To date, genocide scholarship has principally been a qualitative enterprise. The majority of these efforts entail discussions about how to define the topic of interest and surveys of the individual cases in detailed, but anecdotal, fashion. Confronted with the behavioral anomaly of these events, researchers attempted to describe what had taken place, compelling them first to define what they had seen, then to seek out similar behavior and, finally, to try to identify and assess the effects of various causal factors.

Defining Genocide, Understanding Mass Killing

The definition of what constitutes genocide is quite contentious, both within the academic and policymaking communities (see Chalk and Jonassohn 1990; Scherrer 1999). First, the importance of physical destruction as opposed to other forms of activity has been a concern. Some argue that the endangerment of a group’s ultimate survival is the defining characteristic that separates genocidal acts from other forms of political violence, others highlight that psychological and cultural factors are just as significant (e.g., Harff 1992; Fein 1993).
The second aspect of definition for genocide that is frequently addressed is the perpetrators’ intent. This aspect is particularly problematic as gauging perpetrators’ interests before the fact is notoriously difficult. Some argue that we cannot hope to identify intentions in an unbiased way and that we should focus on behavioral indicators such as body counts [Rummel 1994]) while others argue the opposite (e.g., Lemkin 1944; Dadrian 1975; Chalk and Jonassohn 1987).

Third, there has been some discussion about the identity and nature of the perpetrators. Scherrer (1999, 15) for example, suggests that only state authorities can engage in this form of mass violence. He argues that “modern” genocide refers to “actions carried out by a state or ruler with the intent to kill systematically a particular community or social collectivity, resulting in destroying the targeted group in whole or in part.” Others appear to be open to the possibility that non-state actors could also be involved. For example, Fein (1993, 24) claims that “(g)enocide is sustained purposeful action by a perpetrator (any actor) to physically destroy a collectively directly or indirectly, through interdiction of the biological and social reproduction of group members, sustained regardless of the surrender or lack of threat offered by the victim.”

For our purposes, we adopt a position where the physical elimination of a group with a cohesive identity is the objective of the activity, and that the perpetrators of this behavior involve the state and/or its affiliates or agents. It should be clear, however, that we are less interested with establishing a narrow definition for genocide, which tends to isolate the phenomenon of interest, than we are with understanding and investigating the dynamics of violent conflict processes more broadly.


The Exploration of Cases

While the literature describing individual events and seeking to develop rigorous definitions is quite rich, the number of scholars that study comparative genocide – identifying cases across time and different locales is small (see Kuper 1981; Harff and Gurr 1988; Chalk and Jonnasohn 1990; Fein 1993; Rummel 1994; Licklider 1995; Krain 1997; Schmeidl 1997; Harff forthcoming). Interestingly, these efforts at generating lists of genocide have tended to identify that most cases are similar to one another (Valentino 2001, 8) – suggesting a certain degree of coherence around what is meant by the label. Additionally, these research efforts tend to reach similar conclusions about what causes this behavior: existing models highlight the structural conditions that compel authorities to take action in a fairly simple path-dependent manner, commonly driven by ongoing ethnic tensions. There are even fewer studies engaging in rigorous, empirical investigations of genocide where the data were systematically collected and subsequently analyzed with some form of statistical technique (Fein 1979; 1993), Rummel (1994), Krain (1997), Valentino (2001) and Harff (forthcoming). In most of this work, utilizing the same structuralist framework identified above and using the nation-year as the unit of analysis, individuals try to ascertain why genocide takes place in some countries but not in others. The answers to this question have been fairly straightforward and uncontroversial.

For example, across four of the investigations grounded in a politically driven rationalist perspective, researchers have found that the likelihood of a genocide occurring increased under autocratic regimes and during domestic crises. The crises include external and internal guerilla threat (Valentino 2001), regime change (Harff [forthcoming]), and civil war (Fein 1993, Krain 1997). Additionally, these studies find that governments employ genocidal activity when they are able to exert large degrees of control over their societies and when they do not have to concern themselves with electoral consequences at the ballot box, and when the existing political-economic order comes under threat – a factor that increases the benefits of mass killing. In addition to these results common to all the investigations, there are several unique findings worthy of note. First, trade openness appears to decrease the likelihood of genocide, but only within Harff’s (forthcoming) analysis. Krain (1997) found that ethnic homogeneity was found to have a negative influence on genocidal onset. Third, and last, Rummel (1994) and Krain (1997) found that the degree of autocracy increases the magnitude but not the onset of mass killing.

Although these insights have advanced our thinking a great deal, there are two problems of note with existing quantitative analyses. First, governments do not evenly penetrate the societies within their territorial jurisdiction, particularly within relatively less developed, rural contexts (e.g., Gurr and McClelland 1971; Eckstein and Gurr 1975; Tilly 1990; Finer 1997; Scott 1998) and that domestic crises (like civil war) do not involve equally all parts of the societies in which they occur (e.g., Wickham-Crowley 1991; Kalyvas 1999; 2001a;b). In these cases, we expect that the causal effects of the accelerants and retardants will vary across the nation-state in question. Third, we believe that genocide should not be viewed as a binary event, but as a compilation of diverse actions undertaken by varying types of actors directed against different types of targets throughout the time and locale in question (e.g., Fein 1979; Goldhagen 1996; Kiernan 1996; Ball et al. 1999); simply put, what the appropriate level of aggregation should be ought to be treated as an empirical question. The limitations within existing research are important because they influence our perceptions of genocide and its causes.

In short, for most quantitative research, genocide appears as a singular event, where whole societies are either engaged in relevant behavior or they are not, represented by a dichotomous variable. Our research departs from this perspective, in particular arguing that genocide is best understand when we study it in a disaggregated manner, where some parts of societies engage in violent behavior while others are at all. This has potentially important implications for intervention as well as reconciliation.

A Case of Disaggregation

Only one quantitative study has attempted to explore the variation of activity that exists within a genocide (Fein 1979). This work investigated why certain countries under the Third Reich experienced a large amount of killing (e.g., Poland and Slovakia) while others experienced comparatively little (e.g., Rumania and Bulgaria). Using a variety of information sources, Fein (1979) employed cross-tabulation and regression analysis to show which factors explained the variation in Jewish victimization. In her analysis, Fein identifies that pre-war anti-Semitism, German control, state complicity with German activity, and Jewish segregation from the rest of society increased the magnitude of victimization. The number of victims was decreased by early warning of impending German activity and the degree of mass as well as elite resistance to Jewish persecution. This research represents a major innovation within the statistical analysis of genocide for three reasons.

First, it attempts to directly address the issue of causal mechanisms/processes that the other research was not able to assess. Second, while seemingly moving in the opposite direction that we advocate, for she considers activities that take place within multiple nation-states simultaneously, Fein’s effort well complements other developments within comparative politics that call for the disaggregation of the nation state in general because it is still interested in disaggregating an event, across space and time as we attempt to better understand why things take place (Juan Linz and Amando De Miguel 1966). The work of more recent scholars supports this view as well (Tilly et al.’s 1975; King et al. 1994). The lesson from this body of work is clear: the entities that we study contain significant amounts of variation within them and when we attempt to understand what takes place within these entities, we must be attentive to the variation of behavior that takes place Scott (1998).

Third, and perhaps most relevant to the present study, Fein’s research also well complements work within the area of conflict studies/contentious politics that focuses on events on a smaller scale than genocide. Much of this work argues that conflict behavior varies significantly across time as well as space within one event (e.g., McPhail and Wohlstein 1983), one campaign (e.g., McAdam 1982; Tarrow 1989), and one territorial domain (e.g., Tilly 1995). As McPhail and Wohlstein (1983, 580) note, (m)acro scholars have been concerned with variations in the complexity or severity of gatherings, demonstrations, and riots whereas micro scholars have focused on various individual and collective behaviors within those events and on interaction among the various participants.”

The promise implied by taking a disaggregated approach (province-day versus country-year) is important for us because it suggests that in an effort to better understand genocidal processes, we would need to break the genocides we study down across both space and time from the common country-year level of aggregation. It is very possible that during a period of genocide, individuals engage in varying activities (e.g., kidnapping, rape, murder, torture, or the harboring of potential victims) at different magnitudes and with differing intentions, depending on the particular histories in their specific locales. To understand better the nature of the whole one must assess the parts.