Ethnic Norms and Interethnic Violence: Accounting for Mass Participation in the Rwandan Genocide
| Date | 01 November 2006 |
| DOI | 10.1177/0022343306069290 |
| Published date | 01 November 2006 |
| Author | Ravi Bhavnani |
| Subject Matter | Articles |
651
Introduction
In his work on the 1994 genocide in
Rwanda, Mamdani (2001: 224) relates how
a respondent named Mectilde described
mass Hutu participation in the violence: ‘She
gave me a rough count: ten percent helped;
30 percent were forced to kill; 20 percent
killed reluctantly; 40 percent killed enthusi-
astically.’1The response illustrates that the
Hutu population was not monolithic in its
propensity to engage in Tutsi-directed
© 2006 Journal of Peace Research,
vol. 43, no. 6, 2006, pp. 651–669
Sage Publications (London, Thousand Oaks, CA
and New Delhi) http://jpr.sagepub.com
DOI 10.1177/0022343306069290
Ethnic Norms and Interethnic Violence:
Accounting for Mass Participation in the Rwandan
Genocide*
RAVI BHAVNANI
Department of Political Science, Michigan State University
This article specifies a simple mechanism – a behavioral norm defined in ethnic terms – to understand
the dynamics of mass participation by reluctant Hutu in the 1994 Rwandan genocide. The mechan-
ism, which has not been analyzed systematically in the scholarly literature, is specified using an agent-
based model of within-group interaction that captures the role of intra-Hutu coercion in precipitating
genocidal violence, yet is general enough to be applied to other group conflicts and contexts. The model
highlights the role of individual heterogeneity, within-group punishment, behavioral adaptation, group
networks, and ethnic entrepreneurs, and generates a set of results on norm formation and change, a
number of which are not intuitive. These findings suggest that (1) norms are not equally likely to form
in groups with similar aggregate preference; (2) a violence-promoting norm can emerge in a group
dominated by moderates, and violence is not the inevitable outcome in a group dominated by extrem-
ists; (3) strong punishments are a prerequisite for the emergence of norms that promote interethnic
violence or cooperation; (4) interaction patterns matter – for example, the segregation of an ethnic
group clearly inhibits norm formation; and (5) an ethnic entrepreneur can effectively amplify norm
formation within a group in the absence of strong punishment.
* I am indebted to Robert Axelrod, Rick Riolo, and Jonas
Nart for their comments on this article, as well as to Michael
Findley for research assistance. David Backer, Zeynep
Bulutgil, Pradeep Chhibber, Raymond Duvall, Lynne
Eden, Stathis Kalyvas, David Laitin, Scott Page, Elizabeth
Radziszewski, Michael Ross, Phil Shively, Carl Simon,
Richard Snyder, Scott Straus, Elizabeth Wood, participants
at Stanford University’s Center for International Security and
Cooperation lecture series, and participants at Yale Uni-
versity’s Order, Conflict, and Violence lecture series all
provided valuable suggestions, as did two anonymous
reviewers. This article makes extensive use of the secondary
source literature on Rwanda – Des Forges (1999) and
Mamdani (2001) in particular. The simulation data used in
this article can be found at http://www.prio.no/jpr/datasets.
All faults remain my responsibility alone. Please address cor-
respondence to bhavnani@msu.edu.
1The quote from Mamdani resonates with the distri-
bution of behavior identified in Browning’s (1993: 168)
study of Reserve Police Battalion 101 during the Holo-
caust: ‘a nucleus of increasingly enthusiastic killers who
volunteered for the firing squads and “Jew hunts”; a larger
group of policemen who performed as shooters and ghetto
clearers when assigned but who did not seek opportunities
to kill (and in some cases refrained from killing, contrary
to standing orders, when no one was monitoring their
actions); and a small group (less than 20 percent) of
refusers and evaders.’
violence, in spite of which, hundreds of
thousands of Hutu did participate in the
genocide.2So whereas for Mamdani (2001:
224), ‘It is the 40 percent, those who “killed
enthusiastically,” who represent the real
moral and political dilemma of the Rwandan
genocide’, I am concerned with explaining
why another 50% of Hutus participated,
albeit reluctantly or under duress. What, in
other words, explains mass participation by
reluctant Hutu in violence against Tutsis? A
related puzzle concerns the scale and inten-
sity of violence, which was simply unprece-
dented. Why were previous episodes of
Hutu–Tutsi violence localized and con-
tained, while the episode in 1994 consumed
the entire country?
I argue that the compulsion and resultant
participation of the reluctant in the killing
constituted a dramatic behavior shift, with
related consequences for the scale and inten-
sity of anti-Tutsi violence. As such, the
unprecedented magnitude of violence in
1994 can, to a large extent, be attributed to
the emergence of a violence-promoting
norm among Rwandan Hutu, a norm that
compelled all Hutu – reluctant or otherwise
– to participate in the violence. Simply put,
Hutu who opposed the genocide and were
reluctant to participate in the killing, killed
in the hundreds and thousands because they
were left with no other choice: ‘kill or be
killed’.3If, as I argue, a violence-promoting
norm did emerge in Rwanda and engender
mass participation, this raises a number of
related questions: Why, for instance, do
norms that engender mass participation in
violence emerge within some ethnic groups
and not others? Can these behavioral norms
vary in strength within a given episode of
violence? Can these norms also promote
nonviolent outcomes such as intergroup
cooperation? Is norm formation equally
likely in ethnic groups with similar aggregate
characteristics? It is equally likely with weak
or no punishments?
In this article, I analyze the emergence of
ethnic norms – rules instituted and enforced
within an ethnic group to shape the behavior
of its members toward rivals.4By delineating
behavioral expectations in times of conflict
and cooperation, norms constitute one
mechanism to shape group behavior by
increasing cohesiveness among co-ethnics
and enlarging the set of participants in group
action. Therefore, I do not claim that ethnic
norms are the only mechanism that matters
in episodes of mass violence. Rather, I seek
to determine the conditions under which
expectations of mass participation will
emerge by analyzing the dynamics of within-
group interaction. Using the Rwandan
genocide as a foil, my framework under-
scores (1) the initial proclivity of group
journal of PEACE RESEARCH volume 43 / number 6 / november 2006
652
2While no conclusive estimates of the number of partici-
pants have been established, the highest estimates cited by
Kagame’s former advisor Dusaidi in Gourevitch (1998:
244) suggest that anywhere between 1 and 3 million Hutu
participated in the genocide. Mamdani (2001: 7) and Des
Forges (1999: 2, 260) estimate the figure to be in the
hundreds of thousands (Des Forges also cites a lower figure,
referring to tens of thousands of participants), while
Straus’s (2004a: 172; 2004b: 93) estimate, supported by
empirical evidence, places the number of perpetrators
between 175,000 and 210,000. One reason for this
variance concerns precisely what it means to be a ‘partici-
pant’. Does it require that one was an actual perpetrator
and directly complicit in the violence, or that one helped
and was indirectly complicit? For work that addresses the
question of why ordinary people willingly participate in
violence, see Goldhagen (1996), Petersen (2002), De
Figueiredo & Weingast (1999), Staub (1989), and Waller
(2002). For work that addresses non-participation by
ordinary people – the deliberate choice to not participate
in the killing – see Davenport (2003).
3Prunier (1995) notes that the compulsion to kill clearly
varied across groups and localities in Rwanda, with the
government account of spontaneous participation holding
true in certain cases, and individual- or community-wide
coercion required to elicit participation in other cases.
4Note that ethnic norms, as defined, differ from social
norms (Elster, 1989) in that they are outcome oriented (if
you want to achieve Ythen do X, versus do or don’t do X),
yet resemble social norms in that they apply to oneself and
to others, are sustained by sanctions, and offer some scope
for choice and manipulation. Note also that this definition
differs from cultural or attitudinal norms used to explain
the existence of deeply ingrained animosity towards ethnic
rivals and explicitly excludes any notion of internalization
as a result of the norm itself.
To continue reading
Request your trial