DEPARTMENT OF POLITICAL SCIENCE TRANSGENDER RIGHTS AND POLICY RESPONSIVENESS: The impact of public opinion on the adoption of transgender policy in the United States João Paulo Hummel Mota Master’s Thesis: 30 credits Programme: Master’s Programme in Political Science Date: 2024-05-21 Supervisor: Jesper Lindqvist and Peter Esaiasson Words: 14962 Abstract: Over the last few decades, the American population's general perception of the transgender community has improved significantly. With this shift, transgender people were expected to be granted greater access to legal rights at the federal and state levels. However, from the mid- 2010s onwards, when trans policy began to be heavily debated, policy implemented moved against the perceived interests of trans people in some states. As a result, questions arose regarding whether there is responsiveness to public opinion for the implementation of policies aimed at the transgender community, given the existence of a theoretical basis for the occurrence of such dynamic. Due to the lack of policy-specific public opinion surveys at the state level, empirical evidence of responsiveness is scarce. To solve this problem, this study used a novel approach to measure public opinion on transgender rights. By estimating state- level feeling thermometers for the trans and gay population and policy-specific public opinion for bathroom bills through multilevel regression with post-stratification, it was possible to identify a slow process of increasing responsiveness to transgender policies in the states, whether through the adoption of inclusive or discriminatory policies. This responsiveness does not occur for all policies or comply with the opinion of a simple majority, but it does suggest that there is a tendency towards greater congruence between public opinion and policy implementation in the future. Keywords: Transgender rights, policy responsiveness, group-centrism, public opinion 2 Table of Contents 1 – Introduction: ......................................................................................................................... 4 2 - Literature review: ................................................................................................................. 6 2.1 - General context of transgender politics in the USA: ..................................................... 6 2.2 – Previous research on responsiveness: ........................................................................... 8 3 – Theory: ............................................................................................................................... 10 4 - Hypotheses: ........................................................................................................................ 13 5 - Methods and data: ............................................................................................................... 15 5.1 - Independent variables: ................................................................................................. 16 5.1.1 – Trans and Gay thermometers:............................................................................... 17 5.1.2 - Policy-specific public opinion: .............................................................................. 19 5.2 - Dependent variables: .................................................................................................... 20 5.3 – Control variables: ........................................................................................................ 21 6 – Results: ............................................................................................................................... 23 6.1 - Main Analysis: ............................................................................................................. 23 Table 1 .............................................................................................................................. 24 Graph 1 ............................................................................................................................. 24 Table 2 .............................................................................................................................. 25 Table 3 .............................................................................................................................. 28 Graph 2 ............................................................................................................................. 30 6.2 - Additional Analyses: .................................................................................................... 30 Table 4 .............................................................................................................................. 32 Table 5 .............................................................................................................................. 34 Graph 3 ............................................................................................................................. 35 Table 6 .............................................................................................................................. 36 7 – Conclusion: ........................................................................................................................ 36 8 – References: ......................................................................................................................... 39 Appendix: ................................................................................................................................. 45 3 1 – Introduction: Policies aimed at the transgender community in the United States, both favorable and unfavorable of their perceived interests, came to the forefront of the public debate in the mid- 2010s (Wang et al., 2016). In light of the high salience of these policies, public opinion stands out as one of the main factors potentially directly impacting the adoption of them (Kreitzer et al., 2019; Lax & Phillips, 2009a; G. B. Lewis & Oh, 2008). Many of these policies fall under the morality-policy model, which are defined as highly salient policies with public opinion strongly based on the moral values of the population (Kreitzer et al., 2019; D. C. Lewis et al., 2014), that receive considerable attention from legislators due to large popular mobilization (D. Haider-Markel et al., 2019). Nevertheless, research on the role of public opinion on trans policies and its impact on the adoption of these policies in the American states remains limited. Responsiveness is a concept that concerns the interaction between public opinion and policy implementation, being centered on the dynamic and causal relationship between change of public preference and change of policy preferences of the legislators (Beyer & Hänni, 2018). Given that public support for the trans community has been rising at the federal level for several years (D. C. Lewis et al., 2022), and is correlated with support for trans rights (Jones et al., 2018), the political scenario from the mid-2010s should be favorable for the implementation of public policies that protect this population. However, a major puzzle arises when examining responsiveness for transgender rights. Despite this increase in support for the trans community, there has also been an increase in the implementation of policies that go against the perceived interests of transgender people in several states (D. P. Haider-Markel & Vegter, 2020; Jones & Brewer, 2020; Krishnakumar, 2021; Wang et al., 2016), including restrictions on access to medical treatment, identity documents changes and the prominent "bathroom laws" (Mezey, 2020). While these policies have a high profile in the current public debate and are expected to be directly influenced by public opinion, little research has been done on the dynamics of responsiveness likely due to the challenges of public opinion data collection at the state level (D. Haider-Markel et al., 2019). This raises the question of whether there is responsiveness to policies for the transgender community. Among the topics covered in the literature on transgender rights, some researchers have examined the congruence between public opinion and policy implementation (Flores et al., 2015), population characteristics relating to support for these policies (Jones et al., 2018), and the different implementation approaches of states (D. C. Lewis et al., 2014). However, 4 responsiveness for trans policies is an under-researched subject. Flores et al. (2015), for instance, only investigate policy congruence, not potential responsiveness, focusing on implementation only for the year of 2010, and well before the rise in relevance of these policies. This provides an opportunity to examine the dynamics of policy responsiveness for a specific period when these policies had already gained prominence and were expected to be more responsive. Therefore, this study aims to explain how responsive trans-policies are to public opinion in the American states. I examine responsiveness in two ways: One, based on the assumption that public opinion towards specific policies impacting a group stem, in large part, from how individuals in general feel about the group(s) involved (Chudy, 2021; Converse, 2006, pp. 38- 41; Kinder, 2006). This approach advances the study of public opinion, providing a new perspective for the study of responsiveness beyond policy-specific public opinion. I utilize multilevel regression with poststratification (MRP) to estimate feeling thermometers at the state level towards trans individuals (inspired by Lax & Phillip's [2009a] work on responsiveness for gay policy). I subsequently pair this with data on the implementation of transgender rights at the state-level, including individual inclusive policies and an index of both inclusive and discriminatory policies. This novel approach makes it possible to analyze responsiveness for individual states, despite the unavailability of state-level public-opinion data. To provide a benchmark, I also estimate feeling thermometers at the state level towards gay and lesbian individuals, similarly paired with policy implementation data. I am thereby able to compare how responsive policy is to public opinion with a very similar case, where we already know that there are significant levels of responsiveness (Lax & Phillips, 2009a). Two, I also estimate policy-specific public opinion for bathroom bills at the state level using a similar MRP technique. The use of thermometers, as well as policy-specific data, is the best available method to estimate responsiveness in this case, while trying to overcome the lack of state surveys on most of the trans policies. This study is structured as follows: the next section provides a review of the literature on trans-policy and empirical research on policy responsiveness. The third part focus theoretical perspectives on responsiveness, morality-policy, and group-centrism, which provides the theoretical basis for the following analyses. The fourth section presents the three central hypotheses based on the theoretical framework, identifying potential implications from the concepts presented in the previous section. Following this, the methods of this study are described, including the operationalization of the variables used as well as a description of the 5 statistical methods used for estimating public opinion and political responsiveness. The sixth section includes the analysis and discussion of the statistical results. The seventh and last section concludes this work, also including limitations of this research and gaps that should be addressed in future research on responsiveness and transgender-policy. 2 - Literature review: 2.1 - General context of transgender politics in the USA: As Haider-Markel and Vegter (2020) argue, the rights of the LGBTQ+ population in the US have been constantly targeted by conservative groups, who since the 1950s have sought to restrict their access to basic rights. Although the LGBTQ+ community's movement grew stronger in the 1960s, the trans community only began to be included in political debates from the 1980s onwards, with slow and gradual progress in the 1990s and 2000s. Initially, it started through the fight for anti-discrimination laws to cover the protection of gender identity, in addition to the current sexual orientation protection (D. P. Haider-Markel & Vegter, 2020; Taylor et al., 2018). Taylor et al. (2018) point out that there was a significant gap in some states between the emergence of anti-discrimination legislation that covered sexuality as well as gender, up to 20 years between them, with most states only implementing gender policies in the mid to late 2000s or in the 2010s. With substantial advances in the adoption of gay rights at the state and federal level, trans-policy was pointed to as the main civil rights movement in American politics in the mid-2010s (Mezey, 2020). Despite some advances during this same period, a so-called third wave of anti-LGBTQ+ legislation arose with the goal of reversing these advances for the community by restricting the rights of this population, especially those of the transgender community (D. P. Haider-Markel & Vegter, 2020; Jones et al., 2018; Wang et al., 2016). At the beginning of the 2020s, this trend intensified, advancing in state legislatures through discussions on various civil rights for the trans population (Krishnakumar, 2021). In the last decade, the topic has reached a new level of relevance in the political debate, with the annual number of bills on the agenda in the states quadrupling between 2021 and 2023 (Trans Legislation Tracker, 2024), pointing to a continuing upward trend for the foreseeable future. The transgender population has traditionally been socially and politically stigmatized in the US (Witt & Medina-Martinez, 2022), and as far as policies are concerned, the trans community is especially vulnerable as this 6 new wave of legislation negatively impacts their physical and mental health, as well as their general quality of life (Case & Stewart, 2013; Du Bois et al., 2018). Among the policies concerning these populations that are currently being discussed are for example those regarding school spaces, recognition of their gender identity, access to transition-specific medical treatment and anti-discrimination rights at public spaces (Kline et al., 2023; Mezey, 2020). Many of the policies are considered discriminatory, as they attempt to provide different or harmful treatment to this minority group that has lower levels of power and influence in society (Pincus, 1996). The most prominent discriminatory policies are those that restrict their access to medical treatment, participation in sports, and access to gender- segregated environments (Kremen et al., 2021; Mezey, 2020; Witt & Medina-Martinez, 2022). Nonetheless, it is worth highlighting that there have been significant achievements in terms of access to basic rights for the transgender population in recent years when compared to other members of the LGBTQ+ community (Flores et al., 2023; D. P. Haider-Markel & Vegter, 2020; Taylor et al., 2018), with the vast majority of discriminatory policies failing in legislative and committee votes or being ruled against in courts (Sharrow, 2021; Trans Legislation Tracker, 2024). In 2022, the American Civil Liberties Union (2022) mapped 180 “anti-transgender” bills in the United States for that year's legislative session, with the vast majority of them, approximately 88%, failing at some point in the legislative process. Despite the wave of policies disfavoring the perceived interests of the trans community, a large part of the wide range of rights for this population only came about in the same decade, even if not uniformly in all American states (Taylor et al., 2020). Among them was the guarantee of protection against discrimination based on sexuality and gender identity in the work environment (since 2020 at the federal level), which was one of the most notable decisions in the Supreme Court case of Bostock v. Clayton County (Kline et al., 2023). This Supreme Court decision is one of the few jurisdictions at the federal level, since the vast majority of policies aimed at the transgender population, whether inclusive or discriminatory, take place at the municipal or state level. As a result of the Federalist system and the lack of other policies at the federal level for trans people, state legislatures have become the main avenue for trans policy discussion and later execution. This federal-level protection was extremely important for the population, since most American states did not have statutory protection against discrimination for LGBTQ+ people in the workplace at the time (Mallory & Sears, 2020). 7 2.2 – Previous research on responsiveness: Democratic responsiveness has several different definitions in the literature, but it can be simply described as the relation between the level of public support for a policy and the likelihood of adoption of this policy (Gilens, 2012), due to the interest of representatives in pleasing voters by supporting and adopting policies that represent them in the run-up to the next election (Beyer & Hänni, 2018). Studies on policy responsiveness in the US suggest that the relationship between public opinion and policy implementation does occur, but there are reservations about how well it works, based on varying results on the subject. The real impact of public opinion is often difficult to measure and circumstantial (Shapiro, 2011), with different approaches to the subject, through the analysis of various policies in different periods and with different data availability and variables to be analyzed increasing this difficulty (Manza & Cook, 2002). It should be noted that many of these studies focus only on policies with high salience and visibility, suggesting that their relevance is one of the main facilitators for such occurrences (Burstein, 2010; Shapiro, 2011), with it having a great impact on responsiveness (Lax & Phillips, 2012; Page & Shapiro, 1983) Page and Shapiro (1983) suggest that there is increased responsiveness to issues of high salience when they examined federal-level surveys from 1935 to 1979, with considerable levels of incongruence for the period under analysis, but nevertheless evidence of responsiveness to changes in public opinion. The authors also suggest that there is a possible impact of the implementation of policies on the opinion of other populations. Lax and Phillips (2012), by analyzing 39 different policies across states, suggest the existence of responsiveness for a variety of policies at the state level, but also of a “democratic deficit”, due to the lack of congruence with public opinion in half of the policies analyzed. Factors such as interest groups and partisanship have a negative impact on responsiveness, increasing levels of incongruence. They also observe a relationship between greater responsiveness and ideological and partisan factors among states. Additionally, for more recent policies there is a lag between public opinion and the implementation of some public policies, suggesting that responsiveness takes an incremental course, even for salient cases. The report also implies that, even when analyzing policy-specific public opinion, the ideology of the rulers and the electorate also has an impact, as the two are often correlated. Druckman and Jacobs (2006), by analyzing White House data during Richard Nixon's administration, found that the availability of public opinion data had a major impact on 8 presidential decision to support these policies, and the absence of this data meant that ideological issues in the government had a greater impact. There was a perceived correlation between data on several issues they had access to and the implementation of policies, but also for other factors such as the moment in the electoral cycle, the salience of the issues and the president's support for them. Furthermore, there is also some degree of responsiveness at the district level, implying that legislators in Congress tend to represent the policy-specific interests of their constituents (Krimmel et al., 2016; Stone, 1982). With advances in the statistical methods employed to measure public opinion, it has been possible to improve the quality of data on issues at different levels of analysis (as seen in Flores et al. [2015]; Krimmel et al. [2016]; Lax & Phillips [2009a, 2012]; Tausanovitch & Warshaw [2013]). For issues related to the gay and lesbian community, there is a history of public opinion polling suggesting a certain degree of political responsiveness, albeit in an imperfect and policy-dependent way (Rigel Hines, 2021). Support for gay-rights and the gay community has intensified since the 1990s (Brewer, 2003a; Kreitzer et al., 2014; Rosenfeld, 2017), although irregularly and regionally (Lax & Phillips, 2009a; Tilcsik, 2011). G. B. Lewis and Oh (2008) point out that the implementation of laws on same-sex marriage in the 2000s was in line with the opinion of the majority of the population, whether to allow or ban the practice. Lax and Phillips (2009a) also find responsiveness for some of the policies towards the gay community, but not for all the policies analyzed, with a tendency towards adoption of more conservative policies in non-congruent cases. Studies on congruence and responsiveness for trans-rights, however, are scarce in comparison to LGB policies. This shortage is mainly due to the lack of pooling data for most policies concerning the transgender community, specially at the state level (D. Haider-Markel et al., 2019). The work of A. R. Flores et al. (2015) on congruence regarding antidiscrimination policies for the transgender population, for example, applies the term responsiveness, even though it doesn't analyze a causal relationship. Rather, Flores et al. only includes public opinion data from 2011 and policy implementation data from 2010, making it impossible to measure how public opinion may affect policy implementation. However, the study is still of great importance due to its use of MRP to estimate public opinion and for being able to portray a trend of non-congruence for the topic at that time. 9 3 – Theory: Responsiveness becomes a mechanism of interest for trans policies since the increase in policy adoption aimed at the transgender community is expected to follow the trends of support at the state level, especially for cases of highly salient policies, given the inclination for greater consistency between public opinion and implementation in these cases (G. B. Lewis & Oh, 2008; Monroe, 1998; Page & Shapiro, 1983). This trend occurs as politicians pursue issues with greater visibility and that are of greater interest to potential voters (G. B. Lewis & Oh, 2008; Shapiro, 2011). Beyer and Hänni (2018) differentiate congruence and responsiveness by stating that responsiveness involves a causal element, based on the principle that policy changes occur due to changes in public opinion. Congruence does not involve a causal relationship, being limited to the analysis of the static agreement between citizens and legislators and better explained by electoral competition. On the other hand, responsiveness would work on the change of tendency of the public opinion for a policy. This means that there is an element of anticipation for future changes in public opinion for responsiveness, as opposed to congruence, which is based on current public perception. This distinction made by the authors is essential to improve the definitions of the two and avoid possible misconceptions, which are common in the literature. Bringing a different perspective on the subject, Rigby and Wright (2013) suggest that responsiveness can affect policy implementation through alignment or influence. Alignment is interpreted as the matching of the perspectives of politicians, through policy enactment, and the views of the population. Influence, on the other hand, would be focused on the response to the opinion of part of the electorate. Most responsiveness studies are based on the idea of responsiveness to specific policies, but we must recognize the possibility that most people may not have concrete opinions on the majority of policies. Taking this into account, the concept of “group-centrism” can be especially useful to address public opinion on trans people and their rights. Group-centrism was described by Converse (2006) and advanced by Kinder (2006), and suggests that policies with high visibility and salience would have their support based on sympathy with the specific groups to which they are aimed. This happens due to voters not being invested in politics enough to have a specific view on each and every policy (Converse, 2006; Kinder, 2006), as the masses have a lower level of information and greater superficiality in the observations in 10 comparison to the political elites. As a result, the views of the general population on specific policies would be influenced by how elites frame these groups, with few nuances between most of these policies, guaranteeing the construction of a group-centered opinion by the population (T. E. Nelson & Kinder, 1996). This means that "social-groupings" are the central objects in belief systems for individuals, instead of more abstract values that would require a level of information that the general public does not have (Converse, 2006). Political elites, in this case, would differ by centering their belief-systems more on abstract concepts and values, which informs how they view related public policies, their implementation and governmental impacts, as described by Converse (2006). Previous research shows that feelings (of sympathy) towards a specific group are a good predictor of their policy views, as many voters likely use their feelings towards a group as a heuristic, which informs their views on specific policies. The perception of the group is also, in general, more stable than policy-specific public opinion, which tends to change often between surveys (Markus & Converse, 1979). The thermometers of the homosexual and transgender community are therefore expected to be effective indicators of support for policies aimed at them, as they represent one of, if not the most determinant component that determines public opinion, (Brewer, 2003a, 2003b; Kinder, 2006). The community's visibility is one of the central interests in studies on group-centrism. For the transgender community, as happens with the black community, the visibility would be defined by strictly physical characteristics, which would increase the visibility of the group to society, as it is not based just on more abstract concepts such as social class (Converse, 2006). Although the transgender community only represents a small percentage of the American population (Flores et al., 2023), its high salience since the mid-2010s should provide the necessary conditions for the general population to perceive a connection between the political dispute and the group. Several studies have already analyzed the general population's perception of the transgender community and support for their political rights (Chudy, 2021; Flores et al., 2023; Jones et al., 2018; Miller et al., 2017; Tadlock et al., 2017, p. 20). Articles such as those by Jones et al. (2018) and Tadlock et al. (2017) show that general perception of the community is strongly, positively, and significantly correlated to support for trans policies, indicating that perceptions about the group is one of the main variables leading to support for these policies. Although distinct, the policy-mood component used by Stimson et al. (1995) is an interesting variable and, in a way, acts in similar manner to the thermometer of a group, providing an effective alternative to the use of policy-specific opinion. Policy-mood is a major 11 dimension for public preferences regarding policy alternatives, in this case support for greater or lesser government action. For this purpose, they employed a support thermometer for government action, which was positively correlated with the implementation of more liberal policies. When it comes to support for the gay population, there is a direct correlation between the substantial change in the "population support thermometer" between the 1990s and 2010s, and the increase in support of gay rights in the same period (Brewer, 2003a; Rosenfeld, 2017). As the thermometers for the trans population have always been cooler than for the gay population (D. Haider-Markel et al., 2019; Norton & Herek, 2013), it is to be expected that this dynamic of change for both community and rights support occurred at a later period, but should follow a similar upward trend (D. C. Lewis et al., 2022). As a result, the perception of the transgender community could be used as a proxy for trans rights in general, but not necessarily for all policies. Morality policy is another main concept used to study policy implementation of LGBTQ-rights. The literature suggests that this policy model has its own characteristics that directly impact implementation odds. Morality policy describes policies whose subject is divisive, salient, and require little to no prior information from the population in order for them to have an opinion (Haider‐Markel, 1999; D. P. Haider-Markel & Meier, 1996). Beliefs regarding these policies are often based on ethical and faith grounds, using an almost Manichean duality of right and wrong, instead of following a more conventional policy opinion centered on issues of convenience, utility or efficiency (Hollander & Patapan, 2017). Hollander and Patapan (2017) also mention the relevance of the involvement of citizen groups in all stages of the policy process, which is supported for LGBTQ+ policies due to the large mobilization of advocacy organizations and groups against these rights (Taylor et al., 2018). Examples of policies that fit the morality model are policies for the LGBTQ+ community, abortion, euthanasia and drug legalization (D. P. Haider-Markel & Meier, 1996, 2008; Hollander & Patapan, 2017; Kreitzer et al., 2019; Lax & Phillips, 2009a; G. B. Lewis & Oh, 2008; Mezey, 2020). The high salience of these policies is particularly relevant as it generates pressure for parties and politicians to follow the interests of voters, as a result of electoral competition (Kosciw et al., 2009). Before the rise to prominence in public debate of these policies, trans rights were seen as not fitting adequately into the morality politics model (D. C. Lewis et al., 2014), probably due to their lack of salience, being explained better by interest groups than 12 public opinion. However, over time, many of these policies have become prominent, similar to policies for the gay population (D. Haider-Markel et al., 2019), and thereby now better fitting into the morality policy model. Given the prominence of these policies in recent years, the role of public opinion should become more relevant to understanding the adoption patterns of these policies. Similar to the long history of movements for and against gay policies (D. Haider- Markel et al., 2019), the transition from a model of interest group politics (which is less centered on salience, involves less partisanship, and where public opinion is less relevant, see D. P. Haider-Markel & Meier, 2008) to that of morality policy was expected for policies concerning the transgender population as they became more mainstream. It is worth considering that not all policies for the trans population necessarily fit into this model or have specific characteristics. There is a diverse set of opinions on trans policies in the US (Parker et al., 2022) and distinctions between them can also arise from whether the policies are considered body politics, meaning politicization and potential regulation of the trans body (Miller et al., 2017), involve dealing with children (Taylor et al., 2012) or if not fitting the morality policy model at all (Kreitzer et al., 2019). Moreover, there will not always be congruence or responsiveness to public opinion in all morality policy instances (Kreitzer et al., 2019; Lax & Phillips, 2009a), as they tend to also have other potential explanatory variables that justify them, such as party competitiveness and partisan factors (D. P. Haider-Markel & Meier, 1996). 4 - Hypotheses: Three hypotheses were developed aimed at better understanding the relationship between public opinion and the adoption of trans policies. Given that certain transgender policies arguably fall into the model of morality policy, public opinion should be one of the main explanatory factors for the adoption of most of these policies. Furthermore, by using the concept of group-centrism as a foundation, we can verify whether the general perception of the transgender community in the states has, an impact on the adoption of policies. I expect, based on the previous research discussed above, that individuals’ feelings towards specific groups is a good indicator of their support for different policies. Given the high salience of policies for the gay community in the past and transgender policies in the present, these variables would be effective in explaining the implementation of these policies, considering that morality policies require little information for the population to have an 13 opinion (D. P. Haider-Markel & Meier, 1996) and the significant correlation between support for policies directed at the community and indicators of sympathy for the group (Jones et al., 2018). This hypothesis should also allow for a comparison of the responsiveness of policies for the transgender population and the gay population using group thermometers for an examination of group-centrism theory. The first hypothesis will deal specifically with the use of the thermometer to measure responsiveness: H1: Transgender policies implementation is correlated with the feeling thermometer of the transgender community at the state level in the United States. The second hypothesis assumes that, by conforming to the standard of morality politics, bathroom bills will be responsive to policy-specific public opinion (Lax & Phillips, 2009a). Since bathroom bills are considered the most salient policies concerning the trans population, these should fit the morality policy model better than most policies. This policy is also relevant to the topic as it has elevated the public relevance of trans-policy debate at the state level, in addition to being one of the few policies that is frequently polled, which have been extensively published since 2016. Therefore, the second hypothesis is structured as: H2: Bathroom laws adoption is correlated with policy-specific public opinion at the state level in the United States. It is to be expected that not all policies will follow the pattern of responsiveness in the same way, similar to what happens with gay-rights (Lax & Phillips, 2009a). That means that anti-discrimination policies, due to their lower salience, should not be as responsive in comparison to other trans rights, given the low legislative visibility in their adoption (G. B. Lewis & Oh, 2008; Rigel Hines, 2021; Shapiro, 2011). Policies such as bathroom bills, and access to medical treatment through private health plans should therefore respond better to public opinion, as they fit more appropriately into the morality policy model. Thus, the third hypothesis is presented as: H3: Anti-discrimination policy implementation should not correlate to public opinion in comparison to other trans rights that better fit the morality-policy model. 14 5 - Methods and data: I analyze the role of public opinion in the implementation of trans-policies at the state level through statistical analysis of responsiveness to policy-specific and group-based opinions. This approach will reveal whether the adoption trends of policies related to the trans community in different states follow potential changes in public perception. It will also allow the comparison with gay and lesbian policies, which fit the morality policy model, and whose responsiveness tendencies are mainly evident but still highly policy-dependent (Lax & Phillips, 2009a). For this purpose, it is necessary to initially identify the differences between measurements of congruence and responsiveness, given that this study only seeks to analyze the latter. Congruence can be identified using data at moment t for both policies and public opinion. In contrast, responsiveness takes into account the trend of change in opinion over a period, due to legislators’ rational anticipation of their electoral performance. For this to happen, policy implementation must take place in a period following the shift in public opinion, i.e., in period t + 1 (Beyer & Hänni, 2018; Stimson et al., 1995). This will correct the approach of Flores et al. (2015), which included implementation prior to the measurement of public opinion and should better represent responsiveness to transgender rights. For the case of trans- policy in the American context, because elections occur every two years, the period chosen to analyze implementation was two years after the public opinion polls. As a result, public opinion, the independent variable, will be lagged by 2 years. Traditionally, public opinion is measured using policy-specific public opinion data (Lax & Phillips, 2009a), but to deal with the lack of policy-specific data and to study this alternative mechanism of group-centric responsiveness, a state level feeling thermometer will be employed as a proxy for public opinion for trans-policy. The use of observational data should hinder the identification of a causal relationship through the mechanism of responsiveness described but it is widely employed for this type of study. Analysis of responsiveness in the states is carried out following similar strategies to those of Lax and Phillips (2009a), assessing the impact of public opinion on the implementation of various policies. However, this study also differentiates itself by having the state as the level of analysis, instead of individual policies and by contrasting the differences in political responsiveness for policy-specific and group-based opinions. Although it is known that public support towards the transgender population has been increasing particularly since 2015 at the 15 federal level (D. C. Lewis et al., 2022), there is no specific data on each individual state to guarantee that this increase is happening among all of them. In order to measure public opinion, Multilevel regression with poststratification (MRP) was chosen to estimate state-level public opinion for both policy-specific opinion and the thermometer. This method will be explained in more detail in the description of the independent variables used. To identify potential responsiveness to public opinion, I use logistic and linear regressions. Logistic regressions are also used to determine the responsiveness for bathroom bills, and a range of different rights for gay and trans people. Linear OLS regressions are employed to measure responsiveness where the dependent variable is an aggregate of policies. When examining responsiveness to trans policies, a number of innovative approaches had to be applied to deal with the absence of data, while still being appropriate to identify the potential correlation between public opinion and the implementation of these policies. I emphasize that there are certain limitations to what can be done with the available data on the subject, for example resorting to MRP to estimate public opinion. As Lax and Phillips (2009b) point out, despite the effectiveness and sufficiency of these measurements as an independent variable in this kind of model, and its superiority to disaggregation, its estimates are usually better for larger states and do not produce identical results to state-level surveys. This method is currently the best for this scenario but was only necessary due to the lack of surveys on the subject at state level. In addition, the use of thermometers, although theoretically valid and relevant for the study of responsiveness, also arose from a need to deal with the lack of policy- specific data. These adjustments were all necessary to allow examination of such a relevant topic, but in no way detract from its scientific validity and explanatory capacity. 5.1 - Independent variables: The independent variables used are policy-specific public opinion, through opinions on "bathroom bills” and public support for the transgender community, which is measured through a group feeling thermometer. To estimate the average level of opinion/feeling at the state level, I use the MRP method. This approach is able to estimate public opinion as a function of demographic characteristics (such as ethnicity, gender, age and education level) and state- specific effects (Lax & Phillips, 2009b). To achieve this, a multilevel regression is used for each relevant question of the survey and includes demographic factors of the participants as individual level-effects and information from the states as state-level effects. This model 16 generates an estimate of the opinion for each potential combination of characteristics included. Using census data, it is then possible to post-stratify and estimate the support of each state using the weight for each group modeled, using the variables included in the model. For the post-stratification stages, all the models used Census data from Ruggles et al. (2024), which provides all the years already harmonized for use at this stage. MRP is capable of using national surveys of typical size to estimate state-level opinion effectively, far better than disaggregation methods, as long as it uses appropriate predictors to answer such questions (Buttice & Highton, 2017). MRP has been tested in the field before and is considered appropriate for estimating opinions, including gay and trans rights, with a low margin of error for the estimations (Flores et al., 2015; Lax & Phillips, 2009a). All surveys have a sample larger than 2000 individuals, which would be more than enough to make an adequate measurement using MRP and considerably larger than the small sample (fewer than 1000 individuals) used by Flores et al. (2015). 5.1.1 – Trans and Gay thermometers: The thermometers were included as a strong alternative to explain responsiveness to transgender policies, being based on the concept of "group-centrism", which assumes that an individual's policy opinions are mostly influenced by their general perception of the groups involved. Historically, the thermometer for the trans population has always scored lower than that for the gay population (Norton & Herek, 2013), but there has also been an upward trend in recent years (D. C. Lewis et al., 2022), which may indicate a new trend in responsiveness patterns, especially given the increased salience on some of the issues related to this group. Although it has not previously been used for this kind of research on responsiveness, it should perform well as an indicator, since it is not necessarily plausible that political elites represent all the individual-policy interests of populations for all trans-policies adequately, given that there is no data on the vast majority of them not even at the federal level (D. Haider-Markel et al., 2019). As a result, in some of these cases, many of the policy-making decisions would be made not based on policy-specific opinions, but rather highly influenced by general forms of support for the community and their respective policies, given the framing of policies in morality policy. Given that part of the general public may not have opinions on some trans policies, tending to remain neutral (Flores et al., 2018), their opinion on specific policies should mostly represent their general perspective about the community. This will make it possible to 17 measure the impact of the thermometer, since it is one of the most relevant indicators for general public opinion, providing a proper analysis of these policies that have not yet been the subject of attention in polls. However, it is essential to note that the thermometer is probably more adequate for measuring support for a larger set of policies and not specific policies. To estimate the thermometer scores, a specific MRP model had to be used, since the vast majority of the studies employ multilevel logistic regressions, making it impossible to examine thermometers, which are traditionally built on a scale of 0 to 100. To solve this, the model employed here is an adaptation of the one used by Hanretty (2020) for constituency- level estimates of ideology in the UK, which, unlike the model for policy-specific public opinion, uses Bayesian methods for MRP and does not restrict itself to a logistic function. The demographical variables used were educational level (in 5 categories: less than a high school education, high school graduate, some college, college graduate and post-graduate), age (in 6 categories: 18-29; 30-39, 40-49, 50-59, 60-69, over 70), gender (binary variable as 0 = male and 1 = female) and ethnicity (4 categories: white, black, Hispanic and others). For the multilevel model, data from each state was included, namely: region, percentage of Republican votes in the state for the last presidential election, percentage of the evangelical population, trans population in the state (only for the trans thermometer functions), urban population, population that identifies as very religious, population that identifies as conservative and population that identifies as Mormon. The variables included are similar to the ones selected by Lax and Phillips (2009b, p. 384) for their MRP model, but also includes additional indicators comparable to the model by Hanretty (2020, p. 639) for his ideology model. The thermometer data at the federal level is from national public opinion polls conducted by American National Election Studies (ANES) in 2012 (only for the gay thermometer), 2016 and 2020 (American National Election Studies, 2013, 2017, 2021). For their thermometer, the ANES employs a scale from 0 to 100 degrees, where 51 to 100 degrees mean favorable and warm feelings towards the group. Ratings between 0 and 49 indicate unfavorable feelings towards the group or no interest in it. 50 is a neutral value. The question asked to measure trans support is “How would you rate Transgender people?”. The results of the state-level estimations are on a scale from 0 to 10, but all scores range from 4.24 to 6.85. For the first table, however, these will be displayed on a scale from 0 to 100, as the thermometer traditionally are shown on this scale. By finding signs of responsiveness with the use of the thermometer for the gay population, as there is already confirmation of responsiveness with policy-specific opinion 18 (Lax & Phillips, 2009a), I can suggest the presence of a valid mechanism of responsiveness, providing at least partial validation for potential responsiveness for trans-policies. However, as only two surveys for each of the American states were available for the thermometer estimations, there is unfortunately a low number of observations which hinders causal inference. 5.1.2 - Policy-specific public opinion: Views on "bathroom bills" were estimated using national public opinion polls conducted by the American National Election Studies (ANES) for 2016 and 2020 (American National Election Studies, 2017, 2021) and by the Public Religion Research Institute in 2021 as part of the American Values Survey (Public Religion Research Institute, 2021). This allowed me to estimate support for this policy at three different points in time: in 2016 (when this issue came to the fore), in 2020, and 2021. This variable coding for each survey is based on the answer to a question referring to support for bathroom laws. The question in the ANES survey is presented as: "Should transgender people - that is, people who identify themselves as the sex or gender different from the one they were born as - have to use the bathrooms of the gender they were born as, or should they be allowed to use the bathrooms of their identified gender?". PRRI's question is based on the following quote: "Laws that require transgender individuals to use bathrooms that correspond to their sex at birth rather than their current gender identity", offering different options for the interviewee to choose whether or not they agree with the policy. Despite the differences between the questions, I treat these as the same as both should adequately measure support for the same policy. Each answer was re-recorded in a binary format (support or reject). I therefore used multilevel logistic regression to estimate state-level opinion, using the MRP approach. All public opinion scores range from 0 to 1, with 0 representing a rejection of bathroom policies, that is, allowing people to use the bathrooms of the gender they identify with, and 1 as support for bathroom bills or support for the use of the bathroom with the gender they were assigned at birth. Coding these policies in a binary format is a necessary simplification for the analysis of them, but it does not perfectly represent the variety of policies and differences between them, which would require case studies for a closer inspection of differences. 19 I used the MRP model by Kastellec et al. (2019, p. 7) as a foundation for the estimation, which applies R's lme4 package, traditionally used for MRP and suitable for this model, but with more simplified functions than fully Bayesian approaches to MRP (Gelman et al., 2018). Adaptations of the original function were made by applying Lopez-Martin et al. (2022) suggestion of categories for each demographic variable. The demographic variables selected were state, educational level, age, and gender (all categorized in the same manner as in the policy-specific model). The multilevel variables selected were the performance of the Republican candidate in the previous presidential election and the region of the state. These variables have already been used in Lax and Phillips (2009a, p. 384) and Flores et al. (2015, p. 3) and are recurrent in those studies. Combinations of variables were also included to increase the number of categories, namely education combined with ethnicity (in the same categories as in the previous model) and education combined with age, following the two reference MRP models previously discussed. All of the selected variables for the multilevel functions have been used in previous research and have been shown to correlate with support for policies for the transgender population, with gender being one of the most significant variables in mediating this effect (Jones et al., 2018; Tadlock et al., 2017). For this research, the lack of accessible data meant that it was necessary to work with a small number of cases. This ends up negatively affecting the validation of some of the relationships presented here, but it is nevertheless necessary as it is the only feasible way to carry out such analyses. Despite the low number of cases, it is still hoped that the quality of the measurements and the theoretical validation will enable this study to be properly conducted. 5.2 - Dependent variables: Different dependent variables for policy adoption are employed to account for three models being analyzed, these being an aggregate of trans and gay policies, individual inclusive trans policies, and bathroom bills. The first and second models analyze the implementation of policies for the trans or gay population in the states, along with thermometers as the independent variables. The third model includes information on the adoption of bathroom policies in American states. All policy implementation data was taken from the Movement Advancement Project (2023) covering policies implemented between 2014 and 2022. For the trans population, it includes protection against hate crimes, discrimination at work and in 20 housing, discrimination in schools, access to medical treatment through private healthcare or healthcare for public servants and changes of name or gender on driver’s licenses. The policies have all been coded in binary format with 1 being implementation and 0 absence of inclusive policy or adoption of a discriminatory policy. MAP, besides providing information on the implementation of specific policies, also provides State Policy Tallies. This is a more nuanced score, which assigns positive scores for inclusive policies and negative scores for discriminatory policies. In the case of an absence of legislation on the topic, the state receives a score of 0 for that policy. The score for each policy is given based on its implementation in the state, with fractional scores in cases of legislation covering only part of the state or being valid for just part of the population analyzed. The maximum scores in 2023 were 23 for the Gender Identity Tally, which will be used in conjunction with the transgender thermometer, and 21 for the Sexual Orientation Tally, for the gay thermometer. This indicator has been used previously in studies on the impact of the adoption of policies for the trans population on their health (Du Bois et al., 2018; C. L. Nelson et al., 2023; Weixel & Wildman, 2022) and was cited by Mezey (2020) as a relevant indicator for policies affecting the community in her study on policymaking for transgender rights. Comparing the implementation of bathroom bills with the two public opinion indicators will also provide a comparison between policy-specific public opinion and the thermometer. Nevertheless, the low rates of implementation of some of these policies may make further generalizations difficult. 5.3 – Control variables: Similar control variables to the ones used by Lax and Phillips (2009a) and Flores et al. (2015) are included as these are adequately able to explain the majority of cases of responsiveness for LGBTQ+ policies. These variables will be the ideology of the population using Warshaw and Tausanovitch's (2022) indicator, an ideology score of state governments created by Shor and McCarty (2023), professionalism of the legislature (Squire, 2023), percentage of Republicans in the legislature, presence of a Republican governor, percentage of religious conservatives and year, which should focus on the variation between states. I also included the MAP implementation score in the same year of the survey (i.e., a lagged dependent variable, or the already existing policy level in the state) as an additional control variable to examine policy change between the two years, as a robustness check of the strength of the 21 thermometer’s predictive capability. This is not often included in similar studies but could be an important control variable if we believe that citizens are influenced by already existing policy. I included general indicators of ideology of the population and legislators as control variables due to the presence of varying degrees of correlation between policy-specific opinion and ideology, but also for the odds of it being a better predictor of responsiveness than public opinion when political elites do not have access data on policy-specific opinions (Erikson et al., 1994; Flores et al., 2015; Lax & Phillips, 2009a). Shor and McCarty (2011) developed an indicator of ideological preferences of the state legislature, using roll call votes at the state level and allowing analyses that capture ideological differences between parties at the state level. It was chosen over the indicator by Berry et al. (1998), for estimating ideology between 2016 and 2020, while the other dataset has not received any recent updates. Tausanovitch & Warshaw (2013) utilize MRP with large-scale surveys to estimate ideological alignment of the population based on policy preferences and are able to estimate citizens’ ideology from 2006 to 2021. Unfortunately, there are no updates to both citizen and government ideology data by Berry et al. (1998), which was used for both Flores et al. (2015) and Lax and Phillips (2009a). There are no updates after the year of 2016 for Citizen Ideology and after 2017 for Government Ideology. For data from the year 2021, in the bathroom bills analysis, it was necessary to keep this ideological variable unchanged from the previous year, due to the lack of ideology estimates for that specific year. The professionalization of the legislature was included due to the possibility of a more professional legislature adopting politically liberal policies on these issues (Lax & Phillips, 2009a). Squire (2007) takes members of congress' salaries, average days in session, as well as average staff per member, into account to measure professionalization of the legislature, with greater values indicating greater professionalism. The percentage of Republican legislators and the presence of a Republican governor were included since it could make it less likely that transgender-inclusive laws will be implemented even when there's a majority opinion (Flores et al., 2015). As previously stated in the literature review, religion possibly has an impact on the implementation of trans policies and is known to impact the adoption of gay rights, so the proportion of evangelicals and Mormons in a state will be included together as the proportion of religious conservatives. Salience could not be added to the models as there is no data available to compare salience for each of the policies at the state level, only at the federal level as Lax & Phillips (2009a) did. Furthermore, the salience of trans policy has mainly shifted at the national level, and 22 controlling for year partially accounts for differences in salience of trans-policy between years at the national level. Of course, there can be state-level differences in salience, but these are arguably less important compared to national changes in salience. Lastly, the presence of direct democracy in the state was not included as a control variable, as it is not generally a significant predictor of the implementation of most policies targeting homosexuals or transgender populations (Lax & Phillips, 2009a; D. C. Lewis et al., 2014), but future studies may want to examine whether including this control variable, as well as accounting for salience at the state- level, influences the results. Unfortunately, it was not possible to include state-fixed effects for this study, as there are not enough observations for it, with no more than 3 state observations per model (in the main model each state only figure twice, resulting in 100 observations). Flores et al. (2015) also did not include fixed effect in their model due to the low number of observations. This is an ongoing limitation for public opinion analyses on trans rights at the state level, due to the low number of state surveys or federal surveys that allow for MRP. 6 – Results: 6.1 - Main Analysis: I will begin by presenting the thermometer estimates for the trans population that were made using two survey years from ANES. These results in Table 1 suggest an improvement in the perception of the transgender community by the general population in all American states during this period, which corroborates the observed increase in positive views towards trans- individuals at the federal level in the last few years (Lewis et al. 2022). This estimate is also of great importance for studies of the perception of the transgender population as it demonstrates that even states that traditionally have more negative perceptions of the LGBTQ+ community, such as the southern states of the United States (Tilcsik, 2011), are also likely undergoing this shift, rather than having an increased polarization level on the topic. The upward trend in the thermometers has also been identified for the gay population thermometer over the last few decades, intensifying from the 1990s onwards (Brewer, 2003a; Kreitzer et al., 2014; Rosenfeld, 2017), a period significantly earlier than this same trend for the transgender community. When using MRP to estimate the thermometer of the gay population in the states between 2012 and 2020, this rising pattern was again observed. Comparing the thermometers of the two communities (Graph 1), the perception of the trans 23 community remains lower than that of the gay community, as expected (Norton & Herek, 2013; Tadlock et al., 2017), but both follow an upward trend during that period. This upward shift is also associated with a tendency for both communities to achieve greater access to social rights during these periods. Table 1 - Thermometer estimates for the transgender community State 2016 2020 State 2016 2020 Alabama 42.4 48.8 Montana 53.5 57.6 Alaska 54.6 60.0 Nebraska 51.8 56.1 Arizona 56.4 58.2 Nevada 58.3 65.2 Arkansas 44.2 51.2 New Hampshire 59.4 68.0 California 59.5 63.5 New Jersey 64.0 65.6 Colorado 59.0 62.2 New Mexico 55.7 60.5 Connecticut 60.5 63.1 New York 60.2 63.4 Delaware 57.8 63.3 North Carolina 48.9 55.6 Florida 52.0 55.9 North Dakota 50.8 55.2 Georgia 50.1 55.0 Ohio 50.8 56.1 Hawaii 58.6 63.0 Oklahoma 45.5 51.3 Idaho 55.5 55.7 Oregon 58.5 62.4 Illinois 55.2 61.5 Pennsylvania 57.7 58.1 Indiana 49.8 56.1 Rhode Island 62.6 66.2 Iowa 52.2 57.5 South Carolina 48.5 53.8 Kansas 51.4 56.3 South Dakota 47.7 53.7 Kentucky 47.1 52.5 Tennessee 45.2 50.8 Louisiana 47.6 50.5 Texas 50.0 55.0 Maine 56.6 62.8 Utah 59.5 61.9 Maryland 56.9 64.6 Vermont 58.5 63.5 Massachusetts 64.2 68.5 Virginia 52.6 60.9 Michigan 55.3 60.1 Washington 57.2 62.9 Minnesota 54.1 60.4 West Virginia 47.7 52.0 Mississippi 44.0 51.7 Wisconsin 50.7 57.6 Missouri 49.7 55.7 Wyoming 54.7 56.1 Mean 53.7 58.6 SD 0.54 0.49 Graph 1 – Thermometer trend for Gay and Transgender communities 24 As it is already known that there is some responsiveness for policies regarding the gay population (Lax & Phillips, 2009a), by estimating the thermometer of this group I was able to analyze whether this relationship was maintained when policy-specific public opinion was not used, replacing it with the thermometer and estimating general policy level using the Sexual orientation Policy Tally produced by MAP. Although I estimated the gay population thermometer in 2012, as there is no MAP tally data for the year 2014, only the estimations from 2016 and 2020 have been included. By using the gay community thermometer, it will be possible to demonstrate its effectiveness in explaining the implementation of these policies, as happens with policy-specific public opinion. This is again due to the relevance of group support indicators in the construction of public opinion in the context of group-centrism theory. In Table 2, which shows the results of a linear regression with the Policy Tally as the DV and the thermometer as the IV, the thermometer proved to be significant in all the models tested, suggesting its correlation with the implementation of gay rights. Table 2 - Linear Regression (IV: Gay Thermometer; DV: Policy Tally) Variables Model 1 Model 2 Model 3 Thermometer 9.26*** 2.98** 3.82*** (0.75) (0.96) (1.06) Government ideology - -4.28*** -3.23*** (conservatism) (0.78) (0.90) Voter ideology - -9.65* -7.04 (conservatism) (4.11) (0.4.96) % Republican legislature - - -5.19 (3.23) Republican governor - - -0.72 (0.78) Professionalism - - -2.06 (2.56) Intercept -47.65*** -7.67** -9.39 (4.63) (6.09) (6.26) R² 0.61 0.78 0.79 AIC 576 522 514 N 100 100 100 * p <.05; ** p <.01; *** p <.001 25 For the first model, an increase of one unit in the thermometer leads to a shift of 9.26 of a standard deviation in the adoption of pro-gay policies. In the second and third models, the impact is reduced, but still results in a change of roughly three standard deviations in the first model and 3.82 in the adoption of these policies. These results are highly favorable for the thermometer test, showing that even when using ideology, it still has a considerable effect, and a p value smaller than 0.001. The great and significant impact of the thermometer is highly pertinent, since the effect of the independent variable is minimized by the correlations with the control variables, which are also affected by that, considering that there is a substantial association between warmer thermometers for the transgender community on the part of the population in certain states and the election of more liberal candidates or the presence of a more liberal electorate. This ultimately demonstrates that the feeling thermometer acts as an explanatory variable beyond what is captured by ideological factors, which are already quite powerful for analyzing responsiveness. The ideology of the government also proved to be significant in all the models it was included in, with a coefficient between -4.28 and -3.23 in the two respective models in which it was included, which indicates that implementation could depend not only on the perception of this population, but also on the ideological factors of the legislators. By conducting a linear regression using data from MAP's Gender identity Policy Tally as the dependent variable, it was possible to identify a significant relationship between the thermometer and the implementation of policies that benefit the trans community, as seen in Table 3, even with the inclusion of control variables. The first model includes only the thermometer, the second adds the ideologies, like the previous regression, the third includes the variables used by Flores et al. (2015) and the fourth adds controls based on the analysis by Lax & Phillips (2009a). A dummy variable for the year was included as it may reflect the salience of the issue in the given year, addressing the measurement concern mentioned in the methods section. Conservative religious groups (evangelical Protestants and Mormons) were also included. In this case, the thermometer is shown to be an effective predictor of implementation for the policy index. For the first model, an increase of one unit in the trans thermometer leads to a shift of around 12 standard deviations in the adoption of the trans policies of the MAP Tally. In all the other models with control variables, the impact is still significant, but smaller, resulting in a change of more than four standard deviations in the adoption of policies. In the absence of policy-specific opinion, ideology and the thermometers are expected to be the main predictors 26 of public opinion (Brewer, 2003b). For this reason, the thermometer is both theoretically appropriate and the most suitable variable for this context where there is no other data for policies. For these models, the thermometer, the ideology of the legislators and the presence of a Republican governor are correlated with the implementation of policies in a significant and substantial way. Since ideology indicates higher values for conservatism and lower values for liberalism, the negative correlation is to be expected. The positive trend of the thermometer is also expected since the Policy Tally gives higher scores for the implementation of inclusive policies. As Druckman and Jacobs (2006) have pointed out, policymakers are regularly inclined to be guided by general ideology data when there is no policy-specific data available, meaning that ideology could offer a much greater explanatory potential than the thermometer. The significance of the thermometer in this case proves to be an important success of this variable in explaining the implementation of these policies, especially when controlling for ideology, given that it retains its significance even after the inclusion of the variable, which could potentially be a superior explanatory variable for policies that are not commonly surveyed. The high level of R² in the models implies that the adoption of policies for the population is largely driven by these variables, accounting for the variance between implementations. The inclusion of institutional variables revealed the significance of the presence of a Republican governor, which is a reasonable observation given that they have a strong capacity to endorse discriminatory policies and higher chances to veto inclusive policies. However, as this is not the focus of this study, more specific studies on the role of governors are needed to specify their involvement in the adoption of these policies. Included in the appendix is a table with the same regression, but using an aggregate of the inclusive policies presented above (mean of the implementation of the 6 specific policies plus the implementation of access to treatment for public employees by the state healthcare), which presents similar results due to the high correlation between the Policy Tally by MAP and the mean implementation of the other policies. As an additional test to evaluate the effectiveness of the thermometer, in model 5 I decided to add policy implementation level in the same year as the thermometer opinion was collected as a control. With this, I hope to be able to specifically analyze the change in these 2-year periods between t and t+2. As can be seen, even in this test which should further undermine the explanatory capacity of the thermometer, it proved to be one of the few significant variables in this instance, albeit with a smaller effect compared to the other models. This result once again demonstrates the strength of the thermometer in this relationship, 27 suggesting its strong correlation with the implementation of trans policies, and provides evidence in favor of the first hypothesis. Despite the positive results, it should be pointed out that inferring causality in this case is technically not possible, but there is statistical significance and a theoretical foundation to suggest the probable presence of this relationship. Table 3 - Linear Regression (IV: Trans Thermometer; DV: Policy Tally) Variables Model 1 Model 2 Model 3 Model 4 Model 5 12.2*** 4.31*** 4.26*** 4.6*** 2.08** Thermometer (0.837) (0.98) (0.95) (1.08) (0.74) Government ideology - -6.11*** -5.38*** -5.26*** -0.36 (conservatism) (0.8) (0.9) (0.94) (0.75) - -8.37* -6.86 -7.54 3.28 Voter ideology (conservatism) (4.11) (4.93) (5.06) (3.43) - - -0.91 -0.84 0.07 % Republican legislature (3.89) (3.99) (2.59) - - -2.43** -2.49*** -0.77 Republican governor (0.72) (0.73) (0.5) - - 0.16 0.13 1.41 Professionalism (2.11) (2.14) (1.39) - - - -0.49 -1.13* Year (2020) (0.77) (0.5) - - - 1.64 -3.32 Religious conservative (3.52) (2.33) 0.94*** Implementation at the year of - - - - (0.08) opinion -62.1*** -16.53*** -14.49*** -16.47*** -8.94 Intercept (4.72) (5.67) (5.55) (6.28) (4.14) R² 0.68 0.86 0.87 0.88 0.95 AIC 600 526 507 510 426 N 100 100 100 100 100 * p <.05; ** p <.01; *** p <.001 Given the similarity between gay rights and trans rights policies, the comparison between these two analyses suggests that there might be comparable levels of responsiveness to public opinion using the thermometers for the two groups. A version of this table with the ideology indicators by Berry et al. (1998) is included in the appendix. In these the ideology of 28 the government was also significant, with p <0.001, as was the thermometer. The comparison between gay and trans rights is fair, as both groups are similar and their policies fit the morality policy model, but it’s important to note that, since most of the policies for the gay population gained prominence in earlier periods than those for the transgender population, their levels of responsiveness do not match in the same periods, as gay policies have been salient since the late 90’s and early 2000’s. It should be noted that there is a possibility of reverse causality for this topic, for both gay and trans policies. This may occur when the implementation of policies directly affects public opinion of the population for specific policies or the community. However, as Lax and Phillips (2009a) explain when analyzing gay policies, this possibility is small, with the impact of the states in the MRP models being very small or non-existent, suggesting that such results in the thermometers should be largely based on the demographics of the individuals who inhabit them and not on the disparities of implementation between states. Even though it was not possible to include state-fixed effects, the results obtained using thermometers are very promising, suggesting a strong correlation between this variable and the implementation of policies. Graph 2 helps to visualize the change in perceptions of the trans population since the mid-2010s, while also illustrating the new pro- and anti-trans tendencies in the implementation of legislation among the states. It plots MAP's Policy Tally z-score values on the Y-axis as policy implementation where higher values indicate the adoption of inclusive policies for the transgender population. With the gain in prominence that trans rights have obtained following the first discussions on bathroom bills, it would be expected that there would be a movement towards greater responsiveness in the medium and long term, until it stabilizes, with lower levels of responsiveness in the first few years, followed by greater alignment with public opinion (Lax & Phillips, 2012). In 2022, several states that had more negative perceptions of the trans population (below 60% on the thermometer) and their policies reflected these perceptions, which is represented by the Tally score for the absence of inclusive policies or adoption of discriminatory policies, with the opposite happening for states with higher thermometer scores. Like other policies, when taking only public support into consideration, the presence of a majority supporting the implementation of these policies, and a thermometer score around 60% is necessary for this to be reflected in the general adoption of inclusive policies. This reinforces the same phenomenon for other LGBTQ+ policies, which tend to require greater support than a majority, with a 29 tendency to adopt conservative policies in other cases (Krimmel et al., 2016; Lax & Phillips, 2009a, 2012; Taylor et al., 2020). However, it should be emphasized that the implementation of these policies cannot be explained exclusively by the tendency of popular support. Graph 2 – Trans Thermometer and Policy Implementation (MAP Score) 6.2 - Additional Analyses: Moving on to specific policies, I employed logistic regressions to determine the thermometer’s relationship with the implementation of specific trans policies, controlling for the ideology of the population and political elites, as Lax & Phillips (2009a) do in their models of individual policies. For this purpose, I used implementation data between 2018 and 2022 for all the trans-inclusive policies mentioned in the previous section as the dependent variable. By employing the thermometer as an independent variable, as shown in Table 4, it was possible to find a significant correlation for most of the policies analyzed in the models that only include the thermometer as an explanatory variable. This suggests that there may be some responsiveness for these policies in a general context, even if it is not necessarily possible to 30 suggest responsiveness for all of them individually. For the models that include ideology of the population and elites as a control, there is a significant relationship between the thermometer and the implementation of policies that involve protecting the community against discrimination in work and housing, but not for the other policies. Compared to Flores et al. (2015), a study focused on anti-discrimination policies, the significance of the thermometer for these policies was greater than in their results. This is probably due to the trend towards increased responsiveness and salience for these policies since the earlier analysis, contradicting the second hypothesis and suggests that there is indeed responsiveness to this type of policy, especially when using the thermometer. One possible interpretation of this phenomenon is that there is a tendency for elites to underestimate the approval of the communities and consequently, policies for this population, leading to a more conservative bias in the interpretation of opinions and consequently in the implementation of these policies (Broockman & Skovron, 2018), weakening the relationship of responsiveness. However, to achieve a proper explanation, this mechanism will have to be analyzed more in future studies. These results end up contradicting the third hypothesis, suggesting that anti-discrimination policies for the transgender community may be sensitive to public opinion, or at least the thermometer. From this policy analysis, in the period analyzed, responsiveness to these policies seemed to be largely impacted by the ideology of the legislators, rather than by the thermometer. This contradicts the model of morality policy, as it suggests that there should be a greater impact from public opinion than from the interests of elites (G. B. Lewis & Oh, 2008). In some cases, the ideology of voters was also significant, what provides evidence for some level of responsiveness, since it represents the ideology of the population (Lax & Phillips, 2012), even if it is not specifically responsive to this type of policy. The negative value of the thermometer for the anti-bullying policy, also called "safe schools", is apparently an anomaly, while none of the variables used were significant in explaining the implementation of this policy, suggesting that it must have other possible driving factors behind its implementation. As the thermometer has a limited capacity to represent individual support for all trans policies individually, being only partially responsible for building specific public opinion for them (Brewer, 2003a), its use in this case appears to be less appropriate. Another reasonable explanation for the results is that the limited number of cases available for study of the feeling thermometer hinders potential generalizations. Given that for all the policies tested, with the exception of safe schools, there is a positive relationship as was expected, it can be suggested 31 that the variable is in line with the theory, but its lack of statistical significance could be attributed to the size of the sample. Table 4 - Logistical regression of Policy Responsiveness (Inclusive Policies) DV = Employment DV = Housing DV = Safe schools Variables Model 1 Model Model 1 Model Model 1 Model 2 2 2 Trans 7.27*** 6.38** 5.94*** 4.69** 2.7*** -0.13 Thermometer (1.59) (2.06) (1.23) (1.59) (0.58) (0.94) Government - -4.71* - -5.69** - -1.03 ideology (2.01) (2.03) (0.66) (conservatism) Voter ideology - -10.98* - -8.75 - -7.91 (conservatism) (5.09) (4.52) (4.19) Intercept - - - -21.83* - 0.63 40.19*** 31.34** 32.91*** 15.87*** (8.47) (5.42) (8.83) (11.0) (6.83) (3.37) PCP% 84 95 83 94 80 83 AIC 57.8 40.7 67.1 45.0 103 90.8 N 100 100 100 100 100 100 DV = Hate crime DV = Driver’s DV = Private license Insurance Variables Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Trans 4.31*** 1.00 4.2*** 0.34 4.69*** 1.82 Thermometer (0.86) (1.16) (0.84) (1.24) (0.944) (1.22) Government - -2.25** - -2.00* - -1.31 ideology (0.85) (0.85) (0.79) Voter ideology - -5.46 - -9.81 - -5.96 (4.82) (5.69) (4.87) Intercept - -5.46 - -1.59 - -10.73 25.01*** 24.38*** 27.54*** (6.73) (7.18) (7.06) (4.96) (4.83) (5.495) PCP% 84 91 84 90 82 89 AIC 82.1 63.7 83.4 61.7 75.8 66.6 N 100 100 100 100 100 100 * p <.05; ** p <.01; *** p <.001 32 An additional version of this same table is also included in the appendix, but using Berry et al. ideology indicators (1998) constant for all years. In these other scenarios, the results with the thermometer were positive and significant for almost all policies, even including control variables, with the exception of safe schools’ policies, which could probably be explained by the focus of this policy on minors, and the right to rectify driver's licenses. The difference in the results depending on the use of different ideology indicators is due to Shor's (2023) dataset having significant distinctions compared to Berry et al. (1998). These differences include not being based on aggregates, using individual-level measures of legislator ideology, and considering ideological positions of parties in Congress as a proxy for positioning at the state level. These factors suggest that Shor's (2023) approach brings improvements in the measurement of ideology. This difference in control variables could explain the differences in significance found here and in Lax and Phillips (2009a), which utilized ideology data from Berry et al. (1998), found that such control variables were not significant for gay rights or had a very small or minor impact compared to policy-specific public opinion. The high impact of ideology is not necessarily surprising, since ideologies might be better suited to explain the implementation of various policies than policy-specific public opinion (Lax & Phillips, 2009a). As roll-call voting for gay rights is also influenced by ideology and not just opinion (Krimmel et al., 2016), this relationship seems most likely. To assess policy responsiveness of bathroom bills, the most prominent of the trans policies and most likely to fit into the morality policy model, two logistic regression models were used, one employing the thermometer and the other policy-specific public opinion. The table with the regression model with the thermometer as the IV is included in the appendix. For this policy, there was no statistical significance for virtually any of the models tested, suggesting that, even though this policy is the most prominent among the trans policies and the one that should best fit the morality-policy model, there was no responsiveness in the period analyzed. However, interpreting this would lead to a misperception about responsiveness in the long term. This is because in 2016, the year with the highest support for the policy and the lowest thermometer average of all the years used, this policy had not been implemented in any of the states. This analysis therefore suggests a probable inexistence of responsiveness. However, as discussed earlier, the trend towards responsiveness for trans policies suggests that the analysis for the first few years of the policy does not adequately reflect responsiveness, since there will be a trend towards increased relevance of public opinion, even for already salient policies. As bathroom bills only began to be implemented in a much later period, the 33 current situation could be seen as one of late responsiveness. Analyses using specific years also do not suggest responsiveness, but this may also be due to the low N used for each year. Table 5 – Logistic Regression for Bathroom Bill responsive (IV: Support for Bathroom bills; DV: Implementation of Bathroom Bills) Variables Model 1 Model 2 Model 3 Bathroom bills support 7.95 -2.81 -5.23 (4.11) (6.85) (7.35) Government ideology - 3.24* 1.28 (conservatism) (1.36) (1.51) Voter ideology - -0.95 -15.1 (conservatism) (4.49) (7.12) % Republican legislature - - 15.8** (5.98) Republican governor - - 0.92 (1.18) Intercept -6.67*** -2.64 -15.6** (2.29) (1.36) (5.9) PCP% 92 92 92 AIC 83.5 77 70.1 N 150 150 150 * p <.05; ** p <.01; *** p <.001 Considering the trend in implementation of bathroom bills between 2016 and 2023, although not shown in the regression analysis, this policy could be increasingly responsive or congruent when the opinion of the majority of the population is factored in. Throughout this period there was a considerable increase in the congruence between the policy-specific public opinion of the majority of the population, and the implementation of this measure in the states. Even with the constant decline in support for bathroom bills throughout the period, their adoption in 2023 represented the political interests of the majority of the population in 32 of the 50 American states, a considerable increase compared to the early days of wide-scale 34 discussions on the topic. This could point to a case of late responsiveness or a slow stabilization of responsiveness, despite the fact that there was already salience on the subject previously, as these policies only began to be implemented in 2021 in Tennessee, with a peak in 2023 with seven states implementing them in that year. As several states still had support for this policy above 50% in 2021, but many of them have not adopted it, the regression analysis is not the most adequate approach to demonstrate responsiveness in the states at this time of slow but increasing responsiveness. Graph 3 – Support for bathroom bills from 2016 to 2021 The assessment of this specific policy complements the previous analysis suggesting that responsiveness could be increasing for trans policies over time, with a consequent stabilization trend in the long term (Lax & Phillips, 2012). Differences between the estimates of support for bathroom bills between 2020 and 2021 are due to the use of different survey data, with 2020 using ANES and 2021 using PRRI, which generates some distortion in the measurements. However, even with these differences between surveys, this trend could be observed by 2022, when there was 50% congruence with the public opinions of 2020, at a time when only three states had implemented bathroom bills (Alabama, Oklahoma, and Tennessee). In 2023, when 10 states had already introduced such policies, congruence reached 64%, with no state without majority support in 2021 having implemented it. This points to a movement of expansion of this policy in states with greater rejection of the transgender population and support for this policy. This development can already be observed with the continuing debate and implementation of bathroom bills in more states by early 2024 (Trans Legislation Tracker, 2024). 35 Table 6 – Congruence between Adoption of Bathroom Bills and Public Support (above 50% of support) 2018 2022 2023 Congruence with 50% of support (two years before) 32% 50% 64% States with bathroom policies 0 3 10 7 – Conclusion: This research, by conducting an analysis of both policy-specific and a group-based public opinion, succeeds in pushing the understudied topic of trans-policy forward by employing a unique approach in order to address concerns about policy responsiveness through the use of group thermometers as a proxy for public opinion. The lack of polls on most of the policies discussed prevents policy-specific public opinion from being used by politicians, but also opens up the debate that possibly a large part of the policies implemented are not based exclusively on policy-specific opinion, but also on the general perceptions of the population about the group they are aimed at. This method, drawing on the theory of group-centrism, is an important option for studying responsiveness when data on policy-specific public opinion is not available, as well as for examining potential mechanisms of responsiveness that are not necessarily based on individual policies, but on support for the groups, specially if they are highly politized in the political sphere. As discussed in the previous section, there is a tendency for policies that fall under the morality policy model to be more responsive to public opinion, but it is not necessarily guaranteed that this responsiveness will occur immediately after these policies gain in salience, or that it will be perfectly grounded in the opinion of the majority of voters. The development of public debate, the need for polling on policies, party dynamics and the very functioning of the federalist system are factors that ultimately render the dynamic between public opinion and policy adoption more uncertain. Following the analysis of responsiveness with data from the feeling thermometer for the trans population and the implementation of policies aimed at this group, this study has identified a potential trend towards increased political responsiveness to trans rights in the United States, albeit imperfectly between policies and states. Although the relationship was not seen in all the inclusive policies analyzed, this tendency could be identified in the trans rights aggregate, especially when comparing opinion patterns between 2016 and 2020. As observed, the thermometer, as well as the legislative ideology indicator, were significant in most of the scenarios tested, both for some of the policies and for the index. As 36 a result, it has been possible to find evidence for the first hypothesis. However, with regard to the second and third, the results were more unclear, suggesting the necessity of more policy- specific public opinion data or time to observe possible changes in implementation. However, it should be emphasized that this study does not set out to compare the use of the thermometer with that of policy-specific public opinion, as both are potentially useful for different settings. As seen, the indicator of legislators' ideology in many cases proved to be an important variable that could potentially explain the adoption of this policies. This would be connected to the belief that in instances where policy polls are not available, the ideology of governors would be used as a main decision-making tool (Druckman & Jacobs, 2006). Even with improvements at the federal and state levels in the perception of the trans population, there is still incongruence between the interests of the population and the ideological preferences of legislators. This trend was also observed in the analysis of gay rights roll call votes (Krimmel et al., 2016), suggesting that even though many of these policies are expected to be responsive due to them falling under the morality policy-category, the ideology of elites could mediate this relationship, taking a more indirect response, through the reflection of the ideologies of the legislators rather than those of the voters. Overall, this study sought to overcome the various limitations of data availability on trans-rights to estimate whether there is responsiveness to public opinion in the United States for these policies. Although the data and statistical methods used to assess this have their shortcomings, such as a small number of observations, lack of policy-specific data and limitations in the inference of causality for responsiveness, my analysis suggests that there is indeed a potential growing responsiveness to the issue. What can be seen at the moment is that states seemingly are managing to better represent the interests of their voters when it comes to trans rights compared to the mid-2010s, whether this is through the implementation of more inclusive or discriminatory policies. This possible scenario can be observed in Graph 2, which illustrates that between 2016 and 2020 states with higher support for the trans community intensified the implementation of more inclusive policies, with the opposite occurring for states with lower thermometers. When focusing on bathroom bills, even though there is no apparent responsiveness in the regressions, there is a rising trend of congruence motivated by the implementation of this measure in states with greater support for it, as seen in Table 6. This may explain the so-called wave of "anti-trans policies" that has intensified since the same period. This phenomenon is to be expected, given the way the federalist system works and the possibility of backlash between states following the implementation of controversial policies 37 (Hollander & Patapan, 2017). This wave of policies possibly represents a tendency towards greater synchronicity between public opinion and policies for highly salient topics. However, with the growing support for inclusive policies and for the trans community as a whole, the trend could lead to the potential reversal of discriminatory policies at the state or even federal level, as has happened with policies for the gay population, including marriage. Once more polls on trans rights become available in the coming years, it is hoped that the thermometer can be used in conjunction with policy-specific data to analyze responsiveness at the state level or perhaps even at district level. 38 8 – References: American Civil Liberties Union. (2022). Legislation Affecting LGBTQ Rights Across the Country. http://www.aclu.org/documents/legislation-affecting-lgbtq-rights-across- country-2022 American National Election Studies. (2013). ANES 2012 Time Series Study Full Release (May 24, 2016 version) [dataset]. https://electionstudies.org/data-center/2012-time- series-study/ American National Election Studies. (2017). ANES 2016 Time Series Study Full Release (September 4, 2019 version) [dataset]. https://electionstudies.org/data-center/2016-time- series-study/ American National Election Studies. (2021). ANES 2020 Time Series Study Full Release (February 10, 2022 version) [dataset]. https://electionstudies.org/data-center/2020-time- series-study/ Berry, W. D., Ringquist, E. J., Fording, R. C., & Hanson, R. L. (1998). Measuring Citizen and Government Ideology in the American States, 1960-93. American Journal of Political Science, 42(1), 327–348. https://doi.org/10.2307/2991759 Beyer, D., & Hänni, M. (2018). Two Sides of the Same Coin? Congruence and Responsiveness as Representative Democracy’s Currencies. Policy Studies Journal, 46(S1), S13–S47. https://doi.org/10.1111/psj.12251 Brewer, P. R. (2003a). The Shifting Foundations of Public Opinion about Gay Rights. The Journal of Politics, 65(4), 1208–1220. https://doi.org/10.1111/1468-2508.t01-1-00133 Brewer, P. R. (2003b). Values, Political Knowledge, and Public Opinion about Gay Rights. Public Opinion Quarterly, 67(2), 173–201. https://doi.org/10.1086/374397 Broockman, D. E., & Skovron, C. (2018). Bias in Perceptions of Public Opinion among Political Elites. American Political Science Review, 112(3), 542–563. https://doi.org/10.1017/S0003055418000011 Burstein, P. (2010). Public Opinion, Public Policy, and Democracy. In K. T. Leicht & J. C. Jenkins (Eds.), Handbook of Politics: State and Society in Global Perspective (pp. 63– 79). Springer. https://doi.org/10.1007/978-0-387-68930-2_4 Buttice, M. K., & Highton, B. (2017). How Does Multilevel Regression and Poststratification Perform with Conventional National Surveys? Political Analysis, 21(4), 449–467. https://doi.org/10.1093/pan/mpt017 Case, K. A., & Stewart, B. (2013). Intervention Effectiveness in Reducing Prejudice Against Transsexuals. Journal of LGBT Youth, 10(1–2), Article 1–2. https://doi.org/10.1080/19361653.2012.718549 Chudy, J. (2021). Racial Sympathy and Its Political Consequences. The Journal of Politics, 83(1), 122–136. https://doi.org/10.1086/708953 Converse, P. E. (2006). The nature of belief systems in mass publics. Critical Review, 18(1– 3), 1–74. https://doi.org/10.1080/08913810608443650 Druckman, J. N., & Jacobs, L. R. (2006). Lumpers and Splitters: The Public Opinion Information That Politicians Collect and Use. The Public Opinion Quarterly, 70(4), 453– 476. 39 Du Bois, S. N., Yoder, W., Guy, A. A., Manser, K., & Ramos, S. (2018). Examining Associations Between State-Level Transgender Policies and Transgender Health. Transgender Health, 3(1), Article 1. https://doi.org/10.1089/trgh.2018.0031 Erikson, R. S., Wright, G. C., & McIver, J. P. (1994). Statehouse Democracy: Public Opinion and Policy in the American States. Cambridge University Press. https://doi.org/10.1017/CBO9780511752933 Flores, A., Boden, D., Haider-Markel, D., Lewis, D., Miller, P., & Taylor, J. (2023). Taking perspective of the stories we tell about transgender rights: The narrative policy framework. Policy Studies Journal, 51(1), 123–143. https://doi.org/10.1111/psj.12475 Flores, A., Herman, J. L., & Mallory, C. (2015). Transgender inclusion in state non- discrimination policies: The democratic deficit and political powerlessness. Research & Politics, 2(4), 2053168015612246. https://doi.org/10.1177/2053168015612246 Flores, A., Miller, P., Tadlock, B., Taylor, J. K., Lewis, D. C., & Haider-Markel, D. P. (2018). Public Opinion about Transgender People and Policies. In The Remarkable Rise of Transgender Rights (pp. 61–86). University of Michigan Press; JSTOR. http://www.jstor.org/stable/10.3998/mpub.9448956.5 Gelman, A., Lax, J., Phillips, J., Gabry, J., & Trangucci, R. (2018). Using Multilevel Regression and Poststratification to Estimate Dynamic Public Opinion. https://stat.columbia.edu/~gelman/research/unpublished/MRT(1).pdf Gilens, M. (2012). Affluence and influence: Economic inequality and political power in America. Princeton University Press. Haider‐Markel, D. P. (1999). Morality Policy and Individual‐Level Political Behavior; The Case of Legislative Voting on Lesbian and Gay Issues. Policy Studies Journal, 27(4), 735–749. https://doi.org/10.1111/j.1541-0072.1999.tb02000.x Haider-Markel, D. P., & Meier, K. J. (1996). The Politics of Gay and Lesbian Rights: Expanding the Scope of the Conflict. The Journal of Politics, 58(2), 332–349. https://doi.org/10.2307/2960229 Haider-Markel, D. P., & Meier, K. J. (2008). Legislative Victory, Electoral Uncertainty: Explaining Outcomes in the Battles over Lesbian and Gay Civil Rights. Review of Policy Research, 20(4), 671–690. https://doi.org/10.1046/j.1541-1338.2003.00046.x Haider-Markel, D. P., & Vegter, A. (2020). LGBT Politics and the Legislative Process. In Oxford Research Encyclopedia of Politics. https://doi.org/10.1093/acrefore/9780190228637.013.1310 Haider-Markel, D., Taylor, J., Flores, A., Lewis, D., Miller, P., & Tadlock, B. (2019). Morality Politics and New Research on Transgender Politics and Public Policy. The Forum, 17(1), 159–181. https://doi.org/10.1515/for-2019-0004 Hanretty, C. (2020). An Introduction to Multilevel Regression and Post-Stratification for Estimating Constituency Opinion. Political Studies Review, 18(4), 630–645. https://doi.org/10.1177/1478929919864773 Hollander, R., & Patapan, H. (2017). Morality Policy and Federalism: Innovation, Diffusion and Limits. Publius, 47(1), Article 1. 40 Jones, P. E., & Brewer, P. R. (2020). Elite cues and public polarization on transgender rights. Politics, Groups, and Identities, 8(1), 71–85. https://doi.org/10.1080/21565503.2018.1441722 Jones, P. E., Brewer, P. R., Young, D. G., Lambe, J. L., & Hoffman, L. H. (2018). Explaining Public Opinion toward Transgender People, Rights, and Candidates. Public Opinion Quarterly, 82(2), 252–278. https://doi.org/10.1093/poq/nfy009 Kastellec, J. P., Lax, J. R., & Phillips, J. (2019). Estimating state public opinion with multi- level regression and poststratification using R. https://jkastellec.scholar.princeton.edu/publications/mrp_primer Kinder, D. R. (2006). Belief systems today. Critical Review, 18(1–3), 197–216. https://doi.org/10.1080/08913810608443657 Kline, N. S., Webb, N. J., Johnson, K. C., Yording, H. D., Griner, S. B., & Brunell, D. J. (2023). Mapping transgender policies in the US 2017–2021: The role of geography and implications for health equity. Health & Place, 80, 102985. https://doi.org/10.1016/j.healthplace.2023.102985 Kosciw, J. G., Greytak, E. A., & Diaz, E. M. (2009). Who, What, Where, When, and Why: Demographic and Ecological Factors Contributing to Hostile School Climate for Lesbian, Gay, Bisexual, and Transgender Youth. Journal of Youth and Adolescence, 38(7), Article 7. https://doi.org/10.1007/s10964-009-9412-1 Kreitzer, R. J., Hamilton, A. J., & Tolbert, C. J. (2014). Does Policy Adoption Change Opinions on Minority Rights? The Effects of Legalizing Same-Sex Marriage. Political Research Quarterly, 67(4), 795–808. https://doi.org/10.1177/1065912914540483 Kreitzer, R. J., Kane, K. A., & Mooney, C. Z. (2019). The Evolution of Morality Policy Debate: Moralization and Demoralization. The Forum, 17(1), 3–24. https://doi.org/10.1515/for-2019-0003 Kremen, J., Williams, C., Barrera, E. P., Harris, R. M., McGregor, K., Millington, K., Guss, C., Pilcher, S., Tishelman, A. C., Baskaran, C., Carswell, J., Roberts, S., & on behalf of the Gender Multispecialty Service (GeMS) Team. (2021). Addressing Legislation That Restricts Access to Care for Transgender Youth. Pediatrics, 147(5), Article 5. https://doi.org/10.1542/peds.2021-049940 Krimmel, K., Lax, J. R., & Phillips, J. H. (2016). Gay Rights in Congress: Public Opinion and (Mis)representation. Public Opinion Quarterly, 80(4), 888–913. https://doi.org/10.1093/poq/nfw026 Krishnakumar, P. (2021, April 15). This record-breaking year for anti-transgender legislation would affect minors the most. CNN. https://www.cnn.com/2021/04/15/politics/anti-transgender-legislation-2021/index.html Lax, J. R., & Phillips, J. H. (2009a). Gay Rights in the States: Public Opinion and Policy Responsiveness. American Political Science Review, 103(3), 367–386. https://doi.org/10.1017/S0003055409990050 Lax, J. R., & Phillips, J. H. (2009b). How Should We Estimate Public Opinion in The States? American Journal of Political Science, 53(1), 107–121. https://doi.org/10.1111/j.1540- 5907.2008.00360.x 41 Lax, J. R., & Phillips, J. H. (2012). The Democratic Deficit in the States. American Journal of Political Science, 56(1), 148–166. https://doi.org/10.1111/j.1540-5907.2011.00537.x Lewis, D. C., Flores, A., Haider-Markel, D. P., Miller, P. R., & Taylor, J. K. (2022). Transitioning Opinion? Assessing the Dynamics of Public Attitudes Toward Transgender Rights. Public Opinion Quarterly, 86(2), 343–368. https://doi.org/10.1093/poq/nfac014 Lewis, D. C., Taylor, J. K., DiSarro, B., & Jacobsmeier, M. L. (2014). Is Transgender Policy Different? In J. K. Taylor & D. P. Haider-Markel (Eds.), Transgender Rights and Politics (pp. 155–188). University of Michigan Press; JSTOR. http://www.jstor.org/stable/10.3998/mpub.4883502.9 Lewis, G. B., & Oh, S. S. (2008). Public Opinion and State Action on Same-Sex Marriage. State and Local Government Review, 40(1), Article 1. https://doi.org/10.1177/0160323X0804000104 Lopez-Martin, J., Phillips, J. H., & Gelman, A. (2022). Multilevel Regression and Poststratification Case Studies. https://bookdown.org/jl5522/MRP-case-studies/ Mallory, C., & Sears, B. (2020). LGBT Discrimination, Subnational Public Policy, and Law in the United States. In C. Mallory & B. Sears, Oxford Research Encyclopedia of Politics. Oxford University Press. https://doi.org/10.1093/acrefore/9780190228637.013.1200 Manza, J., & Cook, F. L. (2002). A Democratic Polity?: Three Views of Policy Responsiveness to Public Opinion in the United States. American Politics Research, 30(6), 630–667. https://doi.org/10.1177/153267302237231 Markus, G. B., & Converse, P. E. (1979). A Dynamic Simultaneous Equation Model of Electoral Choice. The American Political Science Review, 73(4), 1055–1070. https://doi.org/10.2307/1953989 Mezey, S. G. (2020). Transgender Policymaking: The View from the States. Publius: The Journal of Federalism, 50(3), Article 3. https://doi.org/10.1093/publius/pjaa009 Miller, P. R., Flores, A., Haider-Markel, D. P., Lewis, D. C., Tadlock, B. L., & Taylor, J. K. (2017). Transgender politics as body politics: Effects of disgust sensitivity and authoritarianism on transgender rights attitudes. Politics, Groups, and Identities, 5(1), 4– 24. https://doi.org/10.1080/21565503.2016.1260482 Monroe, A. D. (1998). Public Opinion and Public Policy, 1980-1993. Public Opinion Quarterly, 62(1), 6. https://doi.org/10.1086/297828 Movement Advancement Project. (2023). Equality Maps Snapshot: LGBTQ Equality By State. https://www.mapresearch.org/equality-maps/ Nelson, C. L., Wardecker, B. M., & Andel, R. (2023). Sexual Orientation and Gender Identity-Related State-Level Policies and Perceived Health Among Lesbian, Gay, Bisexual, and Transgender (LGBT) Older Adults in the United States. Journal of Aging and Health, 35(3–4), 155–167. https://doi.org/10.1177/08982643221116762 Nelson, T. E., & Kinder, D. R. (1996). Issue Frames and Group-Centrism in American Public Opinion. The Journal of Politics, 58(4), 1055–1078. https://doi.org/10.2307/2960149 Norton, A. T., & Herek, G. M. (2013). Heterosexuals’ Attitudes Toward Transgender People: Findings from a National Probability Sample of U.S. Adults. Sex Roles, 68(11–12), 738– 753. https://doi.org/10.1007/s11199-011-0110-6 42 Page, B. I., & Shapiro, R. Y. (1983). Effects of Public Opinion on Policy. American Political Science Review, 77(1), 175–190. https://doi.org/10.2307/1956018 Parker, K., Juliana Menasce Horowitz, & Anna Brown. (2022, June 28). Americans’ Complex Views on Gender Identity and Transgender Issues. Pew Research Center’s Social & Demographic Trends Project. https://www.pewresearch.org/social- trends/2022/06/28/americans-complex-views-on-gender-identity-and-transgender- issues/ Pincus, F. L. (1996). Discrimination Comes in Many Forms: Individual, Institutional, and Structural. American Behavioral Scientist, 40(2), 186–194. https://doi.org/10.1177/0002764296040002009 Public Religion Research Institute. (2021). PRRI 2021—American Values Survey (August AVA Wave 3) [dataset]. prri.parc.us.com/client Rigby, E., & Wright, G. C. (2013). Political Parties and Representation of the Poor in the American States. American Journal of Political Science, 57(3), 552–565. https://doi.org/10.1111/ajps.12007 Rigel Hines, T. (2021). Variables Which Influence the Early State Adoption of Pro-LGBT Legislation. Journal of Homosexuality, 68(3), Article 3. https://doi.org/10.1080/00918369.2019.1657751 Rosenfeld, M. J. (2017). Moving a Mountain: The Extraordinary Trajectory of Same-Sex Marriage Approval in the United States. Socius, 3, 2378023117727658. https://doi.org/10.1177/2378023117727658 Ruggles, S., Flood, S., Sobek, M., Backman, D., Chen, A., Cooper, G., Richards, S., Rodgers, R., & Schouweiler, M. (2024). IPUMS USA: Version 15.0 (15.0) [dataset]. [object Object]. https://doi.org/10.18128/D010.V15.0 Shapiro, R. Y. (2011). Public Opinion and American Democracy. Public Opinion Quarterly, 75(5), 982–1017. https://doi.org/10.1093/poq/nfr053 Sharrow, E. A. (2021). Sports, Transgender Rights and the Bodily Politics of Cisgender Supremacy. Laws, 10(3), Article 3. https://doi.org/10.3390/laws10030063 Shor, B. (2023). Aggregate State Legislator Shor-McCarty Ideology Data, April 2023 update [dataset]. [object Object]. https://doi.org/10.7910/DVN/WI8ERB Shor, B., & McCarty, N. (2011). The Ideological Mapping of American Legislatures. The American Political Science Review, 105(3), 530–551. Squire, P. (2007). Measuring State Legislative Professionalism: The Squire Index Revisited. State Politics & Policy Quarterly, 7(2), 211–227. Squire, P. (2023). Replication Data for: A Squire Index Update: Stability and Change in Legislative Professionalization, 1979-2021 [dataset]. UNC Dataverse. https://doi.org/10.15139/S3/2VORBN Stimson, J. A., Mackuen, M. B., & Erikson, R. S. (1995). Dynamic Representation. American Political Science Review, 89(3), 543–565. https://doi.org/10.2307/2082973 Stone, W. J. (1982). Electoral Change and Policy Representation in Congress: Domestic Welfare Issues from 1956–1972. British Journal of Political Science, 12(1), 95–115. https://doi.org/10.1017/S0007123400002830 43 Tadlock, B. L., Flores, A., Haider-Markel, D. P., Lewis, D. C., Miller, P. R., & Taylor, J. K. (2017). Testing Contact Theory and Attitudes on Transgender Rights. Public Opinion Quarterly, 81(4), 956–972. https://doi.org/10.1093/poq/nfx021 Tausanovitch, C., & Warshaw, C. (2013). Measuring Constituent Policy Preferences in Congress, State Legislatures, and Cities. The Journal of Politics, 75(2), 330–342. https://doi.org/10.1017/S0022381613000042 Taylor, J. K., Haider-Markel, D. P., & Lewis, D. C. (2020). Federalism and LGBT Politics and Policy in the United States. In Oxford Research Encyclopedia of Politics. https://doi.org/10.1093/acrefore/9780190228637.013.1201 Taylor, J. K., Lewis, D. C., & Haider-Markel, D. P. (2018). The Remarkable Rise of Transgender Rights. University of Michigan Press. https://doi.org/10.3998/mpub.9448956 Taylor, J. K., Lewis, D. C., Jacobsmeier, M. L., & DiSarro, B. (2012). Content and Complexity in Policy Reinvention and Diffusion: Gay and Transgender-Inclusive Laws against Discrimination. State Politics & Policy Quarterly, 12(1), 75–98. https://doi.org/10.1177/1532440011433589 Tilcsik, A. (2011). Pride and Prejudice: Employment Discrimination against Openly Gay Men in the United States. American Journal of Sociology, 117(2), 586–626. https://doi.org/10.1086/661653 Trans Legislation Tracker. (2024). 2024 anti-trans bills tracker. https://translegislation.com Wang, T., Geffen, S., & Cahill, S. (2016). The Current Wave of Anti-LGBT legislation. The Fenway Institute. https://fenwayhealth.org/wp-content/uploads/The-Fenway-Institute- Religious-Exemption-Brief-June-2016.pdf Warshaw, C., & Tausanovitch, C. (2022). Subnational ideology and presidential vote estimates (v2022) [dataset]. [object Object]. https://doi.org/10.7910/DVN/BQKU4M Weixel, T., & Wildman, B. (2022). Geographic Distribution of Clinical Care for Transgender and Gender-Diverse Youth. Pediatrics, 150(6), e2022057054. https://doi.org/10.1542/peds.2022-057054 Witt, H., & Medina-Martinez, K. (2022). Transgender Rights & the Urgent Need for Social Work Advocacy. Social Work in Public Health, 37(1), Article 1. https://doi.org/10.1080/19371918.2021.1970685 44 Appendix: In this appendix I provide a summary of the variables used in the models as well as complementary regressions that were not included in the main paper. These include tables with alternative measurements of ideology (similar to the indicators used by Lax and Phillips [2009a]) for individual policies, an alternative dependent variable to the MAP Tally using an index of inclusive policies and the use of the thermometer for the analysis of bathroom laws implementation. These results are discussed in the analysis section. Table A1 - Descriptive statistics of the variables: Trans Thermometer N 100 Mean 5.61 Std. error mean 0.0573 Median 5.61 Standard deviation 0.573 Minimum 4.24 Maximum 6.85 MAP Trans Tally N 100 Mean 6.24 Std. error mean 0.843 Median 3.75 Standard deviation 8.43 Minimum -6.75 Maximum 21.8 Gay Thermometer N 150 Mean 5.79 Std. error mean 0.0612 Median 5.83 Standard deviation 0.749 Minimum 4.17 Maximum 7.55 Support for Bathroom Bills N 150 Mean 0.510 Std. error mean 0.00679 45 Median 0.520 Standard deviation 0.0832 Minimum 0.257 Maximum 0.699 MAP Gay Tally N 100 Mean 9.46 Std. error mean 0.674 Median 8.25 Standard deviation 6.74 Minimum -2.00 Maximum 20.3 Implementation of policies: Employment N 100 Mean 0.560 Std. error mean 0.0499 Median 1.00 Standard deviation 0.499 Minimum 0 Maximum 1 Housing N 100 Mean 0.550 Std. error mean 0.0500 Median 1.00 Standard deviation 0.500 Minimum 0 Maximum 1 Safe schools N 100 Mean 0.380 Std. error mean 0.0488 Median 0.00 Standard deviation 0.488 Minimum 0 Maximum 1 Hate crime N 100 Mean 0.400 Std. error mean 0.0492 46 Median 0.00 Standard deviation 0.492 Minimum 0 Maximum 1 Drivers N 100 Mean 0.400 Std. error mean 0.0492 Median 0.00 Standard deviation 0.492 Minimum 0 Maximum 1 Bathroom bills N 150 Mean 0.0800 Std. error mean 0.0222 Median 0.00 Standard deviation 0.272 Minimum 0 Maximum 1 Private Insurance N 100 Mean 0.360 Std. error mean 0.0482 Median 0.00 Standard deviation 0.482 Minimum 0 Maximum 1 Control Variables: Professionalism (Squire) N 150 Mean 0.270 Std. error mean 0.0142 Median 0.234 Standard deviation 0.174 Minimum 0.0480 Maximum 0.840 Republican legislature N 147* Mean 0.548 Std. error mean 0.0161 47 Median 0.585 Standard deviation 0.195 Minimum 0.0592 Maximum 0.900 Republican governor N 150 Mean 0.560 Std. error mean 0.0407 Median 1.00 Standard deviation 0.498 Minimum 0 Maximum 1 Government ideology (Shor-McCarty) N 100 Mean 0.158 Std. error mean 0.0749 Median 0.520 Standard deviation 0.749 Minimum -1.25 Maximum 1.17 Citizen ideology (Warshaw-Tausanovitch) N 100 Mean 0.0570 Std. error mean 0.0157 Median 0.0630 Standard deviation 0.157 Minimum -0.289 Maximum 0.367 Government ideology (Berry et al.) N 50 Mean 39.7 Std. error mean 1.70 Median 33.1 Standard deviation 17.0 Minimum 18.1 Maximum 70.0 Citizen ideology (Berry et al.) N 50 Mean 52.3 Std. error mean 1.61 Median 51.4 Standard deviation 16.1 Minimum 24.0 48 Maximum 97.0 Percentage of religious conservative groups N 100 Mean 0.199 Std. error mean 0.0125 Median 0.169 Standard deviation 0.125 Minimum 0.0404 Maximum 0.704 MAP Tally at the year of the opinion survey N 100 Mean 4.98 Std. error mean 0.746 Median 1.63 Standard deviation 7.46 Minimum -4.50 Maximum 19.0 *The state of Nebraska has a non-partisan state legislature; therefore, it was not included for the measurement of partisanship of the state. Table A2 – Logistical regression of Policy Responsiveness (Inclusive Trans Policies) with alternative ideology dataset DV = Employment DV = Housing DV = Safe schools Variables Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Thermometer 7.27*** 6.47*** 5.94*** 4.94*** 2.7*** 0.78 (1.59) (1.68) (1.23) (1.31) (0.58) (0.72) Government - 0.04 - 0.04 - 0.03 ideology (0.04) (0.03) (0.03) Voter ideology - 0.07 - 0.08 - 0.09* (0.04) (0.04) (0.04) Intercept - - - - - -10.77 40.19*** 40.11*** 32.91*** 32.41*** 15.87*** (3.5) (8.83) (9.58) (6.83) (7.55) (3.37) 49 PCP% 84 88 83 88 80 87 AIC 57.8 53 67.1 58.4 103 88.2 N 100 100 100 100 100 100 DV = Hate crime DV = Driver’s license DV = Private Insurance Variables Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Thermometer 4.31*** 2.41*** 4.2*** 1.83 4.69*** 2.98** (0.86) (0.97) (0.84) (0.99) (0.944) (1.04) Government ideology - 0.08* - 0.07* - 0.08** (0.03) (0.03) (0.04) Voter ideology - 0.03* - 0.08 - 0.02 (0.04) (0.05) (0.04) Intercept -25.01*** -19.11*** -24.38*** -18.17*** -27.54*** -21.76*** (4.96) (5.07) (4.83) (5.07) (5.495) (5.61) PCP% 84 88 84 90 82 88 AIC 82.1 67 83.4 62.5 75.8 64.2 N 100 100 100 100 100 100 * p <.05; ** p <.01; *** p <.001 Table A3 – Transgender inclusive policy index as an alternative to the MAP Tally Model 4 Variables alternative Thermometer 0.18*** (0.05) Government ideology (conservatism) -0.18*** (0.04) Voter ideology (conservatism) -0.74** (0.22) % Republican legislature -0.05 (0.17) Republican governor -0.10*** (0.03) 50 Professionalism -0.14 (0.09) Year (2020) -0.08 (0.03) Religious conservative 0.18 (0.15) Intercept -0.31 -0.27 R² 0.87 AIC -106 N 100 * p <.05; ** p <.01; *** p <.001 Table A4 - Logistic Regression for Bathroom Bill responsiveness (IV: Trans thermometer; DV: Implementation of Bathroom Bills) Variables Model 1 Model 2 Model 3 Thermometer -1.96 -1.64 -5.35 (1.17) (2.75) (5.10) Government ideology - 7.79 13.99 (conservatism) (4.24) (9.08) Voter ideology - -7.67 -47.87 (conservatism) (15.08) (35.38) % Republican legislature - - 54.98 (32.15) Republican governor - - 17.23 (5658) Intercept -6.98*** 0.146 0.09 (5.95) (15.87) (13.3) PCP% 97 97 99 AIC 27.7 26.1 21.5 N 100 100 100 * p <.05; ** p <.01; *** p <.001 51