Migration and the Demand for Transnational Justice

Domestic courts sometimes prosecute foreign nationals for severe crimes—like crimes against humanity, genocide, torture, and war crimes—committed on foreign territory against foreign nationals. We argue that migrants can serve as agents of transnational justice. When migrants move across borders, as both economic migrants and refugees, they often pressure local governments to conduct criminal investigations and trials for crimes that occurred in their sending state. We also examine the effect of explanatory variables that have been identified by prior scholars, including the magnitude of atrocities in the sending state, the responsiveness of the receiving state to political pressure, and the various economic and political costs of prosecutions. We test our argument using the first multivariate statistical analysis of universal jurisdiction cases, focusing on multiple stages of prosecutions. We conclude that transnational justice is a justice remittance in which migrants provide accountability and remedies for crimes in their sending states.


INTRODUCTION
I n 1998, an unprecedented event occurred: the UK House of Lords debated whether to extradite Augusto Pinochet from the UK to face criminal trial in Spain for torture committed in Chile. As Chile's leader during [1973][1974][1975][1976][1977][1978][1979][1980][1981][1982][1983][1984][1985][1986][1987][1988][1989][1990], Pinochet oversaw killings, torture, and disappearances of political opponents by Chile's military and secret police. Some observers viewed the Spanish case as a triumph of international justice (Roht-Arriaza 2005). Other observers described it as "judicial tyranny" that threatened Chile's political autonomy (Kissinger 2001, 86). Both sides recognized that Spain was administering justice in a way that Chile could not or would not do. And both sides recognized a broader trend: the spread of universal jurisdiction, which occurs when a state uses its domestic law and courts to regulate behavior that occurs outside of its domestic territory, does not involve its nationals, and does not have systematic or important effects on its national interests.
The Pinochet trial is not an isolated example. Beginning in the mid-1990s, Spanish courts investigated serious international crimes committed in many foreign states, including Argentina and Cuba. These states had high political repression in the 1970s and 1980s. Yet Spain did not investigate leaders from other states with similar repression during the 1970s and 1980s, including Bolivia, Honduras, and Nicaragua. How can we explain this variation? What makes states more or less likely to assert universal jurisdiction, punishing international crimes using domestic courts?
We argue that migrants-who move across state borders as economic migrants or refugees-serve as agents for transnational justice. For Spain, more migrants in the 1990s came from Argentina, Chile, and Cuba than from Bolivia, Honduras, and Nicaragua. More generally, when individuals migrate from repressive societies, they often have grievances that were not adequately remedied in their sending state. These grievances can motivate migrants to seek transnational justice when they arrive in a receiving state. 1 Migrants have numerous tactics for achieving transnational justice including reporting crimes to authorities, providing evidence and witness testimony, increasing public awareness of atrocities, and lobbying government elites to take action. In some states, they can even launch private prosecutions. However, these demands will only yield transnational justice if enough migrants demand that the receiving state take action. Our main theoretical claim is that higher migrant stocks from a sending state increase the likelihood of criminal cases by the receiving state for acts that occurred in the sending state. We additionally argue that universal jurisdiction cases are more likely when there are more severe atrocities in the sending state; the government of the receiving state is more responsive to political demands; and the economic, legal, and political costs of prosecutions are lower. In sum, transnational justice is a justice remittance, in which migrants provide accountability and remedies for crimes in their sending states.
We test our theory using, to our knowledge, the first multivariate statistical analysis of universal jurisdiction cases. Our measures include the initiation of a case, investigation by state entities, formal proceedings, issuance of a warrant or arrest, and whether a trial was held. Each of these outcomes is subject to measurement error and bias, but they collectively capture universal jurisdiction cases. As our main explanatory variable, we use multiple measures of migrant stocks from the sending state in the receiving state. Additionally, we use several measures of atrocities in the sending state, responsiveness in the receiving state, and prosecution costs. Our statistical analysis provides compelling support for our theoretical argument and highlights the importance of migrants as protagonists with agency.

FROM TRANSNATIONAL COMMUNITY TO TRANSNATIONAL JUSTICE
The most serious violations of international human rights and humanitarian law-including crimes against humanity, genocide, torture, and war crimes-are called international crimes. At least since 1945, experts have argued that leaders can have individual criminal responsibility for such crimes, meaning that they can face criminal prosecutions.

International and Domestic Justice
States and advocacy groups uphold international human rights and humanitarian law using diverse tools. At the international level, states frequently use naming and shaming to publicize legal violations and generate pressure for policy changes (Hafner-Burton 2008;Terman and Voeten 2018). Numerous international bodies also allow individuals to file complaints against states that have allegedly violated their rights (Johns 2019). Previous scholars have highlighted how economic and political relationships-like military alliances-affect whether states name and shame each other (Terman and Voeten 2018). Additionally, the effectiveness of these tools in improving human rights practices depends on a state's economic and political capacity to uphold human rights and humanitarian rules at the domestic level (Hafner-Burton 2008). However, none of these tools assigns individual criminal responsibility by prosecuting political and military leaders.
At the international level, some international criminal tribunals have prosecuted individuals for international crimes. Many of these tribunals are ad hoc institutions that were created to address crimes committed at specific times and places, like the International Criminal Tribunal for Rwanda. The international community is more likely to create tribunals for more severe human rights and humanitarian violations (Rudolph 2001), yet these tribunals are often hindered by powerful states (Bass 2000). The International Criminal Court (ICC) has broader jurisdiction to prosecute individuals connected to its member states. Some scholars argue that states join the ICC to credibly tie their own hands and prevent atrocities, whereas others argue that states only join the ICC when it imposes minimal costs (Chapman and Chaudoin 2013;Simmons and Danner 2010). Regardless of this debate, subsequent cooperation with the ICC appears to be affected by a state's level of democracy (Kelley 2007).
Finally, domestic courts often prosecute international crimes. Since World War II, many states have drafted domestic criminal laws that allow the prosecution of international crimes (Berlin 2020). This trend was amplified by the Rome Statute, which created the ICC and provided detailed definitions of international crimes and modes of responsibility. All members of the Rome Statute are required to implement these rules into their domestic laws to enable domestic prosecutions, but states vary in whether and how they write such laws. 2 Domestic prosecutions often occur after political transitions, when new democracies punish atrocities committed by prior autocratic governments (Teitel 2000). Therefore, democracy therefore plays a primary role in punishing international crimes at the domestic level. Human rights nongovernmental organizations (NGOs) also shape transitional justice institutions (Zvobgo 2020), with some judicial systems allowing individuals and/or NGOs to initiate private prosecutions against alleged criminals (Brinks 2008;Michel and Sikkink 2013). However, transitional justice is fundamentally an internal process, considering only crimes that occurred within a state's own territory.
In contrast, universal jurisdiction cases often appear to involve powerful states asserting authority over economically and politically weaker states (O'Sullivan 2017). Many governments argue that universal jurisdiction is biased against African officials and reflects neoimperialism by powerful states over their former colonies (Geneuss 2009;Jalloh 2010;Mennecke 2017). Domestic political power can also affect universal jurisdiction cases. For example, Langer (2011) argues that the political branches of government employ cost-benefit analysis when deciding whether to prosecute. Greater executive power over judicial proceedings can increase the influence of cost-benefit analysis. This argument is supported by evidence that only low-cost defendants were brought to trial and that legislatures are more likely to amend universal jurisdiction statutes when the costs of universal jurisdiction formal proceedings and trials outweigh their political benefits.
Finally, universal jurisdiction prosecutions should be affected by government preferences, as reflected in international legal commitments. Many scholars argue that regime type affects whether states join and comply with human rights treaties (Hathaway 2007;Simmons 2009). Similarly, support for international criminal tribunals may indicate that a government supports transnational justice.
These trends in international and domestic justice have been enabled by elite transnational actors. A large literature documents how human rights advocates spread norms across borders (Keck and Sikkink 1998;Kim 2013;Simmons 2009). Similar dynamics are apparent in human rights enforcement, with transnational advocates pressuring new democracies to punish prior legal violations (Lutz and Sikkink 2001;Sikkink 2011). Finally, many legal scholars argue that transnational communities of bureaucrats and judges can transmit new ideas into domestic legal systems (Slaughter 2000(Slaughter , 2003. However, these scholars have largely overlooked an important community of nonelite transnational actors: migrants.

Migrants and Diasporas
Many social scientists have studied transnational communities, which are groups of individuals that participate in activities across state borders. When individuals migrate across borders, they can create long-term links between states. These links can consist of economic, sociocultural, and political activities enabled by modern technology (Portes, Guarnizo, and Landolt 1999).
Throughout history, ethnic, national, and religious communities aided development by trading goods across borders (Grief 1993;Pirenne 1970). In the modern economy, migrants often cross borders to gain and provide access to capital, consumers, suppliers, and income (Leblang 2010). Social networks based on ethnicity, nationality, and religion can enable such trustbased activities (Larson 2021). In many developing states, migrants are an important source of economic remittances.
Migrants also spread sociocultural ideas and practices across borders. For example, sociologists have documented how migrants frequently encounter new ideas and practices about childrearing, education, and gender roles when they arrive in a receiving state (Levitt 2001, 73-124). These ideas and practices are communicated back to family and friends in the sending state, creating social remittances (Levitt 2016, 225).
Finally, migrants are often political actors in both the receiving and sending states. Migrants are involved in civil society, unions, and, where eligible, vote in the elections in receiving states; although, ethnicity, nationality, and race affect levels of participation (Greer 2013;Pantoja, Ramirez, and Segura 2001). Migrants increasingly vote, contribute to political parties, and even run for office in their sending state as well (Guarnizo, Portes, and Haller 2003;Levitt and de la Dehesa 2003;Waldinger 2015;Wellman 2021). Many sending and receiving states encourage such transnational political activism using dual citizenship laws, which allow migrants to participate in political life in multiple states (Jones-Correa 2001;Leblang 2017). Additionally, migration appears to influence foreign aid flows, suggesting that migrants influence political outcomes in both receiving and sending states (Bermeo and Leblang 2015). This influence can continue for future generations who claim membership in a diaspora, even if they are not migrants themselves (Shain 1994(Shain -1995. In short, migrants create political remittances back to their sending states (Faist 2008;Piper 2009).
Our argument and evidence suggest that migrants play a complementary role that has been largely overlooked by social scientists: migrants serve as agents for justice. Economic, sociocultural, and political ties can allow migrants to use domestic courts in the receiving state to punish severe international crimes that occurred in their sending state. Simply put, universal jurisdiction cases serve as justice remittances because they allow migrants to use receiving state institutions to remedy international crimes committed in the sending state.

Universal Jurisdiction as Transnational Justice
When states create and enforce domestic laws, they must establish their jurisdiction over the behavior that they seek to regulate. States usually assert jurisdiction based on territory, nationality, or systematic and important effects on their national interests (Johns 2022). However, sometimes a state will assert universal jurisdiction and consider cases that do not have a tangible link between the regulated behavior and the enforcing state when the alleged crime was committed. Under international law, states may assert such universal jurisdiction to prosecute serious international crimes, including crimes against humanity, genocide, torture, and war crimes (Langer 2015a). Universal jurisdiction is included in some international treaties, like the 1984 UN Convention Against Torture. States asserting universal jurisdiction may also rely on customary international law, which is formed by the combination of state practice and acceptance of law (Langer 2015a).
Universal jurisdiction is fundamentally a unilateral, domestic act (Langer 2013). States that assert universal jurisdiction sometimes ask other states for assistance (in arresting defendants, collecting evidence, etc.) and may inquire about whether the territorial state has exercised jurisdiction. But they do not ask for permission to prosecute. Therefore, universal jurisdiction differs from international prosecutions, which are based on explicit international cooperation. For example, international tribunals for the former Yugoslavia and Rwanda were both created by the UN Security Council, whereas the ICC operates based on the Rome Statute, a treaty with broad membership.
States that prosecute universal jurisdiction cases vary dramatically in their domestic laws (Langer 2004). These differences make it extremely challenging to compare criminal cases across different states. For example, formal investigations in common law jurisdictions usually start with an arrest, a grand jury indictment, or an information issued after a preliminary hearing. In contrast, in civil law jurisdictions, investigating judges or prosecutors often begin formal investigations before anyone is arrested or indicted (Langer 2004). In the US, the police may informally investigate a case for long periods provided that they do not arrest an individual or the person is not indicted, whereas in other states, like Argentina, such prolonged informal police investigations are not allowed.
We overcome this difficulty by using the new data that significantly expand on data published in Langer Leslie Johns, Máximo Langer, and Margaret E. Peters (2011) and subsequently updated in Langer and Eason (2019) and in 2020-2021. Our universal jurisdiction data contain information on every known criminal complaint (or case considered by public authorities on their own motion) that involved the alleged commission of one or more of the four core international crimes-crimes against humanity, genocide, torture, and war crimes-by physical individuals; was filed or initiated between 1957 and 2019; and fully or partially relied on universal jurisdiction. The data thus do not include information on civil lawsuits, which are another tool for enforcing human rights law (Johns 2018 Our expanded data contain additional information about the multiple stages of criminal proceedings. First, we measured INITIATION, which is the year in which a universal jurisdiction case began. This requires a complaint filed by the alleged victim or on their behalf by an NGO or other groups or by state authorities by their own motion. Second, we coded INVESTIGATION, which is the year (if any) in which the receiving state took investigative measures or inquired about whether the territorial state (or other jurisdiction) investigated or prosecuted the case. Third, we coded FORMAL PROCEED-INGS, which is the year (if any) in which formal charges or proceedings were brought by a prosecutor or judge. Fourth, we measured ARRESTS, which is the year (if any) in which an arrest warrant was issued for the defendant by the receiving state or the defendant was arrested. Finally, we coded TRIAL, which is the year (if any) in which a criminal trial began.
Each of these variables comes with potential measurement error and bias. For example, INITIATION is a broad measure that includes all allegations of crimes against humanity, genocide, torture, and/or war crimes against an individual. This variable almost certainly underestimates migrant demands because we only observe criminal complaints that are documented by the sources above. Additionally, it does not distinguish between migrant demands for transnational justice and the receiving state's willingness to supply such justice because of variation across states in domestic criminal procedures. For example, this measure includes complaints that may be ignored or overlooked in some states. In contrast, we are extremely confident in our measurement of trials, which are well documented and publicized. Trials occur when there is both demand for and supply of transnational justice. However, TRIAL is an underinclusive measure of transnational justice because it does not include situations in which a state may genuinely want to assert universal jurisdiction but is constrained by limited forensic evidence, witness intimidation, and difficulty arresting defendants who are still mentally and physically competent to face trial. None of these measures perfectly capture demand and/or supply for transnational justice. But we believe that they collectively capture the practice of universal jurisdiction, which is our substantive interest.
One empirical question is whether universal jurisdiction cases are aberrations or common events. Figure 1 shows the number of universal jurisdiction cases plotted over time using our INITIATION measure. 3 Universal jurisdiction cases are rare: there are only 2,162 cases from 1957 to 2020. However, these cases are not isolated or unique. They vary tremendously over time.
After the first case in 1957, there were no cases until the early 1980s. Cases then occurred relatively infrequently until the early 1990s. Since then, there have been cases every year, and although these cases go in waves, they seem to be increasing on average.
A second empirical question addresses which states assert universal jurisdiction. Table 1 lists all prosecuting states by their total INITIATION count. It includes the year of first initiation, the year of first trial (if any), and number of TRIALS (if any). The top 20 states are mostly advanced industrial states. However, middle and lower income states (like Argentina, Senegal, South Africa, and Turkey) have received universal jurisdiction complaints or had cases initiated by officials. Among advanced economies, there are the Scandinavian states and Canada, which are often described as strong human rights advocates, and states like Australia, the UK, and the US, which tend to oppose interventionist international law. Prosecuting states also include many non-European states, including states from Africa, Asia, and Latin America. Although many prosecuting states are former colonial powers (such as France, Germany, and Spain), others are former colonies (such as Argentina, Canada, and Senegal). We also see that trials are much less common, but states that have more initiations tend to have more trials.
A final empirical question is whether universal jurisdiction is driven exclusively by Nazi prosecutions or used more broadly to punish serious crimes. Universal jurisdiction cases come from a variety of times and places. Many cases involve civil wars with large-scale human rights abuses, like the Syrian Civil War, the Rwandan genocide, and conflicts in the Democratic Republic of Congo. Cases also involve authoritarian repression. For example, Table 2 shows that Argentina has the fourth highest number of defendants stemming from its Dirty War, defendants from China have been the subject of several complaints, and complaints have been lodged against defendants complicit in right-wing violence in El Salvador. Finally, some defendants are nationals of powerful states that practice foreign intervention, like France, the UK, and the US.

THEORY: HOW MIGRANTS PROMOTE UNIVERSAL JURISDICTION
Our argument is driven by the strategic interactions of two groups. The first group is migrants who are fleeing atrocities and violence in the sending state. Migrants are diverse actors. Many migrants or their family members are themselves victims of atrocities. Others are individuals who witnessed atrocities and were harmed economically by violence. Finally, some may themselves be the perpetrators of atrocities. These perpetrators can include individuals who are remorseful for their past actions and individuals who are not. 4 For our argument, we assume that more migrants support transnational justice (e.g., victims, witnesses, and remorseful perpetrators) than oppose it.
The second group that is relevant to our argument is government officials of the receiving state.
Transnational justice usually involves diverse government officials in the receiving state. Within the justice system, transnational justice requires the cooperation of police officers, public prosecutors, and judges. Politicians also matter because transnational justice sometimes requires intervention by executives to oversee law enforcement and/or by legislators to provide resources or revise criminal codes. These officials can themselves vary in their resources and attitudes toward transnational issues. Although some government officials may support transnational justice on ideological grounds, others may oppose it.
For example, qualitative accounts of Spanish prosecutions highlight the diverse viewpoints of government officials involved in prosecuting international crimes in Argentina and Chile. Some judges-like Baltasar Garzón-were more inherently supportive of transnational prosecutions than other judges-like Manuel Garcia Castellón. Though these two Spanish judges were initially in charge of the Argentine and Chilean investigations, respectively, only Garzón issued an arrest warrant against Pinochet while he was still in London (Roht-Arriaza 2005, 33-5). Among politicians, the United Left party strongly favored transnational prosecutions, even becoming a private prosecutor in the Argentine and Chilean investigations. In contrast, right-wing Prime Minister José María Aznar was not inherently supportive of these investigations, although he had incentives against interfering because Spanish popular opinion strongly supported these prosecutions (Roht-Arriaza 2005, 10-6).
When migrants and government officials pursue transnational justice, they face numerous legal obstacles (Ferdinandusse 2006). Although some states have incorporated terms like "crimes against humanity" and "war crimes" into their domestic criminal law, other states have not. Additionally, universal jurisdiction cases often target high-level military and political officials. Under international criminal law, such individuals can be responsible for crimes using concepts like "superior responsibility." Yet not every domestic legal system recognizes this mode of criminal responsibility. Next, while some states have domestic legislation that explicitly allows universal jurisdiction (like Germany), other states do not (like Chad). Finally, even within states that do have universal jurisdiction legislation, many states place procedural or temporal limits on these cases. These various legal obstacles mean that migrants and their lawyers must devise complex strategies, like developing novel legal arguments, placing public pressure on government officials to use resources on "tough" cases, and even lobbying legislators to revise laws that limit universal jurisdiction.
After migrants arrive in a receiving state, they have several ways of mobilizing for transnational justice. The first (and most direct) path to transnational justice is to report crimes to police and prosecutors in the receiving state and assist in investigations (Langer 2015a). 5 Human rights scholars have extensively documented how limited state capacity can hinder human rights protections (Cole 2013;Lupu 2013). These same  arguments also apply to universal jurisdiction prosecutions. Migrants have specialized knowledge that is needed for a successful investigation, arrest, and trial, like knowledge of where and how crimes were committed and which individuals were responsible for these crimes. Additionally, migrants can use the various technologies-like airplane travel, international phone service, video teleconferencing, and television broadcast-that enable social remittances for new purposes (Levitt 2001, 23). Finally, the same cultural and social connections that build trust in transnational economic exchanges may persuade victims and witnesses to confront criminal perpetrators.
Returning to the example of Spain, Argentine and Chilean migrants helped define the legal strategy, provided evidence, and testified as witnesses. They also used their local knowledge, personal travel, and communication technologies to help Spanish authorities to collect testimony from individuals still living in Argentina and Chile (Roht-Arriaza 2005, 8-31). In particular, two Argentine migrants, economist Gregorio Adonis and labor lawyer Carlos Slepoy, proposed strategies for the investigation and helped collect evidence (Roht-Arriaza 2005, 8-10).
In some states, this process of reporting crimes and assisting in legal proceedings is an explicit part of the immigration process for new arrivals. In Germany, a Syrian perpetrator was tracked down by authorities through statements made by another Syrian immigrant in an asylum application and was convicted in 2020 for war crimes in Syria. 6 In Finland, two Iraqi twins were tried for war crimes, and ultimately acquitted, in a case that started based on information in their own asylum application (Langer and Eason 2019). In Norway, the police looked for 20 suspected war criminals from Syria, following tips from refugees and local immigration authorities. 7 Second, one human rights activist noted, "A major challenge in any universal jurisdiction case is creating the necessary political will in the forum state" (Brody 2017, 23). Migrants can help to generate such will by mobilizing public opinion in the receiving state. By informing the public about crimes that occurred elsewhere, migrants can gain valuable allies, particularly for interactions with politicians. Such tactics are likely to be most effective in democracies, where public attitudes are influenced by appeals to international law (Wallace 2013). As described above, migrants routinely participate in political activities in receiving states. These activities include using the news media and protests to highlight their grievances in receiving states (Shain 1994(Shain -1995. Many scholars have documented how the cultural, political, and social practices of migrants can be viewed as threatening in the sending state (Østergaard-Nielsen 2003, 8). Transnational justice offers migrants a unique opportunity to demand political change that is consistent with the receiving state's political institutions and values: by advocating for transnational justice, migrants and diasporas reinforce their public commitment to the institutions and values. 8 For example, during and after the Rwandan genocide of 1994, many Rwandan migrants-including victims, witnesses, and perpetrators-fled to Belgium. In July 1994, numerous Rwandan victims filed complaints with the Belgium police about crimes committed by perpetrators who had also migrated to Belgium. When Belgian authorities were reluctant to get involved, migrants and their lawyers organized protests and filed legal complaints against prominent Belgian politicians. These events were widely reported in Belgian newspapers, raising public awareness and creating pressure on the government to appoint an examining magistrate (Human Rights Watch 1999).
Third, migrants also frequently meet directly with elite officials-including judges and legislators-to persuade them to act. In democracies, migrants and diasporas can often successfully pressure politicians to change policies, particularly when these policies are unlikely to be observed and opposed by a majority of voters (Bishin 2009). Migrants frequently testify in bureaucratic and legislative hearings as experts on economic, legal, and political conditions in the sending state (Shain 1994(Shain -1995. Such lobbying and testimony can be particularly effective when it is based on personal narratives about individual experiences, which can only be provided by the victims and witnesses of international crimes (McEntire, Leiby, and Krain 2015).
For example, fearing that the public prosecutor would oppose the Argentine case in Spain, two people in exile in Spain, including the grandmother of a baby who had been kidnapped by the military, met with the Spanish chief prosecutor. Through their personal narratives, they persuaded him to remain agnostic about the legal merits of the case (Roht-Arriaza 2005, 14-5). Similarly, growing public concern over the Rwandan genocide led Belgium legislators to amend their domestic laws in 1999 to allow Belgian courts to have universal jurisdiction over genocide and crimes against humanity (Keller 2001). This legal change facilitated the Belgian trial and conviction of four Rwandan citizens-including two nuns, a government official, and a university professor-for complicity in the Rwandan genocide. 9 Finally, many states provide a fourth tactic for transnational justice: private prosecutions. This power is mostly unknown in the United States, where only public prosecutors can file charges and serve as the prosecuting party in criminal proceedings. However, many other states allow private actors-including the alleged victim of a crime, a private citizen, certain NGOs, or other organizations-to be a party in the criminal process together with or instead of public prosecutors (Langer and Sklansky 2017, 333). The powers of private prosecutors vary across states. But they may include the power to file criminal charges, sue for civil damages in the criminal case, plead before the court, present evidence, interrogate witnesses, appoint and present their own expert witnesses, appeal, and move the case toward trial. Some scholars have highlighted how private prosecutors push forward domestic human rights cases (Brinks 2008;Michel and Sikkink 2013) and transnational prosecutions of international crimes (Langer 2011). Others have underlined the link between private prosecution and migration (Mégret 2015).
Private prosecutions have been important in many transnational cases by giving migrants a way to file these cases and move them forward even when prosecutors and judges are reluctant to do so. For instance, the Argentine Human Rights Association of Madrid was a private prosecutor in Spain, ultimately leading to the arrest of August Pinochet (Roht-Arriaza 2005, 10). 10 In France, an NGO cofounded by a Rwandan immigrant and her French husband has filed multiple complaints and become a civil party in criminal proceedings against Rwandans living in France. 11 Which migrants will use these tools? Like any political activity, advocating for transnational justice is costly. Individuals who advocate are therefore likely to have more resources than individuals who do not. Although news broadcasts often emphasize the poverty and despair of refugees who flee violence, many migrants who flee atrocities are highly educated, wellinformed about the institutions and laws of their receiving state, and have economic resources (Holland and Peters 2020). Such individuals are often given work visas and are labeled as economic migrants, yet they can still be victims and witnesses of atrocities. Qualitative research shows that such migrants can become key actors in building transnational political networks (Østergaard Nielsen 2003).
Migrants are also more likely to engage in political activity if they arrive in the receiving states through legal processes that allow them to establish long-term residency. In contrast, migrants who fear deportation or only intend to live in the receiving state temporarily (as, for example, seasonal workers) are less likely to provide justice remittances. Finally, migrants are mostly likely to advocate for transnational justice when they have been directly affected by atrocities. Many, but not all, of these migrants may come into the state as Convention refugees or gain refugee status through asylum proceedings. There may be a larger effect of refugees because these individuals have more clearly been victims of international crimes. Further, as noted above, the state has information about the international crimes perpetrated against these individuals from the refugee status determination hearings. Nevertheless, we expect that the desire for justice will, on average, diminish over time as victims die and are replaced in the migrant stock measures by new individuals, who did not experience the atrocities themselves.
Migrants are often assisted in all of these activitiesreporting and assisting, mobilizing public opinion, persuading elite actors, and pursuing private prosecutions -by activists and nongovernmental organizations in the receiving states. For example, organizations like Amnesty International and Human Rights Watch provide important expertise and resources to migrants who seek justice. Yet the migrants themselves are key actors with agency and importance in their own right.
Government officials must respond to migrant actions by making their own decisions about whether to pursue a universal jurisdiction case. As shown in Figure 2, the first step is the initiation of a case. Cases often begin when an alleged victim files a complaint with local police in the receiving state. In some states, NGOs can bring cases on behalf of a victim, independent of support from prosecutors, and/or government officials have authority to initiate a case. The second step is an investigation by public authorities. Third, is the initiation of formal charges or proceedings by a prosecutor or judge. Fourth is the issuance and execution of an arrest warrant. The final step is a criminal trial. Each of these steps is a distinct and ordered outcome of a universal jurisdiction case. We believe that each of these steps cumulatively increases the overall amount of transnational justice that a prosecuting state provides to atrocity victims and the justice remittances that migrants may produce.
Four primary explanatory variables affect the outcomes in our theory. Our first-and most importantexplanatory variable is migrant stocks. This variable can have both direct and indirect effects on transnational justice. For the direct effect, as more migrants from a sending state arrive in the receiving state, more individuals are able and willing to advocate for transnational justice. As more migrants take these costly actions, government officials will receive more benefit from taking the various steps in a universal jurisdiction case. Migrant stocks can also have an indirect effect on government officials by reducing the cost of a universal jurisdiction case. 12 All else equal, more victims and witnesses can generate more criminal charges against more perpetrators and provide better and more compelling evidence for trials. Additionally, government officials will find it easier to secure and execute arrest warrants when more perpetrators are present on their territory.
Careful readers will notice that we have not made any specific assumptions about whether transnational justice provides private or public benefits for victims. We also have not explicitly described any strategic interactions among migrants about who exactly should invest their time and effort in securing transnational justice. Perhaps activism is hindered if successful cases provide public benefits, rather than private benefits. Although large groups of actors are hindered by collective action problems, they are also helped by expanded opportunities for individual action and an increase in the collective benefit from prosecutions (Johns 2019). To ensure that our account of transnational justice is not derailed by a collective action problem, we constructed numerous formal models of these strategic processes. 13 These formal models show that our explanatory variables clearly and consistently shape outcomes across many different assumptions about the nature of private versus public benefits. Overall, political pressure by migrants generates our first empirical hypothesis: Hypothesis 1 (H1): Higher migrant stocks from the sending state in the receiving state will yield more transnational justice.
Our second explanatory variable is the level of atrocities in the sending state. Once again, this variable can have both direct and indirect effects. When atrocities are larger in number and/or more severe, more migrants are likely to be more motivated to demand transnational justice. We also expect that more numerous and severe atrocities will make government officials more sympathetic to the demands of migrants. Activists and NGOs are well aware of this effect and often ask victims and witnesses to share their experiences privately with government officials to persuade them to act (Roht-Arriaza 2005). Atrocities can indirectly affect transnational justice by increasing migrant flows from the sending state, thereby reinforcing the effect in H1. 14 Overall, we expect that Hypothesis 2 (H2): A higher magnitude of atrocities in the sending state will yield more transnational justice.
Our third explanatory variable is the responsiveness of the government in the receiving state. This concept includes a government's views on human rights and internationalism, which affect its willingness to prosecute universal jurisdiction cases. It also includes a government's overall responsiveness to public pressure of any kind. Therefore, responsiveness can be measured by the government-specific measures of ideology and state-specific measures of public accountability. We expect that more responsive governments are more likely to provide justice, which should make migrants more likely to demand it. These reinforcing effects yield our third hypotheses: Hypothesis 3 (H3): A higher level of responsiveness in the receiving state will yield more transnational justice.
Finally, we expect that prosecution costs will influence government officials. Many different factors can influence this cost. Poorer states are likely to have less capacity and fewer resources to investigate and prosecute crimes. Similarly, prosecutors are likely to find it more difficult to pursue a state's first universal jurisdiction because such cases require them to form novel legal arguments before skeptical judges. We believe that migrants who anticipate this concern are likely to invest less time and effort to secure justice, thereby lowering the likelihood of justice even further. This yields our final hypothesis: Hypothesis 4 (H4): A higher prosecution cost in the receiving state will yield less transnational justice.
We expect that all four of our hypotheses will hold across all stages of a universal jurisdiction case. Yet we acknowledge that the magnitude and significance of these effects can vary at different stages in a legal proceeding. For example, migrants will probably have a greater influence during the initiation of case, when they have the most agency. We expect that migrant stocks will continue to be a positive force during subsequent stages of a case. That is, we don't expect that larger migrant stocks will ever decrease the likelihood that a case progresses. Nonetheless, it may be more difficult to observe these effects in statistical analyses for later stages of prosecutions, especially because these outcomes are rarer and thus generate more uncertainty in statistical analysis.

DATA AND METHODS: ANALYZING THE EFFECT OF MIGRANTS ON UNIVERSAL JURISDICTION
We now turn to our statistical tests of how migration, atrocities, responsiveness, and costs affect the likelihood that a state will take up a case under universal jurisdiction. We use a directed dyad model to understand when a state takes up a case involving defendants from another state. 15

Outcome Measures
Our outcome variable is universal jurisdiction prosecutions. These data include variables for both the receiving state (which pursued the prosecution) and the sending state (which is the home state of the defendant). Although we are able to identify the defendant's nationality for most cases (68%), some universal jurisdiction cases provide insufficient information to precisely identify the sending state. For example, many prosecutions for crimes under the Nazi regime involved defendants who came from states like Ukraine and Belarus rather than Germany. Similarly, many prosecutions against ISIS members potentially include individuals who were born outside of Syria and Iraq. Finally, for crimes during the dissolution of Yugoslavia, it is often difficult to identify the defendant's nationality. We deal with this challenge in two ways. First, in our most expansive dataset, we assign these cases to the major perpetrator state, thus assigning Nazi cases to Germany, Yugoslav cases to Serbia, and ISIS cases to Syria. We call this the "All" dataset in our tables. Second, in our most conservative dataset, we drop all instances in which we do not know the subject nationality. We call this the "No Doe" dataset. All of our models are estimated separately on both datasets to ensure that our results are robust.
Our main analysis uses the dyadic measure of cases, which measures how many universal jurisdiction complaints a receiving state considered against a defendant from the sending state in a given year. All directed dyads in the world system are included in this analysis. Each state that is a member of the United Nations or has a population of at least 500,000 and receives diplomatic missions from two states is included in the dataset. Directed dyads consist of each pair of states as both the receiving state and the sending state; for example, Spain (receiving state)-Argentina (sending state) 1993 and Argentina (receiving state)-Spain (sending state) 1993 are included as two separate observations. Cases only appear in the data for the year in which they are initiated. We use this variable in two different ways. First, the variable ANY INITIATION measures whether the receiving state considered any complaint against defendants from the sending state in a given year. This measure takes a 1 if there was an initiation of one or more cases by a receiving state against a defendant in a given year and 0 if no cases were brought.
Second, the variable FIRST INITIATION examines the first time the receiving state considers a complaint under universal jurisdiction. This variable takes a 1 in the first year that the receiving state initiates a case against a sending state. We begin by using the same dyadic data that we use to construct the ANY INITIATION variable. However, once states have initiated their first case, they cannot have another "first case," so they are dropped from the dataset. For example, Austria initiated its first case in 1994. For all years prior to and including 1994, all dyads between Austria and all potential sending states are included. All dyads with Austria as the receiving state are dropped from 1995 onward. This data strategy allows us to examine only the first instance of a universal jurisdiction initiation. This test helps relieve some concerns about reverse causation: given that this is the first initiation, migrants are not moving because the receiving state has heard past universal jurisdiction cases.
We also examine later stages of the process-INVES-TIGATION, FORMAL PROCEEDINGS, ARREST, and TRIALas described above. For each of these stages, we examine whether there are ANY of these escalations in the dyad-year and, similar to above, the FIRST (STEP), which is the first time that the receiving state takes one of these actions.

Explanatory Variables
Our main explanatory variable is MIGRANT STOCK, which is a dyadic measure of the total number of individuals from the sending state that are residing in the receiving state for 1960 through 2017. 16 The data provide the migrant stock for each pair of states every 10 years from 1960 to 2010 and for 2013 and 2017. Because these data are reported at three-year, fouryear, and ten-year intervals, we linearly interpolate the data. 17 The data come from each state's census data, and include temporary and permanent migrants, those who came as voluntary and forced migrants, and those who came through regular and irregular channels. We expect by H1 that larger MIGRANT STOCK will correspond to an increase in prosecutions.
Our second main explanatory concept is atrocities in the sending state, which generate demand for prosecutions. We measure this in two ways. First, we include the variable PTS (SS), the political terror score for the sending state (Gibney et al. 2020). The score ranges from 1 (states under secure rule of law) to 5 (states where leaders routinely murder, disappear, and torture the general population). Due to data coverage, this variable is only available starting in 1976; our regressions thus cover 1976-2017. This covers most of the universal jurisdiction cases; our results are robust to dropping this variable.
Second, we include a measure of the sending state's regime type. Because the PTS variable is based on State Department reports, there may be undercounting of human rights violations, as states may try to hide these violations. We think that human rights violations are both more likely and less likely to be reported in autocracies. We use the coding of regime type from Polity (Marshall, Gurr, and Jaggers 2016). The variable DEMOCRACY (SS) has higher values when the sending state is more democratic. We expect that higher values of this variable will correspond to fewer atrocities in the sending state and that higher DEMOCRACY (SS) will decrease the likelihood of a prosecution.
We expect there to be a delay between the period in which an atrocity occurs and the actual filing of a universal jurisdiction case. In addition, we want to account for conflicts and human rights abuses that may go on over a longer period. In our main analysis, we use the average value of both of these variables over the last 10 years.
Our third explanatory concept is government responsiveness in the receiving state, which affects the extent to which government preferences are affected by public pressure from migrants. We examine two types of responsiveness: political responsiveness and legal responsiveness. Our first measure of political responsiveness is the recipient state's regime type using Polity (Marshall, Gurr, and Jaggers 2016). The variable DEMOCRACY (RS) has higher values when the receiving state is more democratic. We expect that a more democratic government should be more responsive to the concerns of individual residents, including migrants. Additionally (or alternatively), democracies maybe more likely to be concerned about reputation costs of not pursuing a case.
Second, we include a measure of the recipient state's population from Heston, Summers, and Aten (2011). We believe that higher values of POPULATION (RS) will make a government less responsive to the concerns of a given set of migrants. 18 Third, we include the left-right orientation of the receiving state because we expect that left parties are more likely to support both human rights and internationalism and thus should be more likely to prosecute universal jurisdiction cases. Specifically, we include a measure of the partisanship of the executive because prosecutions often fall under the executive branch of government. The variables on partisanship comes from Scartascini, Cruz, and Keefer (2018). We include an indicator for CENTER (RS) and LEFT (RS) governments; Right governments are the excluded category.
Our final measure of political responsiveness is whether the recipient state has ratified the Rome Statute, which created the International Criminal Court. The variable ROME STATUTE (RS) is coded as 1 for years in which the receiving state is a member of the treaty and 0 otherwise. States that are members of this treaty have an international legal obligation to give full effect to the treaty by implementing it into their domestic legislation. This implementation includes defining international crimes in their domestic criminal code and may include giving universal jurisdiction to the state's laws and courts. Membership in the Rome Statute signals that a state is willing, at least in the abstract, to be responsive to these cases. Therefore, ratifying this treaty should increase the likelihood that crimes against humanity, genocide, torture, and war crimes are defined as crimes under domestic law and that domestic courts have universal jurisdiction over them, thereby enabling universal jurisdiction prosecutions. 19 16 The data come from several World Bank sources: the Bilateral Migration Database, 1960Database, -2000Database, (Özden et al. 2011, the Bilateral Migration Matrix 2010, the Bilateral Migration Matrix 2013, and the Bilateral Migration Matrix 2017 (The World Bank 2021). 17 Data were originally collected by Miller and Peters (2022). We log all migration data due to the right-skew of the data. 18 Population is logged due to right-skew of the data. 19 Ratification may also indicate government responsiveness and a commitment to advance international justice, as it shows that at least some past government thought that respecting human rights is important.
We include two measures of legal responsiveness: PRIVATE PROSECUTIONS and BRITISH LEGAL ORIGIN. PRI-VATE PROSECUTIONS, coded by Michel and Sikkink (2013), is a measure of whether an individual or organization can be a party in the criminal process and, as such, open an investigation and formal proceedings, plead before the judge, request the production of evidence, interrogate witnesses and appoint its own expert witnesses, appeal judicial decisions, etc. States that have private prosecution should be more responsive to universal jurisdiction cases because alleged victims and NGOs have more legal tools to press charges and to move these proceedings forward toward trial (Langer 2011). In contrast, we would expect British Legal Origin to be associated with fewer universal jurisdiction cases because in British Legal Origin systems the decision on whether a case should be opened or move forward to trial is typically exclusively made by public prosecutors that may be affiliated with the Executive Branch and thus be more responsive to the costs that universal jurisdiction prosecutions may entail (Langer 2011).
Our final explanatory concept is the cost of prosecutions for the receiving state. First, we examine whether the receiving state's wealth affects prosecutions. We expect that the variable GDPPC (RS) (logged; from World Bank 2015) will correspond to more state capacity to investigate and prosecute crimes. This capacity should lower the relative cost of prosecution for the receiving state.
Next, we include a measure for whether the receiving state is an OECD member, OECD (RS). Such states are both democratic and highly industrialized, suggesting that they have the resources to conduct universal jurisdiction prosecutions.
We then control for whether the receiving state has had a prior initiation or successful prosecution of a universal jurisdiction case. The variable PRIOR INITIA-TION (RS) is coded as a 1 for all years after the first initiation of a case from any state in a receiving state. The prior initiation could lead to learning and make the next case easier. PRIOR SUCCESS (RS) is coded as 1 for all years in which the receiving state has previously held a universal jurisdiction trial and the domestic court found the defendant guilty. Such a successful prosecution demonstrates to victims (and to us as researchers) that the recipient state's domestic laws are amenable to claims of universal jurisdiction. Additionally, a Prior Successful Prosecution creates judicial precedent in states with common law system and enables officials to learn how these cases should be prosecuted and tried (Langer and Eason 2019), lowering the cost of future prosecutions. 20 Finally, we include some dyadic measures that may make the receiving state less likely to want to prosecute someone from the sending state or give the sending state the power to use diplomatic measures to stop a prosecution (Langer 2011). We include two measures of economic power: the sending state's wealth, which is the variable GDPPC (SS) (logged; from World Bank 2015) and TRADE, which is the percentage that dyadic trade makes up of the total trade of the receiving state. We also include measures of military power: an indicator for ALLIANCE between the receiving and sending states (Correlates of War Project 2013) and major power status with an indicator variable for MAJOR POWER (RS) and MAJOR POWER (SS) states, which equals 1 if the receiving or sending state, respectively, is one of the permanent five members of the UN Security Council or, after 1991, Germany or Japan. Finally, we also include an indicator variable for SHARED LANGUAGE between the sending and receiving state, as this likely reduces the cost of collecting evidence and interviewing witnesses (Melitz and Toubal 2014).
We include numerous control variables in our analysis. We include a measure of how strong the NGO network is in the receiving state. Larger, more vibrant NGO networks should make it easier for migrants to connect with NGOs that can help them bring their case. The variable NGOs (RS) measures the number of NGOs that are headquartered in the receiving state coded from Smith, Wiest, and Hughes (2020). For the sending state, we include a measure of whether international prosecutions are occurring during a given year. The variable INTERNATIONAL PROSECUTIONS (SS) is coded as 1 for years in which individuals from the sending state are either under investigation by the ICC or subject to the jurisdiction of an international or mixed criminal tribunal. 21 We are agnostic about the expected effect of this variable. Although international prosecutions may encourage receiving states to respond to migrant demands for justice, they may also result in the transfer of defendants to international tribunals. 22 We also include a measure of whether there has been an amnesty in the sending state in the past. The variable AMNESTY (SS), coded from Bell and Badanjak (2019), takes a value 1 if there has been an amnesty and zero otherwise. Migrants might be more likely to seek justice in another state if they cannot get justice at home due to an amnesty.
As dyadic-level controls, we include FORMER COLONY (DYAD), which is an indicator variable that equals 1 if the sender-receiver state pair had a prior colonial relationship, as former colonies tend to send many migrants to the former colonial power. FORMER COLONY (ANY) measures whether the sending state was a colony in any empire. We also include DISTANCE from Gleditsch and Ward (2001) and SHARED BORDERS from Correlates of War Project (2007), as both migration and attention to crimes abroad may decrease with distance. We also include a variable for the COLD WAR and the WAR ON TERROR as these periods may affect the likelihood of universal jurisdiction cases, with 20 Both of these variables are dropped when we examine the first initiation. 21 These include tribunals for the former Yugoslavia, Rwanda, Sierra Leone, Cambodia, and Lebanon, and the Extraordinary African Chambers. 22 See Schabas (2003) for examples of both possibilities. fewer cases during the Cold War and more during the war on terror.

Estimation
We test our hypotheses in four different ways. For the all models, we run an ordinary least squares regression of the dependent variable on our variables of interest, as it requires fewer assumptions about the data and is easier to interpret. For the models on the first case in the receiving state, we capture "onset" of universal jurisdiction within a dyad by dropping observations for the years after the first case, similar to the conflict literature (Beck, Katz, and Tucker 1998). We also include measures of years since 1957 (the start of the dataset), years since squared, and years since cubed. These variables correct for temporal dependence between observations in dyads, as suggested by Carter and Signorino (2010). The data are still dyadic data, with all dyadic variables applying to a single sending state, but we drop all dyads-year observations after the first case. All models were run with robust standard errors clustered by dyad to account for dependance within dyads.

Main Results
We begin by testing our hypotheses on whether there is any initiation of a case in a given year and on the first initiation of a case in a receiving state. Figure 3 displays the coefficient on migrant stocks from regressions that sequentially add control variables. The first coefficient is simply the bivariate correlation between universal jurisdiction initiations and MIGRANT STOCK, the second includes dyad and year fixed effects; 23 the third removes the dyad and year fixed effects to allow us to add controls beginning with our variables measuring atrocities in the sending state; and subsequent models add in our variables for responsiveness, costs, and additional controls. Figure 3 shows that while the coefficient size changes-because we are including variables that are correlated with both migrant stocks   and universal jurisdiction cases-it is always highly statistically significant. Figure 4 displays the marginal effect of migrant stocks on whether there is an initiation in a dyad-year and on the first initiation in the receiving state from the regression models that include all the controls and the Doe cases. Universal jurisdiction cases are quite rare: between 1960 and 2007 (where we have full data coverage) there are only 84 dyad-years or about 0.05% of cases in which there is at least one universal jurisdiction case and only 0.01% of cases in which the receiving state first considers a universal jurisdiction case. As we can see in the graphs, MIGRANT STOCK has a substantively important effect: going from the 25th percentile (no immigrants from the sending state to the receiving state) to the 75th percentile (about 300 immigrants) leads to an increase in the probability of any case from statistically 0 (95% confidence interval of -0.0003 to 0.0004) to 0.0008 (95% confidence interval of 0.0006 to 0.001) and for first case in the receiving state leads to an increase in probability from 0 (95% confidence interval of −0:0002 to 0.00003) to 0.0002 (confidence interval of 0.0005 to 0.0004). Table 3 shows the results of the regressions from the models that include all the variables and allows us to examine our additional hypotheses. We find support   that the level of atrocities in the sending state increases the likelihood of a universal jurisdiction case (H2), just as atrocities can affect the creation and use of international criminal tribunals (Rudolph 2001). Sending states that were less democratic or experienced more political terror in the prior 10 years are more likely to have universal jurisdiction cases against their nationals. The coefficients on PTS (SS) and DEMOCRACY (SS) are not always statistically significant at conventional levels for a sending state's first initiation, but they do have the hypothesized sign.
Next, we find some support for the responsiveness hypothesis (H3). There is not strong evidence that a receiving state's level of democracy or government ideology matter. The vast majority of universal jurisdiction cases have occurred in democracies, yet democratic states are less likely to initiate a universal jurisdiction case once we control for other factors. 24 Similarly, left and right governments are about equally likely to pursue these cases. These findings challenge prior arguments made about the general role of democracy and ideology in promoting support for domestic and international justice (Kelley 2007;Teitel 2000).
In contrast, we find smaller states, states with non-British legal origin, and states that allow private prosecutions are more likely to have universal jurisdiction cases. These findings reinforce the arguments of prior scholars who focus on the details of domestic legal systems within democracies (Brinks 2008;Michel and Sikkink 2013). We additionally find that states that have ratified the Rome Statute also are more likely to initiate a case. Although ratification can indicate general societal support for international human rights, it can also result in the removal of some of the domestic legal obstacles to universal jurisdiction. Our evidence suggests that Rome Statute ratification can have important externalities on other states, an effect that has been largely overlooked by scholars who focus on the domestic political determinants of ratification (Chapman and Chaudoin 2013;Simmons and Danner 2010).
We also find support for the idea that the cost of pursuing these cases matters (H4), although not necessarily for the first case. These results reinforce prior research (Bass 2000;O'Sullivan 2017;Terman and Voeten 2018). Prior cases increase the likelihood of future cases, suggesting that actors learn about effective activism over time. SHARED LANGUAGE also increases the likelihood of a case initiation, likely because it is easier to gather evidence in the same language. In contrast, receiving states are unlikely to receive cases against a defendant from a military ally or a MAJOR POWER (SS), but they do target wealthier states (higher GDPPC [SS]). 25 Finally, GDP per capita of the receiving state, the major power and OECD status of the receiving state, and trade relationships seem to have no effect.
In terms of control variables, we find some surprising results. First, we find that larger numbers of NGOs headquartered in the receiving state increase the likelihood of having any initiation, but not the first initiation. 26 This finding adds nuance to the broader literature on the role of NGOs in promoting domestic and transitional justice (Keck and Sikkink 1998;Kim 2013;Lutz and Sikkink 2001;Sikkink 2011;Simmons 2009;Zvobgo 2020). Second, having an AMNESTY law in the sending state lowers the likelihood of universal jurisdiction cases, perhaps because amnesty signals broad support for moving beyond past conflicts. Third, sending states from farther away are more likely to see a case brought against one of their citizens, perhaps reflecting that these cases are generally brought by a Global North state against the citizen of a state in the Global South.
Other variables that have been hypothesized to have an effect seem to have no effect. INTERNATIONAL PROS-ECUTIONS (SS) do not affect universal jurisdiction cases. Neither do indicator variables for the COLD WAR or the WAR ON TERROR. Finally, former colonies of the receiving state are not more likely to have a case brought against one of their citizens and former colonies in general are less likely to have a case brought against one of their citizens. These findings challenge the claims made by some states that universal jurisdiction cases reflect neo-imperialism over former colonies (Geneuss 2009;Jalloh 2010;Mennecke 2017). We speculate that our findings on colonialism may be affected by two factors. First, many migrants move from former colonies to their former metropole. Thus any possible appearance of neo-imperialism may be the result of omitted variable bias that we have now corrected. Second, receiving states may be unwilling to hear universal jurisdiction cases against defendants from their former colonies to avoid the appearance of neo-imperialism.

Robustness Checks
How robust are our findings? Recall that our main analysis uses case initiations as the dependent variable. Yet many cases proceed from initiations through investigations and formal proceedings all the way to trials. We replicate Models 1 and 3 from Table 3 27 We also replicate these models with refugee stocks instead of migrant stocks. 28 Figure 5 displays the coefficients on migrant stocks for each of these variables. 29 For each measure, a marker represents the point estimate, and the thin (thick) bars represent the 95% (90%) confidence  interval. Each panel includes a dotted vertical line at 0. So coefficients whose confidence interval does not cross the 0 line are statistically significant at conventional levels. Panels (a)-(c) collectively suggest that migrants help to drive the escalation of cases. As shown by Panels (a) and (b), the effect of migrants appears to be strongest and most robust for explaining the initiation of universal jurisdiction cases. Across later stages, the coefficients for migrant stocks are mostly positive for both the ANY ACTION and FIRST ACTION variables, although they are often not statistically significant at the 90% or 95% confidence level. Panel (c) shows similar results for most of the PROPORTION variables. Panels (d)-(f) of Figure 5 collectively suggest that refugees can play a particularly important role in pushing universal jurisdiction cases forward. The coefficients on refugees are always positive and often significant at least at the 90% confidence level for the later stages in a case. Refugees are, by definition, persecuted by their home state and likely are the victims of crimes that could be prosecuted under a universal jurisdiction statute. Thus, the greater number of victims in the state may help push the case along. The evidence in Figure 5 suggests that migrants have the most agency when initiating a case. During later stages, migrants rely more on public authorities to conduct investigations, open formal proceedings, issue arrest warrants, and conduct trials. Additionally, there is usually more chance involved in when and whether later stages of a case occur. For example, a perpetrator's identity may be unknown or he may be impossible to locate and extradite, thereby hindering an arrest and trial. Thus, we are not surprised that the magnitude of the effects of migrant stocks appear to decrease in size as a case progresses and that the confidence intervals become wider.
Our Online Appendix contains many other robustness checks. 30 First, we use 10 alternative measures of migration. We estimate the separate effect of each migration variable on the ANY and FIRST INITIATION variables, both without and with the full set of control variables. This yielded 40 separate models. Overall, the effect of migration is almost always positive and significant at conventional levels.
Recall that our main analysis uses 10-year averages of democracy and atrocities in the sending state. These averages represent delays between the period in which atrocities occurred and the time at which migrants complained to the receiving state, as well as the duration of atrocities. We estimated our results using alternative lag structures, calculating the average values of democracy and atrocities over the last 5 and 20 years. We also examined minimum and the maximum values over those same periods. Overall, we find highly consistent results across all these different measures, with almost all coefficients in the hypothesized direction and significant at conventional levels.
Also recall that our main analysis highlighted the role of some domestic legal factors. To examine the robustness of migrant stocks to alternative measures of domestic legal systems, we first include receiving state fixed effects, which control for all time invariant aspects of a state that might affect universal jurisdiction cases. These effects could include aspects of the legal system that do not change over our time span. We find that migrant stock continues to have a positive and statistically significant effect on universal jurisdiction case initiation. Second, we use data on whether and what types of international crimes are defined in a receiving state's domestic law (Berlin 2020). We find that if domestic laws include at least one definition of an international crime, then the receiving state is more likely to have a universal jurisdiction case. Yet none of these variables help to explain the first case. This finding reinforces qualitative accounts of migrant pressure to change domestic laws in receiving states (Roht-Arriaza 2005).
We also examine whether the effects of migrant stocks are driven by data imputation. We regress an indicator of whether there were any initiations in the decade on the values from the start of the decade. For example, DV takes a value 1 if there was any initiation from 1970 through 1979. We then regress that on migrant stocks and the rest of the variables from 1970 and drop all the years in between. We find highly consistent results.
Next, we examine whether universal jurisdiction cases diffuse across states. We use five measures of diffusion, including global time trends, spatial lags, and counts based on shared regions, language, and legal systems. We find little evidence that diffusion explains universal jurisdiction cases.
Finally, we perform a sensitivity analysis to help understand additional threats to inference. The details of this analysis and its interpretation are included in our Online Appendix. In short, an unobserved confounder would have to have quite a large effect on both universal jurisdiction cases and migrant stocks to reduce the effect of migrant stocks to zero. It is unlikely that such a confounder exists.

CONCLUSION
Our argument and evidence collectively suggest that migrants can serve as agents of transnational justice. Many migrants who flee repression and war arrive in receiving states with grievances about prior crimes that occurred in the sending state. These grievances can motivate them to seek justice in their new state. Many prior human rights scholars have highlighted the important role of transnational activists in upholding international law. Yet they have largely overlooked another key transnational actor: migrants. Similarly, many scholars of migration have documented the economic, social, and political remittances from migrants to their sending states. Our evidence highlights another important transnational effect: migrants can provide justice remittances by pressuring receiving states to invoke universal jurisdiction, thereby punishing severe international crimes that occurred in the sending state.
Our argument and evidence also examine various economic and political factors that affect whether a receiving state responds to this pressure. All else equal, we argue that higher atrocities in the sending state are more likely to lead to universal jurisdiction cases, as proxied by a state's polity score or political terror score. Similarly, higher responsiveness to public pressure in the receiving state, as proxied by variables like private prosecutions, will also increase universal jurisdiction cases. Finally, variables that affect economic and political prosecution costs, like the wealth and sending state Great Power status, will affect whether universal jurisdiction occurs.
Although our focus in this article has been on explaining when universal jurisdiction occurs, our findings come with two important policy implications. First, universal jurisdiction may influence democratization and the termination of armed conflict. When states democratize and end conflicts, they frequently implement amnesties that limit criminal prosecutions for crimes committed by the prior regime. According to some authorities, under international law, amnesties do not apply to serious international crimes because a state cannot unilaterally invalidate individual criminal responsibility under international law (Johns 2022). Therefore, many scholars debate whether the prospect of international criminal prosecutions may prolong armed conflicts and prevent democratization (Dancy 2018;Gilligan 2006;Prorok 2017). Universal jurisdiction cases may have the same effect: perhaps political and military leaders who commit severe international crimes will be hesitant to give up power if they fear universal jurisdiction cases. We remain agnostic on this point, noting only that actual universal jurisdiction trials remain relatively rare, they can be avoided if perpetrators reside in states that limit extradition, and our analysis shows that amnesties reduce universal jurisdiction cases.
Second, we believe that our evidence counters the claim that universal jurisdiction is a form of "judicial tyranny" by domestic courts seeking to intervene in foreign states (Kissinger 2001, 86). Although domestic courts may not be able to base their jurisdiction on traditional bases of jurisdiction-like territory or nationality-they do usually have links to the alleged crimes because victims are living in the prosecuting state. Additionally, most states do not allow actual trials to occur unless the defendant is physically present, meaning that prosecuting states often deny "safe harbor" to an international criminal within their borders, rather than to merely project neo-imperial power abroad (Langer 2015b). Indeed, the UK's refusal to extradite Pinochet ultimately stymied the Spanish prosecution. Extradition proceedings therefore serve as a check on overzealous domestic prosecutions (Roth 2001).
Our findings suggest two avenues for future research. First, we did not examine who is targeted in universal jurisdiction cases. The economic and political costs of cases will almost certainly depend on who is accused of a crime. We expect that these costs will be highest when an alleged perpetrator is a sitting high-level government or military official. In contrast, economic and political costs are probably lowest for nonstate actors. In between these two extremes are former officials or members of paramilitary groups with some state support. There is ample information about some highprofile universal jurisdiction cases, but privacy laws often limit public information about alleged perpetrators. Therefore, collecting perpetrator information for all universal jurisdiction complaints remains an extremely time-intensive, albeit important, task for future research.
Finally, many attributes of domestic legal systems hinder universal jurisdiction cases. In our empirical analysis, we control for these factors in numerous ways. However, future research can include collecting more information about laws that enable universal jurisdiction cases including the definitions of international crimes, whether universal jurisdiction is explicitly allowed under domestic laws, procedural constraints for nonterritorial cases, and modes of criminal liability. Such information would be inherently valuable for comparing how different states implement international criminal law. It would also provide another possible measure of migrant influence on the receiving state because migrants often pressure legislators to revise such laws to enable universal jurisdiction prosecutions (Roht-Arriaza 2005). For now, we believe that we have identified an important phenomenon in transnational politics: how migrants provide justice remittances by pressuring states to pursue universal jurisdiction cases.

SUPPLEMENTARY MATERIALS
To view supplementary material for this article, please visit http://doi.org/10.1017/S0003055422000302.

DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Reivew Dataverse: https://doi. org/10.7910/DVN/PATI3W. Eason for sharing his coauthored dataset and to Francesca Allegra, Abbey Farison, and Bruce Tsai for their excellent research assistance.