Why do some national governments in East and Southeast Asia receive more transnational scrutiny and pressure on their domestic human rights practices than others? Since the United States’ human rights turn in foreign policy in 1977 under the Jimmy Carter administration, human rights international nongovernmental organizations (international NGOs or INGOs)Footnote 1 have monitored, publicized, and criticized governments’ human rights practices in their effort to leverage the influence from the US government, international organizations, the mass media, and multinational corporations for political change (Cmiel Reference Cmiel1999; Moyn Reference Moyn2010, 120–75). This effort has often shaped the targeted states’ politico-economic dynamics in important ways by increasing the Northern media's news coverage, reducing the inflows of foreign direct investment, raising the costs of government repression, or even enhancing the prospect of humanitarian intervention (Barry, Clay, and Flynn Reference Barry, Clay and Flynn2013; DeMeritt Reference DeMeritt2012; Krain Reference Krain2012; Murdie and Peksen Reference Murdie and Peksen2014; Ramos, Ron, and Thoms Reference Ramos, Ron and Thoms2007). Indeed, during the past four decades transnational human rights reporting has increasingly become salient across Asia. According to my new data, in 1977 Amnesty International, a London-based human rights INGO, released a total of only seven special country reports on four Asian states. In 2008 it issued 176 such reports targeting 21 states in East and Southeast Asia alone.
However, beneath this growing transnational attention to Asian states’ human rights issues lie considerable cross-national variations in who gets targeted when and how much. Figure 1 illustrates this by showing the number of how many special country reports were issued by Amnesty International for China, Myanmar, North Korea, the Philippines, and South Korea in each year from 1977 to 2008, as well as the global and Asia regional averages.
Annual Number of Amnesty International's Special Country Reports, 1977–2008
While both the Philippines and South Korea transitioned to democracy around 1987, South Korea continued to receive the far greater coverage from Amnesty International throughout the 1990s as well as in 2008 when freedom of expression started deteriorating (Haggard and You Reference Haggard and You2015). Also, although China, Myanmar, and North Korea remained highly repressive dictatorship during the period, North Korea's human rights violations received little transnational attention (even below the world average), compared to those of the other two. This poses a puzzle since, although the conventional wisdom has concentrated on “the Asian values” as the dominant explanation of the region's human rights issues, Asian states, despite their shared culture, vary greatly in the extent to which they become a target of transnational human rights reporting.
This article argues that the government of an Asian state receives more transnational scrutiny and pressure on their domestic human rights practices if and when local NGOs and activists in that state have been connected with a greater density of human rights INGOs through a local membership base during the preceding period. Specifically, this operates through two mechanisms. First, human rights INGOs can help strengthen local members’ capabilities to leverage human rights and international solidarity as an advocacy strategy, which in turn increases domestic expectations and demands for transnational human rights reporting. Second, human rights INGOs can mobilize local members and volunteers for monitoring and information collection on the ground, thereby increasing opportunities for transnational human rights reporting. Thus, human rights INGOs’ prior engagement with local members and volunteers in an Asian state positively relates to the extent to which the government of that state becomes the target of transnational human rights reporting in the subsequent period.
To test the argument, this article conducts event count analyses of all 25 Asian states for the period from 1977 to 2008, using new data on Amnesty International's special country reports directly supplied by the organization's International Secretariat in London. Controlling for government protection of human rights, regime type, military power, security linkages to the US, and other factors, I find that human rights INGOs’ local engagement has strongly significant positive relationships with the extent to which national governments become a target of transnational human rights reporting by Amnesty International across Asia. This finding is highly robust against three possible sources of endogeneity bias (namely, selection bias, reverse causation, and unobserved country-level heterogeneity), control variable bias, omitted variable bias, and alternative operationalization of control variables. In particular, to preview a robustness check, this article utilizes a two-stage estimation method and explicitly accounts for the possibility that human rights INGOs may be non-random and selective in building their local membership base in Asian states. I find strong evidence that human rights INGOs’ local engagement not only has no endogeneity problem, but also exerts strongly significant positive effects on Amnesty International's coverage. One major contribution of this article is to create the most accurate and comprehensive new data to date on human rights INGOs’ local engagement and to provide rigorous and robust evidence of its effects on transnational human rights reporting in Asia.
The research is important for several reasons. First, by offering the first systematic analysis of the determinants of cross-national variations in the degree of transnational human rights reporting in Asia, this article contributes to the literatures in East Asian Studies and human rights. The existing scholarship has concentrated on the Asian values (that is, Asian states’ shared culture, history, and religion) as the dominant perspective for understanding and explaining the region's human rights issues (for example, Kim Reference Kim1994; Sen Reference Sen1997; Zakaria Reference Zakaria1994; see also Svensson Reference Svensson2002, 47–70). However, it has not been obvious why Asian states, despite their shared values, vary greatly in terms of who gets targeted when and how much by transnational human rights reporting, and whether the existing cultural explanation is effective for explaining this increasingly salient aspect of Asian politics. I seek to fill this theoretical gap and empirical anomaly in the literatures by advancing an actor-centered, political theory of the determinants of transnational human rights reporting in Asia, and by offering rigorous empirical tests, using the most fine-grained quantitative measurement to date of human rights INGOs’ local engagement and Amnesty International's human rights reporting.
Second, this article offers a new way of thinking about transnational–domestic linkages in Asian politics. While I accept that domestic politics is key to understanding Asia's political economy and foreign policy, I also demonstrate that Asian states vary significantly in terms of the density of their domestic actors’ network with human rights INGOs, as well as the extent to which national governments become a target of transnational scrutiny of their human rights record. By specifying the theoretical mechanisms through which human rights INGOs relate to local actors and transnational reporting, and by establishing their relationships empirically, I provide theoretical reasoning and empirical evidence suggesting the crucial importance of paying sustained analytic attention to transnational–domestic linkages in the analysis of Asian politics.
PATTERNS OF TRANSNATIONAL HUMAN RIGHTS REPORTING IN ASIA
This section explores and describes patterns of transnational human rights reporting in Asia by focusing on Amnesty International's so-called special country reports as a crucial case, given the organization's key role in pioneering and leveraging various techniques to investigate, expose, and shame human rights-violating states since its founding in 1961. It should be noted that Amnesty International's special country reports are distinct from its “Annual Report.” Annual reports document and summarize the state of human rights in most countries in the world once in every year (Spry Reference Spry2007, 25). In contrast, unlike its regular annual reports, Amnesty International issues special country reports on an irregular basis to condemn human rights abuses committed by a particular government and to launch a campaign against that government during a specific period, based on in-depth country research and with policy recommendations (Spry Reference Spry2007, 25–28). As a country-specific advocacy tool, these special country reports are sent to Western government policy makers, United Nations (UN) officials, human rights professionals, academics, and journalists (Ron, Ramos, and Rodgers Reference Ron, Ramos and Rodgers2005, 561).
Table 1 compares the level of state repression with the volume of Amnesty International's human rights reporting for all 25 states in East and Southeast Asia. Specifically, the left half of the table lists the states according to the average level of government repression of physical integrity rights over 1976–2007 on a –3 (the most repressive) to + 3 (the least repressive) scale, using Fariss’ (Reference Fariss2014) data. The table's right half rank-orders all Asian states in terms of the total number of Amnesty International's special country reports targeting them throughout 1977–2008. In the left column, the values of government repression are lagged by one year to consider that it is likely to take time for Amnesty International to produce research-intensive special country reports in response to states’ human rights situations.
State Repression and Amnesty International's Human Rights Reporting in Asia
Note: In the left column, the level of state repression is measured annually on a –3 (the most repressive) to +3 (the least repressive) scale based on Fariss’ (Reference Fariss2014) data and then averaged over 1976–2007. The right column is constructed from my new data on Amnesty International's human rights reporting. See this article's Research Design section for more details on both data.
In essence, there exist considerable discrepancies between the level of governments’ physical integrity rights violations and Amnesty International's reporting coverage on such infractions. For instance, although Afghan people suffered, on average, the worst domestic human rights violations across Asia, Afghanistan was merely the thirteenth most-targeted state by Amnesty International, receiving total 75 special country reports. Here, Amnesty International focused on such issues as incommunicado detention, torture, unlawful killings, political prisoners, the death penalty, the Taliban's abuses against citizens, US custody in Afghanistan, and the impunity of the US-led International Security Assistance Force in Afghanistan. In contrast, even though the Japanese enjoyed the best human rights protection by their government, Japan ranked sixteenth in Asia in terms of Amnesty International's coverage, garnering total 65 reports. In this case, Amnesty International covered a very different set of issues beyond physical integrity practices, including North Korea's forcible abductions of Japanese citizens, secret executions of death row inmates, ill-treatment of foreigners, inadequate protection for refugees and asylum seekers, women survivors of Japan's military sex slavery system, and ex-President of Peru Alberto Fujimori.
These variations across Asian states show that the volume and issue coverage of Amnesty International's special country reports is determined by more than just “a meritocracy of suffering” (Bob Reference Bob2005, 6). This suggests the need to go beyond the targeted states’ human rights situations and to unpack the underlying political process for human rights INGOs’ transnational human rights reporting.
EXISTING EXPLANATIONS OF TRANSNATIONAL HUMAN RIGHTS REPORTING
Although transnational human rights reporting has in practice emerged and operated since the late 1970s, it is only in relatively recent years that political scientists have begun to analyze this phenomenon. Keck and Sikkink's (Reference Keck and Sikkink1998) pioneering work, Activists Beyond Borders: Advocacy Networks in International Politics, serves as a starting point for understanding and explaining transnational human rights reporting or so-called “information politics,” which they define as “the ability [of nongovernmental actors] to quickly and credibly generate politically usable information and move it to where it will have the most impact” (16). Drawing on constructivist International Relations (IR) theory and the Latin American cases, they argue that transnational networks, primarily consisting of domestic and international NGOs, “are organized to promote causes, principled ideas, and norms, and they often involve individuals, advocating policy changes that cannot be easily linked to a rationalist understanding of their ‘interests’” (Keck and Sikkink Reference Keck and Sikkink1998, 8–9). Specifically, they identify several conditions for successful transnational reporting. First, transnational reporting is more likely to focus on dictatorships than democracies. According to the so-called boomerang model of human rights change, the lack of domestic remedies and voice opportunities in dictatorships incentivize domestic groups to seek a remedy for their social grievances by circumventing domestic blockage and reaching international allies for external help (Keck and Sikkink Reference Keck and Sikkink1998, 12–13). Second, transnational human rights reporting is more prone to target states where domestic groups’ social grievances are driven by government infraction of individuals’ physical integrity rights (Keck and Sikkink Reference Keck and Sikkink1998, 27).
Recently, however, IR scholars have begun to challenge Keck and Sikkink's (Reference Keck and Sikkink1998) constructivist explanation and to instead focus on political economy factors in transnational human rights reporting. According to this political economy explanation, the conventional constructivist focus on human rights principles, victims’ needs, and altruism alone cannot explain why human rights INGOs like Amnesty International cover human rights-violating states unevenly despite the worldwide prevalence of human rights violations and these organizations’ purported goal of universal coverage (Bob Reference Bob2005; Hendrix and Wong Reference Hendrix and Wong2014; Ron, Ramos, and Rodgers Reference Ron, Ramos and Rodgers2005). In particular, Ron, Ramos, and Rodgers’ (Reference Ron, Ramos and Rodgers2005) pioneering global statistical analysis of Amnesty International's human rights reporting shows that such non-human rights factors as state power and US military aid account for coverage by Amnesty International, above and beyond what the level of government respect for physical integrity rights can predict. Also, Hendrix and Wong (Reference Hendrix and Wong2014) find that Amnesty International is more likely to target states with higher security linkages to the US in terms of arms transfers and voting similarity within the UN General Assembly.
Although the above scholarship has provided some possible explanations, the question of what determines Asia's transnational human rights reporting has so far escaped scholarly attention. The existing literature has either disproportionately represented the Latin American experience (for example, Keck and Sikkink Reference Keck and Sikkink1998) or concentrated on global average statistical associations. As such, little remains known about the generalizability of the existing findings to the historically specific context of East and Southeast Asia. This omission in the field of East Asian Studies is both important and surprising, since Asian states are now routinely subject to transnational scrutiny and pressure on their domestic human rights practices. This article fills this void by offering the first systematic analysis of the determinants of cross-national variations in the extent to which the governments of Asian states become a target of transnational human rights reporting.
In this article, I advance a variant of constructivism, one that underscores the role of human rights INGOs and their local engagement, by both building on and going beyond Keck and Sikkink (Reference Keck and Sikkink1998). I argue that human rights INGOs, through their local membership base, can enhance advocacy repertoire and monitoring capacity at the domestic level, thereby fostering transnational human rights reporting. To begin with, I build on Keck and Sikkink's (Reference Keck and Sikkink1998) insight that the ties between human rights INGOs and local groups play a key role for transnational human rights reporting. Yet, while Keck and Sikkink (Reference Keck and Sikkink1998) and especially Sikkink (Reference Sikkink2008, 17) concentrate on the agency of local groups in the Global South in initiating “boomerangs” for human rights change, my argument (specifically, the second mechanism in the next section) emphasizes the human rights INGO side of the same coin and explains why human rights INGOs choose to return those boomerangs that local groups have thrown to them. Moreover, my theory (particularly, the first mechanism in the next section) advances a new theoretical proposition that the very human rights actorhood of local groups in the Global South, often assumed and unproblematized in the existing scholarship, is in good part a social construct by human rights INGOs through their local membership base. For instance, although Keck and Sikkink (Reference Keck and Sikkink1998) and Sikkink (Reference Sikkink2008) emphasize that local activists and victims have strong preferences for (net)working with human rights INGOs for domestic social change, it remains taken-for-granted and unspecified where such preferences (and even their human rights agency) come from in the first place. By providing a human rights INGO-centered theory of local groups’ identity and interest formation, I seek to fill this important theoretical gap within the existing literature.
Furthermore, I go beyond the conventional focus on the Asian values (that is, Asian states’ shared culture, history, and religion) as the dominant explanation of Asia's human rights issues (Kim Reference Kim1994; Sen Reference Sen1997; Zakaria Reference Zakaria1994). Although much of the existing human rights scholarship has predominantly prioritized this cultural explanation, it is puzzling why Asian states, despite their alleged shared Asian values, vary greatly in terms of who gets targeted when and how much by transnational human rights reporting.Footnote 2 In contrast, I aim to fill this theoretical gap and empirical anomaly in the literature by advancing an actor-centered, political explanation of the determinants of transnational human rights reporting in Asia.
THEORETICAL EXPECTATIONS
This article argues that human rights INGOs, through their local membership base, have played a key role in catalyzing transnational human rights reporting across East and Southeast Asia. Human rights INGOs operate through two theoretical mechanisms. The first mechanism is that human rights INGOs can expand local members’ advocacy repertoire, which in turn increases domestic expectations and demands for transnational human rights reporting. Although Keck and Sikkink (Reference Keck and Sikkink1998, 28–29, 206–207) emphasize the existence of local NGOs and activists as a precondition for successful transnational reporting, they assume local actors’ strong preferences for human rights and transnational activism as given and fixed. However, as recent studies in anthropology, history, and political science demonstrate, local people’ capabilities to understand grievances in human rights terms and to leverage transnational allies for domestic contention are neither innate nor pre-ordained but learned (Bob Reference Bob2009; Merry Reference Merry2006; Moyn Reference Moyn2010; Simmons Reference Simmons2009). In the absence of such learned capabilities, local activists—even in Europe and the Americas—have historically relied on such non-human rights frameworks as Marxism, the decolonization movement, or traditional patriarchal kinship for understanding social problems and their solutions without also searching transnational allies in the Global North (Merry Reference Merry2006, 179–217; Moyn Reference Moyn2010).
In view of this, for local activists and victims to utilize transnational networks and contacts for domestic human rights change, they need to learn about, and value, human rights and international solidarity in the first place as strategically useful for framing and redressing their social grievances. Human rights INGOs, with a local membership base, can serve as an important channel for increasing and reinforcing their local members’ capabilities to leverage human rights and international solidarity as an advocacy strategy in several ways. They increase affiliate NGOs and individual members’ awareness-raising and learning about human rights and international solidarity at the domestic level through training, education, action programs, and meetings (Kim Reference Kim2013, 512–514; Kim Reference Kim2016, 604–608). Human rights INGOs also engage in “field-building activities” on the ground, helping a wide variety of local NGOs and activists connect with one another and rally behind one common cause across different ideologies, preferences for movement strategy, and regional identities (Murdie and Bhasin Reference Murdie and Bhasin2011, 171–172). Furthermore, by establishing permanent national branches within states, human rights INGOs supply local members and volunteers with a steady flow of human, material, and organizational resources (Kim Reference Kim2016, 605–607; Murdie and Bhasin Reference Murdie and Bhasin2011, 173). In doing so, human rights INGOs help local activists and victims understand their grievances and situations as human rights issues of international concern. They thus increase domestic expectations and demands for transnational human rights reporting as a useful advocacy strategy.
The second mechanism is that human rights INGOs can build and strengthen their monitoring capacity on the ground through a local membership base, thereby increasing opportunities for transnational human rights reporting. Building and strengthening monitoring capacity within states is key to human rights INGO's successful transnational reporting for several reasons. First, human rights INGOs need to collect timely information on domestic human rights abuses committed by national governments from many states at low costs. Working under resource constraints, human rights INGOs do not have “all-seeing eyes.” Also, transnational reporting is not proactive or preventive but reactive in nature, primarily designed to provide a remedial action to government repression that has already occurred (Rodio and Schmitz Reference Rodio and Schmitz2010). As such, human rights INGOs need to receive a steady and reliable flow of human rights information from local activists and victims that may otherwise go unnoticed.
Second, human rights INGOs need to collect accurate information on the details of governments’ repressive actions and victims’ suffering to increase the effectiveness of transnational human rights reporting as an advocacy strategy. Given that transnational reporting involves the threat and use of negative publicity against the targeted governments, the burden of proof for human rights INGOs is very high. As Thakur (Reference Thakur1994) noted in the case of Amnesty International, “The long-term credibility of AI [that is, Amnesty International] would be badly damaged if its reports and statements could be shown to be false. The entire structure of the AI movement is designed to collect, distribute, and use information that has been cross-checked and will withstand determined efforts by governments to discredit” (150). Thus human rights INGOs need the necessary details for assessing the veracity and urgency of human rights situations and for proposing specific policy solutions to the targeted governments (Wong Reference Wong, Gourevitch, Lake and Stein2012). Given these requirements of successful transnational reporting, human rights INGOs can mobilize local members and volunteers as low-cost monitors on the ground in many states for information collection and verification. In doing so, human rights INGOs increase opportunities for transnational human rights reporting.
In sum, human rights INGOs should have positive relationships with the extent of transnational human rights reporting by increasing advocacy repertoire and monitoring capacity at the domestic level. Thus, this article hypothesizes that, all else equal, human rights INGOs’ local engagement in a state is positively associated with the extent to which the government of that state becomes the target of transnational human rights reporting in the subsequent period.
RESEARCH DESIGN
This article uses an event count framework to test the hypothesized positive relationships between human rights INGOs’ local engagement and the extent of transnational human rights reporting across Asian states. The analysis begins in 1977, the year in which the inauguration of the pro-human rights Jimmy Carter administration in the US gave an impetus to transnational human rights reporting worldwide (Moyn Reference Moyn2010, 120–175), and ends in 2008, the last year for which accurate data are available. The unit of analysis is the country-year, and the full data set includes 25 independent states in East and Southeast Asia and 766 country-year observations. Due to missing data, the main statistical model includes 740 observations.
The Model
The statistical analysis employs the generalized linear model (GLM) with the negative binomial (NB) probability distribution and the log link function (Gill Reference Gill2001; Hilbe Reference Hilbe2012, 193–199). The ordinary least squares regression model is inadequate because its linear functional form often produces nonsensical, negative predicted counts, and because discrete, nonnegative integers in the count data violate the normality of residuals (King Reference King1988). In contrast, the NB GLM is effective for analysing count data. Within the event count framework, the NB model is preferable to the Poisson model. The Poisson model imposes the mean-variance equality, assuming that each state is expected to receive the same number of special country reports from Amnesty International in each year. However, this assumption is violated since in the data the sample mean is 2.56 and the sample variance is 20.00. This over-dispersion indicates the presence of unobserved heterogeneity, temporal dependence, or both across Asian states. The NB model solves this problem by incorporating a unit-specific random error term into the Poisson variance (Hilbe Reference Hilbe2012, 141–184). Substantively, the NB GLM predicts how many special country reports are issued by Amnesty International for an Asian state in a given year. In the following analysis, the values of all independent variables are lagged by one year to reduce endogeneity bias, that is, to ensure that the independent variables temporally precede the dependent variable. To address cross-national heteroskedasticity, the Huber-White robust standard errors clustered on state are used.
The Dependent Variable
The dependent variable is nonnegative integers that count the number of special country reports—not regular annual reports—issued by Amnesty International for a state in a given year for the period from 1977 to 2008. To compute this variable, I create the most accurate data set of its kind to date by obtaining the new data directly from Amnesty International's International Secretariat in London. I utilize Amnesty International's data for several reasons. First, since it won the Nobel Peace Prize in 1977 for its work on Argentina's “Dirty War,” Amnesty International has been at the forefront of transnational human rights reporting, and its special country reports have been widely regarded as one of the most accurate, comprehensive, and credible sources of information on national states’ human rights practices around the world (Ron, Ramos, and Rodgers Reference Ron, Ramos and Rodgers2005, 559–560; Spry Reference Spry2007). Second, Amnesty International's human rights reporting has recently received sustained analytic attention from IR scholars (Hendrix and Wong Reference Hendrix and Wong2014; Rodio and Schmitz Reference Rodio and Schmitz2010; Ron, Ramos, and Rodgers Reference Ron, Ramos and Rodgers2005), so that it can serve as an appropriate test case for theory evaluation. Third, direct data support from Amnesty International's Research Department within its International Secretariat enhances the quality of quantitative data and statistical inference because it greatly reduces the risk that measurement errors (for example, undercounting Amnesty International's reports) bias the statistical results.
The Independent Variable of Theoretical Interest
The independent variable of my theoretical interest, Human Rights INGO Ties, represents the density of human rights INGOs’ local membership base. It is the natural log of the total number of human rights INGOs that are connected with local NGOs or activists in a state in a given year through a local membership base. The data source is the Yearbook of International Organizations annually published by the Union of International Associations from 1948 to 2009 (Union of International Associations various years). I count as human rights INGOs only those INGOs that have membership in at least two different states and pursue, as their organizational aim, the internationally recognized human rights codified in the Universal Declaration of Human Rights, the International Covenant on Civil and Political Rights, and the International Covenant on Economic, Social and Cultural Rights.
While this variable is not a perfect measure of human rights INGOs’ concrete capability-building and monitoring activities at the local level, there are several reasons why it is a reasonably good proxy for my theoretical mechanisms. First, unlike much of the existing scholarship that has used the same data source, this article measures human rights INGOs specifically, that is, only those INGOs who are relevant for my theory. This is an important measurement progress over much prior quantitative research lumping all types of INGOs together regardless of their issue areas (such as Elephant Aid International, the International Society of Urology, and the World Greyhound Racing Federation) and asserting that this generic measure captured the role of human rights INGOs.
Second, the Human Rights INGO Ties variable's natural-log specification helps capture human rights INGOs’ local-level activities (including those that are not necessarily documented or documentable) for a large number of country-year observations, while minimizing the risk of overestimating the human rights INGO effect. While one may wish to directly measure human rights INGOs’ local-level activities based on their documents, this can create measurement errors that arise from falsely equating absence of human rights INGOs’ documentation with absence of their local engagement. For instance, it appears too heroic to assume that all human rights INGOs in my data (with the sample mean of 63) should have documented each and every aspect of their local engagement in 25 Asian states for all the 32 years from 1976 to 2007. Some forms of human rights INGOs’ local engagement are not readily documentable, such as behind-the-scene public interest lobbying. Also, their other on-the-ground activities are often documented in an indigenous language that is inaccessible to outsiders. In view of these measurement challenges, Human Rights INGO Ties serves as a plausible and replicable proxy for human rights INGOs’ local engagement covering a large number of country-year observations, especially given the ample evidence that these organizations genuinely work with local activists and victims at the domestic level (Keck and Sikkink Reference Keck and Sikkink1998, 39–120; Kim Reference Kim2013, Reference Kim2016; Murdie and Bhasin Reference Murdie and Bhasin2011). Furthermore, specifying Human Rights INGO Ties as logged—not raw—numbers allows it to have a decreasing marginal effect by constraining each additional human rights INGO to contribute less to the statistical association with the dependent variable. As such, it minimizes the risk of overestimating the human rights INGO effect.Footnote 3
Control Variables
The statistical analysis includes a number of control variables to account for other determinants of the extent of transnational human rights reporting in Asia. The first two variables, Human Rights Protection and Democracy, account for Keck and Sikkink's (Reference Keck and Sikkink1998) argument that human rights-violating states and dictatorships are more likely to be targeted by transnational human rights reporting. Human Rights Protection measures the level of government protection of physical integrity rights on a –3 (the least protection) to + 3 (the most protection) scale, using Fariss’ (Reference Fariss2014) data. Unlike the existing alternatives, this provides a new dynamic measurement of governments’ physical integrity practices by explicitly accounting for the fact that the standard of government accountability about human rights abuses has become more stringent over time when international monitors like the US State Department interprets information about those abuses. Democracy is coded 1 if a state is a democracy in a given year and 0 otherwise, based on Cheibub, Gandhi, and Vreeland's (Reference Cheibub, Gandhi and Vreeland2010) Democracy and Dictatorship data. To overcome problems with the existing continuous indicators of democracy (for example, the arbitrary additive aggregation rule of the Polity IV data's democracy score), Cheibub, Gandhi, and Vreeland (Reference Cheibub, Gandhi and Vreeland2010) classify a polity as a democracy if and only if the executive is popularly elected directly or indirectly; the legislature is popularly elected; more than one party competes in the elections; and there must have occurred “an alternation in power under electoral rules identical to the ones that brought the incumbent to office” (69).Footnote 4
The next three variables, Military Power, US Military Aid Share, and US Arms Transfer Share, account for political economy factors in transnational human rights reporting. Military Power considers Ron, Ramos, and Rodgers’ (Reference Ron, Ramos and Rodgers2005, 563, 571) claim that Amnesty International covers powerful states more heavily. Given the military's importance as a repressive state apparatus, I disaggregate the sources of state power and focus on its military dimension.Footnote 5 Specifically, this variable measures a state's share of the world system total of military personnel in a given year, using the Correlates of War (COW) Project's National Material Capabilities 5.0 data set (Singer, Bremer, and Stuckey Reference Singer, Bremer, Stuckey and Russett1972) and following the COW Project's own operationalization of national material capabilities (Greig and Enterline Reference Greig and Enterline2017, 7–8).Footnote 6 US Military Aid Share controls for Ron, Ramos, and Rodgers’ (Reference Ron, Ramos and Rodgers2005, 563–564, 571–572) assertion that US military aid recipients are more likely to be targeted by Amnesty International's human rights reporting. It measures a state's percentage share of total US military aid in a given year, using the United States Agency for International Development's (2016) US foreign aid data. US Arms Transfer Share takes into account Hendrix and Wong's (Reference Hendrix and Wong2014) argument that states with security linkages to the US are prone to be under a greater extent of transnational human rights reporting. It computes a state's percentage share of total US arms export in a given year, based on the Stockholm International Peace Research Institute's (2016) SIPRI Arms Transfer Database. I measure Asian states’ receipt of US military aid and US arms transfer in relative share rather than in absolute amount for two reasons. First, because a substantial number of country-year observations receive no military aid or arms transfer from the US at all, using the logged absolute amount will cause the missing data problem and confound statistical inference. Also, adding a small arbitrary constant (say, 0.001 dollar) to all observations is problematic. Second, relative share is likely to remain more stable and less vulnerable to short-term disturbances (for example, US domestic politics) than absolute amount.
Civil War controls for an additional possible source of human rights violations. It equals 1 if an internal or internationalized internal armed conflict occurs in a state in a given year and 0 otherwise, based on the Uppsala Conflict Data Program/Peace Research Institute Oslo (UCDP/PRIO) Armed Conflict Dataset v.4–2015 (Gleditsch et al. Reference Gleditsch, Wallensteen, Eriksson, Sollenberg and Strand2002).
Finally, ICCPR Ratification accounts for Simmons’ (2009, 88–96) assertion that human rights INGOs may consider governments’ human rights treaty ratification in selecting a target for transnational reporting. ICCPR Ratification is coded 1 if a state is a party to the International Covenant on Civil and Political Rights in a given year and 0 otherwise. Table 2 reports the hypotheses and summary statistics for all the independent variables.
Hypotheses and Summary Statistics
Collinearity Diagnostics
This section diagnoses the degree of the correlation among the independent variables in this article's main statistical model to assess whether multicollinearity (also known as collinearity) may pose a problem to my statistical inference. This article's multivariate statistical analysis enters Human Rights INGO Ties and all control variables simultaneously into the same equation and estimates their relative importance for the dependent variable. If two or more independent variables are highly correlated with one another (that is, not independent enough from one another), multicollinearity becomes a problem because variables’ individual effects cannot be reliably separated out, and the statistical estimates become unstable. For instance, given that democracy is known to create a favorable environment for civic associationalism, one may wonder how much Human Rights INGO Ties and Democracy are correlated with each other and whether their relative relationships with Amnesty International's special country reports can be reliably estimated. However, if and when researchers work with observational—not experimental—data, multicollinearity, indeed, becomes a matter of degrees because certain correlation among independent variables is inevitable in observational data, whose data-generating process usually is not under researchers’ full control. In view of this, collinearity diagnostics verify whether multicollinearity in my data is acceptably low enough for valid statistical inference.
Table 3 presents five different diagnostic tests of multicollinearity for all the independent variables in my main statistical model: namely, the Variance Inflation Factor (VIF), the tolerance, R2, the condition index, and the condition number. As a general rule of interpretation, multicollinearity becomes a problem if the VIF is over 2.50; the tolerance is below 0.40; R2 exceeds 0.60; the condition index is greater than 10; or the condition number is 15 or more (Williams Reference Williams2015). In essence, all the independent variables pass all five collinearity diagnostics, suggesting that multicollinearity is not a problem in my analysis and that one can have confidence in the results of my statistical inference.
Testing Multicollinearity: Collinearity Diagnostics
Note: The VIF denotes the Variance Inflation Factor. As a general rule of interpretation, multicollinearity becomes a problem if the VIF is over 2.50; the tolerance is below 0.40; R2 exceeds 0.60; the condition index is greater than 10; or the condition number is 15 or more (Williams Reference Williams2015).
RESULTS AND DISCUSSION
The statistical results offer strong support for my main hypothesis on the positive relationships between human rights INGOs’ local engagement and the extent to which the governments of Asian states become a target of transnational human rights reporting. Table 4 presents this article's main statistical model while robustness checks are discussed in full detail in the Appendix and the Supplementary Material. In the first column on statistical significance, a positive coefficient means that the independent variable increases the expected number of Amnesty International's special country reports on an Asian state. Conversely, a negative coefficient indicates that the independent variable decreases the number of Amnesty International's reports. The greater number of asterisks indicates that the independent variable is the more systematic part of the underlying political process for Amnesty International's human rights reporting. The second column on substantive significance shows the effect size of each independent variable, that is, whether the relationship between each variable and the dependent variable is large enough to be meaningful.
Determinants of the Extent of Transnational Human Rights Reporting in Asia
Note: The first column reports the sign and statistical significance of the independent variables. Numbers in parentheses are robust standard errors clustered on state. All independent variables use a one-year lag. The second column presents their substantive significance. Changes in the baseline predicted number of Amnesty International's special country reports are computed by shifting one independent variable at a time while holding all the others constant at mean level and modal category, specifically, by increasing a continuous variable from its mean value by one standard deviation and a categorical one from 0 to 1. *** p ≤ .01; ** p ≤ .05; * p ≤ .10, in two-tailed tests.
In the first column of Table 4, Human Rights INGO Ties is positively associated with the expected number of Amnesty International's special country reports. Furthermore, it is highly statistically significant. Thus, controlling for government protection of human rights, regime type, military power, security linkages to the US, and other factors, human rights INGOs’ engagement in a state through a local membership base significantly and positively relates to the extent to which the government of that state becomes the target of transnational human rights reporting across East and Southeast Asia.
To illustrate the substantive impact of human rights INGOs’ local engagement, the second column of Table 4 presents the predicted numbers of Amnesty International's special country reports, based on the coefficient estimates in the first column. The baseline prediction is the number of Amnesty International's reports for the hypothetical average Asian state, for which all the continuous and categorical variables are held constant at their mean value and modal category. The first line of the second column shows the shift in the baseline prediction when Human Rights INGO Ties increases by one standard deviation from its mean value (that is, from about 40 to 120 human rights INGOs) in a given year. The result is strong. The predicted count of Amnesty International's special country reports targeting that state increases by 148.3 percent in the following year. Thus, human rights INGOs’ local engagement has significant and strong positive relationships with the extent to which national governments become under transnational scrutiny and pressure on their domestic human rights practices across East and Southeast Asia, even while controlling for government protection of human rights, regime type, military power, and other factors.
Many of the control variables are of the expected sign, but not all of them are statistically significant. Both Human Rights Protection and Democracy are of the expected negative sign and highly statistically significant. These findings support Keck and Sikkink's (Reference Keck and Sikkink1998) claim that repressive states and dictatorships are more likely to be subject to transnational reporting. Military Power and US Arms Transfer Share are of the expected positive sign and highly statistically significant. These results provide support for the argument that powerful states and US allies are more prone to get covered by Amnesty International's reports (Hendrix and Wong Reference Hendrix and Wong2014; Ron, Ramos, and Rodgers Reference Ron, Ramos and Rodgers2005). US Military Aid Share, Civil War, and ICCPR Ratification all are positive but insignificant. The lack of their statistical significance indicates that they have little direct impact on determining which government in Asia gets targeted by transnational human rights reporting.
Robustness Checks
While I find firm empirical evidence for my original statistical model, I take several further steps to ensure that the positive relationships between human rights INGOs and Amnesty International's human rights reporting are robust against various confounding factors. My robustness checks are reported in full detail in this article's Appendix and Supplementary Material.
First, I consider the possibility that common background factors (for example, regime type) may influence both human rights INGOs’ location decisions and Amnesty International's target selection. I address this so-called selection bias by using the two-stage estimation method and purging out such confounding effects from the statistical association between human rights INGOs’ local ties and Amnesty International's coverage, and the result remains robust. (See the Appendix.) Second, because it takes time for human rights INGOs’ local engagement to influence Amnesty International's reporting, I ensure to get their temporal ordering correct by using a one-year lag for all independent variables in this article. Additionally, I employ a five- and ten-year lag for Human Rights INGO Ties, and the result is unchanged. (See Table OA1 in the Supplementary Material for robustness checks against reverse causation.)
Third, I take into account the possibility that unobserved or unmeasurable background factors (such as historical legacy or the Asian values) may confound the relationships between human rights INGOs’ local ties and Amnesty International's reporting. I address this problem of unobserved country-level heterogeneity by employing fixed-effects models, and the finding is robust. (See Table OA2 in the Supplementary Material.) Finally, I estimate a number of additional statistical models that, for instance, include other possible determinants of transnational human rights reporting, such as economic development and globalization, or employ different measures of democracy. This is to demonstrate that my key finding about the role of human rights INGOs’ local engagement is not an artifact of particular model-specification choices in my original model. (See Tables OA3 and OA4 in the Supplementary Material for robustness checks against control variable bias, omitted variable bias, and alternative operationalization of control variables.)
CONCLUSION
This article has argued that human rights INGOs’ local engagement has been key to determining cross-national variations in the extent of transnational human rights reporting in East and Southeast Asia. Specifically, human rights INGOs increase social demands and opportunities for transnational reporting by expanding local members’ capabilities to leverage human rights and international solidarity as an advocacy strategy and by mobilizing local members and volunteers for monitoring and information collection on the ground. The statistical analysis provides robust evidence that human rights INGOs’ local engagement has systematic positive relationships with the extent to which the governments of Asian states receive transnational scrutiny and pressure on their domestic human rights practices. This is strong support for my theory because the statistical tests explicitly control for government protection of human rights, regime type, military power, and other factors and demonstrate the robustness of the key statistical result for the role of human rights INGOs’ local engagement.
Why should one care? First, transnational human rights reporting has urgent and utmost importance in Asia. For local human rights activists and victims, transnational scrutiny and pressure by human rights INGOs like Amnesty International can be an important way to externalize their social grievances and claims and to pressure their governments to uphold human rights. In contrast, for rights-violating governments and their supporters in society, human rights INGOs’ effort to monitor, publicize, and condemn domestic human rights practices can be an undue interference in the internal affairs of sovereign states and even a form of Western imperialism. Thus, transnational human rights reporting has been and continues to be so controversial across Asia, and it should be of interest to scholars of East Asian Studies.
Second, this article offers a much-needed theoretical framework for explaining when and why national governments in East and Southeast Asia receive transnational scrutiny and pressure on their domestic human rights practices. By explicitly theorizing the role of human rights INGOs’ local engagement in transnational reporting across Asia, this article demonstrates the crucial importance of considering transnational–domestic linkages in the analysis of Asian politics. The human rights INGO mechanism that this article theorizes may play a role for other human rights phenomena in Asia. Future research should examine my theory's generalizability to other cases of Asia's human rights politics, such as government ratification of, and (non-)compliance with, international human rights treaties or de facto and de jure abolition of the death penalty.
Last but not least, this article suggests that well-intentioned human rights INGOs may have a perverse effect, producing the uneven geography of transnational human rights reporting that is often detached from local victims’ grievances and needs around the world. As my research has demonstrated, transnational human rights reporting is a function of human rights INGOs’ ties to local activists and victims, as much as governments’ human rights infractions. The paradox of this finding is that states with good human rights records may actually receive more transnational reporting coverage than they deserve, while highly repressive states attract far less attention and help from human rights INGOs.Footnote 7 As such, selectivity in transnational human rights reporting goes to the heart of questions about the legitimacy of human rights INGOs’ claims to represent human rights principles and victims’ needs and the effectiveness of transnational reporting as a tool for social change.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/jea.2018.6.
APPENDIX
The Appendix presents the procedures and results of the two-stage NB GLM to demonstrate that the key finding about Human Rights INGO Ties is robust against endogeneity bias caused by the possible selection bias that human rights INGOs may be non-random in building their local membership base across Asian states. Here, the source of endogeneity bias is cross-sectional, not temporal as in the case of endogeneity bias arising from reverse causation. That is, common background factors in a given state (for example, regime type) may influence both human rights INGOs’ local engagement in that state and Amnesty International's coverage of that state, which in turn confounds the estimated relationships between Human Rights INGO Ties and the dependent variable. I solve this potential endogeneity problem by using the “control function” or two-stage residual inclusion approach (Terza, Basu, and Rathouz Reference Terza, Basu and Rathouz2008). I choose this method since in nonlinear models (like this article's NB GLM) the conventional two-stage predictor substitution approach is inconsistent and biased whereas the two-stage residual inclusion approach is not (Terza, Bradford, and Dismuke Reference Terza, Bradford and Dismuke2008).
Specifically, in the first-stage equation, I regress Human Rights INGO Ties on both instrumental variables and all of the second-stage equation's control variables (that is, exogenous predictors), using the GLM equivalent to the ordinary least squares regression model, that is, the GLM with the Gaussian probability distribution and the identity link function (Baltagi Reference Baltagi2011, 265; Terza, Basu, and Rathouz Reference Terza, Basu and Rathouz2008; Wiggins Reference Wiggins2013). Based on my work in progress, I select three instruments that influence human rights INGOs’ local engagement but have no clear and direct link to Amnesty International's issuance of special country reports: namely, Urbanization, Judicial Independence, and Distance from the US. Urbanization considers that in the Third World human rights activism has been mainly driven by the urban middle class (Odinkalu Reference Odinkalu2000; Schwarz Reference Schwarz2002). It measures the percentage of total population living in cities with population greater than 100,000 in a state in a given year, based on National Material Capabilities 5.0 data set (Singer, Bremer, and Stuckey Reference Singer, Bremer, Stuckey and Russett1972). Judicial Independence taps into the claim that the rule of law environment critically shapes citizen participation in human rights activism (for example, Stork Reference Stork, Beinin and Vairel2011). It is a continuous variable that measures a state's latent judicial independence in a year on 0 (the least independent) to 1 (the most independent) scale, using Linzer and Staton's (Reference Linzer and Staton2015) data. Distance from the US accounts for the argument that US human rights politics emerging since the mid-1970s has had a ripple effect on the growth of human rights activism in the Global South (Dezalay and Garth Reference Dezalay and Garth2006). It computes the natural log of the minimum distance in kilometers from Washington, DC to each Asian state's capital city, using Gleditsch and Ward's (Reference Gleditsch and Ward2001) data.
These three variables meet the criteria of both instrumental relevance and instrumental exogeneity that are required by the two-stage estimation method. First, Urbanization, Judicial Independence, and Distance from the US explain the large variation in Human Rights INGO Ties. When I regress the latter on the three instruments only, R2 as the goodness-of-fit statistic is 0.33, well above the common minimum threshold of 0.10. Also, when Human Rights INGO Ties is regressed on both the three instrumental variables and all of the second-stage equation's control variables in the first-stage equation, the F-test statistic for the joint significance of the three instruments is 306.61. Worth noting, Baltagi (Reference Baltagi2011, 267) emphasizes that the first-stage F-test statistic should be at least 10 to be acceptable for a two-stage least squares model where the second-stage dependent variable is linear. While this threshold may not readily extend to my nonlinear two-stage model, it can serve as a guide. Thus, Urbanization, Judicial Independence, and Distance from the US are highly relevant strong instruments for Human Rights INGO Ties.
Second, none of Urbanization, Judicial Independence, and Distance from the US appears to significantly relate to Amnesty International's special country reports. It should be noted that the existing diagnostic tests of instrumental exogeneity (such as the Sargan-Hansen test) are only available where the second-stage dependent variable's error process follows the normal distribution and/or the two-stage predictor substitution approach is employed. To the best of my knowledge, there is no readily available, equivalent diagnostic test for the two-stage residual inclusion approach to limited dependent variables (like my two-stage NB GLM). For this reason, I estimate four NB GLMs that regress my second-stage dependent variable (that is, the count of Amnesty International's reports) on the three instrumental variables individually and then jointly. As Table A1 shows, none of them is statistically significant in any models, suggesting that Urbanization, Judicial Independence, and Distance from the US have no direct relationship with Amnesty International's human rights reporting. The fact that all of the second-stage equation's control variables are properly included in the first-stage equation (Baltagi Reference Baltagi2011, 265; Wiggins Reference Wiggins2013) minimizes the risk that Urbanization, Judicial Independence, and Distance from the US have an indirect relationship with Amnesty International's reports by being correlated with an omitted predictor of those reports. Thus, the part of variation in Human Rights INGO Ties captured by the three instrumental variables can be regarded as exogenous.
Testing Instrumental Exogeneity
In the second-stage equation, I estimate the NB GLM that contains Human Rights INGO Ties, all the control variables, and First-Stage Residuals (that is, the residuals from the first-stage equation). As Terza, Basu, and Rathouz (Reference Terza, Basu and Rathouz2008) emphasized, the inclusion of First-Stage Residuals controls for endogeneity caused by unobserved confounders. The second-stage equation uses the Murphy-Topel corrected standard errors for two-stage models (Murphy and Topel Reference Murphy and Topel1985). Table A2 presents the results of my two-stage NB GLM. In essence, First-Stage Residuals testing the endogeneity of human rights INGOs’ local engagement is never statistically significant. This lack of statistical significance indicates that Human Rights INGO Ties is not endogenous and that my original main model in the article's Table 4 is preferable to the two-stage NB GLM on the ground of statistical efficiency. Furthermore, the main result for Human Rights INGO Ties remains unchanged. In conclusion, my key finding about the role of human rights INGOs’ local engagement in transnational human rights reporting is robust against selection bias.
Robustness Check against Selection Bias: Determinants of the Extent of Transnational Human Rights Reporting in Asia
