32,400 Nebraska $46,700 $37,600 $39,700 $40,200 $37,800 $30,700 $36,300 Nevada $44,000 $37,700 $43,500 $40,200 $34,000 $32,400 $38,000 New Hampshire $56,200 $44,300 $48,100 $48,400 $41,700 $36,900 $43,600 New Jersey $63,500 $51,200 $57,800 $54,300 $44,600 $39,700 $50,300 New Mexico $41,900 $34,700 $39,200 $43,600 $31,100 $27,800 $35,300 New York $53,300 $47,600 $59,400 $54,200 $40,600 $40,900 $50,200 301 North Carolina $44,500 $37,000 $41,200 $39,400 $34,100 $31,200 $36,500

North Dakota $51,100 $39,000 $42,000 *** *** $32,500 $38,900 Ohio $50,000 $39,800 $43,800 $41,700 $37,500 $31,500 $38,600 Oklahoma $44,400 $33,900 $41,100 $35,800 $33,600 $27,400 $34,300 Oregon $49,700 $40,500 $48,700 $46,000 $35,100 $33,300 $40,700 Pennsylvania $52,000 $41,900 $48,000 $45,600 $40,100 $34,800 $42,000 Rhode Island $52,900 $45,000 $51,800 $50,600 *** $36,200 $44,500 South Carolina $43,700 $35,000 $38,400 $36,100 $34,300 $29,200 $34,100 South Dakota $42,300 $35,000 $35,900 $33,200 *** $30,000 $32,700 Tennessee $42,300 $35,900 $38,700 $38,300 $32,300 $29,500 $34,700 Texas $47,100 $38,100 $44,700 $41,900 $34,000 $31,400 $38,500 Utah $52,300 $36,900 $42,800 $42,600 $35,100 $31,000 $37,900 Vermont $47,600 $41,000 $42,600 $46,200 *** *** $41,500 Virginia $56,400 $45,000 $55,700 $50,500 $42,700 $38,900 $47,300 Washington $59,200 $45,500 $55,000 $52,000 $42,600 $37,200 $47,200

Table B-12. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia $45,500 $33,500 $38,700 $36,900 $32,700 $28,900 $34,500 Wisconsin $50,000 $40,100 $43,400 $41,600 $39,500 $33,300 $39,300 Wyoming $52,500 $38,100 $42,900 $33,300 *** $30,900 $34,700

302

Table B-13. Percent Below Poverty Line by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 12.64% 17.31% 17.93% 22.86% 23.39% 33.84% 25.27% (0.1192) (0.1247) (2.0047) (1.5968) (3.4526) (2.1535) (0.9551) Alaska 9.15% 9.91% *** *** *** *** 15.80% (0.2921) (0.3115) *** *** *** *** (2.8938) Arizona 12.26% 14.73% 14.51% 16.70% 24.31% 28.64% 21.04% (0.0997) (0.1037) (1.1650) (1.0862) (2.5403) (1.4751) (0.7474) Arkansas 12.84% 17.26% 19.90% 21.20% 26.87% 35.79% 26.51% (0.1532) (0.1635) (2.4141) (2.4552) (4.7236) (2.6932) (1.5043)

303 California 10.70% 13.59% 11.50% 14.34% 19.46% 26.16% 18.08% (0.0394) (0.0426) (0.2915) (0.4879) (1.0366) (0.6069) (0.2899)

Colorado 8.55% 10.78% 11.14% 13.05% 19.46% 23.34% 16.78% (0.0926) (0.1040) (1.0335) (1.1608) (2.6196) (1.7526) (0.7229) Connecticut 7.09% 9.83% 11.08% 12.48% *** 21.78% 15.91% (0.1072) (0.1156) (1.4248) (1.4738) *** (2.1243) (1.0577) Delaware 8.93% 11.98% *** 17.35% *** *** 18.55% (0.2346) (0.2469) *** (2.6100) *** *** (2.1358) District of Columbia 12.23% 15.79% 8.19% 12.80% *** 23.80% 15.54% (0.3486) (0.3333) (1.5684) (2.6107) *** (4.2107) (1.6326) Florida 11.13% 14.22% 12.53% 15.79% 20.07% 26.01% 18.78% (0.0543) (0.0565) (0.5092) (0.6746) (1.4181) (1.0828) (0.4543) Georgia 11.40% 15.51% 13.55% 17.48% 21.40% 29.91% 21.09% (0.0809) (0.0861) (0.8801) (1.1633) (2.3405) (1.8512) (0.8300) Hawaii 7.80% 9.68% 9.20% 12.00% 16.33% 20.15% 14.45% (0.1787) (0.1906) (1.6265) (2.3055) (4.7155) (2.7136) (1.1426)

Table B-13. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 10.85% 13.79% 19.79% 17.13% 24.78% 29.94% 23.20% (0.1960) (0.2111) (3.4248) (3.4029) (6.2006) (3.3706) (2.1838) Illinois 9.45% 12.84% 11.46% 15.36% 19.73% 26.74% 18.54% (0.0632) (0.0705) (0.7310) (1.0367) (1.6581) (1.2852) (0.5635) Indiana 10.05% 13.84% 14.17% 16.72% 22.97% 30.96% 21.63% (0.0900) (0.1004) (1.3177) (1.8033) (2.5547) (1.7525) (0.7821) Iowa 8.70% 11.82% 15.22% 16.75% 21.88% 31.79% 22.02% (0.1253) (0.1400) (2.2072) (2.5498) (4.4959) (2.9290) (1.5395) Kansas 9.84% 12.33% 15.00% 16.45% 24.68% 28.52% 21.16% 304 (0.1358) (0.1545) (2.1741) (2.7854) (4.1162) (3.2620) (1.6727)

Kentucky 13.34% 17.63% 17.22% 19.42% 25.51% 35.90% 25.20% (0.1264) (0.1345) (2.0022) (2.0466) (3.3362) (2.5000) (1.0762) Louisiana 13.63% 19.15% 17.57% 22.56% 26.18% 35.66% 26.04% (0.1307) (0.1387) (1.5322) (1.7352) (4.8346) (2.6734) (1.3171) Maine 9.63% 12.67% 11.91% *** *** 30.66% 19.83% (0.2004) (0.2246) (2.4606) *** *** (5.2244) (2.4018) Maryland 6.96% 9.51% 9.30% 11.13% 14.74% 21.41% 14.57% (0.0827) (0.0897) (0.9576) (1.1363) (2.5407) (1.6671) (0.7860) Massachusetts 7.86% 10.84% 10.48% 11.31% 17.95% 22.86% 15.83% (0.0829) (0.0871) (0.9399) (0.9297) (2.4320) (1.4988) (0.6719) Michigan 11.19% 14.38% 16.29% 17.71% 24.17% 31.79% 22.89% (0.0780) (0.0830) (1.2267) (1.4787) (2.6679) (1.5793) (0.6943) Minnesota 7.70% 9.94% 10.00% 10.55% 19.06% 24.69% 16.24% (0.0892) (0.1016) (1.4196) (1.1944) (3.4015) (2.6420) (1.2264)

Table B-13. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 14.66% 20.53% 20.89% 24.13% 27.12% 37.38% 28.11% (0.1680) (0.1761) (2.3830) (2.6174) (5.3104) (3.8619) (1.8649) Missouri 10.29% 13.71% 14.74% 17.16% 21.70% 30.27% 21.47% (0.0972) (0.1061) (1.4127) (1.5788) (2.7114) (2.2299) (1.2076) Montana 11.25% 13.29% *** *** *** 29.61% 24.47% (0.2470) (0.2629) *** *** *** (4.6554) (2.3155) Nebraska 8.49% 12.00% *** *** *** 28.65% 19.55% (0.1565) (0.1825) *** *** *** (3.8889) (1.6609) Nevada 10.37% 13.16% 12.15% 16.19% 19.34% 25.76% 18.44% 305 (0.1414) (0.1548) (1.3930) (1.8953) (3.2599) (2.3231) (1.0657)

New Hampshire 5.52% 8.02% *** *** *** *** 12.78% (0.1473) (0.1746) *** *** *** *** (1.3815) New Jersey 7.21% 9.90% 9.50% 11.69% 14.93% 20.37% 14.37% (0.0672) (0.0745) (0.6591) (1.0060) (2.5065) (1.6794) (0.6834) New Mexico 15.40% 19.26% 18.88% 18.73% 27.17% 34.17% 25.04% (0.2026) (0.2181) (1.7983) (2.0948) (4.8006) (3.8039) (1.4938) New York 10.64% 14.03% 11.36% 14.32% 20.56% 26.63% 18.39% (0.0546) (0.0594) (0.5794) (0.7609) (1.5075) (1.0980) (0.4633) North Carolina 11.38% 15.02% 15.71% 17.43% 22.53% 30.86% 22.20% (0.0798) (0.0838) (1.0383) (1.1601) (2.3396) (1.6186) (0.6597) North Dakota 8.59% 11.81% *** *** *** *** 20.06% (0.2497) (0.2865) *** *** *** *** (3.0794) Ohio 10.43% 13.97% 15.31% 17.30% 22.61% 31.28% 22.14% (0.0683) (0.0755) (0.8468) (1.1664) (1.6386) (1.5042) (0.6353)

Table B-13. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 11.51% 15.54% 14.89% 17.12% 23.25% 32.13% 22.35% (0.1322) (0.1426) (1.5528) (1.6881) (4.2428) (2.3486) (1.0660) Oregon 10.91% 13.51% 15.61% 14.81% 25.14% 29.32% 21.29% (0.1196) (0.1295) (1.3660) (1.2820) (2.4901) (1.9691) (0.9536) Pennsylvania 9.38% 12.25% 14.30% 16.03% 21.65% 28.47% 20.47% (0.0645) (0.0685) (1.0107) (1.0426) (2.5816) (1.5877) (0.6795) Rhode Island 9.62% 12.84% *** *** *** 26.45% 18.33% (0.2251) (0.2423) *** *** *** (5.0149) (2.1337) South Carolina 11.38% 15.46% 16.25% 21.44% 23.35% 31.53% 23.63% 306 (0.1138) (0.1207) (1.5947) (1.8953) (3.0938) (2.2940) (1.1449)

South Dakota 9.60% 12.95% *** *** *** *** 21.18% (0.2469) (0.2816) *** *** *** *** (3.3319) Tennessee 11.74% 15.60% 15.23% 19.33% 23.34% 32.20% 23.04% (0.0976) (0.1063) (1.1880) (1.5677) (2.8381) (2.1733) (0.9283) Texas 10.67% 14.92% 11.99% 16.33% 18.96% 27.93% 19.25% (0.0475) (0.0524) (0.5864) (0.6526) (1.1999) (0.9624) (0.3932) Utah 8.13% 10.64% 12.20% 15.46% 19.48% 25.32% 18.28% (0.1289) (0.1492) (1.7760) (2.2601) (3.6342) (2.7996) (1.4919) Vermont 8.79% 10.87% *** *** *** *** 17.84% (0.2706) (0.2921) *** *** *** *** (2.5905) Virginia 8.29% 11.06% 11.18% 12.99% 18.63% 24.40% 17.07% (0.0762) (0.0809) (0.9201) (0.9785) (2.8509) (1.6990) (0.8090) Washington 8.90% 11.43% 12.09% 12.85% 20.14% 26.18% 18.07% (0.0814) (0.0921) (0.8086) (0.9924) (2.3090) (1.5133) (0.6629)

Table B-13. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 14.07% 17.92% 21.21% 21.77% *** 39.45% 28.66% (0.1963) (0.2124) (2.6202) (3.1621) *** (4.3274) (1.8216) Wisconsin 8.72% 11.38% 14.26% 15.14% 21.07% 28.00% 19.98% (0.0911) (0.0990) (1.2178) (1.6368) (3.8026) (1.8653) (0.9826) Wyoming 8.54% 12.31% *** *** *** *** 20.15% (0.2833) (0.3279) *** *** *** *** (2.9751)

307

Table B-14. Percent with Insurance by State and Sexual Identity Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Alabama 86.18% 89.33% 82.53% 87.36% 75.92% 81.32% 82.45% (0.1253) (0.1024) (2.2088) (1.4876) (3.2640) (1.7098) (0.9496) Alaska 82.74% 85.66% 80.81% 86.18% 74.82% 81.33% 81.29% (0.3720) (0.3505) (4.0966) (4.7459) (8.8736) (4.8500) (2.1666) Arizona 86.60% 89.73% 86.39% 91.24% 78.53% 84.51% 85.77% (0.1018) (0.0895) (0.9515) (1.1225) (1.9948) (1.4163) (0.6674) Arkansas 87.72% 91.04% 84.86% 89.30% 78.47% 84.48% 84.94% (0.1482) (0.1256) (2.2454) (1.8599) (4.2001) (2.7125) (1.2870)

308 California 88.38% 91.56% 89.05% 92.90% 82.46% 87.72% 88.52% (0.0404) (0.0343) (0.2955) (0.3710) (0.9577) (0.5155) (0.2383)

Colorado 89.15% 92.18% 88.40% 93.68% 83.17% 89.13% 89.14% (0.1042) (0.0881) (1.1694) (0.9376) (2.5338) (1.3015) (0.6369) Connecticut 92.20% 94.81% 91.39% 94.19% 86.79% 90.10% 91.01% (0.1105) (0.0879) (1.1761) (1.0168) (3.4885) (1.9523) (0.9276) Delaware 91.93% 94.80% 92.05% 96.51% 85.80% 91.86% 92.26% (0.2212) (0.1702) (1.9120) (1.7986) (6.7120) (2.8119) (1.5005) District of Columbia 93.96% 96.87% 96.80% 96.27% 91.85% 95.85% 95.71% (0.2429) (0.1592) (0.9058) (1.3920) (4.7297) (2.1790) (1.0715) Florida 83.14% 86.59% 82.51% 86.88% 72.28% 78.12% 80.61% (0.0644) (0.0554) (0.5629) (0.5768) (2.2424) (1.1145) (0.4770) Georgia 82.46% 85.52% 81.54% 84.25% 71.03% 75.97% 78.93% (0.0958) (0.0825) (1.0425) (1.0275) (2.4698) (1.4079) (0.6883) Hawaii 94.77% 96.38% 94.20% 95.46% 91.15% 94.05% 93.99% (0.1538) (0.1211) (1.7495) (1.4063) (4.1755) (1.9386) (1.0332)

Table B-14. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Idaho 86.22% 87.75% 82.89% 84.15% 76.65% 80.29% 81.44% (0.2141) (0.1974) (2.9055) (3.1084) (6.2283) (3.3144) (1.7059) Illinois 89.82% 92.69% 88.87% 92.84% 82.53% 88.51% 88.75% (0.0659) (0.0552) (0.7291) (0.8784) (1.7173) (1.0758) (0.5559) Indiana 88.76% 91.00% 86.15% 88.86% 80.88% 83.75% 85.31% (0.0950) (0.0835) (1.2529) (1.3560) (2.6632) (1.6172) (0.6328) Iowa 93.69% 95.60% 90.88% 94.61% 88.30% 92.77% 92.03% (0.1068) (0.0878) (1.7263) (1.3831) (3.7549) (1.7268) (0.7769) Kansas 88.73% 90.69% 85.44% 92.08% 80.49% 84.38% 86.00% 309 (0.1469) (0.1280) (2.2777) (2.0493) (4.2372) (1.5594) (1.1681)

Kentucky 91.75% 94.39% 90.35% 93.59% 85.41% 90.59% 90.59% (0.1018) (0.0809) (1.5493) (1.1617) (2.8223) (1.4825) (0.7076) Louisiana 85.27% 89.04% 84.06% 87.86% 75.62% 82.85% 83.45% (0.1346) (0.1085) (1.4454) (1.2781) (4.2445) (1.9692) (0.9678) Maine 89.68% 92.47% 88.68% 92.61% 82.68% 87.22% 88.50% (0.2030) (0.1715) (2.2443) (1.9893) (6.3949) (3.1231) (1.2984) Maryland 91.21% 93.97% 89.78% 94.66% 84.86% 89.89% 90.40% (0.0918) (0.0726) (0.9506) (0.8808) (3.3831) (1.4577) (0.6958) Massachusetts 95.91% 97.62% 95.88% 97.41% 92.33% 95.97% 95.81% (0.0591) (0.0427) (0.5321) (0.4471) (1.6769) (0.8083) (0.3202) Michigan 91.87% 94.50% 90.08% 93.12% 85.90% 91.03% 90.56% (0.0671) (0.0541) (0.8894) (0.7904) (1.7372) (0.9657) (0.5998) Minnesota 93.97% 95.82% 93.76% 95.49% 89.91% 93.06% 93.40% (0.0805) (0.0681) (1.3191) (1.0300) (2.2805) (1.6935) (0.8494)

Table B-14. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Mississippi 82.86% 86.59% 78.37% 82.06% 70.14% 77.34% 77.86% (0.1761) (0.1471) (2.3996) (1.9552) (5.2720) (3.2584) (1.3896) Missouri 87.88% 90.24% 86.72% 90.83% 79.15% 83.97% 85.76% (0.1019) (0.0895) (1.2303) (1.3009) (2.8881) (1.5209) (0.7980) Montana 87.60% 90.37% 83.58% 87.49% 78.84% 84.29% 84.10% (0.2519) (0.2241) (3.5982) (3.2386) (7.8540) (3.5451) (1.8023) Nebraska 89.58% 91.35% 86.25% 90.20% 83.00% 86.02% 86.69% (0.1703) (0.1549) (2.2875) (2.5936) (4.4877) (2.7161) (1.3065) Nevada 85.23% 88.12% 84.80% 89.14% 76.94% 82.35% 83.72% 310 (0.1637) (0.1468) (1.4727) (1.5678) (3.2667) (2.1258) (1.0318)

New Hampshire 91.28% 93.70% 88.89% 94.42% 83.44% 89.51% 89.78% (0.1864) (0.1536) (2.5800) (1.5380) (6.1156) (3.0780) (1.6048) New Jersey 88.70% 91.71% 86.59% 93.09% 79.83% 86.10% 86.99% (0.0826) (0.0677) (1.0397) (0.7736) (2.6334) (1.4335) (0.6567) New Mexico 85.77% 89.48% 87.54% 90.64% 79.99% 86.45% 86.90% (0.1960) (0.1696) (1.6269) (1.7779) (4.5594) (2.6732) (1.3229) New York 90.79% 94.12% 89.96% 93.98% 83.89% 90.32% 90.17% (0.0515) (0.0391) (0.4980) (0.4950) (1.4807) (0.6950) (0.3314) North Carolina 85.19% 88.57% 82.96% 87.69% 75.89% 81.08% 82.59% (0.0890) (0.0739) (1.0057) (0.8026) (3.0039) (1.4511) (0.7130) North Dakota 91.24% 93.27% 87.66% 91.15% 84.47% 87.52% 87.82% (0.2451) (0.2244) (3.8933) (4.5311) (9.6839) (5.1474) (2.3935) Ohio 91.47% 94.20% 89.42% 93.75% 85.06% 90.40% 90.22% (0.0634) (0.0505) (0.8018) (0.5913) (1.7171) (0.9756) (0.4661)

Table B-14. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB Oklahoma 82.56% 85.23% 78.01% 81.19% 71.90% 75.16% 76.99% (0.1572) (0.1400) (1.9167) (2.1292) (5.0162) (2.6651) (1.2603) Oregon 90.25% 93.00% 90.23% 94.89% 83.66% 89.59% 90.30% (0.1118) (0.0943) (0.9450) (0.7155) (2.3032) (1.2826) (0.6006) Pennsylvania 91.96% 94.44% 90.93% 94.39% 86.34% 91.08% 91.21% (0.0591) (0.0467) (0.8603) (0.5787) (1.9240) (1.0006) (0.4407) Rhode Island 92.85% 95.77% 92.70% 95.01% 87.65% 93.68% 92.92% (0.1969) (0.1429) (1.8996) (1.7990) (5.7092) (2.1976) (1.0209) South Carolina 85.27% 88.86% 81.95% 85.78% 76.46% 80.43% 81.65% 311 (0.1262) (0.1054) (1.5207) (1.3872) (3.3728) (2.2889) (0.9796)

South Dakota 88.25% 91.41% 83.24% 88.77% 74.92% 84.65% 84.03% (0.2712) (0.2333) (5.8323) (3.8024) (7.9967) (4.9414) (2.5419) Tennessee 86.22% 90.17% 81.99% 89.23% 76.76% 84.00% 83.72% (0.1058) (0.0867) (1.6254) (1.4164) (2.9514) (1.5650) (0.8130) Texas 78.69% 81.07% 78.76% 83.31% 68.88% 73.11% 76.57% (0.0629) (0.0575) (0.6042) (0.6802) (1.4593) (0.9596) (0.4847) Utah 88.08% 89.90% 85.33% 91.90% 79.94% 84.82% 85.89% (0.1510) (0.1403) (1.7545) (1.8070) (3.6844) (2.3296) (1.0651) Vermont 94.23% 96.43% 90.52% 96.55% *** 95.36% 93.84% (0.2235) (0.1706) (2.9229) (1.4013) *** (2.8247) (1.7113) Virginia 88.28% 90.67% 86.89% 90.77% 80.90% 84.83% 86.32% (0.0873) (0.0747) (1.0771) (0.9237) (2.3869) (1.4347) (0.6206) Washington 90.76% 93.46% 90.73% 94.70% 85.16% 90.34% 90.76% (0.0824) (0.0688) (0.7494) (0.6875) (1.9457) (1.0504) (0.4872)

Table B-14. (continued) Heterosexual Heterosexual Gay Lesbian Bisexual Bisexual Total State Men Women Men Women Men Women LGB West Virginia 91.30% 94.29% 90.06% 92.73% 84.75% 90.56% 90.13% (0.1633) (0.1277) (2.4310) (1.7666) (4.9113) (2.8307) (1.5528) Wisconsin 92.66% 94.66% 91.23% 94.66% 88.67% 90.77% 91.56% (0.0843) (0.0712) (1.6193) (1.2216) (2.4552) (1.3616) (0.7725) Wyoming 86.36% 89.54% 82.32% 89.07% *** 80.93% 82.66% (0.3468) (0.3087) (4.7217) (4.5444) *** (5.5329) (2.8593)

312

APPENDIX C

FULL-STATE DEMOGRAPHIC AND ECONOMIC CHARACTERISTICS

Table C-1. Demographic Characteristics by State White, Other Education, Has State Age Married Black Hispanic Non-Hispanic Race BA+ Child Alabama 48.13 51.28% 68.34% 25.33% 4.33% 3.26% 23.01% 35.59% Alaska 45.14 51.47% 65.50% 3.12% 28.21% 5.67% 26.96% 38.98% Arizona 47.94 50.63% 60.19% 4.06% 16.27% 27.11% 26.58% 36.26%

313 Arkansas 48.10 53.77% 75.84% 14.23% 6.33% 5.92% 21.08% 36.46%

California 46.29 50.07% 41.03% 5.63% 32.93% 35.14% 30.77% 40.92% Colorado 46.40 53.98% 72.10% 3.83% 10.40% 18.51% 37.63% 36.11% Connecticut 48.80 51.82% 70.79% 9.84% 11.31% 13.73% 37.06% 37.69% Delaware 48.94 51.33% 66.60% 20.43% 7.83% 7.43% 29.68% 35.66% District of Columbia 43.85 30.64% 39.55% 44.81% 11.37% 9.84% 56.42% 23.97% Florida 49.80 49.16% 57.58% 14.55% 7.61% 23.98% 27.42% 33.89% Georgia 46.53 50.64% 56.46% 30.25% 8.34% 7.82% 28.53% 38.98% Hawaii 49.07 52.96% 23.11% 1.40% 73.16% 7.86% 30.68% 39.86% Idaho 47.50 58.52% 84.92% 0.58% 7.77% 10.26% 24.43% 37.02% Illinois 47.31 51.37% 64.80% 13.43% 12.96% 14.88% 32.24% 38.35% Indiana 47.60 53.24% 82.26% 8.58% 5.84% 5.54% 24.48% 37.34% Iowa 48.34 57.09% 88.65% 2.86% 5.25% 4.62% 26.91% 35.29% Kansas 47.53 56.53% 79.60% 5.32% 8.31% 9.54% 30.64% 37.18% Kentucky 48.01 53.52% 87.05% 7.37% 3.76% 2.80% 22.36% 37.16% Louisiana 47.07 46.90% 61.87% 30.22% 4.85% 4.50% 22.07% 36.33%

Table C-1. (continued) White, Other Education, Has State Age Married Black Hispanic Non-Hispanic Race BA+ Child Maine 50.53 54.35% 94.80% 0.92% 3.35% 1.27% 29.62% 32.22% Maryland 47.60 50.73% 54.01% 29.15% 12.69% 8.47% 37.31% 39.10% Massachusetts 47.99 50.55% 75.09% 6.94% 12.52% 9.93% 41.16% 37.21% Michigan 48.32 51.53% 77.91% 12.90% 6.50% 4.05% 26.78% 36.06% Minnesota 47.72 56.04% 83.80% 5.05% 8.71% 4.21% 33.55% 36.22% Mississippi 47.57 48.30% 59.74% 35.74% 3.13% 2.39% 20.39% 37.66% Missouri 48.16 53.37% 81.99% 10.72% 5.16% 3.30% 27.02% 36.15% Montana 49.09 55.22% 88.78% 0.46% 8.66% 2.99% 28.34% 31.79% Nebraska 47.51 57.21% 82.95% 4.18% 6.47% 8.52% 29.62% 37.40% 314 Nevada 47.23 48.30% 54.20% 8.36% 23.34% 24.81% 22.35% 36.53%

New Hampshire 49.28 55.95% 92.11% 1.18% 4.55% 2.87% 34.68% 35.68% New Jersey 48.19 53.07% 58.54% 12.86% 17.46% 18.24% 37.05% 41.72% New Mexico 48.04 47.87% 41.85% 1.94% 22.13% 45.11% 24.61% 35.27% New York 47.67 48.27% 58.04% 14.97% 19.50% 17.48% 34.47% 37.76% North Carolina 47.89 52.42% 66.88% 20.72% 8.12% 7.22% 28.77% 36.62% North Dakota 46.60 56.93% 87.70% 2.36% 8.10% 2.74% 27.46% 34.20% Ohio 48.42 51.53% 81.72% 11.37% 4.97% 3.01% 26.20% 36.19% Oklahoma 47.34 53.64% 70.41% 6.74% 17.72% 8.28% 23.52% 37.24% Oregon 48.11 52.79% 79.59% 1.71% 11.92% 10.30% 30.68% 34.27% Pennsylvania 49.07 52.18% 79.85% 10.01% 6.83% 5.79% 29.43% 36.44% Rhode Island 48.35 48.48% 76.25% 5.75% 10.70% 12.84% 31.50% 36.29% South Carolina 48.48 50.84% 66.77% 25.84% 4.54% 4.44% 25.82% 35.11% South Dakota 48.15 56.80% 86.13% 1.53% 10.53% 2.83% 26.95% 35.10% Tennessee 47.86 52.64% 76.93% 15.72% 4.50% 4.10% 25.08% 36.42%

Table C-1. (continued) White, Other Education, Has State Age Married Black Hispanic Non-Hispanic Race BA+ Child Texas 45.46 52.92% 46.12% 11.72% 12.90% 35.80% 26.90% 41.67% Utah 43.92 60.57% 80.71% 1.06% 11.17% 12.28% 29.22% 44.70% Vermont 49.72 53.84% 94.27% 0.89% 3.71% 1.51% 35.49% 31.67% Virginia 47.58 53.69% 64.88% 18.46% 11.51% 7.95% 35.93% 37.86% Washington 47.18 54.29% 72.83% 3.44% 18.22% 9.97% 32.91% 36.64% West Virginia 49.76 53.26% 93.52% 3.20% 2.39% 1.18% 19.54% 33.90% Wisconsin 48.36 54.79% 84.91% 5.36% 6.45% 5.24% 28.00% 35.19% Wyoming 47.76 57.72% 86.60% 0.80% 6.38% 8.29% 24.88% 34.24%

315

Table C-2. Economic Characteristics by State Not in Fulltime, Year-Round Poverty Has Median State Unemployed Labor Force Workers (FTYR) Rate Insurance Earnings (FTYR) Alabama 40.30% 3.81% 41.08% 15.32% 87.72% $40,200 Alaska 29.68% 5.06% 42.87% 9.71% 84.09% $53,400 Arizona 38.08% 3.79% 41.23% 13.75% 88.14% $42,000 Arkansas 39.10% 3.19% 42.67% 15.38% 89.34% $37,000 California 33.74% 4.27% 42.38% 12.37% 89.96% $48,500 Colorado 29.46% 3.11% 47.98% 9.90% 90.63% $48,500 Connecticut 30.67% 4.25% 45.21% 8.68% 93.49% $57,600 Delaware 35.25% 3.56% 44.43% 10.76% 93.42% $47,600 District of Columbia 26.36% 5.06% 53.35% 14.08% 95.58% $70,000 316 Florida 39.11% 3.69% 40.81% 12.88% 84.82% $39,200

Georgia 34.24% 3.97% 45.31% 13.74% 83.93% $42,300 Hawaii 33.70% 2.66% 46.02% 8.95% 95.54% $45,100 Idaho 35.45% 2.93% 42.04% 12.62% 86.85% $39,000 Illinois 32.04% 4.28% 45.33% 11.40% 91.22% $49,200 Indiana 33.47% 3.41% 45.13% 12.25% 89.78% $42,300 Iowa 30.13% 2.49% 49.18% 10.57% 94.59% $42,300 Kansas 31.11% 2.87% 48.26% 11.35% 89.64% $42,300 Kentucky 38.50% 3.51% 41.79% 15.77% 93.06% $40,800 Louisiana 37.29% 4.14% 42.76% 16.75% 87.15% $41,600 Maine 35.39% 2.79% 42.35% 11.38% 91.06% $42,300 Maryland 29.40% 3.78% 49.37% 8.45% 92.60% $56,600 Massachusetts 30.26% 3.48% 45.81% 9.59% 96.76% $59,200 Michigan 36.48% 3.94% 40.66% 13.10% 93.15% $45,500 Minnesota 28.20% 2.66% 48.44% 9.01% 94.86% $50,000

Table C-2. (continued) Not in Fulltime, Year-Round Poverty Has Median State Unemployed Labor Force Workers (FTYR) Rate Insurance Earnings (FTYR) Mississippi 39.61% 4.64% 41.21% 17.99% 84.68% $36,900 Missouri 34.43% 3.21% 45.27% 12.29% 89.01% $41,900 Montana 34.62% 2.64% 42.05% 12.61% 88.86% $39,200 Nebraska 27.88% 2.24% 51.47% 10.49% 90.37% $42,300 Nevada 34.14% 4.46% 43.46% 12.01% 86.56% $41,000 New Hampshire 29.78% 2.75% 47.69% 6.93% 92.44% $51,000 New Jersey 31.45% 3.95% 46.34% 8.74% 90.17% $57,100 New Mexico 39.80% 4.14% 39.36% 17.62% 87.67% $38,100

317 New York 34.04% 3.76% 44.37% 12.58% 92.46% $51,200 North Carolina 35.44% 3.83% 43.65% 13.51% 86.86% $41,000

North Dakota 27.46% 1.84% 51.04% 10.40% 92.15% $45,000 Ohio 34.24% 3.59% 44.20% 12.53% 92.80% $44,500 Oklahoma 36.14% 3.21% 44.99% 13.80% 83.76% $39,200 Oregon 35.65% 3.61% 40.46% 12.55% 91.58% $45,000 Pennsylvania 34.71% 3.59% 44.21% 11.10% 93.19% $47,100 Rhode Island 32.68% 3.94% 43.83% 11.48% 94.32% $50,000 South Carolina 37.36% 3.87% 42.67% 13.76% 87.03% $39,800 South Dakota 28.83% 2.33% 51.03% 11.50% 89.71% $38,900 Tennessee 36.45% 3.61% 43.73% 13.97% 88.18% $39,500 Texas 32.38% 3.48% 47.62% 12.99% 79.83% $42,300 Utah 29.84% 2.54% 45.79% 9.63% 88.90% $45,000 Vermont 31.48% 2.58% 45.59% 10.04% 95.30% $44,400 Virginia 31.90% 3.22% 47.38% 9.90% 89.45% $51,200 Washington 33.58% 3.34% 43.78% 10.44% 92.07% $52,400

Table C-2. (continued) Not in Fulltime, Year-Round Poverty Has Median State Unemployed Labor Force Workers (FTYR) Rate Insurance Earnings (FTYR) West Virginia 44.42% 3.61% 37.45% 16.34% 92.76% $39,800 Wisconsin 31.03% 2.68% 46.78% 10.32% 93.61% $45,000 Wyoming 30.82% 3.24% 46.95% 10.65% 87.80% $45,600 318

APPENDIX D

COMPARISONS OF CSMI STATE ESTIMATES TO EXISTING DATA

Table D-1. State-by-State Comparisons of CSMI Estimates to Existing Data State Gallup (LGBT1) Estimated Trans2 CSMI (LGB) Difference Alabama 3.10% 0.61% 2.67% 0.18% Alaska 3.70% 0.49% 3.34% 0.13% Arizona 4.50% 0.62% 4.07% 0.19% Arkansas 3.30% 0.06% 2.85% -0.39% California 5.30% 0.76% 4.75% 0.21% Colorado 4.60% 0.53% 4.23% 0.16% Connecticut 3.90% 0.44% 3.56% 0.10% Delaware 4.50% 0.64% 4.08% 0.22% District of Columbia 9.80% 2.77% 7.81% 0.78% Florida 4.60% 0.66% 4.14% 0.20% Georgia 4.50% 0.75% 3.96% 0.21% Hawaii 4.60% 0.78% 4.05% 0.23% Idaho 2.80% 0.41% 2.49% 0.10% Illinois 4.30% 0.51% 3.91% 0.12% Indiana 4.50% 0.56% 4.07% 0.13% Iowa 3.60% 0.31% 3.37% 0.08% Kansas 3.30% 0.43% 2.97% 0.10% Kentucky 3.40% 0.53% 3.01% 0.14% Louisiana 3.90% 0.60% 3.46% 0.16% Maine 4.90% 0.50% 4.51% 0.11% Maryland 4.20% 0.49% 3.83% 0.12% Massachusetts 5.40% 0.57% 4.97% 0.14% Michigan 4.00% 0.43% 3.67% 0.10% Minnesota 4.10% 0.59% 3.68% 0.17% Mississippi 3.50% 0.61% 3.08% 0.19% Missouri 3.80% 0.54% 3.40% 0.14% Montana 2.90% 0.34% 2.66% 0.10% Nebraska 3.80% 0.39% 3.50% 0.09% Nevada 5.50% 0.61% 5.07% 0.18%

1 Williams Institute (2019) 2 Flores et al. (2016) 319

Table D-1. (continued) State Gallup (LGBT3) Estimated Trans4 CSMI (LGB) Difference New Hampshire 4.70% 0.43% 4.40% 0.13% New Jersey 4.10% 0.44% 3.75% 0.09% New Mexico 4.50% 0.75% 3.99% 0.24% New York 5.10% 0.51% 4.71% 0.12% North Carolina 4.00% 0.60% 3.57% 0.17% North Dakota 2.70% 0.30% 2.36% -0.04% Ohio 4.30% 0.45% 3.95% 0.10% Oklahoma 3.80% 0.64% 3.34% 0.18% Oregon 5.60% 0.65% 5.13% 0.18% Pennsylvania 4.10% 0.44% 3.75% 0.09% Rhode Island 4.50% 0.51% 4.11% 0.12% South Carolina 3.50% 0.58% 3.10% 0.18% South Dakota 3.00% 0.34% 2.69% 0.03% Tennessee 3.50% 0.63% 3.06% 0.19% Texas 4.10% 0.66% 3.65% 0.21% Utah 3.70% 0.36% 3.45% 0.11% Vermont 5.20% 0.59% 4.77% 0.16% Virginia 3.90% 0.55% 3.51% 0.16% Washington 5.20% 0.62% 4.74% 0.16% West Virginia 4.00% 0.42% 3.63% 0.05% Wisconsin 3.80% 0.43% 3.49% 0.12% Wyoming 3.30% 0.32% 3.06% 0.08%

3 Williams Institute (2019) 4 Flores et al. (2016) 320

APPENDIX E

DESCRIPTIVE STATISTICS BY SEX AND SEXUAL IDENTITY

Table E-1. Descriptive Statistics by Sex and Sexual Identity Men Heterosexual Gay Bisexual Variable Mean SE Mean SE Mean SE Description Total 96.30% (0.0149) 2.54% (0.0584) 1.16% (0.0760) Percent of Total Population 321

Annual Income Wage/Salary Income (Median) $41,000 $38,100 $25,900 from Salary and Wages Wages (Median) $19.47 $18.58 $13.30 Hourly Wages Log of Wages (Mean) 2.81 (1.1219) 2.83 (1.0979) 2.46 (1.0258) Natural Log of Hourly Wages

Age (Mean) 42.54 (14.1058) 40.21 (13.4451) 32.49 (11.9515) Age in Years Work Age 83.14% (0.0203) 83.68% (0.2225) 68.06% (0.5622) Percentage Age 25-64 Experience (Mean) 21.97 (14.1831) 19.14 (13.3761) 12.32 (11.9789) Work Experience in Years

Education No High School Less than High School 10.21% (0.0164) 6.93% (0.1111) 10.98% (0.4642) Diploma/GED High School High School 57.26% (0.0267) 54.06% (0.2895) 63.58% (0.6791) Diploma/GED

Table E-1. (continued) Men Heterosexual Gay Bisexual Variable Mean SE Mean SE Mean SE Description BA 20.69% (0.0219) 23.12% (0.2481) 17.87% (0.4604) Bachelor's Degree Graduate or Grad+ 11.84% (0.0174) 15.90% (0.2071) 7.58% (0.2997) Professional Degree

Racial/Ethnic Identity White 75.64% (0.0232) 72.38% (0.2319) 69.59% (0.6809) White Racial Identity Black 10.14% (0.0166) 10.33% (0.2056) 7.92% (0.3441) Black Racial Identity All Other 322 Other 14.22% (0.0191) 17.29% (0.2375) 22.49% (0.5974) Racial Identities Hispanic 17.98% (0.0209) 20.99% (0.2079) 23.01% (0.6691) Hispanic Ethnicity

Family Structure Number of Own Children Number of Children (Mean) 0.7979 (1.1434) 0.1597 (0.5645) 0.2961 (0.7644) in Household Number of Own Children Children Under 5 (Mean) 0.1652 (0.4693) 0.0300 (0.2051) 0.0681 (0.2930) under 5 in Household Married 57.15% (0.0274) 29.31% (0.2760) 21.14% (0.5946) Percentage Married

Region Northeast 17.72% (0.0206) 18.42% (0.2282) 15.07% (0.4384) Percentage in Northeast Midwest 21.54% (0.0222) 16.78% (0.2158) 21.68% (0.5978) Percentage in Midwest South 36.71% (0.0263) 35.35% (0.2955) 31.71% (0.5645) Percentage in South West 24.03% (0.0234) 29.45% (0.2755) 31.53% (0.6413) Percentage in West

Table E-1. (continued) Men Heterosexual Gay Bisexual Variable Mean SE Mean SE Mean SE Description Within Metro Area 81.51% (0.0210) 87.81% (0.1855) 83.61% (0.4969) Percentage in Metro Area Percentage Part-Time 14.03% (0.0188) 17.86% (0.2200) 24.45% (0.5373) Part-Time Workers

Percentage in Occupation Occupational Categories Management, Business, Science, and Arts 12.06% (0.0174) 12.34% (0.1459) 7.53% (0.3014) 323 Business Operations Specialists 2.27% (0.0080) 3.02% (0.0918) 1.97% (0.1458)

Financial Specialists 1.93% (0.0074) 2.33% (0.0660) 1.39% (0.1390) Computer and Mathematical 4.29% (0.0108) 4.63% (0.0983) 3.86% (0.1967) Architecture and Engineering 2.53% (0.0084) 1.87% (0.0704) 1.99% (0.1444) Technicians 0.49% (0.0037) 0.33% (0.0321) 0.42% (0.0663) Life, Physical, and Social Science 0.89% (0.0050) 1.10% (0.0481) 0.84% (0.0935) Community and Social Services 1.13% (0.0056) 1.31% (0.0499) 0.88% (0.0952) Legal 1.04% (0.0054) 1.51% (0.0550) 0.76% (0.1007) Education, Training, and Library 3.05% (0.0093) 4.11% (0.1084) 2.52% (0.2033) Arts, Design, Entertainment, Sports, and Media 1.95% (0.0074) 3.30% (0.0945) 2.69% (0.1771)

Table E-1. (continued) Men Heterosexual Gay Bisexual Variable Mean SE Mean SE Mean SE Description Healthcare Practitioners and Technicians 2.87% (0.0089) 4.05% (0.0906) 1.92% (0.1532) Healthcare Support 0.58% (0.0041) 0.93% (0.0420) 0.70% (0.0811) Protective Service 3.17% (0.0094) 2.63% (0.0877) 3.22% (0.2096) Food Preparation and Serving 4.58% (0.0117) 8.91% (0.1704) 10.96% (0.4071) Building and Grounds Cleaning and Maintenance 4.38% (0.0111) 4.10% (0.1173) 5.16% (0.2479) 324 Personal Care and Service 1.50% (0.0065) 2.41% (0.0717) 2.43% (0.1643)

Sales and Related 9.81% (0.0160) 10.81% (0.1682) 11.23% (0.3377) Office and Administrative Support 6.83% (0.0136) 9.21% (0.1523) 8.86% (0.2950) Farming, Fishing, and Forestry 1.01% (0.0053) 0.45% (0.0377) 0.40% (0.0676) Construction 9.36% (0.0155) 4.37% (0.1111) 7.65% (0.2940) Extraction 0.27% (0.0028) 0.09% (0.0174) 0.16% (0.0501) Installation, Maintenance, and Repair 5.77% (0.0126) 3.39% (0.1173) 5.13% (0.2211) Production 8.05% (0.0146) 5.01% (0.1492) 7.75% (0.3020) Transportation and Material Moving 10.19% (0.0163) 7.77% (0.1804) 9.54% (0.4098) Military Specific 0.01% (0.0006) 0.02% (0.0069) 0.01% (0.0127)

Table E-1. (continued) Women Heterosexual Lesbian Bisexual Variable Mean SE Mean SE Mean SE Description Total 94.93% (0.0221) 2.13% (0.0604) 2.94% (0.0794) Percent of Total Population

Annual Income Wage/Salary Income (Median) $30,697 $34,720 $20,969 from Salary and Wages Wages (Median) $16.00 $17.66 $12.14 Hourly Wages Log of Wages (Mean) 2.64 (1.0052) 2.76 (1.0121) 2.36 (0.9754) Natural Log of Hourly Wages

Age (Mean) 42.43 (14.0636) 40.72 (13.8146) 32.60 (12.2506) Age in Years 325 Work Age 83.03% (0.0215) 82.59% (0.2297) 66.45% (0.3654) Percentage Age 25-64

Experience (Mean) 21.45 (14.3470) 19.27 (13.7464) 11.98 (12.2612) Work Experience in Years

Education No High School Less than High School 6.56% (0.0141) 4.52% (0.1338) 7.46% (0.1734) Diploma/GED High School High School 56.22% (0.0279) 52.61% (0.2934) 61.60% (0.3084) Diploma/GED BA 23.35% (0.0238) 23.34% (0.2321) 20.83% (0.3163) Bachelor's Degree Graduate or Grad+ 13.88% (0.0193) 19.53% (0.2112) 10.11% (0.1962) Professional Degree

Racial/Ethnic Identity White 73.43% (0.0252) 71.24% (0.3012) 69.28% (0.4083) White Racial Identity Black 13.12% (0.0191) 14.16% (0.2301) 10.03% (0.2655) Black Racial Identity

Table E-1. (continued) Women Heterosexual Lesbian Bisexual Variable Mean SE Mean SE Mean SE Description All Other Other 13.44% (0.0198) 14.61% (0.2051) 20.69% (0.3223) Racial Identities Hispanic 15.33% (0.0206) 16.90% (0.2415) 19.88% (0.2986) Hispanic Ethnicity

Family Structure Number of Own Children Number of Children (Mean) 0.8282 (1.0947) 0.3878 (0.8172) 0.4772 (0.9069) in Household Number of Own Children 326 Children Under 5 (Mean) 0.1484 (0.4351) 0.0673 (0.2913) 0.0991 (0.3429) under 5 in Household Married 50.80% (0.0289) 32.22% (0.2758) 20.43% (0.3422) Percentage Married

Region Northeast 18.40% (0.0218) 18.82% (0.2276) 15.74% (0.2622) Percentage in Northeast Midwest 21.83% (0.0233) 17.77% (0.2358) 22.11% (0.2901) Percentage in Midwest South 36.95% (0.0273) 36.11% (0.2583) 32.05% (0.3239) Percentage in South West 22.82% (0.0241) 27.30% (0.2814) 30.10% (0.3609) Percentage in West

Within Metro Area 81.78% (0.0218) 86.52% (0.2131) 83.59% (0.2438) Percentage in Metro Area Percentage Part-Time 26.90% (0.0249) 22.87% (0.2286) 35.80% (0.3244) Part-Time Workers

Table E-1. (continued) Women Heterosexual Lesbian Bisexual Variable Mean SE Mean SE Mean SE Description Percentage in Occupation Occupational Categories Management, Business, Science, and Arts 8.95% (0.0159) 11.03% (0.1627) 6.83% (0.1653) Business Operations Specialists 3.15% (0.0098) 3.52% (0.1030) 2.81% (0.1091) Financial Specialists 2.66% (0.0090) 2.30% (0.0950) 1.76% (0.0820) Computer and Mathematical 1.69% (0.0073) 2.22% (0.0871) 1.39% (0.1038) Architecture and 327 Engineering 0.50% (0.0040) 0.62% (0.0518) 0.50% (0.0493)

Technicians 0.11% (0.0019) 0.14% (0.0181) 0.11% (0.0232) Life, Physical, and Social Science 0.88% (0.0053) 1.24% (0.0700) 0.92% (0.0803) Community and Social Services 2.39% (0.0086) 3.16% (0.1003) 2.03% (0.0906) Legal 1.25% (0.0063) 1.77% (0.0806) 1.13% (0.0763) Education, Training, and Library 9.37% (0.0164) 8.31% (0.1590) 6.63% (0.1673) Arts, Design, Entertainment, Sports, and Media 2.00% (0.0079) 2.54% (0.0881) 2.66% (0.1061) Healthcare Practitioners and Technicians 9.74% (0.0166) 8.51% (0.1398) 6.29% (0.1719) Healthcare Support 4.35% (0.0115) 3.42% (0.1164) 3.96% (0.1615) Protective Service 0.95% (0.0055) 1.90% (0.0725) 1.20% (0.0855)

Table E-1. (continued) Women Heterosexual Lesbian Bisexual Variable Mean SE Mean SE Mean SE Description Food Preparation and Serving 6.03% (0.0140) 8.28% (0.1672) 12.34% (0.3107) Building and Grounds Cleaning and Maintenance 3.35% (0.0100) 2.75% (0.0954) 3.58% (0.1207) Personal Care and Service 6.04% (0.0136) 4.75% (0.1426) 7.19% (0.1935) Sales and Related 10.62% (0.0175) 10.70% (0.1936) 14.93% (0.2894) Office and

328 Administrative Support 19.26% (0.0223) 15.62% (0.2187) 17.08% (0.2740) Farming, Fishing,

and Forestry 0.34% (0.0033) 0.18% (0.0265) 0.16% (0.0331) Construction 0.32% (0.0031) 0.62% (0.0380) 0.35% (0.0399) Extraction 0.01% (0.0005) 0.02% (0.0068) 0.01% (0.0053) Installation, Maintenance, and Repair 0.24% (0.0028) 0.55% (0.0420) 0.30% (0.0401) Production 3.53% (0.0103) 3.10% (0.0935) 3.47% (0.1422) Transportation and Material Moving 2.29% (0.0084) 2.76% (0.0956) 2.38% (0.1014) Military Specific 0.00% (0.0003) 0.00% (0.0042) 0.00% (0.0031)

APPENDIX F

MEDIAN EARNINGS BY SEX AND OCCUPATIONAL CATEGORY

Table F-1. Median Earnings by Sex and Occupational Category Median Earnings Occupation Men Women Management, Business, Science, and Arts $77,200 $58,200 Business Operations Specialists $63,500 $52,000 Financial Specialists $73,300 $52,900 Computer and Mathematical $83,600 $68,800 Architecture and Engineering $86,100 $72,000 Technicians $55,000 $42,300 Life, Physical, and Social Science $63,500 $52,400 Community and Social Services $41,900 $40,000 Legal $100,000 $58,200 Education, Training, and Library $51,200 $38,700 Arts, Design, Entertainment, Sports, and Media $40,000 $26,400 Healthcare Practitioners and Technicians $73,300 $52,400 Healthcare Support $26,200 $23,300 Protective Service $49,700 $35,600 Food Preparation and Serving $17,400 $13,800 Building and Grounds Cleaning and Maintenance $23,000 $12,700 Personal Care and Service $16,900 $11,600 Sales and Related $41,900 $20,000 Office and Administrative Support $33,300 $30,700 Farming, Fishing, and Forestry $23,800 $15,000 Construction $33,900 $25,600 Extraction $52,900 $46,000 Installation, Maintenance, and Repair $43,000 $36,600 Production $40,000 $25,400 Transportation and Material Moving $31,700 $21,200 Military Specific $36,900 $32,800

329