Obstacles and momentum on the path to post-genocide and mass atrocity reparations: A comparative analysis,1945–2010

Introduction

Since World War II, genocides have been studied in the hope of deterring them (e.g. Genocide Prevention Task Force, 2008) – but new outbreaks of the phenomenon since the Holocaust make it painfully evident that we still know far too little about genocides to intervene effectively. Perhaps intractable social inequalities combined with a mercurial human nature augur the continuance of such horrors (see Charney, 1982; Powell, 2011; Waller, 2007). But if we cannot suppress the wrongs of genocide, can we at least right them? Can we rectify past injustices, possibly even foster goodwill between victims and perpetrators? Can we, as some have asked, ‘make whole what has been smashed’ (Torpey, 2001: 333)? That, to be sure, is the principal aim of the reparations politics that has proliferated in various settings around the world – a means of reconciling antagonists, perhaps even building something new out of the shards of violent conflict. In this quest, reparations politics is much more than a politics of the past weakly substituting for an absence of progressive vision; rather it constitutes a search for justice in the present that can heal past iniquities and animate the construction of a better future.

In the spirit of that endeavor, we contribute to the illumination of reparations politics, as well as to the study of transitional justice. We have begun our efforts to do this by delineating what we call a sociography of genocide reparations. This entails development of an analytical grid that helps to identify obstacles to repair, as well as the resources that when mobilized allow a claim to gain momentum on the path toward settlement, in turn enabling a comparative assessment of the life-course of reparations claims. The concept of momentum has heretofore been insufficiently conceptualized and operationalized in the resource mobilization theory (RMT) literature that is the basis of our approach. Our research presents a unique effort to measure the force of momentum on social movement success. ¹

Reparations, social movements, and resource mobilization theory

We use the term ‘reparations’ to describe a variety of formal and informal responses to an unsavory past. These responses are directed toward the goal of societal recovery and stabilization, but are also often intended to foster a level of individual healing for those victimized by genocide and mass violence. Thus, the ‘field’ of reparations politics, as capsulized by Torpey (2006), is comprised of mechanisms that range from justice processes (e.g. tribunals and truth commissions) to institutional reforms (e.g. lustration and preferential hiring policies), as well as from material (e.g. restitution of property and monetary compensation) to symbolic (e.g. commemoration and apologies) forms of redress. Torpey’s characterization of reparations politics is admittedly broad; however, given the wide variety of post-atrocity demands issued by victim groups and their supporters, attempts to delimit such a category would be to prejudge or prescribe allowable forms of reparative settlement. Moreover, when forming their claims, reparations movements stitch together a variety of processes, reforms, and forms of redress into a cohesive reparations frame, and this frame is the focus of our analysis when mapping the obstacles faced and momentum achieved by a particular claim on the pathway to repair. Therefore, we utilize this broad understanding of reparations politics, as it better captures the multiplicity of reparative desires expressed by victims of atrocity.

Today, such mechanisms are typically addressed through the language of ‘transitional justice’, an encompassing term intended to capture the various legal and non-legal mechanisms implemented to assist a society in establishing more peaceable future relations between former antagonists (see Arthur, 2009; Bell et al., 2007; Boraine, 2006). For some authors, the transitional future is one defined by democratization or liberalization (Teitel, 2000). In contrast, our conceptualization of reparative success is focused less on assuming a specifically desired political outcome and more on questions of whether a reparative mechanism was implemented in a former conflict zone, and, if so, whether it is perceived to be successful by those giving, receiving, experiencing, and observing these reparative efforts. We do, however, follow in the footsteps of the transitional justice literature by seeking to construct a comparative model for evaluating reparative processes (Kritz, 1995; Minow, 1998; Offe, 1997). Likewise, it is widely acknowledged in the genocide studies literature that comparative approaches are essential to obtaining ‘conceptual comprehension of the phenomenon’ (Huttenbach, 2000: 7; see also Chalk and Jonassohn, 1990; Jonassohn and Bjornson, 1998; Walliman and Dobkwoski, 1987). However, there is still a dearth of broadly comparative research on post-genocide and mass atrocity reparations claims (although see De Greiff, 2006). Much of the early comparative research on reparations focuses on only one reparative gesture, such as truth commissions or restitution (Hayner, 2002; Long and Brecke, 2002). Other comparative studies ignore or pay inadequate attention to reparations demands that have not achieved any discernible form of resolution (Barkan, 2000; Brooks, 1999). Nonetheless, efforts continue to bring the tools of empirical social science to bear on transitional comparisons (Fletcher et al., 2009; Kim, 2008; Pham and Vinck, 2007; Thomas et al., 2010). We contribute to this latter trend through a comparative examination of multiple factors that may obstruct or facilitate a reparations social movement in achieving its demands for redress. In particular, we are interested in how these social movements build momentum toward repair through the accumulation of resources and supports.

Social movements theory has to date been under-utilized in the study of genocide and post-genocide transitional justice (for notable exceptions, see Hagan and Rymond-Richmond, 2008; Owens and Snow, 2013; Owens et al., 2013; Stein, 1998). Indeed, much of the focus in the early transitional justice literature was on questions related to international relations and the pragmatics of transition (Huyse, 1995; Pion-Berlin, 1994). The objective was to explain why transitional justice emerged when it did, but also to identify potential sources of friction within transitional justice, particularly imbalances of power between various actors, such as soldiers and citizens, or new and old elites (Huyse, 1995; Pion-Berlin, 1994). Other scholars were most interested in the various forms transitional justice might take (e.g. trials or truth commissions) and the advantages and disadvantages of each justice type (Barkan, 2000; Bass, 2000; Brahm, 2007; Hayner, 2002; Minow, 1998; Osiel, 1997; Van Zyl, 2002). However, the literature on transitional justice has expanded to consider the roles played by victim groups and advocacy networks in compelling states to engage in transitional justice and international actors to support these demands (Grodsky, 2007; Keck and Sikkink, 1998; McEvoy and McGregor, 2008; Skaar, 1999). Social movement theorization offers to this literature a conceptual language for examining how victim movements and advocacy organizations frame their justice demands, mobilize resources to pressure state and international power holders, and take advantage of political openings when there exists heightened public and political receptivity to transitional justice demands.

The term ‘social movement’ is not used here to refer to individual organizations. Instead, it refers to a broader, structural organizing principle under which any number of groups may be placed (Diani, 1992: 14). In this sense, a social movement is ‘a network of informal interactions between a plurality of individuals, groups and/or organizations, engaged in political or cultural conflict on the basis of a shared collective identity’ (Diani, 1992: 13). A social movement network is not a static entity; rather, it is the product of ongoing negotiations between movement leaders, constituents, governing agents, and other external actors, who contribute to processes of collective identity formation, grievance and justice framing, as well as collective action. With respect to reparations movements, we see in most cases a transnational network of individuals and organizations who are actively engaged in building a justice narrative around an atrocity event, using this narrative to promote a common set of demands for symbolic and material reparation.

This conceptualization of a social movement requires that such movements be understood as dynamic, process-based entities. Such an understanding is consistent with recent trends in the study of genocide that focus on the complex international, national, and local processes that provoke crimes of group destruction (Bloxham, 2009; Hinton, 2005; Levene, 2005; Moses, 2000). Likewise, some scholars have also examined genocide as a form of societal breakdown that destroys community networks and resources (Fletcher and Weinstein, 2002; Snow et al., 1998; Useem, 1998). The flipside of such analyses, and the logic that drives demands for transitional justice and reparations, is that processes and resources need to be accumulated to make possible the rebuilding of post-genocidal societies (Fletcher and Weinstein, 2002). In such circumstances, several authors identify valuable strategies for framing transitional justice demands, resources that help groups advance these demands (Keck and Sikkink, 1998), and political opportunity structures that increase receptivity to these demands (Irvine and McMahon, 2013; Kim, 2008). All three of these aspects of reparations movements can be understood through an RMT framework.

RMT is a theoretical framework for the study of social movements that seeks to identify the resources and opportunities that make social movement achievements, such as grievance articulation, message resonance, and political traction, possible.

Proponents of RMT argue that social movements do not simply arise in response to a particular grievance or social uncertainty, since these two factors are ubiquitous in any societal period; instead, they suggest that social movements develop and increase in efficacy as the resources necessary for mobilization (e.g. human, economic, and political resources) become available (see Gamson, 1975; Marx and McAdam, 1994; Tilly, 1978). These resources, broadly conceptualized, include not only the economic and other material resources that make the day-to-day functioning of a movement possible, but also the mobilizing structures and networks formed by actors engaged in a common project of collective action (McAdam et al., 1996), the political opportunity structures that remove constraints that might otherwise obstruct this movement (Meyer and Minkoff, 2004; Tarrow, 2005), and the ‘organizational’ and ‘master’ frames whereby movements both organize their constituents and communicate their sense of injustice including demands for reparation to a broader audience (Benford and Snow, 2000). Given these three areas of focus, RMT is a useful perspective for examining the collective action that makes possible genocide and mass atrocity reparations claims, because it directs our attention not solely to internal movement dynamics, but also to how movements mobilize to articulate their messages to broader publics within either facilitative or obstructive social and discursive contexts. The RMT perspective, therefore, is ideally suited for examining the obstacles that confront movements as they seek to achieve their goals.

A major assumption of RMT is that a social movement’s success in accessing resources facilitates further success in doing so. McCarthy and Zald (1977), for instance, hypothesize that ‘As the amount of discretionary resources of mass and elite publics increases, the absolute and relative amount of resources available to the social movement sector increases’ (p. 1224). Jenkins (1983) furthermore notes that a group’s pre-existing degree of organization will optimize its ability to mobilize. Supporting RMT’s emphasis on political process, resource acquisition, and goal-oriented bonds across a myriad of networks of actors in a society, we use the construct of ‘momentum’ to facilitate our analysis in the later, quantitative, portion of this article. While many authors discuss the momentum that social movements achieved at critical points in history, they do so only casually. Few formally discuss and operationalize this seminal metaphor as it applies to social movements. Similarly to Kelly and Amburgey (1991), we conceptualize momentum as the aggregate number of successes a social movement (i.e. genocide repair claim) has in attracting resources unto itself. However, whereas they apply their analysis to a sample of business organizations, following Clark (1966), we conceive momentum to have a much larger compass in relation to genocide repair, gaining force beyond the efforts of individual actors and organizations and taking on a life of its own, so to speak, in the movement toward fulfillment of a reparations claim.

Analytical strategy

The overarching goals of this article are to identify specific obstacles that threaten the possibility of genocide repair and to understand more holistically whether and how a claimant group can overcome them through our concept of reparative ‘momentum’. Meeting these goals involved developing categories to describe the obstacles, and then arranging them into a logical sequence that we use to depict the behavior of each claim as it progresses toward ultimate success or failure. In the first phase of our analysis, we begin by describing in detail our model of the process of genocide repair (i.e. the sequence of obstacle ‘categories’ that we refer to in this article as our ‘analytical grid’). We then describe the overall performance of our sample of cases in terms of those categories and we evaluate the likelihood of a claim’s ultimate success. Included here is a graphic depiction of each case’s performance as well as a discussion of the histories of a small set of claims as case examples.

In the second phase of the analysis, we use a quantitative approach to characterize how the different obstacles may affect the progression of a claim. The conceptual obverse of each obstacle is naturally the success that a claim has in clearing that obstacle. Statistical analysis demanded that we reshape our data so that the obstacle categories could be represented numerically. As noted above, we take momentum to be a claim’s ongoing success in overcoming these obstacles and we measure it as the aggregation of these successes. Operationalized as such, we presume momentum to make up the substance of a claim’s ability to mobilize resources around itself. Our expectation is that the more resources a reparations movement accumulates in the process of seeking justice, the more likely it will achieve some modicum of satisfaction. In essence, we demonstrate how the multiplication of resources could be imperative for a reparations movement to overcome the obstacles in its path. We conclude with an analysis of three seminal research questions that should prove useful in directing further inquiry. These questions ask whether reparative momentum is apparent for some cases, whether it can grow as a claim progresses, and whether it can falter.

Methodology

Our data collection to identify a set of resources necessary to a successful reparations claim occurred in two phases: a first phase or ‘pilot study’ in which we garnered 30 case studies of mass atrocities, and a second phase or ‘primary study’ on which our findings are based, and which encompassed 47 cases, including most of the original 30.

At the time we undertook the pilot study, we had only a vague notion of a reparations continuum and a merely implicit conception of the reparations process functioning as a social movement. Two researchers (one of the co-authors and a sociology doctoral student) chose 30 cases at random from the 88 cases listed in the 2002 Genocide Watch List, ² aiming for a globally stratified sample. Four of those cases were subsequently discarded when it was found that either no reparations claim had been actuated in those instances or there was insufficient information for developing a reliable case study. ³ Rough narrative summaries of the cases were composed for the purpose of identifying obstacles to genocide repair. The investigation was guided to some extent by salient and recurrent themes drawn from the genocide/reparations literature ⁴ and resources/obstacles on the path to reparations were consensually determined through the construction of those narratives based on a review of various online and library resources. This enabled the extraction of ‘categories’ (or variables) that established the particular data-points for the study. At that time in the pilot study, we were primarily concerned to identify potential obstacles and were agnostic about what we might discover through the later statistical analysis. By the end of the exercise, however, it was possible to conceive a logical–temporal sequence across categories and to design the analytic grid in conjunction with RMT. The code-sheet was then developed in preparation for the primary study.

Based on this initial review, as well as a review of the reparations and transitional justice literatures, we isolated 23 categories (variables) that reflect central issues and common themes in the reparations literature, and can be conceptualized as resources or obstacles (when the resource is absent) in relation to a reparations claim. We do not argue that these categories, though wide-ranging, are fully comprehensive; other factors could doubtlessly be added to our list, such as the nature of the conflict (e.g. civil war, government oppression) or the degree of atrocity (e.g. numbers and regions affected). ⁵ As well, closer qualitative examination could be made of sub-categories within each of our selected categories. For example, under ‘societal conditions’ we consider many factors that could be subject to more intensive investigation, such as the level of institutional development within the post-atrocity society, power relations between the new and old regimes, and levels of social disorganization. Our objective, however, was not to provide a complete roadmap and itemize all of the many potential obstacles met and resources required on the pathway toward reparations; instead, our goal is to contribute an understanding of how certain obstacles can slow or completely upend reparative processes, and how momentum can build as resources accumulate, allowing reparations movements to overcome such obstacles. Thus, we order these categories in a linear trajectory of the stages through which a reparations claim ideally progresses – from the lodging of the claim to its resolution (see Box 1). This is not to imagine that each and every reparation claim proceeds in a linear manner. Rather this linearity is an analytical device that allows for the quantification of reparative momentum and assumes, for our purposes, the additive nature of resource mobilization.

More specifically, the categories in Box 1 relate to RMT in the following manner. Category 1, ‘Actionable Claim’, determines whether a specific case should be included in our sample by assessing whether the case fits the United Nations Convention on the Prevention and Punishment of the Crime of Genocide (1948) (UNGC) or subsequent amplifications and modifications of the definition of genocide (adopted by Resolution 260 (III) A of the UN General Assembly on 9 December 1948. Entry into force: 12 January 1951). The primary criterion for inclusion in our sample was the extent to which a case could be argued to fit the UNGC (1948), which remains the principal basis for identifying genocide within international law. Section II of the UNGC reads,

In the present Convention, genocide means any of the following acts committed with intent to destroy, in whole or in part, a national, ethnical, racial, or religious group as such:

However, we are cognizant of the fact that genocide is an ‘essentially contested’ concept (Powell, 2011) and therefore we also included cases of mass and/or systematic violence perpetrated with the intent to destroy in whole or in part a group specified not in the 1948 UNGC but in subsequent amplifications (political, sexual identity, gender) of the definition of genocide. In addition, some cases (though most are also included on the Genocide Watch List) are more technically described as mass atrocities, massacres, or flagrant human rights violations perpetrated against the group in question, because there is no evidence of sustained genocidal action in terms of the UNGC. These latter inclusions made possible comparisons between cases depending on the extent to which they fit formal UNGC genocide criteria, an issue upon which we touch later in the article.

While Category 1 situates the case in relation to the genocide concept, Categories 2, 3 4, and 5 address the extent to which the reparations movement engaged in practices of organizational and master framing: Category 2 assesses whether a claim was publicly made; Category 3 examines whether the movement has made explicit its reasons for demanding a claim (and to what degree); Category 4 observes whether the term ‘genocide’ was used to frame past injustices; and Category 5 focuses on whether the movement gave precise substance to the nature of its reparations claim. Categories 6 through 11 examine elements of the mobilization and networking structures available to the reparations movement. ⁶ Categories 6 and 7 address the involvement of the victim group in launching and sustaining the claim. Categories 8 and 9 monitor the level of resistance to the claim within both the victim and perpetrator groups. Categories 10 and 11 measure the support of third parties for the reparations movement, resistance from third parties, as well as broader institutional auspices that may have lent their support (or provided resistance) to the movement. Finally, Categories 12 to 18 reflect the political opportunity structures available to the reparations movement. Such opportunities include whether the reparations process is conducted under appropriate auspices (Category 12), whether an appropriate forum was/is proposed for purposes of repair (Category 13), whether the government takes a facilitative role in processing the reparation demand (Category 14), whether there exist sufficient governmental or international resources to meet the demands of the claim (Category 15), whether there exist other societal conditions (Category 16) or reinforcing discourses (Category 17) that generated reparative opportunities for the victim group, and whether there exists widespread popular agreement as to the validity of the claim (Category 18). The remaining categories, 19 to 23, are our ‘success measures’ whereby we examine whether there was/is any form of definitive response to the reparation demand (Category 19), the reaction to this response from both the victim (Category 20) and perpetrator (Category 21) groups, whether there exist any discernible reparative effects that resulted from the response (Category 22), and whether international opinion is such that the claim is frequently referred to as a ‘success’ (Category 23). It is important to note that we do not prescribe success in the form of a concrete outcome, but rather adjudge it in accordance with how the reparations process was generally received.

Each of the categories is coded in a way that allows us to identify whether a given reparations claim meets the criteria of a specific stage of the reparations process (for two examples of category coding, see Box 2). ⁷ A positive response to a given category (‘yes’ or ‘qualified yes’) indicates that the claim advanced through the specified stage, whereas a negative response (‘no’ or ‘qualified no’) marks an obstacle on the reparation path. The ‘qualified yes’ and ‘qualified no’ codings allowed us to better capture the complexity of reparative processes, acknowledging instances where moderate progress was made, even though the reparative outcome suggested by the category is/was not yet fully realized. The coding procedure, therefore, yields a trajectory for each case across the 23 categories, highlighting junctures in the reparative process where claims are stalled and aborted and do not proceed to resolution, as well as offering a sense of how the accumulation of successes builds momentum toward definitive reparative outcomes.

In filling out the set of cases for the primary study, nine further cases (beyond the 26 cases retained from the pilot study) were randomly chosen from the Genocide Watch List and cross-checked for regional representativeness, retaining only those cases in which there was an actuated post-1945 reparations claim. To mitigate the possibility of Western bias, we added four settler colonial cases (i.e. Canada, Australia, New Zealand, the United States) that are not included in the sample frame of Genocide Watch. As well, at the urging of various colleagues, and based on our own immersion in the literature, we added eight prominent genocide cases occurring before 1945 (and not on the Genocide Watch List – that is, Germany/Roma, Germany/Jewish groups, Belgian Congo, Turkey/Armenians, Germany/Namibia, Italy/Ethiopia, Japan/Nanking, and USSR/Ukraine) but with ongoing post-1945 reparations claims. Our primary study sample therefore consisted of 35 of the 88 cases on the Genocide Watch List, plus 12 cases not on the list. All of those cases denote reparations claims that have adequate public traction to be known to the non-governmental organization (NGO), online, and human rights media universe. Our persistent effort was to assemble a reasonably representative set of cases throughout the sampling procedure; we chose and included cases without prejudice toward any given case and without presumption about what we would find during the later quantitative phase of the study.

Three trained doctoral student research assistants then served as coders on the project, first constructing their own more intensive narratives of all 47 cases. ⁸ After completing their narratives of a given case, the research assistants then coded the case across the 23 categories of the analytic grid, utilizing the project code-sheet devised by the principal researchers. They also wrote succinct rationales for each coding decision. The principal researchers reviewed the case study documents (i.e. the narrative and category coding rationales) and made suggestions/revisions for the research assistants to consider aimed at ensuring the integrity of the coding schema across the range of cases. ⁹ The analytical grid was itself open to change during this process if, indeed, any new or modified dimensions to the reparational continuum emerged. ¹⁰ Any changes resulted in a revised set of documents for each case study and summary code sheets were then entered into a master chart showing coding trajectories for the complete sample of case studies. ¹¹

This article represents a first-stage comparison of the cases. In so doing, it offers a relatively parsimonious framework for further case study analyses of the course and fate of post-genocide and mass atrocity reparations claims. Such a framework is useful in considering how movements for reparations draw upon available resources and what obstacles present the most detrimental impediments to reparative momentum.

In sum, our research sample for the primary study comprised 47 cases, some occurring in Africa, Latin America, Asia, Europe, the Middle East, as well as examples of colonial assaults on indigenous peoples (see Table 1). The timeline for the claims-making activity periodizing most of our cases – 1945 to 2010 – corresponds with the surge of reparations claims and the vitalization of international law that has occurred since WWII. Three conceptually distinct sets of cases ultimately comprised our overall sample (i.e. genocide cases that fit the 1948 UNGC, cases that rely on an amplified definition of genocide, and cases more often described as cultural genocide, mass atrocities, or massacres, but for which an argument could conceivably be made to apply a further modified genocide definition). ¹² As noted, early in our pilot study, we also considered cases where genocide or mass violence claims could be made, but no discernible claim had emerged. These cases (e.g. North Korea) were eliminated from the sample because our emphasis is on how the accumulation of resources, or momentum, bolsters a claim after it has been launched.

OPEN IN VIEWER

Table 1.

Post-genocide and mass atrocity reparations by case type.

Country	Place	Period	Victim Group	Claim Start	Reparations Start
(a) Australia	Country-wide	18th–20th centuries	Indigenous	1972	n/a
(a) Bangladesh (CHT)	Southeast	1971–1993	Non-Bengali Indigenous	1997	n/a
*(a) Bangladesh [Pakistan]	East Pakistan	1971	Bengalis and Hindus	1972	n/a
(a) Belgian Congo	Congo	1880–1920	Indigenous Congolese	1904/2002	n/a
*(a) Bosnia	Country-wide	1962–1995	Muslims and Croats	1993	2001
*(a) Burma	Country-wide	1962-present	Ethnic minorities and dissidents	1996	n/a
*(a) Burundi	Country-wide	1961–2003	Tutsis and Hutus	1995	n/a
*(a) Cambodia	Country-wide	1975–1979	Ethnic and religious	1997	2009
(a) Canada	Country-wide	19th–20th centuries	Indigenous	1968	1973
*(a) Congo [DRC]	Country-wide	1994–2010	Hutu, Tutsi, and other ethnicities	1999	n/a
*(a) Croatia	Vukovar	1991–1994	Croats	1999	n/a
(a) Germany	Europe	1938–1945	Jews	1939/1945	1952
(a) Germany	Namibia	1904–1908	Indigenous	2001	n/a
(a) Germany	Europe	1933–1945	Roma/Sinti	1979	n/a
*(a) Guatemala	Country-wide	1960–1996	Pol. Opponents and Indigenous	1996	2003
*(a) Iraq	Kurdistan	1987–1988	Kurds	2005	n/a
*(a) Nigeria	Biafra	1967–1970	Igbo-Biafrans	1999	n/a
(a) Ottoman Turkey	Armenian regions	1915–1920	Armenians	1965/2005	n/a
*(a) Rwanda	Country-wide	1994	Hutu and Tutsi	1995	1997
*(a) Russia	Chechnya	1940-present	Muslims	2003	n/a
(a) Sudan	South	1955–2010	Southerners and political opponents	2005	n/a
*(a) Sudan	Darfur	2003-present	Tribal opponents	2006	n/a
(a) Uganda	North	1996-present	Acholi	2005	n/a
(a) US	Country-wide	16th–20th centuries	Indigenous groups	1910/1961	1990
(a) USSR	Ukraine	1921–1956	Ukranians, Peasants, Kulaks and political opponents	1996	n/a
(b) Angola	Cabinda Prov.	1975–2010	Cabinda separatists	2002	n/a
*(b) Argentina	Country-wide	1976–1983	Political opponents	1984	1985
*(b) Chilé	Country-wide	1973–1990	Political opponents	1991	1992
(b) China	Country-wide	1999–2010	Falun Gong	2001	n/a
*(b) East Timor	Country-wide	1975–1999	Political opponents	2000	n/a
*(b) El Salvador	Country-wide	1980–1991	Political opponents	1992	n/a
*(b) Ethiopia	Country-wide	1974–1991	Political opponents	1991	1994
*(b) Indonesia	Country-wide	1965–1966	Political opponents	2000	n/a
(b) Sri Lanka	North & East regions	1956–2009	Tamils and political opponents	2003	n/a
(c) China	Tiananmen	1989	Students	2000	n/a
*(c) China	Tibet	1950-present	Tibetans	2005	n/a
(c) Colombia	Country-wide	1964-present	Political opponents	2005	n/a
(c) Italy	Ethiopia	1935–1939	Ethiopians	1947	n/a
(c) Japan	Nanking	1937–1938	Chinese	1997	n/a
*(c) Kosovo	Region-wide	1996–1999	Muslims	2007	n/a
*(c) Mexico	Chiapas	1994–2001	Indigenous group	1994	n/a
(c) Mexico	Talateloco Square	1968	Students	2005	n/a
(c) New Zealand	Country-wide	19th–20th centuries	Indigenous Maori	1840/1967	1975
*(c) Peru	Country-wide	1980–2000	Indigenous and political opponents	2001	2004
*c) Sierra Leone	Country-wide	1991–2002	Civilians and political opponents	2002	2004
*(c) South Africa	Country-wide	1961–1990	Blacks and coloreds	1998	2003
*(c) Zimbabwe	Country-wide	1979–1987	Ndbele, citizens and political opponents	1985	n/a

*= case also used in the pilot study.

(a) = case fits 1948 UN Genocide Convention.

(b) = case fits subsequent non-ratified amplification.

(c) = case is generally acknowledged as a ‘crime against humanity’ or ‘mass atrocity’, but not as a genocide.

At times, cases may lend themselves to more than one classification; for example, if both ethnic and political groups were targeted by a perpetrator. In such cases, we have listed the country under the more serious classification (e.g. genocide convention rather than non-ratified amplification).

Notable methodological caveats in this exercise are the following:

Despite these (and other) challenges, we believe that comparative research of the type we are undertaking can yield a clearer picture of the conditions conducive to reparative settlements. Ideally, this would entail mapping the reparative process in ways that anticipate some of the conditions that thwart or facilitate the resolution of reparative claims, generating ideas for model-building and theorizing about reparative processes, and ultimately providing direction and guidance to scholars, activists, and government agents committed to these important objectives.

Mapping obstacles to reparations claims

Basic findings

Our initial coding of the cases, following the literature review, was converted into numeric scores and examined to determine major trends; in particular, we sought to identify the major points of breakdown in the life-course of reparations claims. In this analysis, Categories 19 through 23 (as noted above) were treated as dependent variables, meaning that the Categories of definitive resolution, favorable reactions by both victims and perpetrator groups, reparative effects, and international opinion designating the case as a ‘success’ were treated as the positive outcomes of the preceding 18 variables (see Box 1). ¹³ Bearing in mind that individual participant perceptions of reparation and success are subjective, and that even the categories of victim and perpetrator can be problematic in such circumstances (see Mamdani, 2001), we nonetheless reasoned that those few reparations claims that achieved success in terms of Categories 19 through 23 could be viewed as a model or accepted standard for other reparations movements that have not yet reached this level of redress (Torpey, 2001). This allowed us to accept these categories as an ideal typical, albeit contestable, desired outcome for reparations claims.

Descriptive frequencies were run at the outset of our analysis to examine how well the cases fit the criteria for our study, as well as to identify initial points of reparations rupture (see Table 2 for a full breakdown). The first five coding categories (Actionable Claim, Claim Actuated, Explanation for Advancing Claim, Genocide Claim Advanced, Reparations Demands Specified) provide a sense of the degree to which the case can be considered a post-genocide or mass atrocity reparations claim. Category 1, Actionable Claim, shows that 49% of our cases conform to the 1948 UNGC. However, it should be noted that we use a fairly liberal reading of the UNGC, and include cases where, for example, Clause 2(e), ‘Forcibly transferring children of the group to another group’, was interpreted to capture cases of forced adoption and boarding schools, even though such cases have not been legally tested in international courts. ¹⁴ In general, then, we read the UNGC in sociological, rather than legal terms, applying it as a grid for determining physical and biological attempts to destroy a group in whole or in part. In addition, a further 19% of our cases fit amplified definitions of genocide that include attempts to destroy groups not included in the UNGC, such as political groups (e.g. see Fein, 1993; Jones, 2010). Finally, a further 32% of our cases fit neither the UNGC nor amplified definitions, but were interpreted instead as cases for which an argument could conceivably be made to apply a modified genocide definition, although such cases are frequently labeled simply as mass atrocities or massacres.

OPEN IN VIEWER

Table 2.

Frequencies and percentages..

Category	Yes	Qualified Yes	Modified Yes	Qualified No	No
1) Actionable claim	23 (48.94%)	9 (19.15%)	15 (31.91%)	n/a	n/a
2) Claim actuated	47 (100%)	0	n/a	0	0
3) Explanation for advancing claim	47 (100%)	0	n/a	0	0
4) Genocide claim	30 (63.83%)	6 (12.77%)	n/a	5 (10.64%)	6 (12.77%)
5) Reparation demand specified	43 (91.49%)	3 (6.38%)	n/a	1 (2.13%)	0
6) Victim gp. Involved initiating claim	20 (42.55%)	15 (31.91%)	n/a	10 (21.28%)	2 (4.26%)
7) Victim gp. Involved sustaining claim	26 (55.32%)	14 (29.79%)	n/a	6 (12.77%)	1 (2.13%)
8) No resistance to claim within victim gp.	19 (40.43%)	23 (48.94%)	n/a	4 (8.51%)	1 (2.13%)
9) No resistance to claim within perpetrator gp.	0	4 (8.51%)	n/a	9 (19.15%)	34 (72.34%)
10) Third-party bolstering of claim	37 (78.72%)	8 (17.02%)	n/a	2 (4.26%)	0
11) No third-party resistance to claim	10 (23.40%)	7 (14.89%)	n/a	11 (23.40%)	19 (40.43%)
12) Appropriate auspices	21 (44.68%)	21 (44.68%)	n/a	4 (8.51%)	1 (2.13%)
13) Appropriate forum	12 (25.53%)	17 (36.17%)	n/a	14 (29.79%)	4 (8.51%)
14) Facilitative processing of claim	5 (10.64%)	17 (36.17%)	n/a	17 (36.17%)	8 (17.02%)
15) Available resources to satisfy claim	6 (12.77%)	21 (44.68%)	n/a	14 (29.79%)	6 (12.77%)
16) Societal conditions reinforcing claim	2 (4.26%)	18 (38.30%)	n/a	14 (29.79%)	13 (27.66%)
17) Discourses reinforcing claim	4 (8.51%)	16 (34.04%)	n/a	25 (53.19%)	2 (4.26%)
18) Popular consensus regarding validity of claim	4 (8.51%)	17 (36.17%)	n/a	17 (36.17%)	9 (19.15%)
19) Definitive resolution of claim	2 (4.26%)	10 (21.28%)	n/a	30 (63.83%)	5 (10.64%)
20) Favorable reaction to current state of resolution from those receiving/expecting reparations	0	7 (14.89%)	n/a	18 (38.30%)	22 (46.81%)
21) Favorable reaction to current state of resolution from those making/ considering reparations	12 (25.53%)	13 (27.66%)	n/a	12 (25.53%)	10 (21.28%)
22) Reparative effects subsequent to resolution status of claim	3 (6.38%)	13 (27.66%)	n/a	16 (34.04%)	15 (31.91%)
23) Claim adjudged a success	1 (2.13%)	11 (23.40%)	n/a	10 (21.28%)	25 (53.19%)

Despite differences with respect to the applicability of the UNGC, common trends were noted in the framing of reparations demands. Reparations claims were actuated in all cases (Category 2) in accordance with our criteria for selecting cases to analyze, and claimants invariably provide explanations as to why reparations are necessary (Category 3). Within those explanations, 64% of claimants framed the injustice committed against them as genocide in making their claim, while more marginal claims of genocide (e.g. where the genocide claim is made only by individual members of the claimant group, but is not officially or widely endorsed by the claimant group) were evident in another 13% of cases (Category 4). In nearly all of our cases, the claimant group clearly articulated their reparations demands (Category 5, 91.5%) rather than simply issuing a general call for repair.

Assembling mobilization and networking structures presented challenges for many reparations movements. In general, members from the victim group were either primarily or secondarily involved in initiating (Category 6, 75%) and sustaining (Category 7, 85%) the claim. Little resistance to the claims-making process was evident among members of the victim group (Category 8), with the majority either fully supporting (40.5%) or mostly supporting (49%) the claim’s demands. But further down in our coding scheme, points of breakdown begin to emerge with respect to these mobilization and networking structures. For example, Category 9 (No Resistance to Claim Within Perpetrator Group) saw 72.3% of cases experiencing significant resistance from the accused perpetrator party, with a further 19% experiencing substantial resistance from, at minimum, a portion of members within the perpetrator group. Such resistance is to be expected in conflict situations, where groups often hold widely divergent versions of past events (see Eltringham, 2004), particularly when the accused group faces punitive charges. Perhaps more significant is that third-party resistance to reparations claims is also often present, with third parties, such as neighboring states or states with economic interests in the nation of conflict, posing problems for the reparations claim, either by challenging the claim (40.5%) or by opposing all but the most basic, symbolic forms of repair (23%), such as an apology (Category 11). Thus, while third parties can play an important role in facilitating claims-making, with 95% of cases receiving some form of third-party bolstering (Category 10), which most often came from international NGOs rather than nation-states, third parties such as nation-states and corporations are able to exert considerable influence in efforts to obstruct reparations demands.

With respect to the available opportunity structures for repair, once claims receive an official response, it is most often the case that the reparations process is conducted under or directed toward appropriate auspices capable of addressing the demands of the claim (90%, Category 12), which is in most instances the acting government responsible for the territory in which the abuses occurred. Problems arise when it comes to selecting the particular forum in which the reparations claim is to be settled (Category 13), in the processing of the claim through said forum (Category 14), and in determining whether sufficient resources are available to satisfy the terms of the claim (Category 15). For example, in many cases, we found evidence of victim or perpetrator groups being wholly (8.5%) or largely (30%) dissatisfied with the selected forum (Category 13), such as when they felt, in different instances, that the selected forum was not sufficiently independent from the parties accused of perpetrating the atrocity. We also found that various processes that began with some momentum stalled soon thereafter (36%) or received no facilitative support whatsoever (17%, Category 14). Finally, a nation’s lack of material and symbolic resources made it impossible to meet the primary demands of the claim in 13% of the cases, and all but the most minor or symbolic aspects of the claim in a substantial number of other cases (30%, Category 15).

The nature of the opportunity structure was also evident in Categories 16 to 18, which focus on the macro-societal conditions in which the claim is made. In most cases, many of the following social conditions (Category 16) either fully eclipsed (27.5%) or largely overshadowed (30%) the reparations claims process:

The society is still locked in civil conflict;

Key societal institutions (military, government, judiciary) are controlled by the offending party;

The society is in the early stages of recovering from the conflict and has other issues to address (e.g. refugees);

The victim claim is depreciated by opposing interests, giving it less political primacy;

There are legal barriers to the pursuit of justice (e.g. an imposed amnesty).

In contrast, very few societies were defined by discursive conditions (Category 17) in which rationalities of human rights and dealing with the past had not made at least minor inroads to provide a potentially receptive space for transitional justice demands. In only 4% of the cases was silence toward the past brutally imposed through dominant narratives espousing the need to leave the past behind, although in several cases there still existed multiple forms of resistance to dredging up the past, and human rights discourses were either embryonic or newly emergent (53%). Finally, Category 18, Popular Consensus Regarding the Validity of the Claim, sought to tap into survey and other available data to determine how much broad public support existed for responding to the victim group’s demands. In most cases, there was either minor recognition of the validity of the claim with support for only symbolic forms of redress (36%), such as an apology or acknowledgement of the validity of the claim accompanied by the view that the state need not make efforts to initiate or support a reparations process (36%) since, for example, the harm upon which the claim is based happened long ago. In other instances, widespread societal disagreement on the validity of the claim was apparent (19%).

In general, it appears very difficult for a claim to progress through all of these stages and to arrive at a definitive positive resolution (Category 19), as in only 12 of our 47 cases was the reparations claim somewhat (21%) or completely (4%) fulfilled. ¹⁵ This accounted for large-scale dissatisfaction among victim groups with respect to the current state of their reparations claim (Category 20), with the majority reporting some level of moderate to profound dissatisfaction (85%). In contrast, perhaps because so many claims were addressed in only an incomplete or superficial fashion, or were completely ignored by governments, groups accused of perpetrating mass violence or genocide were more likely to react favorably to the current state of the reparations claim (Category 21, 53%). Moreover, because of the lack of definitive resolutions to existing reparations claims, and likely because of widespread victim group dissatisfaction, we could discern few reparative effects resulting from the current status of our sample of reparations claims (Category 22), with 65% of cases showing little to no reparative outcomes from the process. This fact is further reflected in the small group of cases that are judged internationally to represent models of reparative success (Category 23); only 25% of cases were deemed to be at least partially successful.

Keeping with our focus on the obstacles to post-genocide and mass atrocity reparations, we also examined our cases in light of whether they fit the terms of the UNGC. For the 23 cases that could be argued to constitute genocide according to the terms of the UNGC, the same breakdown points arose as those discussed above; namely, reparations cases were blocked or obstructed by perpetrator (Category 9) or third-party group resistance (Category 11) and inappropriate reparative forums (Category 13), as well as by insufficient facilitative processing (Category 14), inadequate reparative resources (Category 15), and lack of popular consensus (Category 18). The same categories also held up as obstacles when we looked solely at those cases that fit an amplified version of the UNGC. Interestingly, for those cases coded under Category 1 as a mass atrocity rather than a genocide, Categories 9 (perpetrator resistance) and 11 (third-party resistance) were magnified, suggesting that these cases provide more opportunities for claim resistance than those that can be formally argued to be genocidal.

While these descriptive frequencies help to pinpoint the likely points of breakdown, they apply to the aggregate sample and tell us little about specific cases. Figure 1 offers a bar chart representation of all of our cases. ¹⁶ While we cannot discuss in any great detail here the characteristics of each specific case, a quick glance illustrates that those cases that achieved some modicum of resolution (i.e. Argentina, Cambodia, Canada, Chile, Germany-Jewish groups, Guatemala, New Zealand, Peru, Rwanda, Sierra Leone, and South Africa) feature fewer dark spaces on the x-axis (numbered 2–18 and arriving at the ‘success’ Categories 19–23). In these cases, the major obstacles to reparative outcomes were either not present, or were overcome along the pathway to repair. In contrast, for those cases that are darkly shaded in Categories 19 through 23, one can observe that the combination of perpetrator resistance (Category 9) and third-party resistance (Category 11) is almost always associated with negative outcomes. The same is true for Categories 13 and 14 (appropriate forum and facilitative processing), as well as for Categories 16 and 17 (social and discursive conditions).

OPEN IN VIEWER

Figure 1.

Visual representation of outcomes for Categories 2–23 among all 47 cases.

Some illustrative examples

To help clarify this complex chart, it is instructive to look at a few individual cases. For example, the Holocaust (Figure 2) offers a case where both perpetrator resistance and third-party resistance had to be overcome, which is different from other successful reparations cases, where there was less evidence of significant resistance. An initial framing challenge was present in the fact that traditionally, reparations were paid to war’s victors, not its victims. Jewish leaders, such as Nahum Goldman of the World Jewish Congress, therefore needed to convince the Allies that if reparation was to be made, it should be made to those who suffered and continued to suffer because of the Nazi’s actions (Woolford and Wolejszo, 2006). Moreover, at the end of World War II, a divided Germany struggled with its own sense of victimization at the hands of the Allies. Aerial bombings, resource shortages, forced evacuations and migrations, among other factors, focused Germans upon their immediate suffering rather than the harms caused by the Nazi regime. These factors resulted in hostility toward both the International Military Tribunal and the 1952 Luxembourg Agreement signed by Konrad Adenauer on behalf of West Germany, which featured a lump sum payment of DM 3 billion to Israel (much of this in the form of goods) and also required West Germany to improve existing compensation and restitution legislation and to make a payment of DM 450 million to be distributed to individual claimants and commemorative projects (Goschler, 2004). Third parties that included powerful nation-states like the USSR also opposed these policies. However, these obstacles were overcome through the well-resourced mobilization of Jewish groups seeking redress, as well as support from other powerful third parties, such as the US government, which maintained a strong economic position during the war, and saw the post-war period as a chance to assert its influence, thereby providing a political opportunity for Jewish claimants (Woolford and Wolejszo, 2006).

OPEN IN VIEWER

Figure 2.

Germany/Jews line graph.

In contrast, other groups making reparations claims, such as the Herero in Namibia (see Figure 3) or the Roma in Europe (see Figure 4), have not been able to surmount perpetrator and third-party resistance to their justice frames. In the case of the Herero, claimants were marginalized within their own country under an Ovambo leadership that was less than sympathetic to Herero demands for repair. This factor, in combination with a lack of knowledge and interest in the Herero from Germany, where political leaders portrayed humanitarian aid disbursements to Namibia as reparations, ¹⁷ left the Herero too isolated to generate momentum for their reparations demands. Similarly, in the case of the Roma (see Figure 4), their status as excluded pariahs throughout much of Europe, as well as their own cultural tendencies toward isolation and disengagement with the formal political world, have made it difficult for them to overcome resistance to their reparations demands and to find or fashion political opportunities for their claims (Margalit, 2002).

OPEN IN VIEWER

Figure 3.

Herero line graph.

OPEN IN VIEWER

Figure 4.

Germany/Roma line graph.

Figure 1 can also be broken down into case subsets for purposes of comparing related reparations claims. For instance, Figure 5 shows reparations claims-making by indigenous groups in English-speaking settler colonial societies. Indigenous groups within settler colonial societies have voiced their demands for reparation since the early days of colonization, although they did not then have the discourses of reparations politics or transitional justice at their disposal for framing purposes. Indigenous groups in Canada, for example, began organizing for the return of their lands and the honoring of treaties soon after implementation of the colonial system of government. A shared feature of all these cases is that they occurred in relatively wealthy, liberal democratic societies. Momentum was gained in each of these cases in the aftermath of World War II and the US Civil Rights movement when a political opening occurred and these nations felt pressure to live up to their liberal democratic ideals and address the claims of indigenous peoples. However, a more recent shift toward neo-liberal policies in each of these nations presents new obstacles to more transformative redress, as opportunities granted for settlements yield only superficial changes intended to preserve the status quo. In Australia and Canada, for example, forms of reparation policy are being designed to assimilate indigenous communities into dominant economic systems (Short, 2003; Woolford, 2005). Factors such as these are incorporated into our coding of the Categories ‘Societal Conditions Reinforcing Claim’ (Category 16) and ‘Discourses Reinforcing Claim’ (Category 17). For all settler societies, we rate each of these cases as a ‘qualified yes’ in terms of there being a modicum of supportive social and discursive conditions within these countries for reparations, although such support is often for an affirmative brand of repair that does not fully recognize indigenous claims to cultural distinctiveness and sovereignty (Short, 2003; Woolford, 2005). For this and other reasons, we see some motion toward resolution in these claims, but little satisfaction among the groups receiving repair (see Category 20). This example cautions that we cannot always assume that reparations policies have unambiguously positive outcomes, since the narrative development of each of the above cases reveals instances in which ‘successful’ repair is accompanied by or leads to new or recurring injustices.

OPEN IN VIEWER

Figure 5.

Visual representation of outcomes for Categories 2 – 23 for the settler societies.

Momentum toward resolution of a claim

The above synthesis highlights especially important moments in the genocide repair process. Our analysis affirms general findings in the transitional justice literature, such as the importance of civil society and non-state actors (Backer, 2003), social and political conditions (Teitel, 2003), and obstructions by internal opponents (De Brito et al., 2001) in impeding or facilitating reparations. Further to this, through the use of our construct of momentum, we now examine how genocide repair, as an artifact of resource mobilization, proceeds across our analytical grid. In so doing, we are able to examine how the mobilization of resources occurs holistically, and how the overall interplay among all categories facilitates or impedes the process of reparations. Our interpretation of momentum in this way encompasses the relevant conceptual dimensions of resource mobilization discussed earlier in this article.

We begin this section by exploring which categories may enhance chances for the successful resolution of a claim (Category 19), by referring to Figure 6 and by employing an odds-ratio probabilistic interpretation. We then use regression models to characterize the momentum that claims achieve as they progress toward possible resolution. In doing so, we directly address three key research questions: (1) Is some kind of reparations-oriented momentum visible for our sample of cases? (2) Can this momentum gain strength and reinforce itself? and (3) Can this momentum falter when there is a recent setback in the reparative process? We ignore Category 1 (Actionable Claim) since it was not interpretable as an opportunity for success or failure, as well as Categories 2 and 3 in the regression models since these were constants in our data. All analysis was carried out in R version 2.15.0 with the help of the packages TraMineR (Gabadinho et al., 2011) and ordinal (Christensen, 2012).

OPEN IN VIEWER

Figure 6.

Average level of success and failure among Categories 2 through 18 in relation to resolved and unresolved claims.

To begin, we take special interest in Category 19 (definitive resolution of a claim) by offering a visual depiction of the different trajectories toward either a positive or a negative outcome for this category. We collapsed each of Categories 2 through 19 into an indicator confirming whether each case was successful for that category (‘Yes’ or ‘Qualified Yes’ = 1) or had failed (‘No’ or ‘Qualified No’ = 0). We then separated into two groups those cases that were successful in Category 19 and those that had failed. Using the numerical representation of success and failure as either 1 or 0, respectively, we next found the average of this number for each of Categories 2 through 18 across the two subsets. We associate each average with a shade ranging between off-white for categories that were successful across all cases, and black to represent 100% failure. Visualized in this way, the average trajectory across the analytical grid for each of the two groups (‘success’ versus ‘failure’ in Category 19) is shown in Figure 6.

In this figure, some categories progress from darker to lighter shades in the change from failure to success for Category 19. Categories 13 (Appropriate Forum for Pursuit of Claim), 14 (Facilitative Processing of the Claim), and 16 (Societal Conditions Reinforcing Claim) show a clear difference, with darker shades appearing for those cases that failed to reach a definitive resolution and lighter shades appearing for those that succeeded. Differences in shading are also visible for Categories 7 (Victim Group Involved in Sustaining Claim), 9 (No Resistance to Claim Within Perpetrator Group), 15 (Available Resources for Satisfaction of Claim), and 17 (Discourses Reinforcing Claim).

To corroborate the patterns visible in Figure 6, we calculated odds ratios (ORs) ¹⁸ for each of Categories 4 through 18 to depict how a ‘success’ in the respective category impacted the actual odds of a ‘success’ in Category 19. Category 13 corresponded with the highest odds ratio at 10.39, which indicates that finding an appropriate forum for adjudication (Category 13) multiplies the odds of an overall successful resolution of a claim (Category 19) more than 10 times. The next strongest odds ratio was calculated for Category 14 (OR = 9.58), followed by Categories 16 (OR = 6.55), 15 (OR = 5.29), 17 (OR = 2.37), 7 (OR = 2.28), and 11 (OR = 2.18). An odds ratio was not possible for Category 9 (No Resistance to Claim Within Perpetrator Group) because it involved a division by zero. A risk ratio ¹⁹ was instead calculated, which, at 5.38, tied a success in this category to more than five times the chances of an overall successful resolution of a claim. These findings reinforce the qualitative examination of the data discussed above, and suggest that those categories that appear most critical to an ultimately successful resolution are by and large those that are closest to outcome Category 19, in turn implying a time-sensitive dimension to the linear progression.

To address the three research questions posed at the beginning of this section, we rearranged the data so that any given set of seven consecutive categories as shown in Table 2 could be used to predict the outcome of the eighth consecutive category. Excluding Categories 1 through 3 from analysis, the models shown in Table 3 predict the outcomes of any of Categories 11 through 19 as the eighth consecutive category based on information about the previous seven categories. This left 9 observations for each of the 47 cases, or 423 total, and no missing data. Each such observation is variously referred to below as an ‘occasion’ or ‘instance’ so as to describe it as a given case’s current status at a given juncture in the reparative process. Up to this point, as shown in Figures 2 to 4, numeric values for each category referred to an increasing degree of failure, culminating with a score of ‘4’ to refer to a coding of ‘No’ or unqualified failure. To properly operationalize, aggregate, and test ‘momentum’ as it occurred across categories, however, in the following analysis, we used reverse-coded versions of these. Codings marked as ‘Yes’, ‘Qualified Yes’, ‘Qualified No’, or ‘No’ were thus instead scored with values of 4, 3, 2, and 1, respectively, to indicate an increasing degree of success rather than failure for each one-unit increase. (As an example, see Table 4 to view these data for the case of Croatia.)

OPEN IN VIEWER

Table 3.

Ordinal logistic mixed-effects regression models testing for the presence of momentum toward genocide repair and its influence on further success.

	Model 1		Model 2		Model 3		Model 4		Model 5
Fixed effects	Odds ratios
Empty model: No fixed effects	n/a
Momentum
Total momentum a			1.16	*
Total in quintiles
Lowest quintile (reference)
2nd quintile					1.66
3rd quintile					1.67
4th quintile					1.82
Highest quintile					3.26	*
Total momentum b							1.07		1.23	***
7th category is a success							0.44
7th category is a success × total momentum b							1.16	*
7th category is a failure									2.27
7th category is a failure × total momentum b									0.87	*
Random effects	Standard deviation
Case ID number	0.992	***	0.539		0.698	*	< 0.01		< 0.01
Category ID number	0.844	***	0.744	***	0.772	***	0.867	***	0.867	***
AIC	1042		1040		1045		1028		1028
−2LL	1034		1030		1029		1014		1014

AIC: Akaike Information Criterion. –2LL: The log-likelihood multiplied by the value of −2.

a

Total momentum is calculated as the simple sum of the first seven of eight consecutive categories (each ranging from 1 to 4 with ‘4’ representing the greatest degree of success) as a predictor of the degree to which the eighth category is successful.

b

Total momentum is calculated as the sum of the first six, rather than the first seven, of eight consecutive categories.

***

p < .001; **p < .01; *p < .05.

OPEN IN VIEWER

Table 4.

Layout of the data used in multilevel logistic regression models: Croatia as a case example.

Case ID number	Category ID number	Coded score a	Previous category b	2nd previous category b	3rd previous category b	4th previous category b	5th previous category b	6th previous category b	7th previous category b
20	11	1	4	1	2	2	2	4	4
20	12	4	1	4	1	2	2	2	4
20	13	3	4	1	4	1	2	2	2
20	14	4	3	4	1	4	1	2	2
20	15	2	4	3	4	1	4	1	2
20	16	3	2	4	3	4	1	4	1
20	17	2	3	2	4	3	4	1	4
20	18	3	2	3	2	4	3	4	1
20	19	2	3	2	3	2	4	3	4

a

This is referred to as the ‘eighth’ category in the text and is the dependent variable in all models shown in Table 3.

b

These are used to create the independent variables used in the models and are discussed in reverse order in the text for the purpose of clarity, for example, referring to the ‘previous’ category as the ‘seventh’ category, the ‘second previous’ as the ‘sixth’, and so on.

The table shows the physical appearance of the data on which the multilevel regression models discussed in Part C are based. Croatia is chosen as an example of a case that, due to the heterogeneity of outcomes among the categories, demonstrates well the characteristic pattern of the data. Whereas Figures 2 to 4 use ordinal values for the categories that correspond with an increasing degree of failure, these scores are reverse-coded so as to represent increasing levels of success instead. The number 4 thus represents confirmation about the claim’s completion of the respective category as ‘Yes’, 3 as a ‘Qualified Yes’, ‘2’ as a ‘Qualified No’, and 1 as ‘No’. The models (Table 3) are also based on data that remove Categories 4 through 10 as observations for the dependent variable and after summing the first seven (‘previous’) of the eight categories as independent variables.

Total momentum was operationalized in the following way. A claim’s total momentum, conceived broadly as traction achieved among contending groups across categories, was defined as the sum total of these ordinal values for the first seven categories. This measure, which was used to answer the first two research questions posed above, had a mean of 16.56, a median of 17, and a standard deviation of 3.07 in our dataset. For the third research question, values for the first six rather than the first seven categories were aggregated to calculate total momentum so that the impact of the seventh could be assessed. This version of total momentum had a mean, median, and standard deviation of 13.90, 14, and 2.68, respectively, and both versions had an approximately normal distribution.

Research question #1: Is some kind of momentum toward the successful resolution of a claim visible for our sample of cases? We addressed this question by asking whether total momentum could be used to predict the outcome of the eighth category. An answer in the affirmative would suggest that total momentum reinforced further success in the reparative process, which would support its validity as a construct. A Welch two-sample t-test was highly significant (t = −7.12; df = 420.43; p < .001) with an average total momentum of 17.58 for those instances when the eighth category was successful and 15.56 for those when it was not.

To more precisely demonstrate the effect of total momentum, we next estimated the typical impact that each unit increase had on the outcome of the eighth category. To do so, we created a cumulative link (ordinal) mixed-effects model using the logit link, total momentum as the independent variable, and the ordinal value of the outcome of the eighth category as the dependent variable. We used case ID number (i.e. the 47 country claims) and another ID number identifying which was the eighth category as the two clustering variables in order to accommodate dependencies across observations. Doing so allows us to interpret the results as if each moment in the genocide repair process were a unique observation. This model tests how well our variable for total momentum predicts further gains in the reparative process, namely, an improvement in the outcome of the eighth category.

In the first (empty) model in Table 3, the standard deviation for Case ID as a random effect is statistically significant, which indicates that some claims had a stronger tendency to positive outcomes than others. The second model includes total momentum as a fixed effect, which corresponds with an odds ratio of 1.16 (p < .05) that ties each unit increase in total momentum to a 16% increase in the odds of improvement in the outcome in the eighth category. Total momentum thus appears to spur further gains in the reparative process, which supports the predictive validity of this construct.

Research question #2: Can this momentum gain strength and reinforce itself? To address this question, we split total momentum into approximate quintiles. Twenty percent of instances had values for total momentum that were as low as 13 or less (‘1st quintile’ as represented in Table 3), whereas the 20% of instances showing the strongest momentum had values of 20 or more. Cutoffs of 14–15, 16–17, and 18–19 in the values for total momentum were used to define those instances with middling scores (the 2nd, 3rd, and 4th quintiles, respectively). Rather than estimating one singular, overall effect of momentum as Model 2 does, characterizing the momentum in this way allows us to compare the distinctive influence of each quintile. Model 3 of Table 3 shows that this variable has an incremental impact. The positive correspondence between total momentum and a successful outcome in the eighth category appears to increase exponentially across quintiles. This suggests that the momentum a claim achieves reinforces itself and gains strength as it grows.

Research Question #3: Can this momentum falter when there is a recent setback? Models 4 and 5 of Table 3 use interaction terms to examine the potential damage of a very recent setback to reparative momentum by comparing the impact of the first six consecutive categories when the seventh is a failure versus when it is a success. For those instances where the seventh category was a failure, Model 4 associates total momentum with only 1.07 times the odds of a more positive outcome in the eighth category. ²⁰ This number compares with an odds ratio of 1.23 (Model 5) for those occasions when the seventh category was a success. Raising these ratios to the power of 16 (the 75th percentile for total momentum) gives overall odds ratios of 3.0 and 27.4, respectively. This difference suggests that even when momentum toward genocide repair is strong, if there is a recent setback, the momentum is likely to falter.

A legacy of previous failure may also lessen the impact of a recent accomplishment. Model 4 associates a success in the seventh category with an odds ratio of 0.44. This figure is similarly interpretable as the extent to which a success in the seventh category increases the odds of success in the eighth category when total momentum among the first six is small. Removing the interaction term from Model 4, however, lifts the same figure to 3.10 (p < .001). Together these results suggest that a string of previous failures totally negates the possible positive impact of a relatively recent success in a claim’s progress toward reparations.

In summary, the prospects for resolution of claims wax or wane as circumstances change along the resource mobilization continuum. While none of the identified obstacles to momentum should surprise readers of the transitional justice literature, Figure 6 and the accompanying discussion above do help to locate momentum-building points on the path to genocide repair. The models additionally help to characterize the sensitivity and nature of that momentum. Results presented in Table 3 support the conclusions that momentum toward genocide repair (1) was evident in our sample of cases, (2) is most forceful when there is a large number of categories having positive outcomes, (3) can be compromised by a recent setback, and (4) is more difficult to achieve after there has already been a legacy of failure in preceding categories. ²¹

Examples illustrating these points can be seen in Figure 1. Supporting the view that some claims experienced momentum toward genocide repair (first research question), Figure 1 shows how cases tended to experience long streaks of either success (‘Yes’ or ‘Qualified Yes’) or failure (‘No’ or ‘Qualified No’) across the categories, rather than moving randomly between the two. Germany, New Zealand, and Guatemala are examples of claims that suffered initial setbacks – this is most strongly felt in the form of resistance from the perpetrator group (Category 9), especially when a repair movement squares off against a well-resourced or powerful perpetrator group. But an accumulation of successes in later categories seems to have facilitated the recovery of the momentum. Other cases illustrate how momentum has a cumulative impact on further prospects for success (second research question). Argentina, Chile, and Sierra Leone showed an especially strong tendency to positive outcomes across the categories, whereas momentum clearly wavered for other claims as they proceeded toward possible resolution. Finally, other examples show how momentum can falter (third research question). Momentum clearly stumbled in Category 16 and never recovered for claims in Bosnia, Chechnya, and Colombia, despite middling success up to that point. In these cases, various mobilization challenges existed in the form of deep societal divisions and imbalances of power, preventing the group from drawing upon its resources to surmount these obstacles. Long streaks of failure also proved to be very damaging for claims in Zimbabwe, Nigeria, and others, indicating that whatever momentum these claims may have had at some point was eventually lost.

From the above discussion of the three research questions, support emerges for the corresponding three hypotheses characterizing movement along the analytical grid: (1) Any accumulation of resources predicts further resource accumulation; (2) sustained increase in resources predicts further resource accumulation exponentially; and (3) failure to sustain resource accumulation negates the impact of prior resource accumulation. These three hypotheses, stemming from our analysis, can be used to examine other instances of movement toward the fulfillment of reparations claims and provide a likely pattern for the traction that is possible as a claim progresses.

Conclusion

In this article, we construct a sociography of genocide reparations grounded in the resource mobilization foundation of social movements. We developed an analytic grid to explain and ideally forecast reparation outcomes among a wide array of mass atrocity events. Our approach relies on identifying and empirically confirming resources related to framing, mobilizing and networking structures, and opportunity structures that hinder or advance momentum along the reparational trajectory. We also offer an operationalized conception of momentum that can be used to characterize the success of reparations as a social movement, that is, as an instance of successfully mobilized resources. Quantitative analysis of reparations claims is an asset to comparative reparations research in that it allows the researcher to identify common challenges to reparation occurring across broadly diverse cases, and to examine how the accumulation of resources contributes to the potential for definitive reparative outcomes. These general findings can then inform more intensive research into the particularities of specific reparations cases, isolating themes that require closer attention when studying or instituting reparations policy.

With respect to the obstacles to reparations success, the data presented above signify that close attention must be paid to how victim groups and their supporters overcome opposition to the claim from perpetrators and their often-powerful third-party allies. Given that reparations claims-making is most often a negotiated, political process, the presence of oppositional elements (Category 9 and Category 11) ideally requires the claimant group to mobilize its own networks and resources (Categories 6, 7, and 10) if their claim is to reach a reparative medium. Even so, new hurdles loom. If the forum is not appropriate to the reparations claim (Category 13), the claimant group is likely to find itself struggling for outcomes not readily available through this means. If the state lacks the political will to facilitate resolution of the claim (Category 14), it has at its disposal numerous ways to forestall the reparations process. If the region does not possess the resources necessary to resolve the claim (Category 15), the claim is likely to encounter official levels of resistance and obstruction. And if the society, at large, in which the claim is being made is not disposed to support the claim (Category 16), and if public discourse about the past and the harms committed is suppressed (Category 17), the claimant group is likely to find that their justice demands do not resonate with wider publics. Our analysis strongly suggests that all of these factors can become serious obstacles to reparations efforts if framing, mobilization and network structures, and opportunity structures are not harnessed to the reparations movement in a timely and effective way. Furthermore, our statistical models suggest that sustaining momentum should be a key goal among claimants as they pursue their case. Although regaining momentum does not appear impossible, both recent failure and a prior history of consistent failure may seriously damage prospects for resolution.

Overall, our aim in this article was to exemplify the potential of a research model for assessing the comparative study of post-genocide and mass atrocity reparations. The data presented here serve to demonstrate the utility of a comparative approach to reparations politics that signals the moments and points wherein particular obstacles may arise upon the path to repair, and that denotes some of the key resources that need to be accumulated in order to sustain a resolution-bound reparations dynamic. Special steps were taken to ensure that our sample of cases is representative of all genocide-related reparations claims that were active after 1945, given that limited resources meant our analysis could not accommodate all cases that potentially fell within our sample frame and our criteria for study. Further research is needed to ensure that our conclusions remain robust for a more exhaustive set of atrocity incidents. Our hope is that the broad scope and suggestive nature of this research effort will stimulate other such innovative investigations in the interest of acquiring vital foreknowledge about the course of transitional justice struggles.