The march toward marriage equality: Reexamining the diffusion of same-sex marriage among states

Abstract

The issue of same-sex marriage and lesbian, gay, bisexual, and transgender equality has received considerable attention from policy scholars. This is unsurprising given the issue is one of the defining social policy battles of the last decade. State governments at the forefront of this battle have responded by proposing a multitude of lesbian, gay, bisexual, and transgender-related policies. In this study, we comparatively assess the diffusion of two of these policies across US states: the legal recognition of same-sex marriages and state constitutional amendments defining marriage as exclusively between one man and one woman. While previous studies have examined the diffusion of same-sex marriage bans across states, none have offered a comparative examination of how both sides of this contentious issue have advanced their policy preferences alongside each other. Using event history analysis, we analyze a unique set of covariates to test two diffusion hypotheses: learning and imitation. We find that for both policies, policy learning is the primary mechanism occurring, suggesting that policymakers learn from one another for the same policy area, even if the policies have different motives or objectives. However, the effect of learning is more prominent for anti-gay policies, suggesting there are differences between policies.

Keywords

Policy diffusion policy adoption gay marriage diffusion same-sex marriage LGBT policy gay marriage bans

Introduction

The issue of same-sex marriage and lesbian, gay, bisexual, and transgender (LGBT) equality has received considerable attention from policy scholars in the last few decades. This is unsurprising given that these issues have been among the defining social policy battles of the past decade. In response, US states have passed several policies regarding marriage rights for same-sex couples, some that give legal recognition to same-sex marriage, and others that limit same-sex marriage. Given the amount of policies that have been adopted by states in a short time span, it is important to understand the specific factors that contribute to the adoption of these policies across states.

Policy diffusion, or the external factors that help explain multiple jurisdictions adopting the same policy, is one framework that can help explain these adoption patterns. In this study, we examine the diffusion of two related policies across US states: one that legally recognizes gay marriage and another that defines marriage exclusively as a union between one man and one woman at the constitutional level. Using a unique set of covariates, we examine both marriage and amendment laws from 2004–2015 and 1998–2015, respectively. Using event history analysis (EHA), we test two diffusion theories: imitation and learning, and compare these two policies on multiple dimensions to determine specific adoption patterns. We find that both policies exhibit similar patterns and that states are learning from one another when adopting these policies. However, the effect of learning is greater for anti-gay policies, suggesting there are some differences. This adds to the growing body of literature that isolates mechanisms of policy diffusion and unravels the complexities of morality policy. In the next section, we provide a background of same-sex marriage in the US.

Same-sex marriage in the US: A background

While the issue of same-sex marriage has been a hotly contested part of the national dialogue for the past decade, it is far from a new issue. Cohabitating homosexual couples have always been a part of American society, albeit one that was not widely acknowledged and the legal recognition of such unions would not become a serious subject of debate for some time. Even though it was not the first time the issue had been raised, a 1970 court case in the state of Minnesota marks the first legal challenge towards marriage equality with Baker v Nelson. While ultimately unsuccessful, the case along with the battle over the controversial equal rights amendment (ERA) in the same decade brought the issue of same-sex marriage into the spotlight. Some social conservatives used the ERA to push states to recognize homosexual marriages as part of their resistance to the amendment, and the threat of such a possibility provided an impetus for several states to proactively ban the practice by the end of the decade (Coontz, 2004).

It would take another 20 years or so before the next legal challenge on the subject of same-sex marriage. While the ultimate result of a similar 1993 court case in Hawaii, Baher v Miike, was also unsuccessful, the differing response and initial successes of the effort showed that times were changing compared to a more dismissive reaction to Baker. Even though little had changed on the marriage front, LGBT activists had managed to achieve some success in terms of repealing sodomy laws in several states, as well as expanding non-discrimination protections in many parts of the country to include gays and lesbians. In terms of gay marriage though, the years after Baher saw several more states ban the practice, eventually prompting federal action in the form of the defense of marriage act (DOMA) in 1996 (Hume, 2013). The final years of the decade also saw the first state constitutional amendments defining marriage as exclusively between one man and one woman.

The turn of the millennium saw these issues reach a new level of prominence in the political climate following a 2003 Massachusetts Supreme Court decision making the state the first in the union to recognize same-sex marriages. This once again brought the issue to the forefront of American politics and played a pivotal role in the 2004 Presidential elections (Lax and Phillips, 2009). This time period also saw the landmark Supreme Court decision in Lawerence v Texas, ruling state sodomy laws unconstitutional. The rest of the decade was much of the same, especially as President Bush's popularity continued to fall and the country swung back towards more progressive politics. The 2006 Midterm Elections saw the Democratic Party come into power in both houses of Congress, and with the election of President Obama two years later the party took control of the White House as well. During this period of time, public support for same-sex marriage continued to increase overall, and more states enacted marriage equality. It is worth noting that this time period also saw a brief rise in an alternative policy position in the form of either civil unions or domestic partnerships. These policies were seen as a way to extend the benefits of marriage to same-sex couples, functionally being different from marriage in name only. While a few states did enact such policies, they ultimately proved to be a short-lived solution that failed to satisfy either side of the issue. Conservatives were not fully swayed by the technical difference and still protested the growing acceptance of LGBT lifestyles in society overall, while LGBT advocates were wary of civil unions becoming a “separate, but equal” situation that could easily be co-opted (Hume, 2013). To combat rising support for marriage equality, the decade also saw more action taken to introduce constitutional definitions of marriage at the state level as a more permanent form of ban, with the most significant victory coming in November 2009 when 11 states ratified such proposals.

Finally, the current decade has shown thus far that the LGBT equality movement is only gaining momentum. Strong signals from the Obama Administration and policy victories including the repeal of DOMA and “Don't Ask, Don't Tell” as well as the passage of the Matthew Shepherd Act which extended hate crime protections to cover sexual orientation and gender identity translated into the spread of same-sex marriage recognition to several states. Prior to the landmark decision in Obergefell v Hodges, 36 states (as well as the nation's capital) already recognized same-sex unions to some degree, leaving only those who had constitutional marriage definition amendments.

Although this was a monumental victory for LGBT equality, those who oppose it have not yet given up the fight. While marriage equality is now a foregone conclusion, social conservatives have started mobilizing for “religious freedom” bills that would exempt those who are acting on religious grounds from discrimination protections. Attempts to enact such policies have started in several states in the wake of Obergefell v Hodges, but at the present time it is too early to assess their popularity or effectiveness.

While morality policy is often tricky, the issue of same-sex marriage in the US is one that also raises questions about who actually has the power to change it. While the 14th Amendment guarantees equal protection under the law, legislating common morality is directly named in the Constitution as a power reserved to the states. In practice, we see that this issue tends to be most commonly dealt with at the state level with national action only coming after public opinion has shifted solidly one way. In trying to understand what factors contribute to the adoption of either same-sex marriage or a method to ban it, focusing on the states allows us to see these factors at multiple stages of the process rather than just one single federal decision and the context in which it occurred.

Policy diffusion

In the last few decades, policy diffusion, or how policies spread among governments, (Butz et al., 2015; Shipan and Volden, 2008) has been examined in many political science subfields, including American Politics, Comparative Politics, and International Relations (Graham et al., 2012). In most subfields, scholars are interested in the specific mechanisms driving this diffusion (Shipan and Volden, 2012). Two of the main theories explaining the diffusion process are policy imitation (Shipan and Volden, 2008; Volden, 2006) and policy learning (Berry and Baybeck, 2005; Mooney, 2001; Valente, 1995). In this study, we test both of these theories to determine which one best explains the US interstate adoption of same-sex marriage policies.¹

Policy learning

Policy learning, an important component of the policymaking process (Radaelli, 1995), refers to policymakers learning from one another about policy ideas (Berry and Baybeck, 2005; Mooney, 2001; Valente, 1995). More specifically, learning is policymakers observing the successes and consequences of a policy in other jurisdictions, assessing the outcomes of the policies (Nicholson-Crotty and Carley, 2015), and then deciding on whether or not to adopt the policy. Policymaking can be a simplistic learning process to one that is highly complex (Blyth, 1997). When a component of this learning is considering the policies of others, such as neighboring lawmakers, this can be defined as a diffusion mechanism (Gilardi, 2010).

While policy learning has been observed in American Politics (Butz et al., 2015; Shipan and Volden, 2008) and has been used to explain international LGBT movements (Ayoub, 2014; Fernández and Lutter, 2013; Hall, 1993; Kollman, 2007; Patternote and Kollman, 2013), it originates from the theory of social learning (Glick 1991).² Social learning theory helps to explain how information is transferred outside of the context of individuals (Rogers, 2010). For example, individuals observe patterns of behavior from others and are more likely to engage in the behavior themselves with some adaptation (Rogers, 2010). This theory contends that communication among individuals does not have to occur, and that individuals learn from observation. For policy learning in diffusion to occur, scholars have observed that that policymakers learn from those closer in proximity, such as a neighboring jurisdiction (Mooney, 2001). Therefore, we believe that adopting a same-sex marriage policy makes other neighboring states more likely to adopt.

Learning hypothesis: A state will be more likely to adopt either a marriage law or amendment as its neighbors adopt these policies.

Conditional learning

With social policies, the effect of learning may be less evident. A cost-saving mechanism or one that improves efficiency has more tangible benefits that can be directly observed, while social policies have a less clear “correct” answer. Furthermore, some have noted that geographic proximity may measure multiple mechanisms, so it is important to use in conjunction with other indicators (Maggetti and Gilardi, 2016). Therefore, we also test an additional learning hypothesis: conditional learning. Conditional diffusion posits that some jurisdictions, particularly larger or smaller jurisdictions, are less susceptible or responsive to the policies of neighboring jurisdictions (Shipan and Volden, 2008). Therefore, to fully explore the learning dynamics we propose that larger urban jurisdictions may be more capable of learning from their neighbors.

Conditional hypothesis: Larger states will be more capable of learning than their smaller counterparts.

Policy imitation

While it seems reasonable to assume that states look to their neighbors when considering gay marriage adoptions and anti-gay amendments, it is less clear what states are learning in regards to the successes and failure of other states. Therefore, imitation may be an alternative explanation to policy learning. Diffusion via imitation is similar to policy learning; but instead of closely examining neighboring policies and learning from their successes and failures, governments adopt policies by mimicking neighboring governments (Shipan and Volden, 2008). Put differently, governments adopt policies hoping that imitation will make their jurisdictions more attractive like a more cosmopolitan or wealthy neighbor (Shipan and Volden, 2008).

Walker (1969) produced an empirical study to model state policy diffusion in political science and observed that diffusion is a complex process and many factors can influence a state's adoption of an innovation. However, he specifically acknowledges the regional effect, or the likelihood of a government adopting an innovation if it has been previously adopted by a government that can serve as a “legitimate comparison,” such as a neighboring state. More specifically, policymakers may look to others they consider leaders (Walker, 1969) or to governments that are larger (Shipan and Volden, 2008). Therefore, imitation has been conceptualized as the mimicking policies of a jurisdiction's larger neighbor (Mitchell and Stewart, 2014; Shipan and Volden, 2008)

Imitation hypothesis: A state will be more likely to adopt either a marriage or amendment law if its largest neighbor has previously adopted.

Before outlining our methodology to test the latter hypotheses, we will first examine the patterns of diffusion these policies exhibit.

Examining the shape of diffusion

Gay marriage policies are unique when compared to other policies such as smoking regulations, welfare benefits, and lotteries among others. They have little direct impact on the vast majority of the country, are largely symbolic beyond the personal level, and contain a dimension of morality. Morality policies have been known to be more salient and more “technically simplistic” than other policies, with the latter meaning that there is a clear winner and a clear loser (Mooney and Lee, 1999: 768). Therefore, morality policy can be expected to diffuse differently. Studying morality policies proves to be beneficial, because these policies move individuals “out of the realm of facts and reason where social scientists and especially policy scholars feel comfortable, and into the realm of values” (Mooney, 2001: 5). While we have a relatively good idea as to what happened and the context of both of the policies discussed in the previous section, we often see them examined in a vacuum. These policies, including the older legislative bans on same-sex marriage, have shown a history of being linked, and it sometimes has seemed as though one of the biggest forces that spurred the enactment of one solution was the advancement of the opposition (Hume, 2013). While this effect has been noted in previous studies, there has yet to be a cohesive effort to establish it. A gap exists in our understanding of what causes an individual state to take a side in this contentious issue, especially in the more recent battles of the past decade. We see patterns similar to those observed with the diffusion of legislative bans throughout the 1990s by scholars like Sarah Soule and Robert Hume, but more recent developments give a chance to see how patterns change when there is a more even balance between opposed forces.

Figure 1 shows the cumulative adoption patterns of the recognition of same-sex marriages and of a constitutional amendment defining marriage between a man and a woman respectively.³ From a preliminary analysis of these figures, we do see that for both policies there is a slow buildup of adoption that picks up steam as more follow suit in line with a standard diffusion model. A complete curve is not realized since diffusion was halted by federal intervention as a result of Obergefell v Hodges. The graphs also show some preliminary support for the idea that the two policies were influenced by the spread of the other, with amendments rapidly adopted between 2003 and 2005, or around when Massachusetts became the first state to recognize same-sex unions. In addition to that, we see recognition of marriage experiencing a surge of growth around the same time period that amendments plateaued.

Figure 1.

Cumulative adoption of marriage and amend.

Methodology

Dependent variables

In this study, we rely on an EHA of two different dependent variables, both binary sequences. The first dependent variable is measured as 0 each year prior to a state enacting a legal recognition of same-sex marriage policy and as 1 the year it was enacted. The second dependent variable is measured as 0 each year prior to an amendment to the state constitution defining marriage as exclusively between one man and one woman was adopted and as 1 the year it was enacted. We rely on the Cox model which tells us the factors that increase or decrease the risk for adoption. The Cox model is most appropriate because each model used in this study does not violate the proportional hazards assumptions. To test proportional hazard violations, we rely on the Grambsch and Therneau (1994) algorithm which tests proportionality using the Schoenfeld residual. The test for each covariate, in addition to the global violation test, was insignificant. We also estimate standard errors clustered at the state level to help alleviate the likelihood of non-independent observations. Additionally, we incorporate a competing risks model to account for the likelihood that our dependent variables impact one another (Brooks, 2008).⁴ A competing risks model poses some difficulty given that both events are occurring at different time points in the initial stages. Therefore, we include a dichotomous variable that indicates whether or not the other competing policy is present in a particular year, and specify it as a competing event in our model.

External determinants

To measure policy learning, we rely on the total number of neighboring states that adopted the respective policy (marriage recognition in the first model and constitutional amendments in the second) in the previous years, which is a commonly used measure in policy diffusion (Berry and Berry, 1990; Mitchell and Stewart, 2014; Mooney and Lee, 1995). This measure has also been used to assess the diffusion of same-sex marriage legislation (Haider-Markel, 2001). Additionally, we also model learning as the proportion of neighboring states, as it is also a commonly used measure (e.g. Neumayer and Plümper, 2015; Plümper and Neumayer, 2010). To measure conditional learning, we rely on an interaction variable between neighboring states that have previously adopted the respective policy and the urban population of that state. To measure imitation, we rely on the Shipan and Volden (2008) measure which is a dichotomous variable that measures whether or not the state's most populous neighbor had already enacted the corresponding policy by that year.

Internal determinants

State ideology variables

As with many political issues, same-sex marriage is one that is divided along ideological lines. Given it has little to do with differences in economic or foreign policy positions, marriage equality has been a major social issue. While an individual's ideology informs personal political opinions, this study is concerned with states. Ideology could be seen as a baseline for which way policymakers would most likely lean, with a caveat being that it is not the only factor influencing the decision. It should be noted though that ideology goes far deeper than just liberal and conservative, and each side of the axis is multifaceted. For this reason, it is important to use multiple variables for state ideology and political culture to encompass this measurement completely.

To measure political culture, we rely on the Sharkansky's measure of political culture (Sharkansky, 1969), which was largely based upon Elazar's (1966) work on political culture (Elazar, 1966). Sharkansky's scale classifies the political culture of states in dimensions of moralist, individualistic, or traditionalist culture. Sharkansky uses a nine-point scale, ranging from one (moralist) to nine (traditional).⁵

Next, we rely on two distinct measures of state ideology. These are from the Berry et al. (2010) study which measures citizen ideology as a function of the interest group scores of both incumbent national legislators from the state, as well as their most supported challenger in the most recent elections, weighted by the relative electoral performance of both candidates (Berry et al., 2010). We also rely on an alternative measure of ideology, which is based on NOMINATE Common Space Indicators. This scale is a compiled profile of a state's congressional delegation using an aggregate of roll-call votes instead of interest group scores.

Political landscape

The issue of same-sex marriage is typically divided along party lines. While there are some on each side like the Log Cabin Republicans who break rank with the mainstream platform, simply seeing the issue as Democrats in support of and Republicans in opposition to same-sex marriage is generally true. Because of this, it is important to understand the political landscape in addition to ideology. A Democratic government in a conservative state, for example, still would likely see some pressure from the national party or want to keep a base of those more liberal than the rest of the state energized. The balance of power between parties is also an important variable to examine for our purposes since a state where parties are more equally represented may be less likely to see elected officials take a bold stance on a controversial issue than one where a party has a more dominant share of the government (Barclay and Fisher, 2003).

We rely on multiple measures of a state's political landscape. First, we rely on the Republican electoral returns, which is the percentage of a state's population that voted for a Republican candidate in the most recent presidential election. Next, we rely on the Republican's percentage share of state legislature, which is the total percentage of seats in both houses of the state's legislature held by a Republican legislator. Finally, we rely on a state's support for same-sex marriage. Taken from Gallup's most recent State of the States report, this measures the percentage of the state's population that holds a favorable opinion toward same-sex marriage.⁶

Demographic factors

Demographic factors can have a significant impact on a state's political reality. Therefore, we include multiple factors that may have an impact on LGBT politics. The first, urban population, is included as a stand-in for a measurement in line with Contact Theory, or the idea that regular interaction between different social groups fosters positive relations. LGBT communities tend to be rather small and concentrated in their own “enclaves” in many major cities (Smart and Klein, 2013). Because of this, we consider it reasonable that those states with higher urban populations would have more people who are regularly exposed to LGBT individuals than more rural states. Race is also a factor that should be considered, with strong support for the idea that both Hispanics and African-Americans are somewhat more opposed to LGBT equality (Sherkat et al., 2010). There have been a few attempts to explain this phenomenon, but it usually comes down to either covariance with lower socioeconomic status or the presence of certain cultural factors in these communities that are less welcoming to gays and lesbians. On that note, socioeconomic status would also need to be considered a factor due to the aforementioned tendency as it tends to affect a wide variety of individual opinions toward various political issues.

Religion is the final demographic factor and would likely be the most important as it has taken a central role in the debate regarding LGBT equality. Religious justifications are quite common among socially conservative viewpoints, and most mainstream religions condemn homosexuality. With marriage specifically, we also see a facet of this movement that challenges what many see as something with a deeply religious connotation.

To measure our demographic factors, urban population measures the percentage of the state's population living in urban areas. Next, per-capita income is the median per-capita income of each state. Additionally, the poverty rate for each state is included. Then, the percentage Black/Hispanic population for each state is measured. Data for these four variables are taken from the most recent US Census. Religion data are taken from Gallup's State of the States series,⁷ which measures the percentage of the state's population that identifies as “deeply religious,” or those that say that religion “plays a significant role in their lives,” and that attend religious services either weekly or almost weekly. While it would be ideal to be able to compare religiosity at the time each policy was enacted, unfortunately these data were not available at the state level. Still, we feel that the available data can serve to illustrate the relative difference between states. Finally, we include a measure that captures a state's regional census designation to control for the potential that specific regions have concentrated views that are more likely to support pro-gay or anti-gay measures.

Institutional factors

An understated effect in policy development is often the set of circumstances under which institutions work. A judge who is subject to re-election, for example, may render a different verdict on a controversial case than one who is safely appointed to a long tenure. This specific example has already been used in studies on diffusion of marriage policy, in both Soule's Going to the Chapel and Hume's Courthouse Democracy. Because these two studies on gay marriage bans bear many similarities to our goals here and found a strong correlation between the method of judicial selection and actions taken on marriage bans, we feel it appropriate to consider it as well. This is not purely an issue that is handled by the courts though, so it also important to account for the effect of how “safe” state legislatures are when it comes to making a tough choice on this contentious issue. Finally, the amount of time that policymakers actually spend in session is something that can vary significantly between states, and it is worth considering that officials who have less time to do the actual work of policymaking may find that there are more pressing issues to deal with or that they can more easily run out the clock on having to make a contentious decision that year.

To measure this effect, we include three institutional factors within each state that may exhibit an influence over the adoption of LGBT policies. First, the length of session is used. This is a dichotomous variable, with states that have a limit on the length of a state legislative session coded as 0, and those with no such limit as 1. Next, the legislator's term length is included. This measures the length of state legislator terms in number of years. Finally, the method of judicial selection is used. This is a binary variable measuring the method by which judges are placed on the state's Supreme Court: 0 denotes a purely appointed position, and 1 represents states that have some form of citizen participation in the process.

Results

Table 1 shows the results of the Cox Proportional Hazards model. The first four columns represent the learning model for the marriage and amendment dependent variables. The next two columns show the learning and imitation variables modeled simultaneously for the marriage and amend dependent variables. The next two columns show the imitation models while the last two columns show the conditional models for learning. According to the learning model Table 1, the number of neighbors variable. Looking at the neighbor model, if a neighboring state adopts a marriage law, a state in question is almost three times more at risk for adopting their own. For the proportion of neighbor's variable, a state is hundreds of times more likely at risk of adoption as the percentage of neighbors adopting increases. The Republican electoral returns variable is also significant. For each percentage increase in the vote for the Republican Party, there is roughly a three percent decreased risk in adoption of same-sex marriage. Figure 2 shows the cumulative hazard rate for the marriage learning model. This depicts the risk for adoption as both time and the percentage of neighbors having adopted increases. This demonstrates that neighboring adopters increases the risk of a state adopting. This is most evident as the number of neighbors adopting reaches 60 percent or above.

Figure 2.

Hazard function for marriage learning model.

Table 1.

Cox proportional hazards model of marriage and amend and learning

	Learning marriage	Learning marriage % neigh.	Learning amend	Learning amend % neigh	Learning and imit. marr.	Learning and imit. amend	Imitation marriage	Imitation amend	Cond. marriage	Cond. amend
Imitation	–	–	–	–	1.74	1.35	7.41*	2.08	–	–
Imitation					(1.36)	(1.01)	(3.49)	(1.28)
Neighbors	2.85*	–	1.67*	–	2.50*	1.62*	–	–	1.15	10.1*
Neighbors	(.460)		(.268)		(.655)	(.323)			(1.08)	(11.44)
% Neighb	–	255.9*	–	5.29*	–	–	–	–	–	–
% Neighb		(212)		(4.40)
Conditionl	–	–	–	–	–	–	–	–	1.01	.973
Conditionl									(.012)	(.016)
Urban	.998	1.00	1.05	1.05	.991	1.06	.983	1.06	.979	1.06
Urban	(.034)	(.033)	(.041)	(.030)	(.035)	(.042)	(.037)	(.043)	(.039)	(.046)
Rep leg.	.368	.258	.007	.010	.272	.008	.309	.009*	.212	.009
Rep leg.	(.960)	(.704)	(.019)	(.025)	(.717)	(.019)	(.939)	(.021)	(.578)	(.025)
Region	1.13	1.10	.400	.402*	1.05	.415	.788	.433	1.24	.431
Region	(.576)	(.548)	(.251)	(.170)	(.546)	(.258)	(.338)	(.259)	(.650)	(.262)
Culture	1.09	1.17	1.12	1.19	1.08	1.16	1.16	1.23	1.14	1.10
Culture	(.278)	(.331)	(.348)	(.289)	(.295)	(.376)	(.353)	(.411)	(.311)	(.346)
Citizen Ideo.	.958	.950	1.07*	1.07*	.959	1.07	.968	1.08*	.952	1.08*
Citizen Ideo.	(.025)	(.024)	(.038)	(.031)	(.026)	(.037)	(.026)	(.037)	(.025)	(.037)
Gov ideo.	1.01	.652	.974	.859	1.01	1.05	1.00	.977	1.02	.973
Gov ideo.	(.014)	(.377)	(.014)	(.773)	(.015)	(.043)	(.014)	(.017)	(.016)	(.013)
Poverty	1.13	1.14	1.00	1.00	1.13	1.00	1.15*	.996	1.14	.989
Poverty	(.084)	(.088)	(.103)	(.073)	(.083)	.111	(.078)	(.111)	(.084)	(.112)
Income	1.00	1.00	.999	.999	1.00	.999	1.00	.999	1.00	.999
Income	(.000)	(.000)	(.000)	(.000)	(.000)	(.000)	(.000)	(.000)	(.000)	(.000)
Black	.005	.001	.175	868922*	.036	769372	5.16	797908	.0004	959372
Black	(.042)	(.010)	(.000)	(5550994)	(.304)	(633650)	(39.8)	(6625885)	(.003)	(7697343)
Hispanic	.963	.408	.253	.205	1.57	.127	1.52	.056	.473	.393
Hispanic	(3.35)	(1.50)	(1.02)	(.577)	(6.0)	(.543)	(5.81)	(.234)	(1.71)	(1.59)
Religion	1.07	1.09	1.00	1.01	1.08	.999	1.05	.989	1.07	.994
Religion	(.070)	(.074)	(.062)	(.055)	(.072)	(.063)	(.067)	(.055)	(.069)	(.059)
Support	1.05	1.07	1.10	1.09	1.06	1.11	1.05	1.11	1.06	1.09
Support	(.0911)	(.086)	(.125)	(.105)	(.093)	(.135)	(.094)	(.142)	(.098)	(.127)
Inst. session	.569	.599	1.80	1.83	.536	2.08	.435	2.55	.603	1.99
Inst. session	(.358)	(.431)	(1.11)	(1.31)	(.357)	(1.46)	(.325)	(1.80)	(.388)	(1.25)
Inst. term	1.04	.900	1.12	1.11	.928	1.23	.600	1.35	1.09	1.17
Inst. term	(.753)	(.707)	(.647)	(.567)	(.650)	(.784)	(.434)	(.769)	(.794)	(.629)
Judicial	1.16	1.89	6.03*	7.09*	1.42	6.07*	2.33	7.90*	1.27	5.77*
Judicial	(1.01)	(1.61)	(3.95)	(6.42)	(1.22)	(3.98)	(1.87)	(4.8)	(1.14)	(3.82)
Rep. elect	.964*	.969*	1.05*	1.05*	.966*	1.05*	.975	1.06*	.964*	1.05*
Rep. elect	(.013)	(.012)	(.020)	(.015)	(.013)	(.021)	(.012)	(.024)	(.013)	(.021)
N	518	518	594	594	518	594	518	594	518	594
Global Ph	.222	.054	.383	.119	.258	.391	.418	.213	–	–
Wald χ²	127.231	169.43	96.17	86.83	121.35	115.42	106	63.98	129.2	105.09
Log Likelihood	−145.6	−143.84	−128.0	−129.83	−145.2	−127.9	−153.7	−131.00	−145.0	−126.6
Prob χ²	.000	.000	.000	.000	.000	.000	.000	.000	.000	.000

p < 0.05, two-tailed test.

Note: Dependent variable is a binary sequence (0 = Prior to adoption, 1 = year policy was adopted). Cox proportional hazards model. Hazard ratios reported; standard errors in parentheses. Models clustered by states. The alpha was calculated for each model (results not reported). Cox proportional hazards assumptions tested for each variable; results not reported.

Turning to the second model in Table 1, the proportion of neighbors' variable is also significant. As the proportion of neighbors having the policy in place increases, the respective state is many times more at risk for adopting the policy themselves. Turning to the learning amend model, a state is 67 percent more likely to adopt an amendment as each additional neighbor adopts, and over 500 times more likely in the proportion of neighbors model. This effect is greater than the previous marriage learning model, which is evidence that the learning effect is greater for anti-gay policies. Additionally, the ideology of a state's citizenry is also significant. For each increase in conservatism among a state's citizens, there is a seven percent increase in risk for adoption for the neighbor model. For the judicial selection variable, a state is many times more likely to adopt this policy if the citizens have a role in the judicial electoral policy. Finally, for the Republican electoral returns variable, a state is roughly at a five percent increased risk for adoption for each additional percentage vote return for the Republican Party. Figure 3 is similar to Figure 1, and shows the cumulative hazard for the amend model, which also shows an increased risk of adoption as the number of neighbors having adopted increases.

Figure 3.

Hazard function for amend learning model.

Next, looking at the learning amend model with the proportion of neighbors measure, as the proportion of a state's adopting neighbors increases a state is many more times likely to adopt. In this model, the region and black variables are also significant, meaning that states in specific regions or states that have more African-Americans a more likely to adopt an amendment that limits gay marriage. For the models that examine learning and imitation simultaneously, we see that learning exhibits much more of an influence. Therefore, in the presence of both variables, we see that the effect of learning is many times more influential than the imitation models. Turning to the marriage imitation model in Table 1, if a state's largest neighbor had adopted a marriage policy, then that state is seven times more likely to adopt. This suggests that states may be imitating the policies of their neighbors if the largest neighbor has passed a policy to allow for gay marriage. In addition, the poverty variable is also statistically significant. For each additional percentage of individuals within a state living in poverty, there is a 15 percent more chance of adopting a marriage law. Figure 4 shows the cumulative hazard function for this model, which shows that if the largest neighbor adopts the policy, there is a sharp increase in risk for adoption.

Figure 4.

Hazard function for marriage imitation model.

Next, for the imitation amend model, the diffusion variable is not statistically significant. However, multiple internal measures are statistically significant, including the Republican's share of a state's legislature, the ideology of a state's citizenry, the judicial selection method, and the Republican electoral returns. Figure 5 shows the cumulative hazard function for this model, which shows similarities to previous models. Finally, the conditional model (labelled cond.) shows the coefficient for the interaction between neighbor × urban population.⁸ According to Table 1, there was little evidence of conditional learning in either model. Finally, for our competing risks model, there was no significant competing event.⁹ Specifically, there was no influence of the amend policy on marriage adoption and vice versa. Overall, the two policies are not homogenous with regard to the external diffusion determinants and the external diffusion mechanisms, although they exhibit some similarities.

Figure 5.

Hazard function for amend imitation model.

Discussion

In this study, we examined the diffusion dynamics of two different LGBT-related policies: a policy that extends marriage recognition to same-sex couples and a policy that establishes state constitutional amendments to define marriage between a man and a woman only. For both policies, the pattern of diffusion was similar. We saw for both cases a couple of states taking the initiative on being the “innovators” followed by a period of early adopters. At some point for each policy, we saw a sharp increase with many more states adopting the policy as it picked up momentum. While it may appear on the surface that federal action in the form of the Obergefell v Hodges Supreme Court decision has put an end to this debate, it may still be too early to tell. The states that did not already recognize same-sex marriages and let the decision pass with little resistance make up the bulk of the “late majority” on a standard diffusion curve. There are many forms of resistance the most staunch holdouts could employ, and we have started to see that these “religious freedom” exemptions are making headway in some states.

Additionally, learning was evident for both policies. This means that for morality policy, learning occurs in a similar fashion compared to other policy spheres, even if they are different ideological orientations. However, learning was more pronounced for the amend model, lending evidence to the notion that the learning effect may be more prominent for anti-gay policies. For imitation, a state was more likely to copy its neighbor if there was a policy that allows for gay marriage, but imitation did not occur for marriage definition amendments. This may suggest that the imitation effect is stronger when a policymaker would have more of an opportunity to appear more cosmopolitan or to imitate major cultural centers, while anti-gay policies are better characterized as a show of solidarity between blocs of neighbors. However, when we examined both models simultaneously, we see that the risk for imitation is diminished significantly, giving evidence to learning being the primary mechanism of diffusion occurring. Finally, there was no conditional effect. While our original hypothesis was that larger states were more likely to be influenced by policy learning, results showed that this had little bearing on actual adoption. The institutional factors of state governments considered could also fall under the scope of conditional learning; but, beyond the already well-documented effect of whether or not a state has an elected judiciary, we saw little evidence that these factors significantly contributed to policy adoption, except for the rather obvious correlation between a stronger Republican Party and adoption of marriage definition amendments.

These findings are consistent with the literature that examines the specific mechanisms of policy diffusion. Particularly, we found that multiple mechanisms may be simultaneously occurring (Shipan and Volden, 2008) and that policymakers engage in both learning and policy imitation. Furthermore, this study adds to the growing body of literature that examines the diffusion of morality policies (e.g. Mooney, 2001). We find that different morality policies may diffuse in a similar fashion. Surprisingly, like Barclay and Fisher (2003), we found that religion, ideology, racial diversity, and population were not significant (Barclay and Fisher, 2003). Also, like Barclay and Fisher (2003), Haider-Markel (2001), and Lax and Phillips (2009), politics did not exhibit a strong influence, with the exception of the electoral returns for the Republican Party. However, there is a regional dimension which is different than Haider-Markel (2001).

In terms of competing innovations, we did see some amount of influence of one policy's adoption on the other. As previously mentioned, marriage definition amendments had plateaued at around the same time that recognition of same-sex marriage truly began to accelerate. From a cursory glance at the rates of adoption, we can see a few more instances where the spread of one policy affected the other. We see that the most rapid period of adoption of marriage definition amendments, for example, occurred between 2003 and 2005, or around the time that Massachusetts became the first state to extend marriage benefits to same-sex couples. Without further controlling for other factors that may influence this interplay between policies, we do not have enough evidence to conclusively say that this is the cause. Furthermore, there was no significant influence in our competing risks model.

While the results of this study do help to further our understanding of the policy diffusion process with regard to moral policy, there are still different parameters that could be considered. With a more firm understanding of how these policies spread among the states and what factors contribute to the adoption of each policy, the next step would be to better model the effects the adoption of one policy had on the other. Additionally, it may be worthwhile to examine the impact of both political and cultural happenings during this time period that influence public opinion towards the issue of same-sex marriage and the social construction of LGBT issues in order to provide a more complete understanding of the climate in which these policies were adopted.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Ayoub

(2014) Contested norms in new-adopter states: International determinants of LGBT rights legislation. European Journal of International Relations. Available at: http://ejt.sagepub.com/content/early/2014/09/01/1354066114543335.full.pdf+html (accessed 10 January 2016).

Barclay

Fisher

(2003) The states and the differing impetus for divergent paths on same-sex marriage, 1990–2001. Policy Studies Journal 31(3): 331–352.

Berry

Baybeck

(2005) Using geographic information systems to study interstate competition. American Political Science Review 99(4): 505–519.

Berry

Fording

Ringquist

(2010) Measuring citizen and government ideology in the US States: A re-appraisal. State Politics & Policy Quarterly 10(2): 117–135.

Berry FS and Berry WD (1990) State lottery adoptions as policy innovations: An event history analysis. American Political Science Review 84(2): 395–415.

Blyth

(1997) “Any More Bright Ideas?” The ideational turn of comparative political economy. Comparative Politics 29(2): 229–250. .

Brooks

(2008) When does diffusion matter? Explaining the spread of structural pension reforms across nations. Journal of Politics 69(3): 701–715.

Butz

Fix

Mitchell

(2015) Policy learning and the diffusion of stand-your-ground laws. Politics & Policy 43(3): 347–377.

Coontz

(2004) The world historical transformation of marriage. Journal of Marriage and Family 66(4): 974–979.

10.

Elazar

(1966) American Federalism: A View from the States, New York, NY: Thomas Y. Crowell Company.

11.

Fernández

Lutter

(2013) Supranational cultural norms, domestic value orientations and the diffusion of same-sex union rights in Europe, 1988–2009. International Sociology 28(1): 102–120.

12.

Gilardi

(2010) Who learns from what in policy diffusion processes? American Journal of Political Science 54(3): 650–666.

13.

Graham

Shipan

Volden

(2012) The diffusion of policy diffusion research in political science. British Journal of Political Science 43(3): 673–701.

14.

Grambsch

Therneau

(1994) Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 81(3): 515–526.

15.

Haider-Markel

(2001) Policy diffusion as a geographical expansion of the scope of political conflict: Same-sex marriage bans in the 1990s. State Politics & Policy Quarterly 1(1): 5–26.

16.

Hall

(1993) Policy paradigms, social learning, and the state: The case of economic policy-making in Britain. Comparative Politics 25: 275–296. .

17.

Hume

(2013) Courthouse Democracy and Minority Rights: Same-Sex Marriage in the States, Oxford: Oxford University Press.

18.

Kollman

(2007) Same-sex unions: The globalization of an Idea. International Studies Quarterly 51(2): 329–357.

19.

Lax

Phillips

(2009) Gay rights in the states: Public opinion and policy responsiveness. American Political Science Review 103(3): 367–386.

20.

Maggetti

Gilardi

(2016) Problems (and solutions) in the measurement of policy diffusion mechanisms. Journal of Public Policy 36(1): 87–107.

21.

Mahajan

Peterson

(1985) Models for Innovation Diffusion, Beverly Hills, CA: Sage Publications, Inc.

22.

Mitchell

Stewart

(2014) Emulation, learning, or competition? Examining inter-county anti-smoking laws in the state of Missouri. Public Administration Quarterly 38(3): 317.

23.

Mooney

(2001) Modeling regional effects on state policy diffusion. Political Research Quarterly 54(1): 103–124.

24.

Mooney

(2001) The Public Clash of Private Values: The Politics of Morality Policy, London: Chatham House Publishers.

25.

Mooney

Lee

(1995) Legislative morality in the American states: The case of pre-roe abortion regulation reform. American Journal of Political Science 39(3): 599–627. .

26.

Mooney

Lee

(1999) The temporal diffusion of morality policy: The case of death penalty legislation in the American states. Policy Studies Journal 27(4): 766–780.

27.

Neumayer

Plümper

(2015) W. Political Science Research and Methods. Forthcoming.

28.

Nicholson-Crotty

Carley

(2015) Effectiveness, implementation, and policy diffusion or “Can we make that work for us?”. State Politics and Policy Quarterly. Available at: http://spa.sagepub.com/content/early/2015/06/26/1532440015588764.abstract. .

29.

Paternotte

Kollman

(2013) Regulating intimate relationships in the European polity: Same-sex unions and policy convergence. Social Politics 20(4): 510–533.

30.

Plümper

Neumayer

(2010) Model specification in the analysis of spatial dependence. European Journal of Political Research 49(3): 418–442.

31.

Radaelli

(1995) The role of knowledge in the policy process. Journal of European Public Policy 2(2): 159–183.

32.

Rogers

(2010) Diffusion of Innovations, New York, NY: Simon and Schuster Publishing.

33.

Sharkansky

(1969) The utility of Elazar's political culture: A research note. Polity 2(1): 66–83. .

34.

Sherkat

De Vries

Creek

(2010) Race, religion, and opposition to same-sex marriage. Social Science Quarterly 91(1): 80–98.

35.

Shipan

Volden

(2008) The mechanisms of policy diffusion. American Journal of Political Science 52(4): 840–857.

36.

Shipan

Volden

(2012) Policy diffusion: Seven lessons for scholars and practitioners. Public Administration Review 72(6): 788–796.

37.

Smart

Klein

(2013) Neighborhoods of affinity: Social Forces and travel in gay and lesbian neighborhoods. Journal of the American Planning Association 79(2): 110–124.

38.

Valente

(1995) Network Models of the Diffusion of Innovations, Cresskill: Hampton Press, pp. 171.

39.

Volden C (2006) States as policy laboratories: Emulating success in the children's health insurance program. American Journal of Political Science 50(2): 294–312.

40.

Walker

(1969) The diffusion of innovations among the American States. American Political Science Review 63(3): 880–899.