BUS 670 Final Paper Already Started, with outline and sources provided

equestrian14

discrimination_against_homosexuals.pdf

Home >Law homework help >BUS 670 Final Paper Already Started, with outline and sources provided

(2013) 42, 213–238

The Review of Regional Studies

The Official Journal of the Southern Regional Science Association

Revisiting Evidence of Labor Market Discrimination against Homosexuals and the Effects of Anti-Discriminatory Laws*

David Christaforea and J. Sebastian Leguizamonb aDepartment of Economics, Yeungnam University, Korea bDepartment of Economics, Vanderbilt University, USA

Abstract: Anti-discrimination laws on the basis of sexual orientation have been adopted by many states to counteract perceived discrimination in the labor market. However, we find the evidence of earnings disparities between homosexual and heterosexual men to be extremely sensitive to the choice of reference group. Relative to married heterosexual men, gay men earn less, and, over time, anti-discriminatory laws lessen this gap. Relative to unmarried, coupled heterosexual men, however, gay men experience similar levels of earnings. The choice of reference group leads to opposite conclusions regarding the effectiveness and necessity of an anti-discriminatory law for homosexual men, which highlights the need to construct reference groups with care. We also find that homosexual women experience similar earnings to their heterosexual female counterparts, and the law has no effect on these relative wages. Keywords: earnings, employment, sexual orientation, policy JEL Codes: J16, J70, J78

1. INTRODUCTION

Gay and lesbian rights advocates have pursued passage of legislation that explicitly prohibits workplace discrimination based on sexual orientation since the 1970s (Rimmerman et al., 2000). They do so on the grounds that this group is the target of discriminatory practices with respect to wages and employment outcomes. Indeed, several studies have found evidence that, on average, gays do earn lower wages than their heterosexual counterparts. 1 Thus, anti- discriminatory laws have the potential to affect the estimated 4 million gays and lesbians currently residing in the United States (Gates, 2011); consequently, the effect of these laws is of notable importance.

Currently the efforts of advocates have resulted in significant successes at the state and local level, but a federal anti-discriminatory law remains elusive.2 Twenty-one states and the District of Columbia prohibit labor market discrimination against gays and lesbians by privately owned companies (HRC, 2009). Additionally, many localities, cities, and in some cases counties

Christafore is International Professor for Education at Yeungnam University, Gyeongsan, South Korea. Leguizamon is Senior Lecturer in the Economics Department at Vanderbilt University, Nashville, TN. Corresponding Author: D. Christafore. E-mail: [email protected] 1 See Klawitter (2011) for a comprehensive overview. 2 “The Employment Non Discrimination Act,” is a bill that is currently being considered by the U.S. Congress to legally prohibit discrimination on the basis of sexual orientation at the federal level.

214 The Review of Regional Studies 43(2,3)

have independently passed their own version of the law.3 Similar to other anti-discrimination provisions (i.e., racial, gender, and disability), these laws attempt to establish equal access and opportunity to employment to those with different sexual orientations. Research on the effect of the other anti-discrimination provisions has not provided conclusive results of the impact they have on the group in question (Acemoglu and Angrist, 2001; Beegle and Stock, 2003; Collins, 2003).

This paper presents a re-evaluation of labor market outcomes for gays and lesbians and the effect of statewide anti-discriminatory laws. Recent studies using U.S. Census data compare the wages of gays and lesbians to their married heterosexual counterparts, and the difference in wages is sometimes interpreted as evidence of discrimination.4

We analyze the effect of anti-discriminatory laws on employment outcomes using a difference-in-difference-in-difference approach (DDD). This approach allows us to capture the relative differences in wages in states that have passed a law compared to states that have not. This method has two well-regarded advantages when considering the analysis of the effect of a law on particular groups. Most importantly, it allows us to control for the established differences in wage differentials of homosexuals and heterosexuals that exist among states over time, independent of any law that was passed. If, for example, a state that had a relatively high earnings differential adopted the law and the earnings differential between homosexuals and heterosexuals decreased, but not to the level of states that did not pass the law, a simple cross- sectional analysis would suggest that the law did not have any impact. Secondly, the DDD controls for state specific time trends that are independent of the passage of an anti- discriminatory law.

To our knowledge, this is the first study to employ a DDD approach to analyze the effect of anti-discriminatory laws on wages of homosexuals and heterosexuals. 5 In addition to the approach, our analysis also differs from previous research with respect to data observation classification. Previous research links an individual’s wages to their place of residence; we link wages to their place of work. Even though most people work in their state of residence, many also do not, as they live in MSAs that cross state borders. Thus a person can reside in a jurisdiction with no law, but work in a place with one. Data suggest that this may not be as important at the state level since many of the states with anti-discrimination laws (ADLs) based on sexual orientation are clustered in the northeast. Yet, since we follow Klawitter (2011) in allowing for the influence of laws passed at the local level, the use of place of work rather than place of residence is necessary.

3 Data gathered from Klawitter (2011) estimate that in 2010 approximately 195 localities protect gays and lesbians from labor market discrimination at private and state agencies, while 137 do so for government entities only. 4 Our data on the classification of an observation as a gay or lesbian individual also come from the decennial U.S. Census. The Census allows individuals to classify their relationship to head of household as a same-sex partner, which is taken as an individual who is in a same-sex relationship. In doing so, our data is limited to same-sex male and same-sex female couples. Although Black et al. (2007a) find that these couples are representative of the general gay and lesbian population, this paper is limited to an analysis of the earning differentials between homosexual and heterosexual (married and unmarried) cohabitating couples. 5 Beegle and Stock (2003) used this approach to analyze the effect of disability laws. The DDD methodology is possible given that the U.S. Census now allows for the identification of households with same-sex couples for a number of years. Our estimates rely on IPUMS micro data from the 1990 and 2000 U.S. Censuses and the 2010 American Community Survey (ACS). For computational ease, and in order to match the density of the ACS sample (one percent), the Census data extracts are re-weighted samples from the five percent sample of the United States population. This allows us to exploit the timing of the state and local laws passed over a period of almost two decades.

CHRISTAFORE & LEGUIZAMON: ANTI-DISCRIMINATION LAWS AND HOMOSEXUALS 215

The results of our analysis suggest that when we control for labor market trends in each state, anti-discrimination laws only have an effect on relative earnings as time since passage of the law increases. Conditional on employment, we find no statistical evidence of contemporaneous earning differences between the relative earnings of gays and lesbians that work in a state with a law and those that work in a state without one. We also find little evidence of an impact of the law on employment of gays and lesbians. These results, however, may reflect two phenomena: a) homosexuals may be reluctant to reveal their sexual orientation in states without the law, making it hard for employers to discriminate in the first place, and b) even in states that adopt the law, homosexuals may not reveal their sexual orientation immediately, resulting in delayed effects.

Overall, however, we find mixed evidence of the existence of earnings differentials between homosexual and heterosexual men. Conditional on an urban location, where the majority of same-sex couples live and work, gay men do appear to have lower earnings, although the results are not statistically significant. Furthermore, the results do seem sensitive to the choice of comparison group (married versus unmarried men). Like Allegretto and Arthur (2001) and Carpenter (2005), we find that some of these economic differences are partly driven by the difference between individuals in same-sex couples and those in conventional heterosexual marriages. These results suggest that in order to better identify the source of these differences and be able to better understand the effect of ADLs, extending the current policies that allow same-sex marriage in some states may be necessary.

In section 2 we provide a general background and summarize the previous literature. In section 2 we also discuss the theory and previous empirical findings. Section 3 describes the data and the empirical method. In section 4 we present the results, and in section 5 we conclude.

2. BACKGROUND

2.1 Discrimination and Labor Market Differences

Extensive research exists on the differences in employment outcomes between gays and heterosexual males and between lesbians and heterosexual females. This research has not been able to provide conclusive evidence for or against the existence of discrimination against gays or lesbians. In general, studies using various types of survey data have found that gays tend to earn lower wages than heterosexual males (Badgett, 1995; Klawitter and Flatt, 1998; Clain and Leppel, 2001; Berg and Lien, 2002; Black et al., 2003; Blanford, 2003; Plugg and Berkhout, 2004; Frank, 2006; Carpenter, 2007; Elmslie and Tebaldi, 2007; Ahmed and Hammarstedt, 2010). In contrast to these findings for gays, similar studies have found that lesbians typically earn the same or more than heterosexual females (Klawitter and Flatt, 1998; Clain and Leppel, 2001; Berg and Lien, 2002; Black et al., 2003; Blanford, 2003; Plugg and Berkhout, 2004; Arabsheibani et al., 2005; Carpenter, 2007; Elmslie and Tebaldie, 2007; Ahmed and Hammarstedt, 2010), although Badgett (1995) did find that lesbians earn less than heterosexual women.

Discrimination in the labor market is well documented in the literature. Herek (2009) uses survey responses from a nationally representative sample of gays and lesbians to determine if they are the victims of various types of discrimination. He finds that one in ten surveyed experience discrimination in the housing or labor market. Badgett et al. (2007) provides a comprehensive overview of many of the surveys that have been conducted of gays and lesbians

216 The Review of Regional Studies 43(2,3)

regarding discrimination. In their analysis they find that in studies conducted from the mid-1980s to mid-1990s, 16 percent to 68 percent of gay, lesbian, and bisexual respondents reported experiencing employment discrimination at some point in their lives, while in the 15 studies conducted since the mid-1990s 15 percent to 43 percent of respondents reported experiencing workplace discrimination.

Wage differentials between gays or lesbians and their heterosexual counterparts cannot be taken as conclusive proof of labor market discrimination, however. Differences in earnings between individuals in same-sex couples and heterosexuals may be partly explained by specializing households or by the marriage premium. Becker (1971) developed a household specialization model that predicts females in heterosexual relationships invest less in obtaining labor market skills as they expect to couple with a high earning male in the future. Meanwhile men tend to invest more in labor market skills because they believe they must compensate for the low earnings of the female they will couple with in the future.6

Becker’s model predicts that discriminatory firms will be driven out of the industry since non-discriminatory ones will benefit from higher profits. Rosen (2003), however, argues that this prediction does not hold under markets with friction and wage bargaining. Applying this household specialization model to same-sex relationships, Becker (1991) predicts that gays invest less in obtaining labor market skills relative to heterosexual males (since they plan on partnering with a male), while lesbians invest more in obtaining labor market skills than heterosexual females (since they plan on partnering with a female). These predictions may seem plausible under the assumption that homosexuals are widely accepted, allowing them to plan their future investments without concern for social norms. Until recently, however, society's acceptance of homosexuality has been limited, making these investment decisions less clear for gays and lesbians. In fact, data show that both gays and lesbians in the workforce invest more in education than their heterosexual equivalents (Carpenter, 2005). This provides little support for Becker's household specialization theory for the case of gays and lesbians, invalidating the notion that the observed earnings differences often found in the literature are in fact due to under-investment by gays and over-investment by lesbians.7

In the absence of household specialization within same-sex couples, differences in earnings may be the result of a marriage premium (Allegretto and Arthur, 2001; Carpenter, 2005). Married men are often believed to be more productive, which is reflected in their wages. Married women, on the other hand, may be a relative liability to some employers given that they may be more likely to interrupt their work because of children. If and only if employers use marriage as a signal of future productivity, the inability of homosexuals to marry their partners, which remains prevalent in many states, may play an important role. Indeed, unmarried women in same-sex partnerships may benefit in the hiring process if employers believe their family commitments will not interfere with their future productivity. On the other hand, unmarried men in same-sex partnerships may be perceived by employers as less productive. If the earnings

6 In the presence of potential discrimination in a competitive market, Becker (1971) predicts that market forces will eliminate wage differentials due to employer tastes for discrimination. 7 An alternative measure of education is considered by Black et al. (2007b). They combine U.S. Census data for individuals in same-sex relationships with data from the National Survey of College Graduates and show statistics in which gay men are more likely to graduate with a “typically female major” than other men. The opposite is true for lesbians. Under the assumption that “typically male” majors yield higher returns, this suggests that wage differentials can be partly explained by education choice. See Brown and Corcoran (1997) for detailed descriptions of ``typical female/male majors."

CHRISTAFORE & LEGUIZAMON: ANTI-DISCRIMINATION LAWS AND HOMOSEXUALS 217

differentials are only due to observed productivity differences, and marriage is simply an outcome that characterizes productive men, any observed differences in earnings between gay men and married heterosexuals could be due to our inability to distinguish between gays that would and would not marry if allowed. Even after accounting for some measure of productivity, the potential for marriage as a signal cannot be ruled out as the source of earnings differentials. Given the small number of same-sex married couples today, employers who dislike homosexuals may assume all (or most) married men are heterosexual, thus increasing demand for married men (Carpenter, 2005).

In an attempt to overcome difficulties in measuring discrimination, some researchers have employed labor market experiments to test for its existence. Ahmed et al. (2013) conducted a national Swedish experiment in the hiring process, randomly assigning sexual orientation to different job applicants. They find that gays experience discrimination in male-dominated industries and that lesbians experience discrimination in female-dominated industries. Using a similar procedure in the United States, Tilcsik (2011) performed an experiment for gays in which the signal for sexual orientation was identified on subject’s resumes—participation in a gay organization. He also finds evidence for discrimination against gay men. These results, however, pertain to openly gay men, who may be different in unobserved ways from the gay population in general. Both experiments test for discrimination in the hiring process rather than wages.

As seen, the state of the literature on labor market discrimination based on sexual- orientation highlights many of the difficulties in measuring the causes of differences in labor outcomes for homosexuals and heterosexuals. Although some studies find evidence of direct discrimination, other studies find that it may be a function of the social environment, including the ability to marry. Even in the absence of support for marriage, some outcomes can depend on how societies perceive homosexuality. These perceptions may not only shape gays’ investments in education, but it may also affect their location decisions. Gays and lesbians may not only locate in places where homosexuality is widely accepted, but they also may locate strictly based on employers’ attitudes. If true, actual potential discrimination may be understated by the data, since we should find fewer “revealed” homosexuals in areas where they are not accepted. Given that all studies depend on individuals’ willingness to disclose their orientations, results to date— including ours—must be interpreted carefully.

2.2 History of Anti-Discriminatory Laws

Although measuring discrimination against gays and lesbians through rigorous statistical analysis is complicated, the documented cases of discrimination has motivated law makers to adopt anti-discriminatory legislation on the basis of sexual orientation. There is not currently a federal anti-discriminatory law but a few states have had such legislation in place for decades. In 1975, Pennsylvania adopted a law that prohibited discrimination on the basis of sexual orientation at all state agencies. Wisconsin was the first state to implement a law that covered the private sector in 1982 (although the District of Columbia adopted a law that covered employment in the private sector in 1973). By 1990, the first year in our sample, only Wisconsin and Massachusetts (1989) had adopted a law that applied to all private companies.

218 The Review of Regional Studies 43(2,3)

Table 1: States with Sexual Orientation Provisions in Anti-Discrimination Law

State Sector Year Statute or Law Source AK Public 2002 Admin. Order No. 195 HRC

AZ Public 2003 Exec. Order No. 2003-22 HRC CA Public 1992 Gov. Code §121940 HRC Private 1992 §12920 and Civ. Code §51 Klawitter (2011) CO Public 1990 Exec. Order D0035 Williams Institute Private 2007 C.R.S 24-34-401, 24-34-402 HRC CT Public/Private 1991 Conn. Gen. Stat. §46a - 81c-m HRC

DE Public 2001 Exec. Order No. 2000-83 DE Registrar of Regulations

Private 2009 Senate Bill 121 (2009) HRC HI Public/Private 1991 H.R.S. 515-2 HRC

IL Public 1996 Admin. Order No. 2 Illinois Code Section 302.790

Private 2005 775 ILCS 5/1-102 HRC IN Public 2001 N/A Klawitter (2011) IA Public/Private 2007 §216.2(14) IA Code, Ch. 216 (HRC) KS Public 2007 Gov. Sebelius Exec. Order HRC KY Public 2008 Gov. Beshear Exec. Order The Equality Party

LA Public 2004 N/A Division of Admn. LAc

ME Public 2001 Gov. King Jr. Revision of Code Service Bulletin 13.4B Private 2005 M.R.S. ANN. tit. 5 §4571-76 HRC MD Public/Private 2001 Md. Ann. Code art. 49B §5 HRC MA Public/Private 1989 MASS. GEN. LAWS ANN. ch. 151B, §§3-4 HRC MI Public 2003 Exec. Order No. 2003-24 Office of the Governor

MN Public 1991 Exec. Order No. 91-4 MN Legislature. 1991-01- 29

Private 1993 MINN. STAT. §363A.01 to §363A.41 HRC MO Public 2010 Exec. Order 10-24 Art. 1 Office of MO Governor MT Public 2000 N/A Williams Institute

NV Public/Private 1999 NV. REV. STAT. 233.010(2); 613.330 HRC NH Public/Private 1997 N.H. R.S.A. §§21-I:42, 354-A:2 , 354-A:6 HRC NJ Public/Private 1992 N.J. STAT. ANN. §10:2-1; 10:5-1 - 49 HRC NM Public 1985 New Mexico Exec. Order No. 85-15 HRC Private 2003 N.M. Stat. Ann. §§28-1-2, 7, 9 HRC NY Public 1983 Gov. Cuomo Exec. Order Office of the Governor Private 2002 NY EXEC. LAW §296, 296-a HRC OH Public 2007 Exec. Order 10S Gov. Strickland’s website OR Public/Private 2008 OR Equality Act 100, Oregon SB 2 HRC PA Public 1975 Exec. Order 1975-5 Equality Pennsylvania RI Public 1985 R.I. Exec. Order No. 11 Governor’s Office website Private 1995 R.I. Gen. Laws §28-5-7 HRC

VT Public/Private 1992

21 §495; 9 §4503; 8 §10403; 8 §4724; 3 §963 Exec. Order 1

HRC

VA Public 2006 Exec. Order 1 Office of Governor WA Public 1985 Exec. Order 85-09 HRC Private 2006 §49.60.130-175, §356-09-020 HRC

WI Public/Private 1982 Relevant: §36.12, §106.50 §106.52 §111.31 §230.18 §224.77.

HRC

a Human Rights Campaign: State Laws and Legislation http://www.hrc.org/laws-and-legislation/state. b Internet Blog site that seeks equal constitutional liberties and freedom for all people regardless of sexual orientation or gender identity. c Repealed in 2008. d Memorandums created by The William Institute of UCLA’s Law School. Note: Gender identity is currently included in private ADLs in CA, CO, CT, HI, IL, IA, ME, MA, MN, NV, NH, NY (in some cases), RI, VT, and WA.

CHRISTAFORE & LEGUIZAMON: ANTI-DISCRIMINATION LAWS AND HOMOSEXUALS 219

Table 1 summarizes the development of the state laws adopted to date. We are particularly interested in laws that cover employment in the private sector.8 As Table 1 shows, there are some states in which discrimination is explicitly prohibited in state agencies, but not in private companies. In other states, neither state nor private employers are required by the law to provide equal opportunities to those with different sexual orientation. Additionally, state provisions differ from each other in regard to the inclusion of gender identity. For example, the Maryland Annexed Code art. 49B §5 of 2001 does prohibit discrimination on the basis of sexual orientation but does not explicitly address the issue of gender identity.

These state laws have largely been designed after other stipulations that prohibit employment discrimination based on race, gender, and, to some extent, disability (Klawitter, 2011). California, for instance, has added the term “sexual orientation” to broader legislation that includes race, religious creed, color, national origin, ancestry, physical and mental disability, medical condition, marital status, sex, and age. It explicitly prohibits discrimination in hiring, firing, compensation package, and employment conditions. It prohibits discrimination against the participation in labor organizations such as unions or other training programs, and also forbids sexual harassment. Many of these laws are not limited to employment and also include housing and other public accommodations.

2.3 Theoretical Effects of Anti-Discriminatory Laws

Logic dictates that anti-discrimination laws should help dissipate the differences in employment outcomes due to discriminatory behavior. For the case of race and sex, this has been partly true. The laws have reduced earning gaps, but not eliminated them.9 However, theory suggests that the effects may not be straightforward. Similar to Beegle and Stock (2003), we consider the theoretical model laid out by Acemoglu and Angrist (2001) and apply it to gays and lesbians in particular. The model predicts that employment outcomes for gays and lesbians are affected by the passage of anti-discrimination laws through two main channels.

The first channel of effect occurs because the passage of the law increases firms’ costs by adding the risk of being sued over not hiring and/or firing a homosexual worker. The higher the cost to the firm, the increased probability of a member of that group being hired. This is often characterized as a “hiring subsidy” for the target group and may result in firms increasing their demand for gay and lesbian workers to avoid incurring the high cost of a potential lawsuit. On the other hand, firms also want to avoid the costs of lawsuits that arise from firing these workers. This raises the cost of hiring gays and lesbians if this is seen as a potential outcome, thereby providing an incentive for firms to decrease their demand for gay and lesbian workers. To the extent that the costs of both firing and not hiring gays and lesbians induce opposing shifts the demand curve, the overall effect of the law on wages and employment levels depends on the relative probabilities of a lawsuit over not hiring versus that of a lawsuit over firing individuals in the protected group. Within this context, the impact of a sexual orientation anti-discrimination law depends on firms’ fears of potential lawsuits. Rubenstein (2001) finds that population- adjusted complaint rates based on sexual orientation discrimination are similar to the population- adjusted complaint rates based on race and gender differences. Rubenstein (2001) provides 8 We also run additional specifications including controls for the existence of laws covering public employees only. States with ADLs covering employment in the private sector usually have additional laws that cover the public sector. However, some states may have laws that apply to public employees but not those in the private sector. These additional results are qualitatively similar and are available upon request. 9 See Klawitter (2011) for a discussion of studies.

220 The Review of Regional Studies 43(2,3)

evidence that firms have a reason to fear a costly lawsuit from gays and lesbians and that law has an impact on firms’ hiring decisions.

The second channel of effect acts through an equal-pay provision that is typically included in anti-discriminatory laws. The equal pay provision aims at reducing the earnings gap between gay and lesbian workers relative to heterosexual workers. Gays and lesbians might receive lower wages for various reasons. Some employers might have a “taste for discriminating” with respect to minority groups (gays and lesbians in this case), implying that firms without such taste will hire a gay or lesbian but at a wage lower than a heterosexual worker with similar characteristics (Black, 1995). It also might be the case that the hiring cost, described above, is larger than the hiring subsidy, driving down wages. Another possibility is that there exist systematic unobserved productivity differences between homosexuals and heterosexuals which explain the earnings gap. However, there is no reason to believe that (beyond sexual orientation) homosexuals are systematically different from heterosexuals. Instead, discrimination against gays and lesbians is more likely due to “differences in constraints that systematically alter the incentives and behavior of heterosexuals” (Black et al., 2007b). Regardless of the reason, we expect to observe a lower equilibrium wage in the market for gay and lesbian workers, relative to the equilibrium wage for heterosexual workers. This implies that an equal-pay provision will act as a price floor in the market for gay and lesbian workers, reducing employment levels but raising the wage.

Given these two effects, the passage of an anti-discrimination law may result in several different employment outcomes for gays and lesbians. We would expect wages to increase if the hiring subsidy is greater than the hiring cost. The opposite occurs if the cost of hiring (or probability of a lawsuit over firing) is relatively higher. In the presence of an equal pay provision, if the hiring subsidy dominates and wages are still less than those for the heterosexual counterparts, then wages increase further and employment decreases. The overall effect of the law in this case is an increase in wages. The effect on employment is ambiguous given that the increase that arises from increased demand is offset by decreases from the equal pay provision. On the other hand, if the hiring costs dominate and wages for homosexuals are indeed lower than those for heterosexuals, then employment decreases further but wages increase from the equal pay provision. Even if there is no evidence of discrimination, we should expect to observe changes in wages and the employment level due to the possibility of lawsuits resulting from the implementation of the law.

2.4 Empirical Findings of Anti-Discriminatory Legislation

To our knowledge there have been three previous papers using data from the 1990 and 2000 U.S. Census to analyze the effect of state anti-discrimination laws on gay and lesbian labor market outcomes. Klawitter and Flatt (1998) collect data on both state and local public and private employment laws and find no evidence that such laws affect wages of gays or lesbians. However, this study was performed using 1990 data, and most of the anti-discrimination laws in place had only been around for a short period. Therefore, the insignificant effects of the laws may have been due to the laws not having a long enough time to be fully implemented and their effects to be observable. Gates (2009) performs a similar analysis using data from the 2000 U.S. Census. He examines only state laws because they allow for an exact geographic match of same- sex couples. State-level policies also provide a more consistent standard of application and enforcement than do local laws.

CHRISTAFORE & LEGUIZAMON: ANTI-DISCRIMINATION LAWS AND HOMOSEXUALS 221

Gates (2009) finds that the presence of an anti-discrimination law increases the relative earnings of gays by 3.0 percent and by 0.3 percent for each year the law is in effect. For lesbians, he finds an increase in relative earnings of lesbians of 2.0 percent in states with an anti- discrimination law and an increase of 0.3 percent for each year the law is in effect. More recently, Klawitter (2011) also uses 2000 U.S. Census data but includes local as well as state laws and uses a multilevel analysis. She finds evidence that anti-discrimination laws decrease the earnings penalty of gay men, primarily by increasing hours worked per week, but finds no evidence that anti-discrimination laws are associated with differences in the earnings of lesbians.

Our paper differs from previous studies of sexual orientation anti-discrimination laws in two ways. First, given the potential for working in one state and living in another, we identify the impact of a law using the place of work rather than the place of residence; and second we use a difference-in-difference-in-difference approach to account for trends in local labor markets. We extend the work of Gates (2009) by also including laws adopted at the local level. Unlike Klawitter (2011), however, in our estimation we only use local laws indicators when state laws have not been adopted. As in Gates (2009), we assume that once a statewide law is passed, individuals will be more likely to use state courts, rather than the local government, in case of a legal dispute. This is partly reaffirmed by Klawitter (2011), as she finds no strong evidence of labor market effects in places with both laws.

3. EMPIRICAL RESEARCH

3.1 Difference-in-Difference-in-Difference

Recent studies that analyze the effect of race and sex anti-discriminatory laws have estimated regressions with pooled data from different periods using the difference-in-difference- in-difference method. Neumark and Stock (2001) employ the DDD approach to analyze the effect of the passage of sex and race anti-discrimination laws at the state level. They find that race anti-discrimination laws generally increase the relative earnings and employment of blacks and that sex anti-discrimination laws increase relative earnings and decrease relative employment for women. Collins (2003) applies a similar DDD framework to study the effect of race anti-discrimination laws and also finds, at least for the 1940s, that the passage of race anti- discrimination laws improved labor market outcomes for blacks. The DDD approach is also used by Beegle and Stock (2003) to analyze the impact that the passage of disability anti- discrimination laws at the state level have on wages, employment, and labor market participation of the disabled. Their findings suggest that disability anti-discrimination laws generally lead to lower relative earnings and lower relative labor-force participation for the disabled, and also do not have an effect on their relative employment.

The difference-in-difference-in-difference approach allows us to identify the net effect of the laws on the employment outcomes for individuals in same-sex relationships. We follow Beegle and Stock (2003) by controlling for differences in employment outcomes across states and time. Like them, we also include control for interaction effects between state and year, thereby controlling for differences in employment outcomes across states for each specific year. Additional interactions include those between time and state with the indicator for “gay/lesbian.” This controls for the presence of individual shocks in the labor market for homosexuals in different states, as well as their time trends. However, using these interactions precludes us from estimating the coefficient on the general effect of the law across all observations, given the

222 The Review of Regional Studies 43(2,3)

generation of a common intercept for all individuals in a state and year. The model is characterized as follows,

(1) yist = α + (slst ∗ ssist)ϕ + ssistγ + (ssist ∗ states)δs + (ssist ∗ yeart)δt + statesηs + yeartηt + (states ∗ yeart)ηst + Xistβ + llistρ + (llist ∗ ssist)θ + durationstϕ + (durationst ∗ ssist)ς + εist

where yist is the employment outcome of interest for an individual i in state s and Census year t. The matrix denoted by X includes information about race, gender, age, age squared, a dummy variable to indicate if the person lives with children, dummy variables for educational attainment, disability status, English proficiency, urban or rural residency, and unearned income. sl corresponds to a binary indicator for the existence of the law in a state in a given year, and ss is an indicator defining whether the person is part of a same-sex relationship. Similarly, ll identifies if the individual lives in a city or town where there exists an anti-discrimination law in that particular year, but no law has been passed at the state level. This is particularly important since there might be regional differences within the states due to the fact that labor markets may be very localized. Finally, duration measures the amount of time since the state law has been passed for the year of measurement.

Previous studies, including Gates (2009) and Klawitter (2011), have reported results for subsets of the population. Beegle and Stock (2003), for instance, estimate the impact of disability laws for groups characterized by age, race, and sex. The purpose of dividing the sample is to account for the passage of other anti-discrimination laws. Like most studies of gays and lesbians, we estimate the model for men and women separately. The combination of state and year interactions, and the relative low number of observations in same-sex relationships when we cut the sample into smaller subgroups (i.e., by race or age) presents problems of perfect predictability and collinearity, precluding us from obtaining separate estimations by race. We, however, control for these characteristics in the main regression.

Finally, it is recognized in the literature that the majority of laws are the result of predisposed conditions, namely political and economic. This implies that the passage of laws is not exogenous and quasi-experimental approaches like the difference-in-difference method need to account for this possibility. Besley and Case (2000) propose two approaches. One is an instrumental variable approach in which the laws are instrumented with political variables, and second uses a third difference by adding a valid comparison group (making this a DDD approach). Though no perfect solution, we opt for the DDD and control for these preconditions by using state and year dummies.

As noted by Beegle and Stock (2003) and Bertrand et al. (2004), there is also the potential that a serial correlation problem exists. Our specification controls for time since the law was adopted. Although this does not eliminate this concern, it does greatly reduce the problem. Additionally, we consider only three periods. Following Beegle and Stock (2003), we argue that this further reduces the threat of serial correlation. As a practical matter, the passage of the laws can increase the incentive to self-identify as a gay or lesbian, making the effect of the law harder to interpret. Although a simple look at the data, described below, suggests that the percentage of people who self-identify as gays or lesbians is small and very similar in states with and without laws, we discuss the implications surrounding the possibility of increased self-identification in states with ADLs in the results section.

CHRISTAFORE & LEGUIZAMON: ANTI-DISCRIMINATION LAWS AND HOMOSEXUALS 223

3.2 Data

The data on gays and lesbians come from the five percent samples of the U.S. Census for 1990 and 2000 and the 1 percent sample of the 2010 American Community Survey (Ruggles et al., 2009).10 These data sets supplied by the U.S. Census are the only U.S. national data sources that enable researchers to identify individual members of same-sex couples as well as their states and localities of residence and of work. While most respondents work in the same state where they live, there is a nontrivial number of respondents that work in another state. The Census also provides information on race, gender, disability status, education, and age for which we can control.

The U.S. Census Bureau does not ask individuals directly about their sexual orientation. Rather, beginning with the 1990 Census, it has been possible to identify same-sex couples that cohabitate with the nature of the relationship inferred from the answer to “relationship to household head” question on the Census.11 A major concern about this way of identifying gay and lesbian couples has been raised by Black et al. (2007a) and Gates and Steinberger (2010). The Census recoded those responses that appeared to be same-sex married couples as same-sex unmarried couples. Black et al. (2007a) find that most same-sex married couples are actually different-sex married couples that misclassified themselves. Gates and Steinberger (2010) suggest excluding same-sex couples for which at least one of the members of the household had their marital status allocated by the Census to correct for this misclassification, which would otherwise overestimate the count of same-sex couples.

To control for labor market conditions, the difference-in-difference-in-difference approach calls for the use of a comparison group. This group must be as similar as possible to the treatment group except that the law does not affect them directly. Since homosexuals could not marry during the period of evaluation, one cannot tell with certainty if those in same-sex partnerships are more similar to married or unmarried heterosexuals. Thus, we first estimate the model using both married and unmarried heterosexuals as the comparison group, and later compare them to the two groups separately and see if the findings vary. Both studies by Gates (2009) and Klawitter (2011) use married couples as the comparison group, but also include different-sex unmarried couples in the sample. Which comparison group is most appropriate is a debated issue that remains unanswered. In part it depends on the type and number of people in same-sex partnerships that would marry if they could. Given the data limitations, we leave this issue for future research as more states legalize same-sex marriage and data on the topic become more available.

The data on state anti-discrimination laws are drawn from the Human Rights Campaign (HRC, 2009) and information for local laws is obtained from Klawitter (2011). As noted above, our indicator for the presence of a local law takes the value of 1.0 only if a state law that prohibits discrimination on the basis of sexual orientation in all private companies has not been adopted and implemented. This is done for each of the three years in the sample. In other words, workers in Los Angeles, California, are subject to the local law in 1990 but no longer in 2000, since thea statewide law was adopted in 1992. We create this variable to control for the fact that

10 These data are publicly available from www.ipums.org. For computational ease, we extracted a re-weighted sample of one percent density for the years 1990 and 2000. 11 Gay and lesbian households are identified from the “relationship to household head” heading on the U.S. Census. The categories under this heading are spouse, child, in-law, unmarried partner, and other nonrelative.

224 The Review of Regional Studies 43(2,3)

workers in Los Angeles had the opportunity to file complaints with the local authorities in 1990, while others in other California towns could not. Local laws often cover cities, counties, and other smaller towns. Using the U.S. Census, we are able to first allocate those laws to workers in each specific city and county. For towns for which we cannot find a match in the Census data, we follow Klawitter (2011) by using PUMAs, although we raise the coverage threshold of the local law to 40 percent of the PUMA.12

The sample is initially restricted to include all people of working age (16 to 64 years).13 In an effort to exclude college students, who are likely to have roommates with the same gender, and also those in the process of retiring, we also examine a sample of individuals between the ages of 25 and 59.14 In the analysis, employment is conditioned on being in the labor force, while earnings are conditioned on being employed. Furthermore, we only analyze the earnings of those who work full-time for most of the year (i.e., 30 or more hours per week and 27 or more weeks per year). We also exclude observations with missing data and those with real hourly wages equal to $1.00 or less. Since weeks and hours worked are reported in ranges for 2010 we use midpoints.

Table 2 is a general description of the data. The share of individuals who identify themselves as gay or lesbian (i.e. in same-sex partnership) is measured as a percentage of the total number of couples, including both married and unmarried heterosexuals. When looking at the number of gays and lesbians as a percentage of unmarried couples only, we find that the share oscillates between 4 and 8 percent, independent of whether or not a law exists. Table 2 shows that there are slightly more same-sex couples in states with a law in any given year. But the numbers are not significantly different from a statistical perspective, and the difference is less than a percentage point.

Overall we do not observe any major abnormal labor outcomes. Labor force participation is slightly lower for those who are married. It is likely in certain cases a member of a married couple chooses to opt out of the labor force as part of their household’s decisions. At the same time, the data suggest, given they participate in the labor force, that married individuals are more likely to be employed. Interestingly, the employed share of individuals in same-sex relationships is higher than that same share for unmarried heterosexuals. This is true for all years whether they locate in a state with an anti-discrimination law or not. Additionally, the employed share of same-sex partners is higher in states with an ADL than in states without one. Although this is true for all years, the difference is consistently small, especially in years where the level of employment around 95 percent (Privately Employed, If in labor force in Table 2).

There are not noticeable differences in other employment outcomes, such as weeks worked or total earnings received.15 Persons in same-sex relationships tend to be more highly educated, younger (in some cases), and more likely to live in metro areas. All such factors undoubtedly play a role in employment success.

12 See Klawitter (2011) for the use of PUMAs for the allocation of local laws. 13 Most studies restrict the sample to exclude those of ages 16 and 17. Legally, these individuals are allowed to work, and we observe that a few identify themselves as living with a partner of the same sex. 14 Note that the U.S. Census does identify “roommates” as a category in the “relationship to the household” variable. Thus, for comparability we also used a sample that included all those between the ages of 18 and 64. The results are nearly identical and are available upon request. 15 Earnings are annualized earnings given that the U.S. decennial Census (1990, 2000) and the ACS (2010) do not report wages.

CHRISTAFORE & LEGUIZAMON: ANTI-DISCRIMINATION LAWS AND HOMOSEXUALS 225

Table 2: Descriptive Statistics for Married and Unmarried Household Couples According to the U.S. Census

1990 2000 2010 Same Sex Opp. Sex Married Same Sex Opp. Sex Married Same Sex Opp. Sex Married States With Law In labor force 0.897 0.900 0.821 0.848 0.817 0.756 0.871 0.853 0.800 Privately Employed Total 0.885 0.817 0.790 0.814 0.752 0.724 0.814 0.753 0.746 If in labor force 0.986 0.908 0.961 0.960 0.921 0.958 0.935 0.883 0.932 Ln(earnings) 10.202 10.033 10.306 10.542 10.189 10.492 10.581 10.144 10.539 Weeks Worked 48.486 48.486 45.535 47.772 47.772 45.765 47.142 44.866 44.866 Hours Worked If Employed 42.986 40.872 40.684 41.964 40.589 41.071 41.132 38.469 39.821 If ‘Full-Time’ 44.952 43.008 43.480 43.925 43.110 43.838 43.699 41.998 43.206 Fully Employed 0.900 0.853 0.845 0.887 0.845 0.858 0.876 0.817 0.847 Age 35.414 32.658 40.678 39.500 34.515 42.728 42.173 36.297 45.442 Percent Male 0.700 0.518 0.548 0.476 0.518 0.546 0.504 0.511 0.535 Percent White 0.914 0.824 0.917 0.829 0.738 0.775 0.860 0.777 0.806 HS Degree Only 0.114 0.364 0.378 0.114 0.275 0.230 0.117 0.270 0.211 Some College 0.186 0.226 0.178 0.211 0.260 0.227 0.206 0.249 0.210 Bachelor’s Degree 0.286 0.107 0.157 0.379 0.246 0.299 0.381 0.285 0.337 Higher Degree 0.286 0.069 0.104 0.237 0.065 0.132 0.265 0.078 0.170 Percent Disability 0.071 0.038 0.034 0.064 0.100 0.093 0.045 0.045 0.041 Speaks English 0.857 0.658 0.592 0.879 0.803 0.803 0.872 0.779 0.781 States Without Law

In labor force 0.897 0.839 0.762 0.846 0.809 0.749 0.865 0.833 0.776 Privately Employed Total 0.844 0.756 0.721 0.805 0.744 0.720 0.804 0.723 0.722 If in labor force 0.941 0.901 0.946 0.951 0.919 0.962 0.929 0.868 0.931 Ln(earnings) 10.290 10.031 10.326 10.313 10.024 10.354 10.351 9.951 10.372 Weeks Worked 46.884 44.499 46.894 47.659 45.697 47.795 47.001 44.152 47.033 Hours Worked If Employed 40.818 40.198 40.744 41.927 41.003 41.552 40.768 38.371 40.400 If ‘Full-Time’ 42.960 42.720 43.417 43.386 43.194 44.033 43.343 42.032 43.411 Fully Employed 0.873 0.823 0.850 0.905 0.851 0.870 0.873 0.807 0.861 Age 35.088 32.705 40.400 37.329 33.928 42.216 41.477 35.430 44.806 Percent Male 0.558 0.529 0.567 0.484 0.518 0.549 0.464 0.514 0.537 Percent White 0.885 0.808 0.875 0.873 0.781 0.857 0.897 0.802 0.858 HS Degree Only 0.138 0.339 0.316 0.171 0.348 0.295 0.170 0.334 0.261 Some College 0.257 0.236 0.213 0.242 0.252 0.236 0.250 0.273 0.226 Bachelor’s Degree 0.273 0.111 0.158 0.338 0.192 0.261 0.371 0.222 0.315 Higher Degree 0.187 0.043 0.092 0.181 0.041 0.105 0.183 0.047 0.130 Percent Disability 0.047 0.042 0.038 0.064 0.108 0.091 0.058 0.057 0.053 Speaks English 0.867 0.703 0.641 0.837 0.685 0.658 0.856 0.703 0.683 Law No Law Law No Law Law No Law Percent Same-Sex 0.0037 0.0025 0.0072 0.0049 0.0085 0.0069 Source: US Census (1990 and 2000) and American Community Survey (ACS 2010). Individuals include the head of household and their respective partners. Individuals are grouped according to their living arrangements as follows. (a) Same Sex: Those couples in which head of household and his/her unmarried partner are coded with the same gender. (b) Opp. Sex: Different-sex individuals in cohabitating households. (c) Married: Head of household and partner are of different sex and legally married. Note: For ln(Earnings), weeks, and hours worked we follow Beegle and Stock (2003) conditioning them on employment.

226 The Review of Regional Studies 43(2,3)

Table 3: DDD Estimates of the Impact of Anti-discrimination Laws on Employment

Variable Men Women Age 0.0519*** 0.0470*** 17.76 17.94 Age Squared -0.0004*** -0.0004*** -13.26 -12.13 Head of Household 0.1449*** 0.0176** 13.3 2.01 Presence of Children 0.0410*** -0.0053 5.62 -0.61 High School Degree 0.1240*** 0.2164*** 7.37 14.80 Some College 0.1637*** 0.3253*** 8.24 20.89 College Degree 0.4293*** 0.5530*** 18.68 29.78 Post College Degree 0.5347*** 0.7000*** 14.81 31.75 Disabled -0.2190*** -0.2070*** -8.33 -5.58 Speaks English 0.0749*** 0.3481*** 3.8 12.79 White 0.2243*** 0.2246*** 10.78 12.97 Resides in Metro Area 0.0095 0.0403*** 0.54 3.39 Unearned Real Income -0.0531*** -0.0633*** -11.23 -14.80 In Same Sex Relationship (SS) 0.3309* 0.0954 1.82 0.46 ADL*SS 0.0725 -0.1228 0.69 -1.11 Time Since Passed -0.0037*** -0.0065*** -12.66 -21.09 Time Since Passed *SS -0.0253*** 0.0080 -3.13 0.88 Local ADL 0.1566*** 0.1191*** 6.56 3.32 Local ADL*SS -0.0655 -0.0979 -0.74 -1.02 Pseudo R-Squared 0.0703 0.0809 No. Observations 1211514 1005408

Notes: Estimates are the raw coefficients of a probit estimation. They do not represent the marginal effects. t-statistics are reported below each coefficient, and are calculated based on standard errors that are corrected for clustering of observations by state and year. State and Time dummies, as well as State and Time interactions, State and SS, and Time and SS interactions are included in order to account for time trends. Level of significance:

10 percent *(t-stat>1.645), 5 percent **(t-stat>1.960), and 1 percent *** (t-stat>2.576).

CHRISTAFORE & LEGUIZAMON: ANTI-DISCRIMINATION LAWS AND HOMOSEXUALS 227

4. RESULTS

The estimated effects of ADLs on employment and earnings are shown in Tables 3 and 4. Controlling for education, location, age, race, and other characteristics, we do not find evidence to support claims that individuals in same-sex relationships are less likely to be employed, relative to heterosexuals in committed relationships (i.e., married and unmarried but living in cohabitation).

In fact, relative to other heterosexuals, men in same-sex relationships are generally more likely to be employed.16 Holding everything else equal, including time and place, we find no evidence that, relative to heterosexuals, men and women in same-sex partnerships who locate in states where ADLs have been passed have better employment prospects than those in states without ADLs (ADL*SS). Moreover, although not statistically robust, we find that relative employment levels of women in same-sex partnerships are worse in states with ADLs. We find, over time, that state ADLs can have small negative effects on overall employment. Indeed, the evidence suggests that gay men are less likely to be employed in states where the law has been in place longer. Additionally, the evidence suggests that overall employment in areas where ADLs have been adopted at the local level is relatively higher. Interestingly, the data do not support the claim that the probability of employment for men and women in same-sex partnerships is significantly different in states with ADLs.

Table 4 shows the estimates of the impact of the law on annualized earnings. While Black et al. (2007b) suggest that college major may be important in explaining wage differences, we do not have data on individuals’ college majors for the years 1990 and 2000. The U.S. Census does identify industry as well as occupations, however. Controlling for occupation presents problems of perfect predictability given the small number of homosexuals in certain occupations in some regions. As in Table 3, we analyze men and women separately. Except for a few variables, the results are mostly comparable to those obtained for the case of employment. Consistent with some previous findings (Klawitter, 2011), we find that men in same-sex relationships do earn less relative to men in heterosexual relationships, while we find no statistical evidence of earnings differentials between women in same-sex partnerships and women in heterosexual ones (married and unmarried).

We find no statistical evidence that the relative earnings are higher for either men or women in same-sex partnerships working in ADL states than their counterparts in states without ADLs. Once again, the Time Since Passed appears to negatively influence the overall average earnings, although the effect is economically small. Contrary to findings on employment propensities, the relative earnings of men in same-sex partnerships do tend to rise as time since the implementation of ADLs passes. This might be due to the fact that these men do earn relatively less, but appear to be more likely to be employed. It may be that the law affects earnings not only by increasing the wages of gay men, but also by decreasing those of heterosexual men. Since the percent of gay men is relatively low compared to that of heterosexual men, this could explain the decrease in overall earnings.

16 The results do not represent the marginal effects, but rather the coefficients from the probit regression, since we are mainly concerned with the signs of these coefficients.

228 The Review of Regional Studies 43(2,3)

Table 4: DDD Estimates of the Impact of Anti-discrimination Laws on Ln(Earnings) – Conditional on Full-Time Employment

Variable Men Women Hours Worked 0.0139*** 0.0207*** 28.70 29.56 Weeks Worked 0.0331*** 0.0340*** 87.70 85.15 Age 0.0680*** 0.0535*** 67.28 53.42 Age Squared -0.0007*** -0.0006*** -58.62 -49.21 Head of Household 0.0893*** 0.0217*** 26.98 8.15 Presence of Children 0.0574*** -0.0186*** 18.14 -5.01 High School Degree 0.1200*** 0.0726*** 14.70 5.10 Some College 0.2265*** 0.2109*** 21.37 13.69 College Degree 0.4924*** 0.5011*** 34.44 27.34 Post College Degree 0.7474*** 0.7689*** 31.66 33.34 Disabled -0.1010*** -0.0786*** -15.29 -11.66 Speaks English 0.3528*** 0.3475*** 20.31 15.97 White 0.1788*** 0.0760*** 30.01 16.28 Resides in Metro Area 0.1295*** 0.1564*** 21.27 23.95 Unearned Real Income 0.0144*** 0.0090*** 24.75 16.33 In Same Sex Relationship (SS) -0.2246* 0.0256 -1.94 0.43 ADL*SS 0.0125 -0.0133 0.42 -0.45 Time Since Passed -0.0009*** 0.0029*** -10.95 29.08 Time Since Passed *SS 0.0074*** -0.0039 3.83 -1.46 Local ADL 0.0283* 0.0804*** 1.73 5.42 Local ADL*SS 0.0056 0.0005 0.26 0.02 R2 0.3925 0.3993 No. Observations 883566 614500

Notes: Samples include individuals that worked 27 weeks or more, and at least 30 hours per week only. Earnings are adjusted for inflation where 1990 is the base year. t-statistics are reported below each coefficient, and are calculated based on standard errors that are corrected for clustering of observations by state and year. State and Time dummies, as well as State and Time interactions, State and SS, and Time and SS interactions are included in order to account for time trends. Level of significance: 10 percent * (t-stat>1.645) 5 percent ** (t-stat>1.960), and 1 percent *** (t-stat>2.576).

CHRISTAFORE & LEGUIZAMON: ANTI-DISCRIMINATION LAWS AND HOMOSEXUALS 229

The results suggest that while there is insufficient evidence to support the claim that employment and earnings of homosexuals in states with anti-discriminatory laws based on sexual orientation are relatively higher or lower than those in states without the law, there seem to be noticeable differences over time, especially for men. Yet, these results must be interpreted carefully. As mentioned before, individuals in same-sex partnerships may be more likely to reveal their sexual orientation in states where they feel more protected, both economically and socially. This would mean that those in same-sex relationships may be more reluctant to show any signs of differences in sexual orientation in states without ADLs. This would decrease the chances of discrimination given that employers who would be inclined to do so would have a hard time identifying these individuals. A lack of discrimination in states without ADLs could then explain why we find no statistical difference between states with and without ADLs. In terms of policy implications, this would suggest that although the law may not immediately change the relative wages of those in same-sex relationships, it may allow them to be open about their orientation without fear of discrimination.

An additional concern is the location preferences of those in same-sex partnerships. Individuals in same-sex partnerships may be more likely to locate in urban areas where wages are higher. Furthermore, they may be more likely to locate in certain regions of the country. In order to address these and other measurement issues, we check the sensitivity of our results by looking at regional variations and urban living. Additionally, we check the robustness of the results by restricting our sample to only working individuals between the ages of 25 and 59. This allows us to discount college students who may work in low-paying jobs for only a brief period. This also allows us to account for possible measurement error arising from college students living with same-sex roommates.17

Tables 5 and 6 show the results for three additional specifications. In columns 1 and 2, we present the results when the sample is restricted to individuals of ages 25 to 59. The results in columns 3 and 4 are found by further restricting the sample to those located in urban areas, and the results in columns 5 and 6 regard the inclusion of regional dummies rather than state dummies. Statistically, the results are robust and very similar to those found above. For the case of employment, we observe some differences when using regional dummies. This suggests that although there might be regional differences, there is also variation within the regions, and once we control for state dummies the results obtained generally change. The same appears to be true for earnings. Of importance is the fact that there is no longer statistical support to the claim that men in same-sex partnerships earn less than other men in urban areas. Additionally, the decrease in the relative wages of women in same-sex partnerships overtime, in states with ADLs, is now statistically significant.

Table 7 highlights the importance of the Difference-in-Difference-in-Difference approach. In it we show the results when using each year of data separately. As shown, some of the results vary from year to year. As expected, the attitudes toward homosexuals evolve and labor market conditions have changed over the 20-year period. Failure to account for these changes can lead us to make an inference based on observations for a particular point in time,

17 Recall that while not immune to coding errors, the U.S. Census does distinguish between roommates and unmarried partners in the “Relationship to the Householder” category.

230 The Review of Regional Studies 43(2,3)

Table 5: Alternative DDD Estimates of the Impact of Anti-discrimination Laws on Employment

Variable Age 25-59 Urban Only Regional Controls Men Women Men Women Men Women Age 0.0361 *** 0.0314 *** 0.0532*** 0.0455 *** 0.0514 *** 0.0465 *** 10.49 *** 10.02 15.83 14.98 12.70 11.72 Age Squared -.0003*** -0.0002 *** -0.0005*** -0.0004 *** -0.0004 *** -0.0004 *** -6.37 -6.03 -12.56 -10.09 -9.88 -7.97 Head of Household 0.1531 *** 0.0183 * 0.1467*** 0.0201 ** 0.1433 *** 0.0175 14.35 1.95 12.47 2.00 7.40 1.57 Presence of Children 0.0552 *** 0.0197 ** 0.0366*** -0.0080 0.0407 *** -0.0066 7.23 2.49 4.20 -0.80 3.73 -0.53 High School Degree 0.1453 *** 0.2070 *** 0.1008*** 0.2132 *** 0.1205 *** 0.2188 *** 9.07 13.27 5.25 12.65 4.72 9.50 Some College 0.1772 *** 0.3016 *** 0.1553*** 0.3194 *** 0.1589 *** 0.3264 *** 9.10 17.73 7.41 16.65 4.45 13.10 College Degree 0.4395 *** 0.5320 *** 0.4183*** 0.5337 *** 0.4252 *** 0.5547 *** 18.88 27.09 17.07 25.33 10.30 19.25 Post College Degree 0.5438 *** 0.6874 *** 0.5201*** 0.6760 *** 0.5303 *** 0.7032 *** 14.68 30.10 13.44 27.91 8.03 19.71 Disabled -0.2321 *** -0.2226 *** -0.2106*** -0.1902 *** -0.2163 *** -0.2064 *** -8.97 -6.06 -6.64 -4.36 -5.36 -3.28 Speaks English 0.0868 *** 0.3660 *** 0.0936*** 0.3627 *** 0.0723 *** 0.3580 *** 4.14 13.04 4.84 13.41 3.19 12.39 White 0.2241 *** 0.2162 *** 0.2123*** 0.2133 *** 0.2308 *** 0.2280 *** 10.83 12.30 9.30 11.27 8.44 10.32 Resides in Metro Area 0.0151 0.0387 *** 0.0118 0.0191 0.88 3.27 0.68 1.06 Unearned Real Income -0.0531 *** -0.0632 *** -0.0528*** -0.0643 *** -0.0534 *** -0.0629 *** -10.88 -14.24 -9.52 -12.68 -6.71 -8.80 In Same Sex Relationship (SS) 0.2188 0.4410 0.7511*** -0.0688 0.1017 -0.0308 0.73 1.49 4.52 -0.38 1.24 -0.24 State ADL 0.0927 *** 0.0440 * 3.44 1.85 ADL*SS 0.0026 -0.1990 -0.0078 -0.0849 0.1442 * -0.0684 0.02 -1.58 -0.07 -0.76 1.67 -0.67 Time Since Passed -0.0022 *** -0.0055 *** -0.0061*** -0.0077 *** -0.0006 0.0019 *** -7.96 -19.02 -23.87 -31.41 -0.29 0.61 Time Since Passed*SS -0.0180 ** 0.0085 -0.0219*** 0.0105 -0.0076 0.0055 -2.11 0.87 -2.62 1.26 -1.51 0.64 Local ADL 0.1540 *** 0.1163 *** 0.1377*** 0.1118 *** 0.1467 *** 0.1100 *** 6.64 3.25 5.12 3.00 4.86 2.77 Local ADL*SS -0.0905 -0.0918 -0.0750 -0.0785 -0.0466 -0.1216 -0.85 -0.74 -0.88 -0.81 -0.46 -1.07 Pseudo R2 0.0657 0.0753 0.0726 0.0824 0.0662 0.0783 No. Observations 1088352 898712 843683 689321 1211581 1005540 Notes: Estimates are the raw coefficients of a probit estimation. They do not represent the marginal effects. t-statistics are reported below each coefficient, and are calculated based on standard errors that are corrected for clustering of observations by state and year. State and Time dummies as well as State and Time interactions, State and SS, and Time and SS interactions are included in order to account for time trends. Level of significance: 10 percent * (t-stat>1.645) 5 percent ** (t-stat>1.960), and 1 percent * (t-stat>2.576).

CHRISTAFORE & LEGUIZAMON: ANTI-DISCRIMINATION LAWS AND HOMOSEXUALS 231

Table 6: Alternative DDD Estimates of the Impact of Anti-discrimination Laws on

Ln(Earnings) – Conditional on Full-time Employment

Variable Age 25-59 Urban Only Regional Controls Men Women Men Women Men Women Hours Worked 0.0137*** 0.0205*** 0.0148*** 0.0209*** 0.0140*** 0.0207*** 28.25 30.18 28.18 27.65 17.35 19.73 Weeks Worked 0.0333*** 0.0343*** 0.0336*** 0.0342*** 0.0331*** 0.0340*** 81.59 83.83 76.87 73.49 48.29 54.18 Age 0.0631*** 0.0438*** 0.0702*** 0.0568*** 0.0683*** 0.0539*** 49.44 32.71 61.68 49.10 40.88 43.76 Age Squared -0.0006*** -0.0005*** -0.0007*** -0.0006*** -0.0007*** -0.0006*** -40.43 -32.17 -54.63 -45.43 -37.25 -44.68 Head of Household 0.0910*** 0.0246*** 0.0885*** 0.0269*** 0.0884*** 0.0228*** 26.86 8.93 22.42 9.28 17.37 6.14 Presence of Children 0.0611*** -0.0208*** 0.0590*** -0.0206*** 0.0588*** -0.0188*** 18.58 -5.49 15.70 -4.97 15.30 -3.19 High School Degree 0.1221*** 0.0734*** 0.1113*** 0.0806*** 0.1177*** 0.0697*** 14.07 4.86 11.17 4.74 8.99 2.88 Some College 0.2324*** 0.2194*** 0.2273*** 0.2230*** 0.2273*** 0.2116*** 20.88 13.55 17.38 12.32 14.04 8.31 College Degree 0.4965*** 0.5083*** 0.5024*** 0.5123*** 0.4942*** 0.5031*** 32.91 26.26 29.83 23.65 24.22 18.49 Post College Degree 0.7471*** 0.7727*** 0.7561*** 0.7739*** 0.7538*** 0.7754*** 31.02 32.06 28.60 29.26 21.20 21.27 Disabled -0.1035*** -0.0790*** -0.1032*** -0.0828*** -0.1019*** -0.0787*** -14.99 -11.31 -13.88 -10.56 -8.50 -6.57 Speaks English 0.3587*** 0.3580*** 0.3580*** 0.3665*** 0.3486*** 0.3388*** 20.25 16.47 21.16 17.95 17.93 13.57 White 0.1810*** 0.0797*** 0.1819*** 0.0755*** 0.1688*** 0.0662*** 29.58 17.00 27.34 16.32 23.87 8.93 Resides in Metro Area 0.1319*** 0.1592*** 0.1403*** 0.1731*** 21.26 23.93 17.35 27.69 Unearned Real Income 0.0153*** 0.0096*** 0.0164*** 0.0103*** 0.0146*** 0.0092*** 26.47 17.01 28.68 19.28 19.26 10.39 In Same Sex Relationship (SS) -0.2660* 0.0793 -0.1927 0.0013 -0.2573*** -0.0494 -1.82 1.18 -1.34 0.01 -7.12 -1.03 State ADL 0.0641** 0.0724*** 2.38 2.95 ADL*SS 0.0015 -0.0208 0.0191 -0.0178 0.0051 0.0383* 0.05 -0.61 0.60 -0.60 0.11 1.89 Time Since Passed -0.0001 0.0034*** -0.0029*** 0.0019*** 0.0017 0.0032 -1.27 32.08 -45.32 21.56 0.71 1.39 Time Since Passed*SS 0.0061*** -0.0033 0.0067*** -0.0065** 0.0024 -0.0036*** 2.99 -1.04 2.89 -2.41 0.72 -2.88 Local ADL 0.0297* 0.0818*** 0.0212 0.0767*** 0.0339* 0.0755*** 1.77 5.25 1.26 5.08 1.88 4.61 Local ADL*SS -0.0063 -0.0107 0.0092 -0.0032 -0.0086 0.0107 -0.25 -0.33 0.40 -0.11 -0.30 0.65 R2 0.3759 0.3817 0.4011 0.3900 0.3871 0.3935 No. Observations 798506 549734 621172 429344 883566 614500 Notes: Samples include individuals that worker 27 weeks or more, and at least 30 hours per week only. Earnings are adjusted for inflation where 1990 is the base year. t-statistics are reported below each coefficient, and are calculated based on standard errors that are corrected for clustering of observations by state and year. State and Time dummies, as well as State and Time interactions, State and SS, and Time and SS interactions are included in order to account for time trends. Level of significance: 10 percent *(t-stat>1.645) 5 percent **(t-stat>1.960), and 1 percent ***(t-stat>2.576).

232 The Review of Regional Studies 43(2,3)

Table 7: Inter-decennial Estimates of the Impact of Anti-discrimination Laws on Ln(Earnings) –Conditional on Full-time Employment

Variable 1990 Only 2000 Only 2010 Only Men Women Men Women Men Women Hours Worked 0.0090*** 0.0147 *** 0.0139*** 0.0199 *** 0.0183*** 0.0255 *** 18.92 13.51 57.66 48.67 71.25 60.43 Weeks Worked 0.0330*** 0.0346 *** 0.0296*** 0.0310 *** 0.0367*** 0.0373 *** 67.69 59.81 51.58 90.03 74.62 44.77 Age 0.0763*** 0.0556 *** 0.0608*** 0.0519 *** 0.0734*** 0.0617 *** 57.43 30.81 39.87 29.09 54.58 39.16 Age Squared -0.0008*** -0.0006 *** -0.0006*** -0.0006 *** -0.0007*** -0.0006 *** -54.28 -30.76 -30.80 -26.17 -44.94 -34.26 Head of Household 0.1016*** 0.0352 *** 0.1010*** 0.0271 *** 0.0684*** 0.0234 *** 12.09 7.67 25.73 9.62 17.38 5.08 Presence of Children 0.0468*** -0.0520 *** 0.0498*** -0.0206 *** 0.0707*** 0.0012 9.20 -7.85 10.80 -5.72 12.35 0.33 High School Degree 0.0735*** -0.0266 *** 0.1663*** 0.1727 *** 0.1725*** 0.1852 *** 9.72 -3.72 21.49 17.96 26.09 15.38 Some College 0.1615*** 0.1028 *** 0.2703*** 0.3151 *** 0.3026*** 0.3293 *** 22.28 9.79 24.80 26.45 29.08 23.00 College Degree 0.4160*** 0.3886 *** 0.5106*** 0.5776 *** 0.5949*** 0.6444 *** 31.95 38.03 30.15 37.06 39.54 31.98 Post College Degree 0.5647*** 0.5658 *** 0.7642*** 0.8327 *** 0.9292*** 0.9668 *** 36.32 36.26 39.00 49.59 55.53 59.12 Disabled -0.1776*** -0.1222 *** -0.0568*** -0.0383 *** -0.1529*** -0.1376 *** -25.89 -10.52 -13.05 -7.03 -33.31 -16.42 Speaks English 0.3325*** 0.2876 *** 0.2709*** 0.2345 *** 0.3436*** 0.3464 *** 16.25 8.39 13.36 7.99 19.89 14.76 White 0.1891*** 0.0627 *** 0.1629*** 0.0599 *** 0.1683*** 0.0678 *** 18.59 4.87 21.30 8.57 16.51 6.68 Resides in Metro Area 0.1733*** 0.1916 *** 0.1526*** 0.1811 *** 0.1307*** 0.1743 *** 8.78 8.74 13.11 15.24 10.50 14.23 Unearned Real Income 0.0181*** 0.0151 *** 0.0166*** 0.0094 *** 0.0124*** 0.0063 *** 29.47 17.49 19.05 15.95 11.40 5.28 In Same Sex Relationship (SS) -0.1386*** -0.0204 -0.1006*** -0.0018 -0.1305*** 0.0316 * -3.71 -0.76 -5.50 -0.12 -4.42 1.70 State ADL -0.2265*** -0.3239 *** 0.1280*** 0.1561 *** 0.0799*** 0.0956 *** -14.53 -22.68 3.31 2.89 3.23 3.83 ADL*SS -0.0237 -0.5349 *** 0.1193 0.1205 *** 0.0748 0.0096 -0.56 -16.84 1.60 3.45 1.28 0.34 Time Since Passed 0.0248*** 0.0348 *** -0.0034 -0.0033 0.0028 0.0041 33.99 26.94 -0.78 -0.47 1.21 1.52 Time Since Passed*SS -0.0136*** 0.0358 *** -0.0052 -0.0153 *** 0.0010 -0.0033 ** -16.21 32.50 -1.48 -3.46 0.24 -2.28 Local ADL 0.0739*** 0.1255 *** 0.0873** 0.1195 *** -0.0054 0.0361 * 4.07 5.79 2.44 3.23 -0.20 1.92 Local ADL*SS -0.0421 0.0022 0.0335 0.0036 0.0916* -0.0051 -0.88 0.03 0.67 0.10 1.96 -0.13 R2 0.3650 0.3388 0.3672 0.3829 0.4264 0.4281 No. Observations 270823 173712 302641 213234 310102 227554

Notes: Samples include individuals that worker 27 weeks or more, and at least 30 hours per week only. Earnings are adjusted for inflation where 1990 is the base year.t-statistics are reported below each coefficient, and are calculated based on standard errors that are corrected for clustering of observations by state and year. State and Time dummies, as well as State and Time interactions, State and SS, and Time and SS interactions are included in order to account for time trends. Level of significance: 10 percent *(t-stat>1.645), 5 percent **(t-stat>1.960), and 1 percent ***(t-stat>2.576).

CHRISTAFORE & LEGUIZAMON: ANTI-DISCRIMINATION LAWS AND HOMOSEXUALS 233

with particular social, political, and economic environments. For instance, studies that only account for a single period (1990 or 2000) may conclude that the law enables changes in the relative earnings of women in same-sex partnerships, when in fact the earnings changes may be due to other factors that are captured by the time trends of the DDD approach.18

In all specifications, ADLs implemented at the local level are associated with increases in overall earnings. Yet, the results show no statistical evidence supporting a difference in the relative earnings of individuals in same-sex partnerships (Local ADL*SS). This suggests that either ADLs are implemented in localities where local labor markets are particularly strong, or that there is little evidence of discrimination in localities without laws due to the reluctance of these individuals to reveal their sexual orientation. Overall, there is insufficient evidence to support the hypothesis that the relative earnings of homosexuals (compared to heterosexuals) are lower in ADL places.

Finally, we look at the relative earnings of homosexuals by using two different control groups (married and unmarried heterosexual couples separately). The literature has found wage differentials between married and unmarried heterosexuals. Although we lack information on the potential marriage outcomes for individuals in same-sex partnerships, we analyze whether the results above can be partly driven by differences in marital status.

We find no statistical difference when we compare the earnings of homosexual and heterosexual unmarried men. Although determining the correct comparison group is difficult, our results suggest that part of the difference in annual earnings found in previous studies of gay men may be partly due to the marriage premium (see Table 8). In terms of the ADL, we find that it increases the relative earnings of gay men to both unmarried and married heterosexual men over time. These results suggest two possibilities for gay men in particular. If men in same-sex partnerships are more similar in unobservable characteristics to married heterosexuals, anti- discriminatory laws are helping to close the earnings gap over time. This is desirable if the source of the marriage premium is firms that discriminate by using marriage as a signal of heterosexuality. However, the labor market may be distorted by the ADL if marriage is used as a signal of productivity by firms. In this case, a desirable labor market policy would be to allow homosexuals to marry, which would enable them to send the same productivity signals as heterosexuals.

If unmarried but coupled heterosexuals are the appropriate reference group, the law may have acted to unnecessarily increase the wages of gay men (unnecessary because all other things equal, we do not observe a statistically significant gap between these groups). If instead it is believed that gay men are more similar to married heterosexual men, the passage of an ADL may act as intended. However, if gay men are deemed more similar to unmarried heterosexual men, the law creates distortionary labor market effects. These results highlight the need to identify the true source of income disparities. This implies being able to find out what share of individuals in same-sex partnerships would marry if allowed. There seems to be little difference from enactment of the laws on labor market outcomes between homosexual women and their heterosexual married and unmarried equivalents.

18 As mentioned before, there are states with ADLs that cover public-sector employees, but none that cover private-sector ones. ADLs still may affect the private sector since gays have an alternative between private and public sector employment when they face discrimination in the one but not the other. We run additional specifications including controls for public-sector ADLs. The results are qualitatively similar and are available from the authors upon request.

234 The Review of Regional Studies 43(2,3)

Table 8: DDD Group Estimates of the Impact of Anti-discrimination Laws on Ln(Earnings) – Conditional on Full-Time Employment

Variable Married Individuals

as Reference Cohabitating Heterosexuals

as Reference Men Women Men Women Hours Worked 0.0136*** 0.0207*** 0.0159*** 0.0203*** 27.70 28.66 35.83 27.26 Weeks Worked 0.0327*** 0.0339*** 0.0330*** 0.0340*** 82.58 83.17 71.75 55.28 Age 0.0674*** 0.0485*** 0.0659*** 0.0643*** 66.41 46.79 33.13 34.95 Age Squared -0.0007*** -0.0005*** -0.0007*** -0.0007*** -58.34 -42.39 -26.92 -28.47 Head of Household 0.0758*** 0.0289*** 0.1239*** 0.0561*** 24.61 9.80 24.06 10.75 Presence of Children 0.0371*** -0.0292*** -0.0258*** -0.0683*** 11.56 -7.68 -3.83 -9.58 High School Degree 0.1188*** 0.0663*** 0.1161*** 0.1116*** 13.59 4.56 18.42 8.83 Some College 0.2270*** 0.2062*** 0.1897*** 0.2279*** 20.42 13.10 19.92 16.07 College Degree 0.4907*** 0.4931*** 0.4239*** 0.4950*** 33.48 26.26 29.34 29.32 Post College Degree 0.7424*** 0.7595*** 0.6873*** 0.7349*** 31.44 32.73 24.73 29.97 Disabled -0.1012*** -0.0767*** -0.0908*** -0.0908*** -15.40 -11.07 -7.98 -8.32 Speaks English 0.3627*** 0.3501*** 0.2692*** 0.3492*** 21.02 16.38 11.19 11.98 White 0.1798*** 0.0745*** 0.1381*** 0.0718*** 28.98 15.89 20.97 10.54 Resides in Metro Area 0.1334*** 0.1590*** 0.0934*** 0.1389*** 21.99 24.07 11.47 16.91 Unearned Real Income 0.0143*** 0.0101*** 0.0111*** 0.0042*** 25.31 18.57 9.25 4.57 In Same Sex Relationship (SS) -0.2577** 0.0067 -0.1050 0.0903 -2.24 0.12 -0.89 1.12 ADL*SS 0.0084 -0.0145 0.0327 0.0038 0.28 -0.49 1.04 0.11 Time Since Passed -0.0013*** 0.0032*** 0.0050*** 0.0024*** -17.00 34.36 24.70 9.63 Time Since Passed*SS 0.0078*** -0.0039 0.0050** -0.0046 4.00 -1.45 2.27 -1.59 Local ADL 0.0300* 0.0839*** 0.0457** 0.0757*** 1.91 5.92 2.05 4.07 Local ADL*SS 0.0031 -0.0026 0.0032 0.0094 0.14 -0.08 0.14 0.31 R2 0.3738 0.3915 0.4029 0.4513 No. Observations 808603 548931 74963 65569

Notes: Samples include individuals that worked 27 weeks or more, and at least 30 hours per week only. Earnings are adjusted inflation where 1990 is the base year. t-statistics are reported below each coefficient, and are calculated based on standard errors that are corrected for clustering of observations by state and year. State and Time dummies, as well as State and Time interactions, State and SS, and Time and SS interactions are included in order to account for time trends. Level of significance: 10 percent * (t-stat>1.645), 5 percent ** (t-stat>1.960), and 1 percent *** (t-stat>2.576).

CHRISTAFORE & LEGUIZAMON: ANTI-DISCRIMINATION LAWS AND HOMOSEXUALS 235

5. CONCLUSION

The influence, and potential influence, of anti-discriminatory laws (ADLs) on the basis of sexual orientation have a rising consideration in the literature. Though other studies have analyzed the effect of these laws in the labor market, we are the first to apply a difference-in- difference-in-difference approach. We find evidence of relative lower earnings for gay men with evidence that the difference may be partly driven by the marriage premium. Relative to married heterosexual men, gay men earn less and, over time, enacted ADLs appear to reduce this gap. Relative to unmarried but coupled heterosexual men, gay men experience similar levels of earnings. Both groups, gay men and unmarried heterosexual men, experience an increase in wages over time after passage of an ADL.

We find do not find evidence that women in same-sex partnerships experience different earnings than their unmarried, coupled and married heterosexual female counterparts. The effect of the law is to increase overall female earnings over time, but the law does not have a statistically significant impact on the relative wages over time. We employ a DDD approach that controls for trends in states and other local labor markets. In addition to allowing for different possible reference groups, we specify the impact of the laws in the state of work rather than the state of residence as previous research has done.

Our results highlight two important policy issues. Statistically, we find mixed evidence with regards to the efficacy of the law to help equalize the wages of gays and lesbians with respect to their heterosexual counterparts. Although we find no statistical differences in the relative wages of homosexuals in states with and without ADLs, we observe statistical differences over time. The lack of differences, however, may be due to the lack of discrimination in states with ADLs. Our results suggest that the observed equalization of earnings may not be due to increases in earnings of those starting with lower wages. Indeed, total earnings those who start with higher wages decrease as a result of the law. Furthermore, the marriage premium seems to contribute to the wage gap between homosexual and heterosexual men. The data, however, do not allow us to identify the source of this contribution. It may be due to unobserved differences in productivity, signaling of sexual orientation, or signaling of productivity. This raises concerns about the efficacy of ADLs in eliminating differences in labor market outcomes between homosexuals and heterosexuals. Policymakers may want to consider alternative solutions.

The increase number of states legalizing same-sex marriages should help us find out whether the differences observed today are indeed due to discrimination or marriage. If marriage is a productivity signal, the ability of only some groups to signal appropriately may result in labor market adjustments. If, instead, marriage is simply a characteristic of productive individuals, we may not see changes in relative earnings, granting opportunities for future research on this subject. REFERENCES

Acemoglu, Daron and Joshua D. Angrist. (2001) “Consequences of Employment Protection? The Case of the Americans with Disabilities Act,” Journal of Political Economy, 109, 915– 957.

236 The Review of Regional Studies 43(2,3)

Ahmed, Ali M. and Mats Hammarstedt. (2010) “Sexual Orientation and Earnings: A Register Data Based Approach to Identify Homosexuals,” Journal of Population Economics, 23, 835–849.

Ahmed, Ali M., Lina Andersson, and Mats Hammarstedt. (2013) “Are Gay Men and Lesbian Women Discriminated Against in the Hiring Process?,” Southern Economic Journal, 79, 565–585.

Allegretto, Sylvia A. and Michelle M. Arthur. (2001) “An Empirical Analysis of Homosexual/Heterosexual Male Earnings Differentials: Unmarried and Unequal?,” Industrial and Labor Relations Review, 54, 631–646.

Arabsheibani, G. Reza, Alan Martin, and Jonathan Wadsworth. (2005) “Gay Pay in the UK,” Economica, 72, 333–347.

Badgett, M.V. Lee. (1995) “The Wage Effects of Sexual Orientation Discrimination,” Industrial and Labor Relations Review, 48, 726–739.

Badgett, M.V. Lee, Honing Lau, Brad Sears, and Deborah Ho. (2007) “Bias in the Workplace: Consistent Evidence of Sexual Orientation and Gender Identity Discrimination,” The Williams Institute: UCLA School of Law.

Becker, Gary. (1971) The Economics of Discrimination, 2nd ed. University of Chicago Press: Chicago.

_____. (1991) A Treatise on the Family, enlarged ed. Harvard University Press: Cambridge, MA.

Beegle, Kathleen and Wendy A. Stock. (2003) “The Labor Market Effects of Disability Discrimination Laws,” Journal of Human Resources, 38, 806–859.

Berg, Nathan and Donald Lien. (2002) “Measuring the Effect of Sexual Orientation on Income: Evidence of Discrimination?,” Contemporary Economic Policy, 20, 394–414.

Bertrand, Marianne, Esther Dufflo, and Sendhil Mullainathan. (2004) “How Much Should We Trust Differences-in-Differences Estimates?,” Quarterly Journal of Economics, 119, 249–275.

Besley, Timothy and Anne Case. (2000) “Unnatural Experiments? Estimating the Incidence of Endogenous Policies,” Economic Journal, 110, 672–694.

Black, Dan A. (1995) “Discrimination in an Equilibrium Search Model,” Journal of Labor Economics, 13, 309–334.

Black, Dan A., Hoda R. Makar, Seth G. Sanders, and Lowell J. Taylor. (2003) “The Earnings Effects of Sexual Orientation,” Industrial and Labor Relations Review, 56, 446–469.

Black, Dan A., Gary J. Gates, Seth G. Sanders, and Lowell J. Taylor. (2007a) “The Measurement of Same-Sex Unmarried Partner Couples in the 2000 U.S. Census,” California Center for Population Research On-Line Working Paper Series.

Black, Dan A., Seth G. Sanders, and Lowell J. Taylor. (2007b) “The Economics of Lesbian and Gay Families,” Journal of Economic Perspectives, 21(2), 53–70.

Blanford, John M. (2003) “The Nexus of Sexual Orientation and Gender in the Determination of Earnings,” Industrial and Labor Relations Review, 56, 622–642.

CHRISTAFORE & LEGUIZAMON: ANTI-DISCRIMINATION LAWS AND HOMOSEXUALS 237

Brown, Charles and Mary Corcoran. (1997) “Sex-Based Differences in School Content and Male/Female Wage Gap,” Journal of Labor Economics, 15, 431–465.

Carpenter, Chris. (2005) “Self-Reported Sexual Orientation and Earnings: Evidence from California,” Industrial and Labor Relations Review, 58, 258–272.

_____. (2007) “Revisiting the Income Penalty for Behaviorally Gay Men: Evidence from NHANES III,” Labour Economics, 14, 24–34.

Clain, Suzanne H. and Karen Leppel. (2001) “An Investigation into Sexual Orientation Discrimination as an Explanation for Wage Differences,” Applied Economics, 33, 37–47.

Collins, Williams. (2003) “The Labor Market Impact of State-Level Anti-Discrimination Laws, 1940-1960,” Industrial and Labor Relations Review, 56, 244–272.

Elmslie, Bruce and Edinalo Tebaldi. (2007) “Sexual Orientation and Labor Market Discrimination,” Journal of Labor Research, 28, 436–453.

Frank, Jeff. (2006) “Gay Glass Ceilings,” Economica, 73, 485–508.

Gates, Gary J. (2009) “The Impact of Sexual Orientation Anti-discrimination Policies on the Wages of Lesbians and Gay Men,” California Center for Population Research On-Line Working Paper Series. Last accessed in March 014 at http://papers.ccpr.ucla.edu/papers/PWP-CCPR-2009-010/PWP-CCPR-2009-010.pdf

_____. (2011) “How Many People are Lesbian, Gay, Bisexual, and Transgender?,” Williams Institute, UCLA School of Law. Last accessed in March 2014 at https://escholarship.org/uc/item/09h684x2.

Gates, Gary J. and Michael Steinberger. (2010) “Same-Sex Unmarried Partner Couples in the American Community Survey: The Role of Misreporting, Miscoding, and Misallocation,” last accessed in March 2014 at http://economics-files.pomona.edu/steinberger/research/Gates_Steinberger_ACS_Miscod e_May2010.pdf.

Herek, Gregory M. (2009) “Hate Crimes and Stigma-related Experiences Among Sexual Minority Adults In the United States: Prevalence Estimates From a National Probability Sample,” Journal of Interpersonal Violence, 24, 54–74.

Human Rights Campaign. (2009) State Laws and Policies, last accessed in October 2009 at http://www.hrc.org/documents/Employment_Laws_and_Policies.pdf.

Klawitter, Marieka 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, 658–686.

Klawitter, Marieka. (2011) “Multilevel Analysis of the Effects of Antidiscrimination Policies on Earnings by Sexual Orientation,” Journal of Policy Analysis and Management, 30, 334– 358.

Neumark, David and Wendy Stock. (2001) “The Effects of Race and Sex Discrimination Laws,” NBER Working Paper 8215.

Plugg, Erik and Peter Berkhout. (2004) “Effects of Sexual Preferences on Earnings in the Netherlands,” Journal of Population Economics, 17, 117–131.

238 The Review of Regional Studies 43(2,3)

Rimmerman, Craig A., Kenneth D. Wald, and Clyde Wilcox, eds. (2000) The Politics of Gay Rights. University of Chicago Press: Chicago.

Rosen, Asa. (2003) “Search, Bargaining, and Employer Discrimination,” Journal of Labor Economics, 21, 807–829.

Rubenstein, William B. (2001) “Do Gay Rights Laws Matter?: An Empirical Assessment,” Southern California Law Review, 75, 65–120.

Ruggles, Steven, Matthew Sobek, Trent Alexander, Catherine A. Fitch, Ronald Goeken, Patricia Kelly Hall, Miriam King, and Chad Ronnander. (2009) “Integrated Public Use Microdata Series: Version 4.0 [Machine-readable database],” Minneapolis: Minnesota Population Center [producer and distributor].

Tilcsik, Andras. (2011) “Pride and Prejudice: Employment Discrimination against Openly Gay Men in the United States,” American Journal of Sociology, 117, 586–626.

Copyright of Review of Regional Studies is the property of Southern Regional Science Association Inc. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.