Institutional Paths to Policy Change: Judicial Versus Nonjudicial Repeal of Sodomy Laws

Abstract

What variables lead judicial and nonjudicial decision-making bodies to introduce policy change? In the theoretical framework proposed, the path-dependent nature of law has a differential impact on courts and legislatures. Likewise, certain political institutions including elections and political accountability lead those bodies to introduce policy change under dissimilar circumstances. Global trends, however, affect both institutional paths equally. We test this theory with data for the repeal of sodomy laws in all countries from 1972–2002. Results from two disparate multivariate models overwhelmingly confirm our predictions. The unique institutional position of courts of last resort allows them to be less constrained than legislatures by either legal status quo or political accountability. Globalization, on the other hand, has a comparable effect on both. This work is path breaking in offering a theoretical framework explaining policy change via different institutional paths, systematically testing the framework comparatively and with respect to a policy issue still on the agenda in many countries.

In 1969, the Canadian Parliament passed the Criminal Law Amendment Act. Among other sweeping changes to the existing criminal code, the Act decriminalized consensual same-sex relations, effectively repealing sodomy laws, which had been a part of Canadian jurisprudence since its inheritance of the English Buggery laws (McLeod 1996). A mere two years earlier, the Supreme Court of Canada had upheld the conviction of Everett George Klippert who had been sentenced to an indefinite prison term as a “dangerous sexual offender” for engaging in consensual sex with another man (Klippert v. the Queen 1967). The Court's decision elicited strong condemnation from the Canadian legal community, causing Justice Minister Pierre Trudeau to table legislation in the House of Commons to amend the criminal code (Kinsman 1995). Heavily influenced by the United Kingdom's 1957 Wolfenden Report, the repeal initiative was added to a larger reform effort that called for the creation of a “zone of legal privacy … for acts committed between consenting adults.” The combination of global trends and legal evolution led to significant policy change emanating from the Canadian legislature and culminated in passage of the 1969 Criminal Law Amendment Act.

Three decades later, the South African Constitutional Court ruled in National Coalition for Gay and Lesbian Equality and Another v. Minister of Justice and Others (1998) that three legal provisions pertaining to sodomy were all “inconsistent with the Constitution of the Republic of South Africa” including Section 20A of the Sexual Offences Act of 1957, which banned all sexual contact between males, Schedule 1 of the Criminal Procedure Act of 1977, which prohibited sodomy, and sections of the Security Officers Act of 1987, which excluded men found guilty of committing sodomy from employment as security officers. The Court referenced Section 9(3) of the 1996 Constitution in its opinion, which explicitly banned discrimination on the basis of sexual orientation (Jagwanth 2004). Demonstrating the potential of a strong judiciary backed by a “modern document designed for the twenty-first century” (Wing 2008), the South African Court found it within its powers to repeal the sodomy provisions.¹

The cases mentioned earlier relating to the rights of lesbian, gay, bisexual and transgender (LGBT) citizens from diverse parts of the globe both resulted in policy change. Yet, the institutional route leading to change differed significantly. In the first case, repeal of sodomy laws took place in the legislature. In the second, a court of last resort altered the status quo (legal status quo is used hereinafter interchangeably with legal state-of-affairs). Given the strong initial resistance to such change, as well as the potential backlash against the advancement of gay rights (D'Emilio 2006; Keck 2009; Klarman 2005; Rosenberg 2008; Sommer & Asal 2013; Asal & Sommer 2013), this area of the law is of particular interest for the study of paths to policy change.

What prompts a policy question to be resolved in courts rather than legislatures or vice versa? Developing a theory explaining why judicial bodies are in some instances the institutions charged with undertaking policy change while in others the location of important policy innovation is in nonjudicial institutions is the primary goal of this study. We then turn to offer a robust empirical test for the theory (we refer only to courts of last resort under the category of judicial institutions; with only one exception, we refer to legislatures as nonjudicial institutions).

The theoretical framework proposed here accounts for three factors that influence policy change including legal path dependence, political accountability, and globalization. We expect on the basis of stark institutional differences that the first two predictors of policy change will have a disparate impact on judicial compared with nonjudicial institutions while the third will have a similar effect on both. We take each of these issues up in turn.

First, courts of last resort are uniquely situated to change policy despite the path-dependent nature of stare decisis (Kahn 2006; Segal & Spaeth 2003). This does not hold for legislative bodies or even lower-level judicial bodies which, compared with courts of last resort, are more constrained by precedent (Segal & Spaeth 1996; Songer, Segal, & Cameron 1994).² This logic leads to a prima facie counterintuitive conclusion that, in certain ways, courts of last resort are less constrained by the legal status quo than other political and legal actors. The way it is used here, legal path dependence refers to the process through which the legal state of affairs, formed by past decisions, influences and informs future legal outcomes.

Second, apart from legal path dependence, we expect certain political institutions and arrangements (i.e., the electoral rules that enhance political accountability) to also affect those decision-making bodies differently. In the context of sodomy reform, we expect the presence of religious constituencies to influence the locus of policy change. Indeed, past research has suggested that these constituencies have a significant impact on the accountability of government officials, as well as the trajectory of public policy decisions (Adserà, Boix, & Payne 2003; Castles 1994; Fox 2001). Such constituencies would constrain legislators but have a more limited effect on justices who are, by institutional design, more insulated from the electorate.³

Finally, we expect globalization trends to similarly influence the introduction of policy change via judicial and nonjudicial bodies. Globalization has a pervasive effect in society. Social, political, and economic globalization affects political entrepreneurs, public opinion, political organizations, as well as social movements (Frank, Boutcher, & Camp 2009). Such trends, therefore, are likely to equally impact courts, which are engaged in transnational judicial dialogues, as well as legislatures, which are immersed in the globalizing political system.

We use data for the repeal of sodomy laws in the period between 1972 and 2002 to estimate two sets of models (generalized estimating equation [GEE] time-series cross-sectional and multinomial logit analyses). Separate analyses for all countries and for democracies only confirm that the predictors with differential effects are legal path dependence and political accountability. Globalization, on the other hand, similarly affects decisionmaking in judicial as well as nonjudicial institutions.

It is important to note from the outset that sodomy reform is only one piece of the puzzle for LGBT rights. Accordingly, we recognize that countries where same-sex relations are legal are not necessarily places where members of the LGBT community are treated equally (Waaldijk 2000). Legalization of sodomy, hence, is not the ultimate yardstick for discrimination against sexual minorities. With that in mind, however, an examination of the repeal of sodomy laws is key to developing a better understanding of the rights of the LGBT community cross-nationally. Although legal change may not always precipitate change on the ground (Epp 1996), the repeal of sodomy prohibitions is still a meaningful policy choice that warrants scholarly attention. While a court decision or a legal measure may fail to translate into full equality for the minority they aim to protect, such legal change has a declaratory value, is educational, and provides members of the minority group with venues to claim redress (McCann 1994; Scheingold 2004; Waaldijk 2000; Zemans 1983).

Can Courts Change Policy?

Scholarship on American politics has extensively entertained questions concerning courts as policy makers and, more specifically, concerning the way in which courts as policy makers relate to and compare with other institutions. A common understanding is that not only do courts rarely initiate policy change against the preferences of national majorities (Barnum 1985; Dahl 1957; Flemming & Wood 1997; Funston 1975; Mishler & Sheehan 1993), but that courts are in fact incapable of initiating such change and are therefore a “hollow hope” for those seeking social change (Rosenberg 2008, 1991). More recent scholarship, however, has suggested that the interactions between courts and legislatures is better understood as one in which the elected branch actively seeks to involve the judiciary in the policy-making process. Specifically, Graber (1993) and Lovell (2003) have argued that when resolving controversial social issues, elected lawmakers wary of alienating constituents with strongly held moral convictions will defer decisionmaking to courts by delaying action on these issues, or by intentionally inviting judicial intervention through vague and ambiguously worded statutes. Most recently, Whittington (2007) has suggested, much as Dahl did fifty years earlier, that courts are often used to reinforce and uphold the policies of the dominant political regime. Given the controversial nature of sodomy reform, this line of reasoning would suggest that policy change is more likely to be initiated by judges in countries where elected officials would prefer to avoid being held accountable for their stance on the issue.

With few exceptions, comparative public law has had little to add to this discussion.⁴ By analyzing different institutional paths to policy change in a large-N framework, we hope to identify systematic effects and thus go beyond case studies to make meaningful contributions to these debates in the literature (Epp 1998; Frank, Boutcher, & Camp 2009). We elucidate whether policy change may emanate from courts of last resort and, if so, how this policy change compares with policy change originating in the elected branches. A key explanatory variable in this context is legal path dependence, which is the process through which past decisions and the legal status quo they produce constrain future legal outcomes.

Legal Path Dependence and the Origins of Sodomy Laws

Although analysts have had difficulty in developing a clear definition of the meaning of path dependence (Mahoney 2000), we refer specifically to the processes involving positive feedback, or increasing returns, that induce further movement in the same direction over time (North 1990; Pierson 2000, 2004; Smith 2008). At the core of this definition of path dependence is the notion that, once out of the gate, institutions stay on a particular path because the costs involved in switching to a new path are prohibitively high (Kahn 2006). The timing and sequence of events in such a theory matters a great deal because future outcomes are shaped by past decisions (Maioni 1998; Pierson 2004; Rose-Ackerman 2010).

While past research indicates that the phenomenon of path dependence is widespread in politics (Ertman 1997; Hacker 1998; Huber & Stephens 2001; Kurth 1979; Pierson 2000, 2004), relatively little has been said regarding the path-dependent nature of legal development. We define precedent as previously defined legal rules. In a Common Law system, the principle of stare decisis dictates that precedent must be upheld and respected in future cases. As an important institutional norm, stare decisis produces the law's path-dependent character (Hathaway 2003). Specifically, stare decisis compels lawmakers to respect the decisions of their predecessors. Once a precedent is set, lawmakers have limited ability to switch paths and induce change.⁵ In sum, by explaining how the legal system has the capacity to lock-in laws and thus generate stability over time, path dependence is critical for explaining the legal state of affairs. This is also true, we argue, in the realm of sodomy laws.

The British Buggery Act of 1533 is the source of criminalization of sodomy not only in British law, but in many countries beyond the British Isles as well. Common Law, which was adopted by and/or imposed on other countries by the British Crown, led to the criminalization of sodomy in countries as diverse as India, New Zealand and Israel. Under the doctrine of stare decisis, decision makers in these countries worked to uphold sodomy prohibitions.⁶ French Civil Law, on the other hand, did not include an equivalent provision criminalizing the act of sodomy. The historically religious source of the proscription on sodomy led the National Constituent Assembly to reject its definition as a crime following the French Revolution (Sanders 2009). French Civil Law was subsequently introduced in many European countries as a result of French occupation during the Napoleonic Wars and in many Latin American countries owing to colonization. The notion of path dependence applies equally to Civil Law systems; this unique point of departure in French law influenced the development of sodomy reform in other countries. Ultimately, sodomy laws were more likely to have been repealed in Civil Law countries.

To summarize, path dependence leads us to expect not only that the likelihood of legal prohibitions on sodomy in countries where Common Law systems is greater, but also that the likelihood of those provisions is diminished in Civil Law countries. How does legal path dependence, however, relate to differences in the institutional origins of policy change? As we elaborate later, legal path dependence has a divergent impact on judicial and nonjudicial branches in Common Law countries.

Theory—Alternative Paths to Policy Change

In the theoretical framework proffered here, legal path dependence and political accountability have differential effects on policy change emanating from courts of last resort compared with change originating in nonjudicial bodies. Courts, legislatures, and executives operate within a system where different institutions compete with one another over spheres of influence (Clinton 1994; Eskridge 1994; Spiller & Gely 1992). While the role of the elected branches of government in policymaking is relatively clear-cut, debates in the literature still exist concerning the extent to which courts function in the capacity of policy changers (Keck 2009; Klarman 2004; Rosenberg 2008). Yet, as the examples from South Africa and Canada (as well as other cases including Chile, the United States, Ecuador, and elsewhere) demonstrate, policy change takes place in both judicial and nonjudicial contexts. The key question, therefore, is not whether courts can initiate policy change—they do. Rather, we are interested in the difference in predictors of change emanating from each type of institution. To examine the alternative paths to policy change, we first examine the motivations underlying the actions of those different decision-making bodies and then analyze the differential effects of legal path dependence on each. In addition, we discuss the importance of political institutions insofar as political accountability dictates a differential effect of constituencies on decision makers in nonjudicial bodies compared with judicial ones. We then examine a range of control variables.

Motivations of judicial and nonjudicial bodies differ. On the one hand, reelection is a major motivation for elected officials and their policymaking unfolds accordingly (Jacobson 2004; Hall 1996; Fenno 1996; Kingdon 1989; Mayhew 1974). On the other hand, the introduction of policy change by courts may add to the court's legitimacy if such policies are respected and upheld (McCloskey 2010; Whittington 2007; Tushnet 2006; Lovell 2003). Although a parallel argument can be made about the introduction of successful policy change by nonjudicial bodies, questions about institutional legitimacy are particularly concerning for judicial bodies. This is because the judiciary lacks, as Hamilton puts it, “influence over either the sword or the purse.” Courts, therefore, face potential challenges to their legitimacy both in terms of their lack of electoral unaccountability, as well as the danger that their pronouncements will either face significant delays in their implementation (Rosenberg 1991) or be ignored altogether (Bigel 1991). On the other hand, by embedding changing social, political, and economic realities in legal precedent, judges increase the likelihood that their decisions will be accorded respect and support by both political elites and the public (Kahn 1999, 43)⁷ .

The process through which law and politics mutually construct each other, combined with the unique institutional position of courts of last resort, lead to the somewhat counterintuitive conclusion that, in some instances, courts of last resort are less constrained by the legal status quo (Kahn 2006; Segal & Spaeth 1996, 2003; Spaeth & Segal 1999).⁸ Key to our argument is the notion that, compared with the decision-making process in nonjudicial institutions, the effects of legal path dependence on decisionmaking in courts of last resort are commensurably weaker.

As Chief Justice Marshall writes for the Court in Marbury v. Madison: “It is emphatically the province and duty of the Judicial Department to say what the law is.” Justices in a court of last resort are in a position to recognize changes taking place in society and, by institutional design, are free to act accordingly. These justices have the (often exclusive) privilege to interpret constitutional and statutory law, typically enjoy long tenures (life-appointments in many cases), have no superior court to overrule their decisions, and sometimes can even set their own agenda (Segal & Spaeth 2003). Furthermore, legal decisionmaking does not amount to a dry application of legal principle. To justify a change in policy, judges must show that existing precedents are unreasonable or mistaken.⁹ Scholars have thus argued that, unlike lower courts, a Supreme Court is not always bound by its own decisions (Hathaway 2003; Thurmon 1992; Cooper 1988). The unique institutional position enjoyed by courts of last resort, combined with judicial detachment from the vagaries of day-to-day political reality, enable judges to recognize when competing claims in the polity make social change possible (Kahn 1999).

The way we conceptualize the legal status quo is not as black letter law, but rather as the set of rules that comprehensively organizes and constitutes the political, social, and economic spheres (Brigham 1999; Ewick & Silbey 1998; Kahn 2006). Granted, constitutional separation of powers and checks and balances work to ensure that precedent operates as a significant constraint on legislative bodies. For instance, this principle forbids elected officials from ignoring precedent simply because they disagree with the Court's decision. Yet, our argument goes deeper; because the legal status quo is entrenched in various aspects of the political, social, and economic systems, path dependence operates as a more significant constraint on legislative bodies. Introducing a controversial policy game-changer is prohibitively costly for legislators who operate in a political system and in a society that are organized according to the existing legal state of affairs. Constituencies, interest groups, social organizations, parties, economic bodies, and financial institutions all work within the framework of existing policy and thus typically favor the status quo. In turn, elected officials are held accountable to such groups because of sunk costs, vested interests, and path dependence.

In the context of a contentious issue such as sodomy reform, introducing a controversial policy change is particularly costly for elected officials. Such officials may avoid the public pursuit of controversial policy goals (Graber 1993). In the context of sodomy reform, it is no wonder that many elected officials have refused to come out in support of “a group of people despised by virtually everyone … condemned by every significant religious tradition, and pathologized by scientific experts” (D'Emilio 2006). In contrast, courts are more autonomous from ordinary politics. Indeed, under certain circumstances, their independence leads judicial decision makers, particularly in courts of last resort, to “ignore, resist and even disregard robust political pressure” (Kahn & Kersch 2006: 18).

The theory canvassed thus far would lend itself to two interesting empirical tests. The paths to policy change in Common Law compared with Civil Law countries will not be the same (Merrryman & Perez-perdomo 2007: 46–47). First, we expect that having a Common Law system would decrease the likelihood of legislative repeal because elected officials are confined to work within the boundaries of existing precedent. Second, a system of Common Law would not have the same effect on courts of last resort as they are less constrained by path dependence and by the legal status quo. In Common Law systems, the constraints generated by legal path dependence on policy change tend to limit nonjudicial rather than judicial institutions, thereby increasing the likelihood of judicial repeal. Because their operation is closely intertwined with the political, social, and economic systems, nonjudicial institutions in Common Law systems are more constrained by the legal state of affairs than the somewhat autonomous judicial bodies. This would translate empirically into a differential effect of the legal system:

H ₁: Ceteris paribus, a Common Law legal system should decrease the likelihood of legislative repeal.

H ₂: A Common Law system may not affect or even increase the likelihood of repeal by a court of last resort.

Political Institutions and Religious Constituencies

While their origins are Judeo-Christian, proscriptions on sexual relations between people of the same sex are found in other religions as well (Sanders 2009). Based on the principles of the Qur'an, the central text of Islam, and Hadith, which are oral traditions determining the Muslim way of life, Islam condemns same-sex intimacy. While it is true that not all countries with a Muslim majority treat same-sex–related activity as a crime, we do expect that there is considerable pressure on public officials not to repeal sodomy laws in countries with a strong Muslim constituency.¹⁰ Akin to Islam, Catholic dogma also condemns sexual relations between members of the same sex as sinful and contrary to natural law (Dempsey 2008). Therefore, we predict that the higher the percentage of Catholics in the population, the greater the likelihood that the state will fail to decriminalize sodomy.

We contend further that the effect of religious constituencies on disparate types of decision-making institutions would be different. The dissimilarities between the two institutional paths relate to the issue of accountability. We expect constituents to have a greater impact on elected officials than on courts of last resort. In the context of gay rights, and in light of the controversial nature of sodomy reform, it is the religious constituencies that we expect to have the most sway. In particular, we expect the two denominational groups discussed earlier, Muslims and Catholics, to have the greatest impact (Adserà, Boix, & Payne 2003). Importantly, though, religious constituencies would affect different government institutions dissimilarly. Those constituencies would have greater influence when decisionmaking in legislatures is concerned. Worried about funding, reelection, or pressure from lobbyists and interest groups, officials in nonjudicial bodies are likely to pay heed to this type of pressure and thus be more constrained in considering the repeal of sodomy laws. Important to our theory is the notion that rather than diffuse public pressure, it is Muslim and Catholic constituencies specifically that would target officials and pressure them against repeal. The larger those constituencies, the stronger the hypothesized effect on legislators would be. Courts, on the other hand, with their relative autonomy from political pressures, would be less affected.

This is not to say that judges are insulated from their own political and cultural environments (see Gibson 1980; Giles & Walker 1975; Vines 1964). These pressures may influence the likelihood of members of the judiciary to alter culturally entrenched views of sexual minorities. However, on balance, we expect that the realities of popular political influence will be felt much more acutely by those least separated from the public at large, that is, officials serving in the political branches. Furthermore, justices' views of their own role typically do not include responsiveness to majoritarian elements such as public opinion or popular will, which in the context of sodomy laws, means the court may be less resistant to change (Kahn 2006). In sum, unconstrained by accountability like its political counterparts, and not perceiving responsiveness as part of its role, the judicial branch is less threatened or influenced by outside pressures.

H ₃: Dominant religious constituencies should not affect the likelihood of repeal by courts.

H ₄: Dominant religious constituencies should decrease the likelihood of repeal by nonjudicial institutions.

Taken together, legal path dependence combined with political accountability create a powerful explanation for the differences between judicial and nonjudicial venues as initiators of policy change. Yet, certain variables should be controlled for in order for our analysis of the repeal of sodomy laws to be valid. Political circumstances change as countries globalize, become more democratic, and as their gross domestic product (GDP) grows.

Globalization

Globalization has grown in scale, speed, and importance (Kinnvall 2004). While some argue that globalization is not an unprecedented phenomenon (Hirst & Thompson 1999; Williamson 1996), current levels of globalization are different than in the past. People move more freely across the globe as tourists, immigrants, refugees, or international students and businesspersons (Appadurai 2000). Globalization has also increased contacts among societies. There is heightened awareness of different political arrangements through mass media. Likewise, there is a major increase both in volume of trade and the financial flows involved as well as in the intensity of these interactions and the key role of information and communication technologies (Giddens 2002). This phenomenon has reduced transaction costs across a range of human interactions (Hollingsworth 1998) with vast political implications. It permits the exchange not only of goods and services, but also of ideas, values, beliefs, and political institutions (Hermans & Kempen 1998).

Scholars contend that globalization leads to more rights and freedoms by diffusing the ideals of freedom and democracy (Tsutsui & Wotipka 2004) or by forcing states to adopt norms of rights and freedoms in response to increasing international pressures. Information about alternative legal arrangements should be more readily available in a globalized state. It becomes easier to identify alternative legal frameworks and the way civil rights and liberties in general and gay rights in particular are organized in other jurisdictions. In this sense, the strengthening of links among countries allows for the diffusion of new human rights and norms of tolerance (Tsutsui & Wotipka 2004). Moreover, the world culture and normative diffusion literature argues that the strengthening of links between countries allows for the diffusion of new human rights and norms of tolerance (Boli & Thomas 1997, 1999; Ramirez & McEnaney 1997). In terms of legal elites, an increasing trend toward communication among members of the judiciaries across national borders also seems to exist. For instance, to support its 2003 decision in Lawrence v. Texas (as well as in more recent cases), the Supreme Court of the United States has cited a number of authorities beyond the Fourteenth Amendment including Dudgeon v. United Kingdom, a decision by the European Court of Human Rights invalidating sodomy laws (Wells 2004).¹¹ In a more formal context, the 1996 Constitution of South Africa, discussed earlier, explicitly requires the consideration of foreign law by members of the Constitutional Court. In the case of sodomy reform, we argue, there is a norm diffusion that happens among countries. More highly globalized countries should thus be more likely to repeal their prohibitions.

Globalization has a pervasive effect in society and should affect both institutional paths. Global trends are likely to equally impact courts, which are engaged in transnational judicial dialogs, as well as legislatures, which are immersed in the globalizing political system. When the country is immersed in the global culture (socially and politically), political entrepreneurs, public opinion, political organizations, and social movements are able to recognize alternative legal arrangements within which to settle gay rights (Barclay et al. 2009). Thus, both courts and nonjudicial bodies should be in a position to repeal sodomy laws with greater social and political globalization.

H ₅: Notwithstanding the institutional path, the likelihood of policy change increases with globalization.

We also control for GDP and for the democratic conditions in the country. To demonstrate the robustness of our findings, as described later, all models are estimated for all nations as well as for democracies only.

Data and Methods

Two sets of regression models are estimated to test the hypotheses. First is a multinomial logistic regression with year dummies, and the second is GEE time-series cross-sectional analysis. The dependent variable for the multinomial logit, Repeal Type, has three levels – 0 for no repeal; 1 for nonjudicial repeal; and 2 for judicial repeal. For the GEE analyses, three dependent variables are coded. Legal indicates whether a country decriminalized sodomy¹² (1 = sodomy is legal, 0 = otherwise). The second dependent variable for the GEE models is Court Repeal, which is coded 1 for countries where repeal happened in court in the three decades following 1972 and 0 otherwise. Nonjudicial Repeal, which is equal to 1 when nonjudicial institutions repealed sodomy, and 0 otherwise, is the third dependent variable for the GEE models. Data for all outcome variables are taken from the May 2009 report of The International Lesbian, Gay, Bisexual and Intersex Association.¹³ From 1972 to 2002, nine countries repealed their sodomy provisions via judicial institutions, while 35 countries repealed their statutes via nonjudicial institutions. Other than in Fiji where repeal originated with the executive, the legal measure against sodomy was revoked by the legislature in all other such cases. All countries are listed in the appendix. Given common wisdom about policy emanating from the political branches, the fact that over 20% of the repeals in the three decades under examination here were via the judicial path further underscores the significance of our research question.

As for our predictors, Common Law is a dummy that equals 1 for countries whose legal origin is Common Law and 0 otherwise.¹⁴ Based on the Central Intelligence Agency (CIA) fact book and Barrett (1982), Percent Catholic indicates Catholics as percentage of the population and Percent Muslim indicates Muslims as percentage of the population. To measure international connections, we use a measure of globalization—the KOF Indexes of Globalization (Dreher 2006; Dreher, Gaston, & Martens 2008). The indexes for the globalization variables are measured in line with the standard in comparative public law (Frank & McEnaney 1999; Tsutsui & Wotipka 2004) and range from 0 to 100. Higher values indicate higher levels of globalization. The index of Political Globalization is measured by the number of embassies and high commissions in a country, the number of memberships the country has in international organizations, participation in United Nations peace-keeping missions, and the number of international treaties signed since 1945 (Dreher 2006; Dreher, Gaston, and Martens 2008). The Social Globalization measure includes three categories of indicators: personal contacts (e.g., telephone traffic and tourism), information flows (e.g., number of internet users), and cultural proximity (e.g., trade books and number of warehouses of Ikea per capita) (Dreher 2006; Dreher, Gaston, & Martens 2008). The Globalization predictor is a weighted average of social, political, and economic globalization. To measure Democratic Conditions, we utilize the POLITY score, which was imputed using Freedom House data. The scale ranges from −10 (least democratic) to 10. GDP per capita in constant U.S. dollars at base year 2000 was used (Gleditsch 2002). See appendix for descriptive statistics and a correlation matrix.

We use time-series cross-sectional data, listing all countries for the years 1972–2002. The multinomial logit with year dummies is the appropriate approach if we wish to include both legislative and judicial repeals in the same analysis, because the dependent variable is categorical. To test the effects of predictors on legislative and judicial repeal separately, we employ a GEE model (Zorn 2001). A marginal approach, such as the GEE, is appropriate because we are interested generally in what variables influence decriminalization of sodomy, rather than the propensity to do so in a particular country, for which a conditional approach would suffice (p. 475). Because the dependent variables are dichotomous and because of the data structure described earlier, we employ a GEE model with first-order autoregressive component and logit as the link function. In both analyses, we use robust standard errors.

For added confidence in our results, we provide two versions of each model estimated. Democratic conditions are potentially critical for the facilitation of legal equality (Wilensky 2002). Hence, along the lines of similar/different systems designs (e.g., PrzeworskiA, Stokes, & Manin 1999; Przeworski et al. 2000; Teune & Mlinar 1978), the first model in each of the two tables presents the analysis of data for both democratic and nondemocratic systems.¹⁵ The second twin-model in each case presents the results for democratic countries only.¹⁶

Results

Table 1 offers a first insight into the predictors of repeal. With Repeal Type as the outcome variable, the predictors for repeal by nonjudicial institutions are largely in support of our hypotheses. In the first column, the results for all countries are presented. Legal path dependence in Common Law countries decreases the likelihood of repeal via the legislature. Religious constituencies (Muslim and Catholic) have a similar effect—the larger those constituencies are, the less likely would legislative repeal be. Globalization increases the likelihood of legislative repeal. Lastly, democratic conditions increase the likelihood of legislative repeal. In the analysis of democracies only (the second column), the effects are very similar. Democratic conditions, however, have an insignificant effect. The reason is little variance in this predictor, when only democratic countries are included in the analysis. As for judicial repeal, in both samples (all countries and democratic countries only), a Common Law legal system increases the likelihood of repeal. Political accountability to religious constituencies does not have a statistically significant influence. Judicial repeal is more likely when the country is more democratic, and more globalized. These effects are true for both the sample of democratic countries only and for the sample of all countries. Possibly because of collinearity with the Social Globalization predictor, the effect of GDP is not significantly different from zero. The goodness of fit measures are reasonably high in both the model for democratic countries and the one for all countries.

Table 1. Multinomial Logistic Regression Models

Dummies for years not presented in the table.

Because of issues of data availability, the following countries were not included in the analyses: Afghanistan, Andorra, Angola, Antigua and Barbuda, Bahamas, Barbados, Belize, Brunei Darussalam, Cape Verde, Comoros, Czechoslovakia, Djibouti, Dominica, Equatorial Guinea, Eritrea, East Germany, West Germany, Grenada, Guinea, Haiti, Iceland, Iraq, Kiribati, North Korea, Laos, Liberia, Libya, Liechtenstein, Luxemburg, Marshall Islands, Mauritania, Micronesia, Monaco, Nauru, Niger, Palau, Samoa, San Marino, Sao Tome and Principe, Seychelles, Solomon Islands, Somalia, St Kitts and Nevis, St Lucia, St Vincent and the Grenadines, Suriname, Taiwan, Tibet, Timor-Leste, Tonga, Tuvalu, USSR, Uzbekistan, Vanuatu, North and South Vietnam, North and South Yemen, Yugoslavia, Zanzibar. Values for the globalization variables are not available for those countries.

^# p < 0.1; ^*p < 0.05; ^**p < 0.01; ^***p < 0.001; one-tailed tests where directionality hypothesized. Bold indicates statistical significance.

The GEE models are presented in Table 2. The first two columns indicate the results for the predictors of repeal (the institutional path notwithstanding) in all countries and in democratic countries only, respectively. The third and fourth columns present the predictors of repeal by the country's court of last resort in all countries and in democratic countries only, respectively. The last two columns on the right have the predictors of repeal by nonjudicial institutions for all countries and for democratic countries separately in each column. With Legal as the dependent variable, the first two columns outline which variables predict repeal generally. The coefficient on Common Law is negative, but does not reach standard levels of statistical significance. The effect of democratic conditions is statistically significant, and so is that of social and political globalization. The differences between the sample of democratic countries and of all countries are in the coefficients for democratic conditions (not statistically significant in democratic countries) and Percent Muslim, which is significant only in democratic countries. The performance of the models is good overall.

Table 2. Generalized Estimating Equation Models

Due to issues of data availability, the following countries were not included in the analyses: Afghanistan, Andorra, Angola, Antigua and Barbuda, Bahamas, Barbados, Belize, Brunei Darussalam, Cape Verde, Comoros, Czechoslovakia, Djibouti, Dominica, Equatorial Guinea, Eritrea, East Germany, West Germany, Grenada, Guinea, Haiti, Iceland, Iraq, Kiribati, North Korea, Laos, Liberia, Libya, Liechtenstein, Luxemburg, Marshall Islands, Mauritania, Micronesia, Monaco, Nauru, Niger, Palau, Samoa, San Marino, Sao Tome and Principe, Seychelles, Solomon Islands, Somalia, St Kitts and Nevis, St Lucia, St Vincent and the Grenadines, Suriname, Taiwan, Tibet, Timor-Leste, Tonga, Tuvalu, USSR, Uzbekistan, Vanuatu, North and South Vietnam, North and South Yemen, Yugoslavia, Zanzibar. Values for the globalization variables are not available for those countries.

^* p < 0.05; ^**p < 0.01; ^***p < 0.001; one-tailed tests where directionality hypothesized.

The results in Table 2, third and fourth columns, lend support to our hypotheses concerning judicial repeal. As expected, the Common Law variable does not decrease the likelihood of judicial repeal of prohibitions on sodomy. The path-dependent nature of legal evolution in Common Law countries has little effect on justices' ability to alter policy status quo when it comes to this type of rights. A trend toward globalization increases the likelihood that the sodomy law is repealed with a court decision. Large religious constituencies seem to have insignificant effects on courts. Lastly, the effects in the subsample of democratic countries largely resemble those in the general sample.

In the fifth and sixth columns of Table 2, the outcome variable is Nonjudicial Repeal. The negative and highly significant coefficient indicates that a Common Law system decreases the likelihood of repeal via legislatures. The path-dependent nature of legal development in Common Law systems influences nonjudicial institutions when they attempt to reform policy related to sodomy law. Furthermore, the electoral link hypothesis wins support. The coefficients on both Percent Muslims and Percent Catholics are negative and significant. The effect of globalization is in line with the theoretical expectations. The effects in democratic countries are comparable. In the subsample of democracies, Democratic Conditions is not statistically significant because of decreased variance in this predictor.

Overall, legal path dependence decreases the likelihood of repeal in Common Law countries and increases this likelihood in systems of Civil Law. Yet, its effect is considerably more substantial where the legislative path is concerned. Between the multinomial logit analyses and the GEE models, the results lend robust support for the theory developed here. Tables 1 and 2 confirm that judicial and nonjudicial institutions introduce policy change under dissimilar circumstances. In Common Law systems, the juxtaposition of the two tables indicates legislatures are less likely to be the venues for policy innovation. Courts of last resort, on the other hand, are either more likely or as likely to introduce policy change. Indeed, an overwhelming majority of Civil Law countries repealing sodomy laws between 1972 and 2002 saw those provisions rescinded in nonjudicial institutions; in 97% of the cases, when a sodomy provision was repealed in a Civil Law country, the decision was nonjudicial. Conversely, courts have been considerably more popular as venues for repeal in Common Law countries where in 64% of the cases the institutional path to policy change was judicial. Religious constituencies are statistically significant predictors of legislative repeal, but have no significant effect on judicial repeal. Political accountability impacts the likelihood of the introduction of policy change in legislatures; conversely, courts as initiators of policy innovation are largely unaffected by accountability to constituencies. Democratic Conditions is statistically significant in both tables. When the country is more democratic, repeal of prohibitions on same-sex sex is more likely via either institutional path. The same is true for globalization. Finally, comparison of a subsample of democracies to the general sample indicates the results are largely the same in both cases, with discrepancies largely accounted for by the theoretical framework we developed. The only consistent difference is that in analyses of democratic countries only, because of limited variance in that regressor, Democratic Conditions does not come out statistically significant.

Discussion and Conclusions

The goal of this study was to analyze the origins of policy change via different institutional paths, comparing judicial and nonjudicial institutions. This question, which has been crucially important in the study of public policy, judicial politics, and social movements, was examined here for the first time systematically in a cross-national framework over a period of several decades and with respect to a question still on the agenda in numerous countries, namely the decriminalization of same-sex intimacy. While the debate in the literature may still be unsettled, the theory developed here suggests that policy change emanates from judicial as well as nonjudicial bodies.

The key message of this study, however, is not limited to the notion that courts create policy change. Rather, we explain theoretically, and then substantiate empirically, that different sets of variables systematically explain policy change via disparate institutional venues. Our findings clearly indicate that legal precedent in a Common Law system limits the introduction of policy change via the political branches more than via courts of last resort, and particularly when such change proves contentious. At the same time, the path-dependent nature of law in civil law countries makes it easier for policy change to emanate from the legislature. Indeed, whereas legislative repeal constitutes 97% of the cases where sodomy laws were revoked in Civil Law countries, 6 in every 10 repeals in Common Law countries were judicial. Such findings are significant not only to our understanding of law, but to a range of topics including legal development, accountability, and the effects of religiosity on policy formation and change.

In addition, our findings lend support to the notion that political actors are more constrained by legal status quo than their judicial counterparts, and accordingly, that the effects of path dependence on decisionmaking in supreme courts are commensurably weaker. More broadly, this finding addresses a major criticism leveled against path-dependence scholars concerning their inability to explain policy change. We contend that the judicial hierarchical structure enables courts of last resort to produce policy change. Indeed, as Kahn (2006) suggests, such courts may serve as important mechanisms of change, a relief valve of sorts, in theories of path dependence.

Furthermore, it is evident that political institutions, such as accountability, entail closer proximity between the will of constituents and decision-making authorities in the political branches. At the same time, we find courts are less affected than legislatures by majoritarian elements including, for instance, political pressures exerted by religious groups. Some forces at the domestic (e.g., democratic conditions) and global (e.g., globalization) levels affect repeal, notwithstanding its institutional venue. In a subsample of democracies, the effect of democratic conditions is diminished, but the effects of the other predictors remain largely unaffected.

There is also an important normative element to this discussion concerning the legitimacy of unelected judges altering the policy made by decision makers who are accountable to the electorate. Indeed, this concern was expressed most famously by Bickel (1962) who wrote of the “counter-majoritarian difficulty” and more recently by Powers and Rothman (2002) who see the judiciary as ill-equipped to resolve issues better left to legislatures. A number of responses have been offered to this critique, ranging from empirical studies demonstrating the reluctance of the court to stray too far from public opinion (Barnum 1985; Mishler & Sheehan 1993), to suggestions that lawmakers themselves create conditions favorable to judicial policy making (Gillman 2002; Rogers 2001). The theory proposed in this study (and the empirical support presented thereafter) directly engages this scholarship. As far as the protection of sexual minorities is concerned, the findings in this study clearly indicate that judicial institutions may well be the ones to extend legal protections to minority populations. It is not always clear in such instances, however, that judges are acting in a counter-majoritarian fashion. When the United States Supreme Court handed down its decision in Brown, for example, only 17 states required segregation of public schools (Balkin 2008). Similarly, when the Court decided on the constitutionality of sodomy prohibitions in Lawrence, only 13 states still criminalized same-sex sodomy. Under the right circumstances, policy change may originate from courts of last resort, but such change does not necessarily run contrary to popular will or to elected institutions.¹⁷

As far as predictors of social change are concerned, and in particular in the context of the rights afforded sexual minorities, the analytical advantage of examining disparate institutional paths is clear. For instance, despite failures to find effects for religiosity in past work (e.g., Frank & McEneaney 1999), our theory and empirical tests illustrate the critical importance of analytically treating disparate institutional paths in order to accurately assess the effects of independent variables such as religious constituencies. The political stars align differently in dissimilar jurisdictions. When Common Law and strong religious constituencies are present in a polity, courts may be the venue of choice for those seeking social change. Indeed, the Canadian and South African cases described earlier are but two examples illustrating these dynamics.

Lastly, this work also offers some empirical predictions to be further developed and tested in future work. Civil law systems tend to hold case law to be subordinate to statutory law, which might also explain the greater reliance on nonjudicial institutions. Testing this theoretical account would complement the findings in this study. In addition, religious constituencies beyond those studied here may influence policy output (Campbell & Monson 2003; Wright, Erikson, & McIver 1987). A thorough treatment of the dynamics of a broader range of religious groups and consequent judicial and nonjudicial policymaking (with respect to gay rights and otherwise) merits further study. With respect to institutional paths to policy change, future work may wish to examine the introduction of other policies (related to sexual minorities or otherwise) via disparate institutional paths. The set of predictors offered in this article may account, for instance, for the introduction of antidiscrimination policies in different countries. While some accounts in the literature claim that, in Europe for instance, the mere decriminalization of same-sex sex inexorably led to the introduction of antidiscrimination measures (e.g., Waaldijk 2000), the theory proposed here offers an alternative analysis. Considering institutional paths of policy change and their respective predictors including type of legal system, special constituencies, democratic conditions, and globalization, our theory offers a rich framework for scholars studying those processes. Moving beyond sexual minorities, the findings here may serve future examinations of policy change relevant to additional minority groups and policy domains.