Support for same-sex marriage among Latin American parliamentarians

Latin American countries differ in the extent to which they have implemented pro-LGBTQ+ laws, how the debates have been framed, the timing of the processes, and whether the judiciaries or legislatures have been responsible for change (Campana and Vaggione Reference Campana and Marco Vaggione2021; Díez Reference Díez2015). Currently, of the region’s eighteen countries, seven have legalized same-sex marriage (Argentina, Brazil, Chile, Uruguay, Colombia, Ecuador, and Costa Rica) and one other (Mexico) offers partial recognition. ^{Footnote 3} Of these, however, court rulings were necessary for the changes in four countries, and only in Argentina, Uruguay, and Chile have legislatures passed favorable laws (2010, 2013, and 2021 respectively). ^{Footnote 4} Legislatures debated the issue but rejected extending such rights in Honduras (2005), El Salvador (2009), the Dominican Republic (2009), Bolivia (2013), and Peru (2018).

To explain variance in legal status, we evaluate legislators’ attitudes toward SSM. As noted, we do this by relying on the PELA-USAL survey, which, after every legislative election, samples a large portion of lower-house legislators for all countries in the region except Cuba. ^{Footnote 5} For this study, in addition to using the survey’s data on partisanship, ideology, gender, and religiosity, we focus on the following question about SSM (labeled VAL1): “on a scale of strongly disapprove (1) to strongly approve (10), how strongly do you approve or disapprove of same-sex partners having the right to marry?”

For most countries, the question was posed twice between 2010 and 2017, and the data indicate that the average support for same-sex marriage across Latin America rose from 4.1 in the first survey wave to 5.0 in the second, with increases everywhere except Argentina, Peru, and the Dominican Republic (details in Appendix Tables A.1 and A.2, and Figures A.1 and A.2). ^{Footnote 6}

Although the regional trendline is upward, the levels of support are inconsistent across Latin America (Figure 1). Unsurprisingly, these levels correlate with the paths toward incorporating new rights addressing LGBTQ+ concerns. The Argentine and Uruguayan cases stand out at one extreme as having particularly high average support scores (8.0 and 7.5) and being the only two legislatures at the time of the data collection that had passed laws to recognize SSM (passage in Chile came several years later). At the other extreme, SSM was banned or voted down in countries where support has been weakest (e.g., Panama 1.9 and Honduras 2.1). The set of countries where SSM was legalized through a judicial process or was partially recognized falls in the middle.

Figure 1. Average support score among legislators for same-sex marriage in Latin America, 2010–2017.

Blue = passed a law; green = approved via court decision; orange = partially recognized; red = illegal. *Chile passed a law in December 2021, but our data does not cover the new Congress.

Confirming that the level of support in a legislature for SSM is clearly tied to whether a country approves legislation justifies the need to search for factors that underlie an individual legislator’s likelihood of being supportive of such laws. Correlates of support that focus on a legislator’s characteristics (gender, ideology, religion) provide a first approximation, but as Table 1 shows, that type of analysis misses intercountry differences such as the varying propensity of religious legislators to support SSM among the different countries. For example, the PELA-USAL survey indicates that even legislators who were “believers” or attended church at least monthly were supportive of SSM (support scores of 7.0 and 5.6, respectively) in countries that approved SSM, while they were staunchly opposed (average support score of 3.0 or 2.7) in countries where SSM is still illegal. The table also shows variance in the propensity of female legislators to support SSM and divisions both among and within the right and the left. We propose that religious context helps to account for these differences.

Table 1. Support for same-sex marriage, legal status, and correlates.

Note: See Table A.3 for variable definitions and averages of included countries, Table A.4 for descriptive statistics by country, Table A.5 for country-level variables descriptive statistics, and Figure A.6 with a related visualization.

* Countries in the respective categories are (1) Argentina and Uruguay; (2) Brazil, Colombia, Ecuador, and Costa Rica; (3) Chile and Mexico; and (4) Venezuela, Peru, El Salvador, Paraguay Bolivia, Dominican Republic, Guatemala, Nicaragua, Honduras, and Panama.

Theoretical framework: Religiosity, context, and contact among legislators

Our suspicion about the important role of religious context on legislators’ support for SSM derives from an amalgamation of theories that highlight exposure to new ideas (Díez and Dion Reference Díez and Dion2018; Riggle, Ellis, and Crawford Reference Riggle, Ellis and Crawford1996), context (Acemoglu and Robinson Reference Acemoglu and Robinson2012; Craig and Cornelius Reference Craig, Cornelius, Almond Gabriel and Verba1989; Flores Reference Flores2014; Putnam, Leonardi, and Nanetti Reference Putnam, Leonardi and Nanetti1993), and social contact (Allport Reference Allport1954; Baker, Ames, and Rennó Reference Baker, Ames and Rennó2020; Becker Reference Becker2012; Dion and Díez Reference Dion and Díez2020; Pettigrew et al. Reference Pettigrew, Tropp, Wagner and Christ2011; Reynolds Reference Reynolds2013). This and other literature we consider combines studies of public opinion (published in journals of psychology and political science) with some on legislative behavior (e.g., Caughey and Warshaw Reference Caughey and Warshaw2018; Erikson, Mackuen, and Stimson Reference Erikson, Mackuen and Stimson2002). We also refer to studies that account for differences across countries and among individuals, although we focus on legislators rather than voters.

In explaining attitudes in Latin America toward progressive rights and whether or not countries adopt related policies, it is essential to consider the prodigious role of the Catholic Church in the region, as well as the growing Evangelical Christian movement (Bohigues et al. Reference Bohigues and Guedes-Neto2022; Boidi Reference Boidi2013; Lodola and Corral Reference Lodola and Corral2010; Maldonado Reference Maldonado2015; Montalvo Reference Montalvo2015; Navarro et al. Reference Navarro, Barrientos, Gómez and Bahamondes2019). Historically, the religious context is at the forefront of explanations for why the region has been slow to adopt changes to divorce and abortion as well as SSM (Blofield Reference Blofield2006; Campana and Vaggione Reference Campana and Marco Vaggione2021; Corrales Reference Corrales2017; Díez Reference Díez2015; Encarnacion Reference Encarnación2018; Merino Reference Merino2013; Olson, Cadge, and Harrison Reference Olson, Cadge and Harrison2006). But while Vaggione (Reference Vaggione2017, 936) notes that “the Catholic Church remains the main challenge to the expansion of sexual rights in Latin America,” he also shows that “dissidences” among Catholic and liberationist evangelical churches are prominent enough that some Argentine legislators cited them when voting on SSM. Overall, he explains that approval in Argentina of divorce laws in the 1980s and SSM in 2010 confirms that the conservative wing of the church is less influential than in the past. Valenzuela, Bargsted, and Somma (Reference Valenzuela, Bargsted and Somma2013) used surveys to show the secularization in Chile, and Corrales (Reference Corrales, Victoria Murillo, Pablo Luna, Falleti and Schrank2022, 22) discussed “the rise of irreligiosity and … light Catholicism” in some parts of Latin America. Although these and other studies do show some decline and variance in the role of religion in the various countries, majorities of legislators across the region maintain religious beliefs. Given the concomitant anti-LGBTQ+ attitudes, few legislatures have approved SSM laws.

In some cases, however, we see growing acceptance of SSM, even among religious legislators. Since Allport’s (Reference Allport1954) articulation seven decades ago, contact theory has been the starting point for studies showing that prejudices can be overcome when in-group members meet and interact with those from the out-group. When applied to the issue of gay or transgender rights (Barth, Overby, and Huffmon Reference Barth, Marvin Overby and Huffmon2009; Barth and Parry Reference Barth and Parry2009; Earle et al. Reference Earle, Mark Romeo Hoffarth, MacInnis and Hodson2021; Herek and Glunt Reference Herek and Glunt1993; Lewis Reference Lewis2011; MacInnis, Page-Gould, and Hodson Reference MacInnis, Page-Gould and Hodson2017; Merino Reference Merino2013), the theory suggests that personal interaction with someone from these communities will decrease biases and increase support for SSM and other progressive policies. Reynolds (Reference Reynolds2013, 271) finds that “even in small numbers, out lesbian, gay, bisexual and transgender MPs in national legislatures encourage the adoption of gay-friendly legislation.” Díez and Dion (Reference Díez and Dion2018) adapt this theory to show that a change in attitude does not require direct contact with members of the out-group, only exposure to ideas about SSM and interactions with supportive fellow citizens, even if through digital forms (see also Hoffarth and Hodson, Reference Hoffarth and Hodson2018).

Theories that focus on contact or other variables to explain attitudes among individuals often struggle to interpret differences in average tendencies among countries or across other types of subpopulations. The primary approach to explaining differences among countries (or subpopulations) is to consider how everyday surroundings infuse individuals with corresponding values. Literature in this vein convincingly shows that context explains why societies differ on themes as diverse as social capital (Putnam, Leonardi, and Nanetti Reference Putnam, Leonardi and Nanetti1993), views on democracy (Acemoglu and Robinson Reference Acemoglu and Robinson2012; Craig and Cornelius Reference Craig, Cornelius, Almond Gabriel and Verba1989), and the nature of economic voting (Singer and Carlin Reference Singer and Carlin2013). To combine the levels of analysis, Hodson (Reference Hodson2009) applied models that interact personal characteristics and contact in explaining prejudices. Díez and Dion (Reference Díez and Dion2018; Dion and Díez Reference Dion and Díez2017, Reference Dion and Díez2020) found support for this approach through a series of publications in which they interact personal (e.g., religion but also media usage and contact with members of the LGBTQ community) and contextual variables (e.g., population density, economic and political development) to explain support for SSM. Merino (Reference Merino2013) added the importance of friend networks as a mediator of contact. Earle and colleagues (Reference Earle, Mark Romeo Hoffarth, MacInnis and Hodson2021, 851) then confirmed the approach through a large-N cross-country study, finding “importance of both individual and contextual factors in predicting support for LGBT communities.” While these studies focused on religion at the individual level, only one—Díez and Dion (Reference Díez and Dion2018)—tested for the role of religion at the contextual level. ^{Footnote 7} They found the national-level variable significant in creating “baseline” differences among countries, but their focus was on individual-level media exposure. We still lack, then, tests that consider how the level of secularism impacts the legislature as a whole and religious legislators in particular.

Translating contact theory and contextual analyses to a study of legislator attitudes is not straightforward, as extensive social and political forces work to influence legislators’ opinions directly and shape their policy positions. Further, as Granovetter (Reference Granovetter1973) and Kirkland (Reference Kirkland2011) explained, legislators are part of a tight network composed of both strong and weak ties. Thus, contact is extensive, and although it can be casual, it is often intended to bring about attitudinal changes. This conception of the legislature is the basis of Fowler’s (Reference Fowler2006) classic study that maps different leaders’ networks, often based on cosponsorship patterns (Kirkland Reference Kirkland2011; Kirkland and Gross Reference Kirkland and Gross2014). In the Latin American context, Alemán (Reference Alemán2009), Alemán and Calvo (Reference Alemán and Calvo2013), Barnes (Reference Barnes2016), and Bonvecchi, Calvo, and Stein (Reference Bonvecchi, Calvo and Stein2016) have also shown complex legislative networks based on factors such as gender, partisanship, committee membership, and region. Our goal, however, is somewhat different; instead of trying to uncover which legislators are the centers of networks, we are interested in how the context—in our case, the level of religiosity within the legislature—generates shifts in attitudes. This application is still related to contact theory, but because we are not evaluating how specific contact with someone from an out-group moderates prejudices from an in-group member, our framework strays from typical usage. We are concerned with how exposure to colleagues who hold a different worldview might lead to changes in a particular policy position. Our review of the literature does not uncover tests of this type, which perhaps fit between context and contact, although our application is not unrelated to Reynolds’s (Reference Reynolds2013) analysis of the relation of LGBT colleagues and policy change, or to Díez and Dion’s (Reference Díez and Dion2018) consideration of media exposure. We contend that, because daily interactions among legislators meet conditions critical to contact theory—they are repeated, lengthy, and extensive (Pettigrew Reference Pettigrew1998)—this type of tertiary contact has the potential to shape policy attitudes. In summary, we expect the following

These hypotheses are not meant to discount the importance of electing more progressive legislators, the critical influence of legislators’ personal religious beliefs, or other variables typically associated with individual or cross-country differences in support for LGBTQ rights. Previous work has found the gender balance and ideological skew of the legislature telling (Becker Reference Becker2012; Gaines and Garand Reference Susan and Garand2010), as well as levels of development or postmaterialism (e.g., Inglehart and Welzel Reference Inglehart and Welzel2005; Navarro et al. Reference Navarro, Barrientos, Gómez and Bahamondes2019). Other studies, such as Friedman (Reference Friedman2009), considered the ideology of the presidents, and Corrales (Reference Corrales, Victoria Murillo, Pablo Luna, Falleti and Schrank2022) highlighted many exceptional cases, especially among leftist presidents not supportive of SSM. At the individual level, as noted, numerous studies find telling the role of ideology, gender, and religion. We accept the premises of these other studies but use our hypotheses to turn attention to how the religious context shapes political attitudes, regardless of a legislator’s personal religious beliefs (or other factors).

Legislative debates and religiosity in three cases

To illustrate the plausibility of our argument, this section examines the progress—or lack thereof—toward the legal recognition of SSM in three countries, focusing on the level of religiosity and considering party politics, gender, and ideology. Our goal is not to provide a full accounting of the policy process (for such an analysis, see Díez Reference Díez2015) or an alternative test that goes beyond our survey data. Instead, we use these examples to provide background and build a prima facie case that common explanations of support for SSM are insufficient but are strengthened by adding religion as a contextual variable.

For these case studies, as a proxy for our main independent variable, religious context, we evaluate a survey question that asks whether respondents are “believers.” ^{Footnote 8} We then chose Argentina, Chile, and Costa Rica as our cases because they illustrate the impact of varying the level of a legislature’s religiosity while also holding relatively constant other potential explanations (the level of democracy and economic development). The discussions of the cases show some role for ideology and partisanship, but they also exhibit evidence for the hypothesis that as the number of secular legislators grows, the propensity of their colleagues to support same-sex marriage also grows, regardless of whether they are religious. Individual views, then, are influenced by the religious context.

Argentina: Low religiosity

In 2010 Argentina’s lower house passed a bill legalizing SSM with a vote of 126–110 (plus four abstentions and sixteen absences), it passed the Senate 33–27 (three abstentions), and then Cristina Fernández became Latin America’s first president to sign such legislation. More important to this outcome than party politics or ideology, the data suggest that the relatively high number of irreligious legislators provided the necessary context.

Historically, Argentina’s political arena has been dominated by two main parties, the center-right Radical Civic Union (UCR) and the populist (but ideologically divided) Peronist or Justicialist Party (PJ). Fernández was the head of the left-wing of the Peronist party (then called the Front for Victory), but neither she nor her party campaigned on the issue of SSM. In 2009, however, Mauricio Macri, the future president who at the time was her center-right political rival and mayor of Buenos Aires, decided against appealing a judicial decision that allowed SSM in the country’s largest city. In response, Fernández decided to actively support a pro-SSM bill, framing her support in terms of democracy, human rights, egalitarianism, and nondiscrimination (Encarnación Reference Encarnación2018).

Although Fernández’s support was critical, the outcome was not decided by partisanship or ideology. The bill passed, but Fernández was unable to bring more than a bare majority of her copartisans to support it (53 percent). Adding ideology does not provide a full accounting, either: although leftists were likely to support SSM, so were many who identified as centrist or even center-rightist. Among the Peronists, thirteen of sixteen (81 percent) who reported being leftist (scores of 2–3 on a ten-point ideological scale) gave full support to SSM, but so did fourteen of twenty-nine (49 percent) who identified as centrist (scores of 4–7). Most UCR members score their own ideology as 4 or 5, and of these, sixteen of twenty-six (62 percent) supported SSM (scores of 8 or higher). Ideology does not seem a good contextual variable either, given that Argentina has a smaller proportion of leftists than other countries. ^{Footnote 9}

Gender also appears relevant, given that Argentina has a high percentage of female legislators (37 percent), 86 percent of whom supported SSM, compared with 62 percent of men. The differential in support levels between the genders is greater than in other countries (average rates for all other countries are 35 percent for men and 40 percent for women). Gender, however, does not seem an adequate country-level explanation, because other countries such as Bolivia, Nicaragua, and Mexico display lower average support scores for SSM and have not approved pro-SSM laws, despite having even more women in Congress.

Religion has a much stronger relationship with attitudes toward SSM in Argentina. For instance, of Peronist believers, 48 percent (15 of 31) of believers were fully supportive (a score of 10 on the scale) of SSM, compared to 86 percent (12 of 14) of nonbelievers in the party (Figure 2).Footnote ¹⁰ Among the UCR legislators, these values were 30 percent (of 33) versus 60 percent (of 5). The divide is similar when analyzing ideology; 50 percent (of 14) of leftist believers were fully supportive versus 78 percent (of 23) of nonbelievers. ^{Footnote 11}

Figure 2. Argentinean legislators’ support for same-sex marriage (2010 and 2012).

Argentina, like other countries, exhibits a secular-religious divide, but the level of support for SSM among Argentine (and Uruguayan) believers, regardless of ideology and partisanship, is unique. Specifically, the percentage of religious Argentine (and Uruguayan) legislators who fully support SSM is four times that found in other countries (40 percent versus 10 percent). Which intercountry variables can explain this variance? Argentina does not stand out in terms of ideology or gender, but with nearly one-third of its members claiming that they are not believers, Argentina (again, along with Uruguay) is qualitatively different from other countries in terms of religiosity. Having a high preponderance of nonreligious colleagues thus appears to be a good candidate for explaining the acceptance of progressive ideals among Argentine legislators.

Chile: Moderate religiosity

Until December 2021, Chile was the only country in the Southern Cone without legalized SSM. The country recognized civil unions in 2015 under President Michelle Bachelet, who headed a center-left coalition (Concertación). In accord with domestic and international pressure favoring LGBTQ+ rights from the Inter-American Commission on Human Rights (IACHR), Bachelet introduced a bill recognizing SSM in 2017. Soon thereafter, however, a conservative president (Piñera) took power and the bill languished until December 2021, when it was approved in the lower house by 82–20 (with two abstentions and fifty-one absentees). Just months before he stepped down, Piñera signed the bill, and it went into effect in March 2022.

This surprising outcome shows the evolution of views, among the right, the left, and importantly the centrist Partido Demócrata Cristiano (PDC). We lack survey data for the recent congress, but comparing the 2020 lower house voting results to those in our 2014 survey of the lower house suggests that the main leftist party (Party for Democracy, or PPD) and the PDC have overcome divisions and sided uniformly in favor of same-sex marriage. ^{Footnote 12} On the right, the two main parties (Unión Democrática Independiente, or UDI, and Renovación Nacional, or RN) aligned against the liberalization at the time of the 2014 survey, but many voted in favor in 2020 (UDI: eight yes and eight no; RN: eleven yes and seven no). ^{Footnote 13} As an illustration of the shift in positions, while the populist José Antonio Kast exclaimed in 2017 that the government had “surrender[ed] to the gay dictatorship,” during his 2021 presidential campaign he stated that he would respect the newly approved SSM. ^{Footnote 14}

The shifting attitudes were already evident in the 2010–2014 data, when, perhaps in response to growing domestic and international pressure to adopt policies friendly to the LGBTQ+ community, the percentage of SSM supporters in the legislature grew substantially, from 19 percent to 55 percent, even though the partisan mix did not change much. Why, however, was the pressure more successful in Chile than in other countries? The levels of development or democracy seem inadequate for the explanation, as we held those variables constant in our choice of cases yet still witnessed variable rates of change. Ideology is not a promising variable either, as the average Chilean legislator was significantly to the right of counterparts in Argentina (average ideology scores of 5.2 versus 4.3). The Chilean legislature is also to the right of other countries that have not seen increasing support for SSM. Next, the likelihood of female legislators in Chile expressing support for SSM grew from 40 percent to 75 percent between 2010 and 2014, but the legislative gender balance fails as a part of a comparative explanation, because other countries with greater proportions of women did not experience such great changes in the propensity to support. ^{Footnote 15} Women’s high level of support for SSM was important to the 2021 vote (providing 29 percent of the favorable votes with 23 percent of the seats), and the ongoing constitutional reform process may lead women to have even greater representation and political weight in the future. Women’s numbers, however, have remained too low to suggest that the gender balance would prove significant as a contextual variable in a comparative test.

Perhaps even more noteworthy than the much increased acceptance of SSM among female legislators, the proportion of religious legislators who were willing to voice support for SSM grew from just 12 to 46 percent in the years under study. Figure 3 shows that many religious members of all parties except the UDI were at least moderately favorable by 2014 (and about a third of the UDI legislators voted in favor of SSM in 2021). Consistent with our theory, these changes occurred in a legislature where the level of religiosity was moderate. In Chile, 80 percent of legislators did claim religious belief, but this is ten percentage points less than any other country except the two where support for SSM was very high (Argentina and Uruguay). Most countries have faced societal and international pressures to liberalize their policies, but only in Argentina, Chile, and Uruguay has there been a significant increase in support. This implies that a country-level variable—which we suggest is the number of secular legislators—facilitated the pressure campaign in moving some oppositionists. Perhaps, then, the Chile case indicates the level of secularism necessary to induce some religious legislators to withdraw their objections to progressive rights policies.

Figure 3. Chilean legislators’ support for same-sex marriage (2010 and 2014).

Costa Rica: High religiosity

In 2020 Costa Rica became the first country in Central America to legalize SSM. The change, however, was possible only via a Supreme Court decision, as almost all legislators (94 percent) were religious and, in consequence, legislative support for SSM remained very low (average support score of 3.6).

Support for SSM has not increased despite having parties that lean left and promote other progressive policies. Figure 4 shows the support of SSM for each of Costa Rica’s five main political parties, averaging responses for 2010 and 2014. The largest party, the National Liberation Party (PLN) has promoted a centrist ideological platform with social democratic policies, but well under 20 percent of its members were supportive of same-sex marriage and more than 60 percent were in complete opposition. Legislators of the next-largest party, the Citizens’ Action Party (PAC) were divided in their support, and this was despite the party’s advertising a progressive, socially democratic platform (average ideology score of 5.0). A final example was the Broad Front (FA); despite claiming a leftist position on the ideological scale (average score of 2.8), seven of its nine legislators declared a religious belief. Of those, two were fully opposed to SSM and only three were fully supportive (the two nonbelievers also supported SSM). Perhaps it is notable that in the 2018 election, the party lost all its seats.

Figure 4. Costa Rican legislators’ support for same-sex marriage (2010 and 2014).

Although the high level of religiosity is consistent with the low rate of support for SSM, other country-level variables fail to show much impact. In comparative perspective, Costa Rica scores high in terms of the number of female legislators, but just 20 percent indicated unequivocal support for SSM, a stark contrast from the 75 percent in Chile. Country-level ideology also seems unimportant, as the average ideological position among Costa Rican legislators in 2014 (5.1) was essentially the same as in Chile (5.2).

Bivariate tests

Before turning to a multivariate model, we discuss bivariate relations in support of our primary hypotheses and show that other standard correlates with progressive policies fail to explain intercountry differences in the support of SSM. We begin by operationalizing religiosity in three alternative ways to show its profound effect—at the individual and contextual levels—on views about SSM.

The first proxy for religiosity is whether respondents answer that they are believers, as we applied in the case studies. This variable is highly skewed, with most countries registering averages of more than 90 percent. In addition to a high correlation at the individual level (r = –.36), believer status is clearly related to intercountry legal status, as the two countries that passed laws to legalize SSM have far fewer believers than others (Table 1). Average support scores for SSM are also (negatively) correlated with degree of religiosity (r = –.37). Operationalizing the variable as those who attend church at least once per month yields a similar outcome. ^{Footnote 16} To illustrate, 50 percent of those who attend church regularly fully oppose SSM, and fewer than 8 percent are supportive. By contrast, among nonattendees, just 25 percent fully oppose and 31 percent fully approve of SSM. Unsurprisingly, there is also a clear relation at the country level, with a correlation coefficient of more than 0.80 (n = 18). Interacting the levels, we also see that many regular church attendees do not take ardent stances against SSM in the countries where attendance is weaker. Where national attendance is 60 percent or less, 49 percent of regular attendees support SSM, whereas in all other countries that percentage is 23 percent. Finally, recognizing that the Evangelical Christian movement has held a powerful veto to counter liberalization (Corrales Reference Corrales2017, Reference Corrales2021), we tested whether legislators’ evangelical identity affects our thesis. The data show that in countries that stand out for high levels of evangelical legislators (Brazil and Guatemala, with 12 percent and 17 percent, respectively, compared with 5 percent in the rest of the region), the rate of opposition to SSM is high. But because it is also high in other countries, the believer variable, which encompasses evangelicals, provides a better demarcation.

Moving to other explanatory variables, we first evaluate gender. Women are more supportive of same-sex marriage than men by a seven-point margin (36 percent versus 43 percent), and there were marginally more women in the countries where same-sex marriage was approved than in other countries. The three-point range (18 percent versus 21 percent) between the size of the female cohort in countries where the same-sex laws were approved and those where it is still illegal, however, is insufficient to explain the different policy outcomes. Further, several countries defy the relation. For example, and in addition to the Chilean anomaly discussed earlier, Uruguay has the highest level of support for SSM (average 8.1) in the region but an exceptionally low percentage of female legislators (22.2 percent in 2014) and Nicaragua presents the opposite relation, with many female legislators (44.6 percent in 2016) but low average support (2.9). ^{Footnote 17} In sum, while women as individuals are more supportive, the gender balance does not seem significantly related to overall support levels.

Next, the data show differences among parties between leftists and rightists in the propensity of legislators to support SSM, but case studies suggested that average ideology cannot explain country-level support, and the bivariate data uphold that finding. In our data, 51 percent of legislators who scored themselves as far left (3 or less on the ideological scale) also support SSM; for the far right (score of at least 8), 80 percent were in opposition. Pitting average ideology versus average support for SSM (Appendix Figure A.7) yields a moderately downward slope that is mostly driven by homogeneity among countries on the right. That is, although legislatures leaning right all have low average support levels, and the countries with the highest level of support for SSM have legislatures that lean left, several other far-left legislatures exhibit more opposition to the policy. Thus, while there is no evidence that a country’s average ideology determines the level of support for SSM, there is evidence that an outsized proportion of rightist legislators does have an effect.

The next standard expectation is that a left-leaning president will pull more legislators toward support for LGBTQ+ rights and a rightist will have the opposite impact. However, as Corrales (Reference Corrales, Victoria Murillo, Pablo Luna, Falleti and Schrank2022) also recognized, a review of patterns in the various countries does not provide support for this idea. The standard expectation is met for Argentina and Uruguay. Leftist presidents were also leading countries that had largely supportive legislatures but incorporated LGBTQ+ rights through judicial decisions (Brazil and Ecuador). But the relationship is not upheld in other countries.. A leftist president was coincident with partial recognition and increased legislative support in Chile in the mid-2010s, but a rightist president supported the policy in 2021. In Mexico the leftist president has opposed SSM, and many of the other countries headed by leftist leaders (e.g., Bolivia, Nicaragua) have remained at the very bottom in terms of legislative support for same-sex marriage.

The final alternative explanatory variables are those that traditionally correlate with progressive policies (Appendix Figures A.7 and A.8). First, age shows a clear relation with support for SSM at the individual level (younger legislators are more supportive), but there is not much variation in average age at the country level. Education has a counterintuitive pattern. It does produce some cross-country variance, but while university-educated legislators are more likely to support SSM (39 percent versus 31 percent), Argentina and Uruguay rank poorly in terms of the percentage of legislators who have completed their university education. The next potential explanation is based on Dion and Díez (Reference Dion and Díez2017), who conjectured that attitudes toward democracy covary with SSM because both variables point toward “social tolerance.” Our survey data does not allow a direct test of this thesis, because, unlike voters, almost all legislators agreed that democracy is always the best form of government. ^{Footnote 18} As an alternative, we follow Inglehart and Welzel (Reference Inglehart and Welzel2005) and Booth (Reference Booth2017), who proxy postmaterialism with attitudes about the environment. PELA asks legislators to weigh the importance of the environment on a ten-point scale, which yields an average of 7.5 (standard deviation = 2.1). Even at the individual level, however, the relation of this variable and views of same-sex marriage is very weak. We also considered the level of democracy and per capita gross domestic product (Inglehart, Reference Inglehart and Welzel2005; Corrales Reference Corrales2017), ^{Footnote 19} but while levels on these variables are much higher in most countries that have approved SSM than in those where it is still illegal, Chile’s change came late and other countries, including Costa Rica (as noted in our case study) and Panama, are severe underperformers. ^{Footnote 20}

In summary, although the data do indicate expected legislator-level correlations between the support for SSM and factors such as ideology, gender, and religious belief, the data show continued intercountry differences. To explain these, we showed limited impacts for the legislative gender balance, the degree to which legislatures skew to the left or have higher economic development (although environmentalism does not increase support). In contrast, the level of religiosity of the legislature has a telling influence, such that a more secular environment increases support for SSM among secular and religious legislators alike.

Multivariate analysis

We now employ multivariate models to validate the small-n analyses and disentangle the influence of individual- and country-level variables. Our goal in these models is to show that religion works as a contextual variable, and when the degree of secularism in the legislature rises, even religious legislators are more likely to show support for SSM. To preview, these tests confirm our hypotheses.

In all our model specifications, our dependent variable is support for SSM, which has a scale of 1 (fully opposed) to 10 (fully supportive). Our models are built to test the role of religiosity on individual legislators and among countries as the prime independent variables, but they also test, at the individual level, ideology, gender, postmaterial views (with environmental concerns as a proxy), and at the country level, age, the ideology of the president and the legislature, gender balance, and per capita gross domestic product. ^{Footnote 21} In the base model we operationalize religiosity using the question about belief and we measure ideology as each legislator’s self-identification on the ten-point scale. At the country level, all variables are the aggregates of the individual level, except ideology, which we measure as the percentage of far right legislators to accord with our bivariate finding. ^{Footnote 22} In robustness tests, we alter these definitions and find similar results. Modeling the impacts of individual and contextual factors allows for several options, depending on how to handle variability at the two levels. Our first option (Equation 1) is a model that clusters errors by country-year. In that equation, y _i captures the position of individual i regarding same-sex marriage. The vector X _i contains the individual-level predictors such as whether the representative is a believer, or female, and the magnitudes of the effects are captured with β coefficients. The impact of the country-level variables (Z _j[i]) is measured by the γ coefficients, with the subscript indexing legislator i in country-year j (notation follows Gelman and Hill Reference Gelman and Hill2006).

Consistent with Arzheimer (Reference Arzheimer2009), we also estimate a hierarchical model that includes a random intercept for country-year (Equation 2), where μ _j[i] indicates the hierarchical random intercept. This allows us to include our context-level variables, which are constant within a country-year, while also addressing concerns about country-year heterogeneity. All models estimate robust standard errors clustered by country to account for heteroskedasticity and correlation of the error terms:

(1) $${y_i}\; = \;\beta {X_i} + \;\gamma {Z_{j\left[ i \right]}}\; + \;{\varepsilon _i}\;$$

(2) $${y_i}\; = \;\beta {X_i} + \;\gamma {Z_{j\left[ i \right]}}\; + \;{\mu _{j\left[ i \right]}} + \;{\varepsilon _i}\;$$

Figure 5 depicts results from the two base models (model results are in Table 2), with the marks representing the coefficients surrounded by 95 percent confidence intervals. The results are substantively in line with our theoretical expectations. While several variables gain statistical significance at the individual level (age, gender, education, ideology), religiosity stands out as the most meaningful. Specifically, the model predicts that legislators who indicate that they are believers will be 2.3 points below others in terms of support for SSM. This implies an impact of more than twice that of gender or university education, each of which add about one point to expected level of support. Regarding age, the coefficient is negative and significant at the 90 percent level, but moving from ages 30 to 80 would still change support by only about one point. ^{Footnote 23}

Table 2. Multilevel analysis of the legislators’ support determinants of same-sex marriage (N = 2,386).

Notes: Dependent variable is support for same-sex marriage. Standard errors in parentheses.

*p < 0.1. **p < 0.05. ***p < 0.01.

Finally, ideology is also negative and significant at the 99 percent level, which means that a legislator on the right extreme (ideology score of 10) is, on average, 1.35 points less supportive than a moderate legislator (ideology score of 5), which is mirrored for a legislator on the extreme left (ideology score of 1). Moving to the country level, average religiosity is the only variable with statistical significance in both models. Substantively, a country where all legislators are religious will have an average SSM support score that is six points lower than one in which the legislature is fully secular. These results, therefore, provide empirical support for Hypothesis 1.

Figure 5. Model results: legislators’ support of same-sex marriage.

Dependent variable: SSM (1 = fully unsupportive, 10 = fully supportive). Robust standard errors clustered by country.

Using the hierarchical version of the model, the last column in Table 2 displays results of a test of Hypothesis 2 by adding an interaction term between individual legislators’ religiosity and the percentage of secular legislators. ^{Footnote 24} Figure 6 illustrates marginal effects to show that as the percentage of religious leaders decreases from about 100 percent (similar to Panama) to 40 percent (as in Uruguay), the average support score of a religious legislator rises from about 3.1 to 8.2. ^{Footnote 25} The change for nonreligious legislators is from about 9.1 to 6.9 over that span. In short, religious context drives down support for SSM, even when accounting for individuals’ own beliefs.

Figure 6. Marginal effects for individual and country-level of religiosity.

Dependent variable: SSM (1 = fully unsupportive, 10 = fully supportive). Robust standard errors clustered by country.

To ensure the robustness of our results, we ran multiple alternative models, all of which continue to support our hypotheses on the role of religiosity at both the individual and contextual levels (Tables A.6 and A.7). ^{Footnote 26} First, to confirm that our outlying case, Uruguay, was not driving the results, we eliminated that case and reran the model. ^{Footnote 27} The next two tests changed the measure of religiosity, first using a dummy for evangelicals and then testing for whether a legislator attends church at least once a month. ^{Footnote 28} Neither specification changed the theoretical conclusions. The next variant changed the individual level ideology variable into a nonlinear format, with dummies for the far left and right. The individual level variables continued to show a small effect (yielding a two-point difference in predicted differences, like the original specification), but the variables testing ideology at the country level are significant in only one of the twenty-two different tests. We then incorporated a dummy for leftist president, which was positive but statistically insignificant. Again, the coefficients for religiosity (individual and contextual) in that test remained statistically and substantively unchanged. Next, we included several interactions, and while two coefficients did reach statistical significance, our primary variables of interest still held up to the stringent test.

In Models 7–11 we retest the hypotheses using our hierarchical model, which also includes all country-level control variables together. We also include there control variables about public opinion, to rule out the possibility that changes among voters were driving the results. To consider this possibility, we included the proportion of citizens who attend church at least once per year and the percentage who support SSM, both of which we drew from the AmericasBarometer. The tests show some significance for the importance of the public’s support for SSM, but not level of religiosity among citizens. Most important, our key variable, level of religiosity in the legislature, continues to be significant. Finally, we added fixed effects for country and year, finding, again, that the change has no significant impacts on our religiosity variable. In those models, the percentage of women in the legislature gains statistical significance, but the substantive impact is still very small.

An additional concern regards the direction of causality. Some of our opinion data are subsequent to legal changes, which could influence legislators’ opinions. We are not generally concerned with this possibility for two reasons. First, Lax and Phillips (Reference Lax and Phillips2009) argue that elites’ opinions shape policies rather than the reverse. There are theoretical reasons to suppose that legislators’ opinions affect policy (they are the policy makers) but limited reasons to claim that legislators will adopt preferences only to match with the state’s policy. As an example, passage of abortion rights in the United States hardened the opposition. Second, the level of religiosity has not changed in response to legal changes. In Argentina, for example, more legislators called themselves believers in 2012 than in 2010, and the same is true for Chile when comparing 2010 and 2014. Even more, the other variables that explain variation in support for LGBTQ+ rights are mostly demographic characteristics, which are relatively fixed over time. Still, we reestimated the main models, dropping cases where the opinion data is subsequent to full or partial legalization (Table A.8) and found only limited substantive differences in the results. ^{Footnote 29}