Transnational advocacy and domestic law: International NGOs and the design of freedom of information laws

Abstract

Can international non-governmental organizations (INGOs) influence domestic policy? This paper offers new quantitative evidence of the impact of INGOs in one specific policy area—Freedom of Information (FOI) laws—as well as highlighting an under-studied mechanism of INGO influence on the design of domestic laws. I test this argument by examining the effect of legal analyses of draft FOI legislation published by the INGO Article 19. These analyses provide expert legal assessments and make normative evaluations—both information politics and symbolic politics. I find that in countries in which Article 19 conducted legal analyses, the design of the subsequently passed FOI laws was significantly stronger than in countries that were not subject to such analyses. I demonstrate that this finding is not an artifact of Article 19’s selection process. I also present suggestive evidence that highlights symbolic politics, not information politics, as the more salient mechanism. Finally, I examine the process of FOI drafting and adoption in Serbia to illustrate the argument and specific mechanisms at work.

1 Introduction

Can international non-governmental organizations (INGOs) influence domestic politics and policy? Scholars of international relations have argued that INGOs can influence state policies through several different mechanisms, including shaping international norms and institutions (Finnemore and Sikkink 1998, Price 1998, Florini 2000, Charnovitz 2006, Cichowski 2007), “naming and shaming” norm-violating states (Keck and Sikkink 1998), empowering domestic social movements (Risse and Sikkink 1999), persuading key gatekeepers to make international contributions or commitments (Busby 2010), monitoring state compliance and providing information to enable action by other actors (Raustiala 1997, Mitchell 1998), and directly undertaking public services under some circumstances (Murdie and Hicks 2013, Lake 2014). More recently, quantitative studies have reached mixed conclusions from empirical evidence focused specifically on the ability of INGO naming and shaming to improve states’ human rights practices (Hafner-Burton 2008, Franklin 2008, Murdie and Davis 2012, Bell et al. 2012, Krain 2012, DeMeritt 2012, Hendrix and Wong 2013, Bell et al. 2014).

Building on this past work, this paper seeks to make two contributions, one empirical and one theoretical. Empirically, it offers quantitative evidence that one specific INGO (the London-based Article 19) can have a measurable impact in one specific issue area (the freedom of information)—a different type of finding from studies of aggregate human rights indicators. Theoretically, it emphasizes a type of INGO influence that has been under-studied in the literature to date: the direct influence of INGOs in the translation of international norms into domestic laws. That is, instead of examining the role of INGOs in the adoption of new policies,¹ this study examines the design of those policies, and how well they meet international standards.

Indeed, norms are often translated² into weak laws in ways that undermine the principles of the norm, potentially becoming less efficacious tools to shape the behavior of domestic actors. While much research has studied the compliance problem—normative slippage on the way from international treaties to domestic practice (Hathaway 2002, Hafner-Burton and Tsutsui 2005, Simmons 2009), less research has addressed the separate problem of normative slippage on the way from international norms to domestic laws. Not all international norms are adopted by the ratification of treaties. Instead, many are adopted when states individually write them into domestic law, opening up an entirely different context in which normative slippage can take place.

Why, when adopting international norms into domestic law, do some states pass stronger laws and others weaker laws? I argue that the process of translation into domestic law can be a key access point for the influence of international NGOs using the tools of transnational advocacy. International NGOs have been recognized as key norm entrepreneurs—actors who contribute to the emergence of new norms by calling attention to issues, creating new issues, putting norms on the international agenda, and convincing initial states to adopt (Finnemore and Sikkink 1998). In this case, however, international NGOs’ role as norm entrepreneurs extends beyond the norm emergence stage to the process of domestic legal design.

Focusing on the specific issue area of Freedom of Information (FOI) laws, this study finds that international NGOs are able to assess draft legislation and influence domestic political actors to pass laws which are stronger, and more in line with international norms, than they would pass otherwise. They are able to do so through both information politics—using their policy expertise to provide credible and useful information—and symbolic politics—using their role as norm entrepreneurs to make principled arguments about appropriate behavior in a particular policy area (Keck and Sikkink 1998).

FOI laws are a policy innovation that has been passed in over 90 countries to date, with well over half of those adoptions taking place since 2000. The global spread of FOI laws reflects the growth of an international norm of transparency. However, some FOI laws make stronger commitments to this norm than others, in terms of their scope, procedures, sanctions, and other dimensions of legal design. Since the late 1990s, the London-based international NGO Article 19, one of the most active and high-profile advocates working on FOI issues, has been conducting and publishing legal analyses of draft FOI laws in many countries. These analyses highlight portions of the draft laws that Article 19 considers to be not in keeping with international norms, and offer public recommendations for changes to strengthen their legal design.

I find that in countries in which Article 19 conducted legal analyses, the legal design of the subsequently passed FOI laws was significantly stronger than in countries that were not subject to such analyses. Further, I demonstrate that this finding is not an artifact of any selection process by which Article 19 chose where to focus its efforts. I also present suggestive evidence highlighting symbolic politics as the more salient mechanism by which legal analyses have an effect, rather than information politics. Finally, I examine the process of FOI drafting and passage in Serbia to illustrate the specific roles that Article 19 played.

The second section of this paper briefly reviews the global spread of FOI laws and details the measurement of their legal design. The third section theorizes both the domestic and transnational determinants of the strength of FOI laws, with specific attention to the role of the international NGO Article 19. The fourth section develops empirical models of the strength of FOI laws and presents their results. The fifth examines the illustrative case of Serbia, and the sixth section concludes.

2 Freedom of information laws around the world

Freedom of Information laws, also sometimes called Access to Information or Right to Information laws, give “citizens, other residents, and interested parties the right to access documents held by the government without being obliged to demonstrate any legal interest” (Ackerman and Sandoval-Ballesteros 2006). By giving individuals the ability to request information or records from government bodies and requiring officials to respond, FOI laws seek to guarantee government transparency, increase public participation and trust, and deter corruption and other forms of official malfeasance. In many countries, FOI laws are used by tens (or even hundreds) of thousands of individuals each year to request government information.

The first FOI law was passed in Sweden in 1766, requiring that official documents “immediately be made available to anyone making a request” (Banisar 2006). After the passage of the United States’ Freedom of Information Act in 1966, other states in both the developed and developing world began passing similar laws at an increasingly rapid rate. Today over 90 countries around the world have passed FOI laws, over half of them since 2000. Research on the global spread of FOI laws has focused on factors including global diffusion, civil society advocacy, and domestic political structures (Ackerman and Sandoval-Ballesteros 2006, Berliner 2014, Berliner and Erlich 2015, Florini 2007, Puddephatt 2009, Michener 2010, Michener 2011, McClean 2011, Relly 2012, Roberts 2006). This paper, however, focuses not on passage but specifically on the design of FOI laws.

Not all FOI laws are created alike. Indeed, there is a great deal of variation in their legal design. In terms of scope, some apply to a broader range of branches of government, agencies, and types of information than others. In terms of exemptions from disclosure, some allow for broader exceptions for reasons such as national security, privacy, and commercial secrets, while others have fewer exemptions, or put in place a public interest test or other restriction to balance those exemptions against the public’s interest in disclosure. Some FOI laws also establish mechanisms for appeals, oversight, monitoring, and sanctions for non-compliance. These in turn vary from reliance on the courts, to ombudsman’s offices, independent information commissioners, or alternative approaches.

Two examples of relatively strong laws come from India and Serbia. Passed in 2005, India’s Right to Information Act applied to state and local governments as well as federal agencies, created an independent Information Commission, and put all exemptions to a public interest test “whereby information may be released if the public interest in disclosure outweighs the harm to the protected interest” (Banisar 2006). Indeed, India’s law has been very successful by many measures. The information commissioner has exercised its authority to fine officials for non-compliance, requests have been used to expose official corruption, and Indian citizens filed roughly two million information requests in the law’s first two and a half years (Roberts 2010). An experimental study found that information requests were almost as effective as paying bribes in enabling poor citizens to receive their ration cards (Peisakhin and Pinto 2010).

Passed in 2004, Serbia’s Law on Freedom of Access to Information of Public Importance contained relatively few exemptions, a public interest test, limits on fees associated with requests, and created an independent commissioner whose decisions are binding on public authorities (Banisar 2006). The Serbian Information Commissioner’s 2011 report found that over 50,000 requests were filed in the previous year, a substantial number for a country of Serbia’s size, and that over 90 % of requests which were appealed to his office were ultimately resolved in favor of the requester (Serbia Commissioner for Information of Public Importance and Personal Data Protection 2012).

Many other FOI laws are considerably weaker, however. The Dominican Republic’s 2004 Law on Access to Information contained “a seriously overbroad regime of exceptions, which recognizes secrecy provisions in other laws and which does not include a harm test or public interest override” (Mendel 2009, 67), provided for no independent appeals or oversight agency, and even required requesters to declare their reasons for seeking information (Banisar 2006, Mendel 2009). In Angola, the 2002 Law on Access to Administrative Documents was almost entirely copied from the 1993 Portuguese law of the same name, and was very limited in the types of documents that it applied to, and provided for no monitoring or sanctions (Banisar 2006). Weak FOI laws are not limited to developing countries, either. In Germany, the 2005 Federal Freedom of Information Act contained so many exemptions of such breadth that one expert said that “from the user’s point of view, this is a disaster” (Global Integrity 2007, 1). The most extreme example of a weak FOI law is Zimbabwe’s 2002 Access to Information and Protection of Privacy Act. While this law used the language of a fundamental right to access information, its exemptions were so broad as to be unusable, and in practice provisions of the law were actually used to shut down independent newspapers (Banisar 2006).

Of course, it is important not to conflate the de jure and de facto dimensions of policy design. Although strong legal design in written law may not necessarily lead to strong implementation and enforcement of laws in practice, Michener (2015b), does find that strong legal design is an important determinant of effective FOI regimes in practice among Latin American countries. While the extent to which these findings hold beyond Latin America is an important topic for future research, this paper focuses specifically on variation in de jure design elements, in order to evaluate how international normative principles were translated into concrete domestic policy in written FOI laws.

2.1 Measuring the legal strength of FOI laws

I measure the design of FOI laws using data from the Global Right to Information Rating (RTI Rating), a “comprehensive comparative analysis of the legal frameworks for accessing information in each of the 89 countries where such a system exists” (Centre for Law and Democracy 2011). The NGOs Centre for Law and Democracy and Access Info Europe co-sponsored the development of this rating system, in consultation with 79 experts from around the world, and released the information in September 2011. Each FOI law was scored on 61 different indicators, most ranging from 0 to 2 points each, but some higher. This coding system yielded a total possible score of 150 points for a country’s FOI law, although in practice the observed scores range from 39 to 135.³ These indicators, in turn, comprise seven different categories, listed in Table 1 below.

Table 1 Categories in global right to information rating

The four categories accounting for the most points are Scope, Requesting Procedures, Exceptions and Refusals, and Appeals. Scope concerns the agencies and branches of government to which the law applies. Requesting Procedures concern procedures such as time limits for response, fees, and assistance for requesters with special needs like illiteracy. Exceptions and Refusals concern the categories exempt from disclosure (such as national security and privacy) and the existence of public interest tests to override them. Appeals concern independent appeals agencies and other avenues of recourse for denied requests. Among the other categories, Right of Access concerns the framing of the law as a human right, Sanctions and Protections concern sanctions for non-compliant officials, and Promotional Measures concern public awareness, trainings for officials, and reporting of statistics on information requests. There is substantial variation in the total strength of FOI laws across countries, ranging from a low of 39 (Austria) to a high of 135 (Serbia).

3 Transnational advocacy and the legal strength of FOI laws

In this paper, I argue that transnational advocacy by international NGOs can shape the design of new laws and policies, through tools of both information politics and symbolic politics (Keck and Sikkink 1998). Information politics refers to “the ability to quickly and credibly generate politically usable information and move it to where it will have the most impact,” while symbolic politics refers to “the ability to call upon symbols, actions, or stories that make sense of a situation for an audience that is frequently far away” (Keck and Sikkink 1998, 16). In the case of advocacy for stronger FOI laws, international NGOs used these tools in attempts to influence domestic policymakers to pass laws that better met international best practices. In some cases this influence was directed specifically at policymakers, by working or communicating directly with them, while in other cases this influence was directed at empowering local civil society actors with new informational and symbolic tools to be involved in the policy process.

To capture the role of international NGOs in influencing the strength of FOI laws, I focus on the activities of Article 19, an international NGO and prominent member of the transnational advocacy network focused on transparency and freedom of information. Article 19, named after the corresponding article of the Universal Declaration of Human Rights, was founded in 1987 and bills itself as “defending freedom of expression and information.” While much of Article 19’s work on the freedom of information has taken the form of global efforts at norm entrepreneurship and agenda setting, in the late 1990s they began publishing detailed legal analyses of draft FOI laws, aimed at assisting policymakers with the drafting process, at persuading them to pass stronger laws, and at empowering local groups to participate in the process. While other organizations have become increasingly active on the FOI issue, especially over the last decade, for most of the period of time under study here Article 19 was the primary international NGO engaged in activities targeting individual countries’ draft legislation. Thus, its efforts are the focus of study here.

Article 19 published several key documents raising the global profile of the FOI issue, and highlighting issues of legal design. These include the 1996 “Johannesburg Principles on National Security, Freedom of Expression and Access to Information,” and the 1999 publication “The Public’s Right to Know: Principles on Freedom of Information Legislation,” both of which were endorsed by the UN Special Rapporteur on Freedom of Opinion and Expression. At the same time, Article 19 began publishing detailed legal analyses of draft FOI laws, aimed both at assisting the drafting process in countries with limited resources, and at persuading countries to pass stronger laws with greater protections and fewer limitations. These analyses were distributed via Article 19’s own email list, email lists of related advocacy organizations and networks, as well as targeted dissemination to politicians, local groups, and other individuals in each country. In 2001, Article 19 also published “A Model Freedom of Information Law,” which was referenced by most subsequent legal analyses as well as becoming a widely accepted standard to which many other INGOs, civil society groups, and international institutions refer.

Experts from Article 19 had access to the tools of both information politics—by virtue of their credibility (Gourevitch et al. 2012) and status as legal experts in an epistemic community (Haas 1989)—and to the tools of symbolic politics—by virtue of their role as norm entrepreneurs (Finnemore and Sikkink 1998). Their legal analyses both provided technical recommendations on the ramifications of different choices of legal language and features of institutional design, and made normative arguments about the principled value of the freedom of information and its recognition by international treaties and institutions.

In terms of symbolic politics, most analyses presented the freedom of information as a fundamental human right, and made principled arguments that it is important for democracy and good governance. They grounded these arguments in references to freedom of information in international treaties, such as the Universal Declaration of Human Rights and the International Covenant on Civil and Political Rights. They also often referenced statements and declarations made by other international and regional bodies—even when the country in question was not in the relevant region (for example, referencing statements by the Inter-American Commission on Human Rights and the African Commission on Human and Peoples’ Rights in an analysis of Nepal’s draft law). This highlights the importance of symbolic arguments in these cases, rather than direct institutional membership. Legal analyses also often pointed out how many other countries have passed FOI laws, and reviewed principles of design considered as best practice by Article 19.

In terms of information politics, analyses often highlighted some sections of draft laws for commendation, while reviewing sections that Article 19 considered problematic in terms of scope, procedures, exceptions, sanctions, or other design features, often making a case that specific features would have negative consequences for the freedom of information in practice. Analyses also often pointed out inconsistencies between different sections of a draft law, and highlighted omissions, such as oversight bodies or proactive disclosure requirements. Most importantly, legal analyses made clear recommendations of changes to draft legislation. Finally, in cases where Article 19 had already published an analysis in the same country, legal analyses often compared positive and negative changes from previous drafts.

Of course, not all legal analyses included all of the aforementioned components, as some were longer and more elaborate than others. Additionally, while most legal analyses were published solely by Article 19, some were in collaboration with domestic civil society groups. Finally, in some cases, the legal analyses were accompanied by physical visits to the country and meetings with key political figures involved in the drafting process. Thus, while some legal analyses differed from others in these regards, all used tools of symbolic politics and information politics in seeking to persuade political actors to pass stronger FOI laws.

While other organizations have also published legal analyses of draft FOI legislation in various countries, including the Commonwealth Human Rights Initiative and the Carter Center, or engaged in more direct legal capacity-building (such as the Organization for Security and Co-operation in Europe and the American Bar Association), Article 19 was both the first and the most consistent advocate engaged in shaping legal design in countries worldwide, and by far the most frequent in its use of legal analyses as a strategy. I measure these activities by recording every legal analysis published by Article 19 of draft FOI laws.

I consider only analyses published prior to the passage of each country’s FOI law, as critical analyses after passage cannot shape the design of those laws.⁴ The main sample of countries includes only those that passed FOI laws through 2008, as nearly every law passed after 2008 was subject to one or more legal analyses by Article 19 or more recently formed organizations, leaving little variation left to consider on the main independent variable. Finally, I exclude laws passed before 1999, as Article 19 was not previously active in conducting legal analyses.⁵ These exclusions leave 53 cases of FOI laws passed from 1999 to 2008, twenty of which received one or more Article 19 legal analyses. Table 2 lists the countries whose draft laws were subjects of legal analyses, along with the years those analyses were published and the year the law was passed in each country.⁶

Table 2 Article 19 legal analyses of draft FOI laws included in the main model. Excludes analyses where no laws had yet been passed as of 2008. Countries in the main model with FOI laws for which Article 19 published no legal analysis prior to passage: Antigua and Barbuda, Albania, Angola, Bosnia and Herzegovina, Chile, China, Czech Republic, Dominican Republic, Ecuador, Estonia, Germany, Greece, Georgia, Honduras, Indonesia, Jamaica, Japan, Liechtenstein, Nicaragua, Panama, Poland, South Africa, Slovakia, Slovenia, Saint Vincent and the Grenadines, Switzerland, Tajikistan, Trinidad and Tobago, Turkey, United Kingdom, Uruguay, Uzbekistan, Zimbabwe

4 Modeling the legal strength of FOI laws

To test whether legal analyses by Article 19 contribute to stronger FOI laws, I model legal strength across every FOI law passed between 1999 and 2008. The small size of this dataset sets a relatively high bar for finding a statistically significant role of transnational advocacy. The independent variables for each observation are, except where noted otherwise, the values observed for the year in which each law was passed. Thus the observation for Angola includes data on the independent variables for the year 2002, while the observation for Germany contains data for 2005.

The dependent variable is the total score on the RTI Rating measure of the legal strength of FOI laws. I use OLS regression, as the dependent variable is continuous and relatively normally distributed. While the measure of legal strength ranges in theory from zero to 150, the observed values range from 39 to 135, with a mean of 83.96. The key independent variable is a dummy variable measuring whether or not Article 19 published one or more legal analyses prior to passage of each country’s FOI law. I also show results using the count of legal analyses published prior to passage. Article 19 published such analyses prior to passage in twenty countries, twelve of which were subject to multiple analyses.

I include several control variables that may also be expected to shape the strength of FOI laws. Indeed, past work such as that of Michener (2010, 2015a) has focused on domestic factors such as independent media and constraints on the executive as shaping the strength of FOI laws in Latin America. It is essential to control for domestic political characteristics, as well as alternative international and transnational factors, in evaluating the role of transnational advocacy.

First, I include a measure of the passage of time, as the evolving global standards may lead to more recently passed laws being stronger than older ones. I measure the passage of time with a logarithmic transformation implying increasing marginal effects over time—that is, that a law passed in 2005 will be stronger than a law passed in 1995 to a greater extent than a law passed in 1995 will be stronger than a law passed in 1985.⁷ Democracies can be expected to pass stronger laws than autocracies for a variety of reasons. Democracies tend to already be more transparent (Hollyer et al. 2011), so the costs of a strong information regime to political actors are lower for democratic leaders than for autocratic leaders. Democracies are more likely than other states to be home to domestic interest groups that advocate for transparency, such as organizations of journalists, lawyers, or environmental or human rights organizations. Finally, democracies are likely to be more responsive to such citizen or interest group demands for transparency. I measure democracy using the Polity2 measure from the Polity IV project, which ranges from −10 to 10.⁸

Economic development should be an important determinant as well, but its effect could work in different directions. Wealthier countries have more resources to devote to legislative drafting and are likely to face stronger citizen demands for post-materialist values like transparency, but at the same time may also be home to more powerful interests that resist disclosure for security or business reasons. I measure economic development using the log of GDP per capita from the World Development Indicators.

Corruption is another important factor that may work in different directions. On one hand, more corrupt political actors can be expected to resist initiatives to increase accountability, as Bussell (2010, 2011) found in the case of public service reforms using new information technologies, both for Indian states and across countries. This would lead us to expect that when more corrupt countries are passing FOI laws, they will tend to pass the weakest laws, in order to appease international or domestic audiences while changing little in terms of the exposure of their activities to scrutiny. On the other hand, corrupt countries which do pass FOI laws may have reason to pass particularly strong laws, either out of a genuine commitment to reduce corruption, or in order to better impress external stakeholders. I measure corruption using the Control of Corruption measure from the World Bank World Governance Indicators, in which higher values reflect less corruption while lower values reflect more corruption. Since this measure is not available for the entire period under study, and given how slowly corruption tends to change over time in each country, I use the average for each country over all available years.

Finally, the design of FOI laws may be shaped by the legal structures that they enter into. A large literature has focused on the different implications of legal systems with common law and civil law origins for outcomes such as financial growth (Beck et al. 2000) and international human rights agreements (Simmons 2009). Indeed, many scholars have noted that common law systems tend towards more transparency (Beck et al. 2000). I include a dummy variable indicating whether a country has a common law legal system. However, most common law countries are also English-speaking former British colonies, so the results for this variable may, in fact, be proxying for the effects of one of those factors.

While many studies include country fixed effects in order to capture any omitted variables at the country-level, such a strategy is not possible in this paper since only one observation exists for each country in the dataset. However, I also show results including region fixed effects and year fixed effects, which capture any unmeasured regional differences or temporal trends. These models are much more conservative, only taking into account variation among laws passed in the same region, or among laws passed in the same year, thereby ensuring that the results are not an artifact of regional differences or of evolving global normative standards over time.

5 Main results

Table 3 presents the main results. Model 1 first presents a base model with no measure of legal analyses. The effect of time is positive but not statistically significant, while democracies and common law countries both tend to have stronger FOI laws. Wealthier countries, on the other hand, tend to have weaker FOI laws. The effect of corruption is not statistically significant.

Table 3 Main results: linear models of the legal strength of FOI laws, over the period 1999 to 2008

Model 2 presents a simple model testing only the bivariate effect of legal analyses with no other control variables. The results show that countries subject to legal analyses tend to have laws 18.46 points stronger on the RTI Rating. In Model 3, which adds back in the other control variables, the results show that countries subject to legal analyses tend to have laws 14.77 points stronger than other countries. Given that the observed range of the RTI Rating is from 39 to 135, with a standard deviation of 21.2, an effect equivalent to roughly 15 points is substantively very large. Comparing this relationship with the coefficient for democracy, simply being exposed to legal analyses has an effect on legal strength equivalent to a difference of 8.79 points on the Polity2 measure of democracy ranging from −10 to 10.

The results of Model 4, using a measure of the number of legal analyses, shows that each additional legal analysis published about a country’s draft FOI law tends to lead to a law 7.65 points stronger. Since the observed range of this measure is from 0 to 4, this implies that a country subject to the maximum observed number of legal analyses would be expected to have a law 30.6 points stronger than a country with none. This is equivalent to 1.44 standard deviations of legal strength, or a shift of 19.78 points on the Polity2 scale—nearly the difference between the most democratic and most autocratic possible countries. Models 6 and 7 include region fixed effects and year fixed effects, respectively. Even in these more conservative models, the findings remain strong. In particular, the results of Model 7 show that even among laws passed in the same year, those that received legal analyses before passage tended to be stronger than those that did not. These results effectively counter any concerns that the results are an artifact of global normative standards evolving over time, or of any region being particularly prone to outside influence.

Additional results (available in the Online Appendix) include numerous robustness checks confirming the main findings, including many additional domestic and international control variables, measures capturing international NGOs beyond Article 19, and variables using alternate samples and measures of legal analyses.

6 Endogeneity concerns

Endogeneity poses a potential concern for modeling the effect of Article 19 legal analyses on the strength of FOI laws. There is no plausible exogenous instrumental variable that might explain the occurrence of legal analyses but not be associated with the strength of FOI laws through any other causal channel. Nonetheless, by closely examining the precise forms of endogeneity that might potentially be at work, I can demonstrate that they are unlikely to be fatal problems for the results of this study. The concern here is not over direct reverse causality, but rather over two potential forms of endogeneity. In the first, some unobserved variable might be causally related both with the publishing of legal analyses and with strong FOI laws. In the second form, Article 19 might strategically choose to issue legal analyses for countries whose laws they expected to be stronger. Either of these situations might result in biased results showing a strong effect for legal analyses when the true effect was smaller. I will address each of these possibilities in turn, and then present alternative results using propensity score matching.

In any quantitative study, some possibility remains that an unobserved variable jointly determines the outcome and a key independent variable, and researchers can only do their best given available data to avoid this possibility. Given that this is not a panel study, it is not possible to include country fixed effects. However, I do show in Model 5 in Table 3 that the main results hold even when including region-specific dummy variables, thus controlling for any unobserved variables at the regional level, such as cultural or historical patterns. I also show in Model 6 in Table 3 that the main results hold when including year-specific dummy variables, controlling for any unobserved changes over time. Beyond these checks, I can only argue that the independent variables that I have included in the models and in the robustness checks (in the Online Appendix) are comprehensive of any factors that could potentially be causally related with both the publishing of legal analyses and the strength of FOI laws.

The second possible form of endogeneity concerns the potential that Article 19 strategically issues legal analyses of draft legislation that it expects will result in strong FOI laws. This would lead to a finding of a strong relationship when the true relationship was not a causal one. While research has shown that international NGOs do act strategically in terms of where they target their activities (Ron et al. 2005), this is a picture clearly at odds with the reality of the analyses provided by legal experts at Article 19 and of the countries where they are directed. Instead, Article 19 targets their legal analyses towards laws that they expect to be particularly weak, a situation which they wish to redress. Most legal analyses focus on rectifying language of draft laws which is seen as not meeting international best practices, such as broad exemptions to coverage of the law, potential barriers to its use, or the absence of independent appeals agencies and sanctions for non-compliance.

A former legal expert for Article 19 noted that countries were selected on the basis of several factors, including Article 19’s capacity at the time, the presence of local partners on the ground or involved international institutions, the “sensitivity of the country to change,” as well as “a degree of ad hoc-ness in the sense of what we found interesting, when we had more time, access to the law in English (or French or Spanish depending on staff).”⁹ Additional patterns can also be seen by comparing the observable characteristics of countries where Article 19 focused attention, and those where they did not. Based on these factors, legal analyses appear more likely to be published in less favorable situations for strong FOI laws to emerge. Among countries which passed FOI laws in 1999 or afterwards, the mean Polity2 democracy score of countries that received no legal analyses was 6.84, while the mean score of countries that received one or more analyses was 3.9. Countries that received analysis were more autocratic, on average, making them more likely to have weak laws absent the intervention of Article 19.

If Article 19 were strategically targeting countries it expected to pass strong laws, it would not be publishing legal analyses of draft laws in countries like Azerbaijan, Pakistan, and Uganda. If Article 19 instead targets countries that it expects to have weak laws, the only direction in which this would bias the coefficient for the legal analysis variable is downwards—if legal analyses had no causal effect on legal strength, we would expect to find a negative relationship. In fact, this makes the finding of a statistically significant positive effect even more noteworthy.¹⁰

It is also possible to formally calculate how great the role of unobserved determinants of such targeting would have to be, relative to the observed controls, in order to fully explain away the results. I follow the coefficient-stability approach of Bellows and Miguel (2009) and Altonji et al. (2005) to compare the coefficients for legal analyses in a model with no other controls (Model 2 in Table 3), and a model with controls for observable factors which might impact Article 19’s selection process (Model 3 in Table 3). The ratio of the coefficient in the model with controls to the difference between the two coefficients is equal to 4. This means that, in order to fully explain away the effect of legal analyses, the role of unobserved factors in driving Article 19 targeting would have to be four times as great as the role of the observed factors time, democracy, economic development, corruption, and legal system.

As one final approach, I present results using propensity score matching. This procedure matches each observation that received one or more legal analyses with another observation that received no legal analyses, yet is as similar as possible on all other independent variables. The resulting “matched” pairs of similar countries can be compared with a simple difference in means test or be employed in a model. Although this approach is limited to matching on observable variables, it can at least provide assurance that the results are not an artifact of observed differences among countries that did, and did not, receive legal analyses.

I use propensity score matching with two datasets, one restricted to the post-1998 time period when Article 19 was publishing analyses and the other unrestricted. Effectively, the first approach matches each country that was the subject of legal analyses with another country which was not and which also passed a FOI law post-1998. The second approach, on the other hand, potentially matches legal analysis countries with other countries that passed FOI laws at earlier points in time, if those countries are the best matches in terms of the other independent variables. Each procedure results in 20 matched pairs, for a total of 40 countries.

For the dataset matched only among post-1998 laws, the difference in means is 17.75, with a p-value of 0.005. For the dataset matched using all FOI laws, the difference in means is 17.25, with a p-value of 0.007. Results from modeling the strength of FOI laws using these matched datasets are presented in Table 1 in the Online Appendix. In both models, the effect of Legal Analyses is positive and statistically significant at a 95 % level or higher. These results are further evidence of the positive effect of Article 19 legal analyses on the strength of FOI laws.

7 Evidence on mechanisms: Information politics and symbolic politics

The models presented in Table 4 use several different approaches to test between information politics and symbolic politics as the primary mechanism by which Article 19 legal analyses have an effect. I approach this in two different ways—by focusing on differences in the legal analyses themselves, and by focusing on differences in the context in which laws are passed. The results of both of these, while primarily suggestive, support symbolic politics as the more salient mechanism than information politics.

Table 4 Linear models of the legal strength of FOI laws, including variables to test between mechanisms of information politics and symbolic politics

To focus on differences in the analyses, I coded every legal analysis in the dataset for the total number of words in the document, and the total number of words in each of the two main sections of each document. The first of these sections, which I refer to as “Recommendations,” counts the number of words in the section containing detailed recommendations pertaining to specific articles and features of the legislation. The second of these, “Principles,” counts the number of words in the section grounding the recommendations in international, regional, or constitutional principles, standards, and obligations. In nearly every analysis, these sections were explicitly marked with distinct headings, and in the remainder I was still able to identify separate, non-overlapping sections of the document. The proportions of words in each of these sections serve as indicators of the salience of the informative content of the analysis versus the symbolic content.

Using this information, I first consider separately the 11 legal analyses that contained no section on international principles at all, alongside the remaining analyses that contained at least some non-zero proportion of words in this section. Model 1 in Table 4 includes two separate count variables, one capturing each of these sets of legal analyses. The results show that the legal analyses containing a principles section have a large and statistically significant effect on legal strength, whereas legal analyses with no principles section have a smaller effect that is not statistically significant. This supports a stronger role for symbolic politics than information politics, as the analyses without any symbolic content have no significant effect.

Next, in Model 2, I interact the dichotomous legal analysis indicator with the total proportion of words in each country’s legal analyses that were devoted to recommendations. In Model 3, I instead interact the legal analysis indicator with the total proportion of words devoted to principles. These models do not include the base terms for the word proportion variables, as these variables take only zero values for countries with no legal analyses. The results show that legal analyses have a greater impact on legal strength as their international principles sections become a greater proportion of their total word length, whereas they have a weaker effect as their recommendations sections become a greater proportion. Neither interaction term is statistically significant, so these results must remain merely suggestive. They are however, consistent with the results of Model 1 highlighting symbolic politics as the more salient mechanism.

Finally, in Model 4 of Table 4, I focus on differences in the context of each country that can help tease apart the two mechanisms. Countries differ in how rich of an information environment policymakers are operating in when designing FOI laws. In some countries, FOI laws were passed when there were few other laws already passed in their region, or even none altogether. In other countries, FOI laws were passed with many other laws already passed in their region, providing a multitude of examples to learn from. In a low-information environment with few other laws in the region, we might expect the information-provision role of legal analyses to be able to contribute the most. On the other hand, in a high-information environment with many other laws in the region, we might expect this role to be much less needed and the symbolic role of legal analyses to predominate. Thus, an interaction term between the legal analysis variable and the number of laws already passed in a country’s region serves as a test between information politics and symbolic politics. The results of Model 4 show that the effect of legal analyses is greater when there are more FOI laws already passed in a country’s region—that is, where policymakers already have access to more information about the consequences of different design choices in similar countries, and the information-provision role of legal analyses is thus less important. This finding is consistent with symbolic politics as the more important mechanism, and inconsistent with information politics as the more important mechanism.

Importantly, while all of these results are limited in how definitively they can support one mechanism over the other, they all agree in supporting symbolic politics as the more salient mechanism. The next section focuses on the specific case of Serbia in order to more closely examine the role that Article 19 played.

8 Illustrative case of Serbia

The quantitative evidence presented in this paper shows that FOI laws that were subject to legal analyses by Article 19 before their passage tend to be stronger in their legal design than those that were not. In this section, I build on this cross-national evidence by focusing on the case of one specific country, Serbia, and presenting evidence from the process of FOI law design and adoption. This section not only illustrates how the argument of this paper played out in one specific country, but also shows how Article 19 played an important role in shaping the design of Serbia’s FOI law in multiple ways and at multiple points in the process.

Serbia is a useful case to examine in more detail. Serbia’s 2004 FOI law has been rated by the RTI Rating as the strongest in the world in terms of legal design. Yet this stands in stark contrast with Serbia’s context of widespread corruption and “illiberal resilience” (Pešić 2007, Edmunds 2009), especially in the period when the law was passed. In 2004, Serbia was tied for 97th place out of 146 countries on Transparency International’s Corruption Perceptions Index: the same placement as Algeria, Lebanon, Macedonia, and Nicaragua. Only three European countries received worse corruption scores: Albania, Moldova, and Ukraine. Thus, while the evidence here cannot conclusively demonstrate that a counterfactual Serbia where Article 19 was not involved would have passed a weaker law, this case can nonetheless help to illustrate the argument and specific mechanisms at work.

In Serbia, Article 19’s role was multifaceted, involving both in-country meetings and the publication of analyses and press releases, the direct provision of expertise and more passive normative standard-setting through their Model Law, as well as direct political pressure at crucial points in time. Article 19 provided crucial information, but was seen as an authoritative source of such information largely because of their symbolic role in setting international normative standards of FOI legal design.

Article 19 assisted, at different points in the process, both civil society drafters of proposed legislation and parliamentary drafters of official legislation (Lilić and Milenković 2004, Otvoreni Parlament 2004). Multiple experts from Article 19 visited the country multiple times, as well as participating in region-wide civil society meetings.¹¹ Article 19 published one analysis of an early draft law prepared by local NGOs, and three separate analyses at different stages of official government draft legislation. These analyses were distributed via press releases and through email lists. One of the analyses even criticized the most recent draft for being worse in many areas than the previous one examined, writing that “Unfortunately, the majority of our recommendations to address these shortcomings have not been addressed by the recent amendments. In fact, some of the positive features we initially identified have been removed from the new version of the Law” (Article 19 2003b, p. 1).

Article 19 also helped to pressure the Serbian government at a crucial stage in the process when the election of a new Prime Minister threatened to wipe out all earlier progress. Zoran Đinđić, Prime Minister from 2001 until his March 2003 assassination, had been relatively open to the many draft laws on multiple topics (including amnesty laws, broadcasting laws, and laws on conscientious objectors in the military) being promoted by civil society groups—indeed a civil society-produced draft FOI law was adopted by the Culture Minister in modified (albeit weakened) form as the official government draft (Lilić and Milenković 2004). Following Đinđić’s assassination and the ensuing state of emergency and series of caretaker Prime Ministers, Vojislav Koštunica was elected Prime Minister after the December 2003 elections, taking office in March 2004. While Đinđić’s social-democratic Democratic Party and Koštunica’s conservative nationalist Democratic Party of Serbia had both been members of the Democratic Opposition of Serbia opposed to Slobodan Milošević, they had since parted ways.

Pešić (2007, p. 2–3) writes of the period after Koštunica’s election:

“Following the change in government, there was a shift in priorities leading to a more old-styled manner of governance, expressed in the political/party control of the police, the security intelligence agency, the media, and the judiciary, bringing back the old cadres to positions in state organs. Whereas the first Djindjic government ambitiously and enthusiastically concentrated on enabling Serbia to integrate with the EU as soon as possible, enthusiasm for the EU integration process noticeably ebbed after the second government came to power. A rightist clerical-nationalist party has played the leading role in the coalition government set up in March 2004. Adverse to western values, it has placed commitment to EU integration on a back burner.”

At the time there was “considerable doubt as to whether the Koštunica government would carry out the liberal reforms proposed by the DOS coalition” (Fagan 2010, p. 114). Freedom House (2004) noted that “in 2003, the direct influence of NGOs and think thanks on the government appeared to diminish as the political consensus among those who ousted Milosevic began to break down.”

The FOI law was one of the many reform bills put in jeopardy in this period, yet it was still passed in 2004. A human rights lawyer involved in the drafting process recalled that of the many draft laws which “didn’t come back” after Koštunica sought a “tabula rasa,” the draft Law on Free Access to Information of Public Importance was “one of the only ones that did come back,” largely as a result of concerted pressure by domestic civil society, Article 19, and the OSCE.¹²

Both before and after Đinđić’s assassination, Article 19 played an especially important role in emphasizing the importance of an independent appeals body as an institutional design choice. This was over the skepticism of both some politicians and of some civil society actors, who all opposed the creation of such an independent body—the civil society groups because they thought its independence would be compromised by political pressure, and the politicians because they feared it would remain too independent.¹³ Indeed, after passage of the law, the Information Commissioner emerged as one of the most important figures in promoting awareness, implementation, and enforcement of the law’s provisions.

On October 26 and 27, 2004, the Serbian Parliament debated the legislation for the final time before adopting it on November 5. Several speakers attested to the important role of NGOs in shaping the law. The Minister of Culture, Dragan Kojadinović, said “the law was developed in cooperation with the OSCE, the Council of Europe and a number of non-governmental organizations” (Otvoreni Parlament 2004).¹⁴ Aleksandar Lazarevic, from the G17 Plus party, said that “ultimately, in this bill are quite involved NGOs… non-governmental organizations have full involvement in the preparation of this law” (Otvoreni Parlament 2004). Snežana Lakićević Stojačić of the Democratic Party went so far as to specifically point out Article 19 by name. After stating that “the International Center Against Censorship, Article 19, formulated a set of international principles to establish the standard by which everyone can assess whether domestic laws actually provide access to official information,” she reviewed each of the nine principles from their 1999 publication “The Public’s Right to Know: Principles on Freedom of Information Legislation”¹⁵ in great detail, before concluding that “I am pleased to note that all these principles are incorporated into the provisions of this law” (Otvoreni Parlament 2004).

Other groups played important roles in Serbia as well, especially the OSCE Mission to Serbia, the Council of Europe, and the Open Society Institute. However, they were much more important as sources of funding and of political pressure than as sources of expertise pertaining to legal design. While the OSCE and Council of Europe also provided civil society members with comments on early proposals, a human rights lawyer involved in the drafting process recalled that Article 19’s comments were more useful and had greater influence on changes made to subsequent drafts, particularly noting the usefulness of the Article 19 Model Law as a model, and the helpfulness of two Article 19 experts who visited Serbia.¹⁶ They also noted that most employees of the relevant OSCE office in Belgrade at the time came from the domestic civil society community, and so had little expertise on the issue beyond that already possessed by other members of that community.¹⁷ The American Bar Association, meanwhile, conducted a legal analysis only after the law had already been passed.¹⁸

The case of Serbia highlights how Article 19 contributed to a much stronger FOI law than could otherwise be expected given the political context at the time. Article 19 played multiple roles, contributing expertise that domestic civil society groups did not have access to, helping to empower local civil society groups, and directly lobbying political actors. Both local civil society actors and individual legislators also saw significant symbolic value in Article 19’s recently-developed international standards. Crucially, Article 19 helped ensure the inclusion of an independent oversight body in the design of the FOI law, which earlier drafts had either not included or included only in weaker form.

9 Conclusion

This paper offers new quantitative evidence that international NGOs can influence domestic politics and policy, leading policymakers to pass stronger laws than they would otherwise. This evidence adds to a growing body of work on the impacts of INGOs, but goes beyond focusing on “naming and shaming” alone to highlight the influence of INGO advocacy on domestic policymaking. This paper also highlights a specific type of INGO influence that has received little attention to date—shaping the translation of international norms into domestic law. INGOs can directly and indirectly pressure policymakers to alter legal design to better reflect international norms, using the tools of both information politics and symbolic politics, as shown in an illustrative case study from Serbia.

While the empirical analyses in this paper are limited by a small sample size, their results are highly robust to alternative modeling approaches, inclusions of additional control variables, and the use of very conservative approaches including regional fixed effects, year fixed effects, and propensity score matching. The results of these models consistently find substantively large effects of transnational advocacy by the INGO Article 19 on the strength of FOI laws. Further, there is little reason to suspect that the results are an artifact of endogeneity or of any strategic selection process by Article 19. I also offer suggestive evidence that favors symbolic politics over information politics as the more salient mechanism.

While strong legal design does not always mean strong institutions in practice, recent research (Michener 2015b) has found that strong legal design is an important determinant of effective FOI laws in practice among Latin American countries. The existence of independent information commissions, an important component of legal design, has been crucial to the success of FOI laws in countries such as India, Mexico, and Serbia. On the other hand, in many other countries Freedom of Information laws are poorly implemented and enforced, regardless of their design. In that sense, the phenomena studied here can be understood as one stage in a “spiral model,” where adoption of a norm is still to be followed by subsequent stages of norm internalization (Risse and Sikkink 1999).

Lastly, while the evidence in this paper is from one specific issue area—the design of Freedom of Information laws—similar dynamics are likely to be at work in other issue areas. International NGOs undertake similar activities to shape new draft legislation on topics including environmental laws,¹⁹ human trafficking,²⁰ and domestic violence.²¹ In all of these cases, transnational advocates can use both information politics—based on their legal expertise—and symbolic politics—making principled arguments based on relevant international norms—to pressure policymakers to pass laws which are stronger and more in line with international principles. Future research should seek to compare across these issue areas, across different international NGOs, and across different mechanisms of influence, in order to more fully understand the mechanisms and scope conditions for transnational advocacy to shape domestic policies.