Genocide - University of Essex

Genocide
Assessing its determinants in civil wars 1
Maria Belen Gonzalez2
University of Essex
[email protected]
March 3rd, 2012
(Draft paper: please, do not circulate)
Abstract
Why does genocide occur during civil war? This article examines genocide taking place
within civil war dynamics. Previous literature has focused on unconditional effects of
particular structural factors; however genocide onset follows from a more complex
framework. To better understand what determines the necessary context for genocide
onset in on-going conflicts, I present an explanation of the pre-genocidal context. I
argue that during this period, (i) the pre-existent structural factors combined with (ii)
destabilising events are triggering animosities between groups. The interplay of these
two sets of elements determines the onset of genocide.
Key words: genocide, perpetrator, victimised group, context, deterioration, civil war.
1
The present draft has been written to support the poster presentation of the workshop “Advancing the
Scientific Study of Conflict and Cooperation: Alternative Perspectives from the UK and Japan” Second
Meeting, Colchester UK 20-22 March 2012.
2
I acknowledge the financial support of the ESRC for this project. The present paper is part of a PhD
process that accounts for the onset of genocide in different frameworks.
Page 1 of 12
1. Introduction
Genocide is an extreme violent event that implies the active motivation of a group to
eradicate or considerably reduce the existence of a victimized group. The intentionality
of the violence gives place to radical measures of killing innocent civilians that evoke
the darker side of humankind. Massacres, killings, labour camps, deportations, forced
emigration and starvation are within the policies of extermination that have taken the
life of millions of people along history.
The understanding of genocide as a devastating event which implies profound
human rights violations was conceived after witnessing of the Armenian Genocide
(1915-1917) and the Holocaust (1935-1945). As a result, the United Nations
‘Convention on the Prevention and Punishment of the Crime of Genocide (1948)
(hereafter CPPCG), conceives genocide as a crime punishable whether it is committed
in time of peace or war. The awareness of the implication of one-sided violence against
civilians combined with the concept of human rights violations and civilians’ protection
triggered the study of genocide with particular detail in cases after World War II.
Consistently, most of the studies have been case specific, providing information on the
historical triggers. Based upon these case specific studies and on the onset of genocide
in multiple frameworks, it is possible now to study genocide across countries, trying to
better understand what makes this event take place.
In this article, I assess the onset of genocide during civil war. I present a model on
its occurrence focused on the pre-genocidal situation. This approach is relevant to
identify the factors that determine the necessary context where the mass killing takes
place in on-going wars3. Previous literature has focused on unconditional effects of
structural factors, neglecting the role of other destabilising elements on the onset of
genocide (Harff 2003, Valentino 2004, Krain 1997, Bhavnani 2003). However, in a
framework of on-going civil war, the structural model to explain genocide is correlated
to the causes of conflict, offering limited explanation to why genocide happens in some
wars and not others. I argue that a more accurate model to explain this has to account
for a context where an identifiable set of variables interact simultaneously, generating a
dynamic that exacerbates the likelihood of genocide onset. Hence, it is the aim of this
paper to find what is needed for genocide to happen during civil wars.
To answer the question why genocide occurs during civil war, I analyse 31 episodes
of genocide that happened during 197 civil conflicts between 1955 and 2010 (data form
Harff, 2003 and UCDP 2010, respectively). I take into consideration the structural
models explained by the literature (Harff 2003, Valentino 2004) and I introduce a set of
elements that represent destabilising shocks to the on-going situation, generating
3
Here mass killing is used as synonymous of genocide. There is a debate on the academic community on
whether mass killing is an attribute of genocide or an independent type of violent event. See Valentino
2004, Porter 1982
Page 2 of 12
tipping points towards genocide onset. This results in a dynamic model to account for
the deterioration process that precedes the event of genocide.
In order to present this theory, the paper is structured in four different sections4.
Firstly, I present the theory on genocide occurrence in civil wars. Secondly, the research
designs followed by a discussion of the empirical finding. A conclusion will recapture
the main points presented.
2. Literature Review
Within the types of one-sided violence, genocide is the one that has caused the higher
number of civilian victims (reference). This violent event consists on acts committed to
eliminate or considerable reduce the population of a group of civilians identified with
specific communal characteristics. The intentional nature of genocide is consistent with
its logic; an intentional and systematic plan to kill large quantities of members of the
target population.
Genocide has taken place along history either as a war tactic (i.e. Carthage in 44
BCE, East Asia by Genghis Khan in the XIII century), as a way of clearing up a
territory claimed by a particular nation or civilization (i.e. the Zulu kingdom 1810-1828,
Argentina in 1870), or even in the epic and religious literature to punish those that did
not respect the rules (The Iliad Trojan War, The Bible Moses on Midianites Number
31:1). However, its study has not been specific until the end of the World War II. It is in
this period when the term genocide is coined for the first time by Raphael Lemkin
(1943) in relation to the violent events in Turkey against the Armenian population
between 1915 and 19175. Based upon this, the United Nation on its Convention on the
Prevention and Punishment of the Crime of Genocide (Adopted by the General
Assembly the 9th of December 1948, General Resolution 260) defines genocide as “acts
committed to destroy, in whole or in part, a national, ethnical, racial or religious group”
(Article II). However, this definition is not safe from criticism, the main one being the
exclusion of specific groups in the definition that are identifiable by their political
ideologies (Kuper 1981, Fein 1988, Gurr and Harff 1992, Jones 2004). 6
Literature on genocide has been normative or case specific, being particularly relevant
to understand the historical events or triggers that originated the favourable conditions
for its occurrence (Totten and Parsons 1997, Jones 2006). Even though this type of
analysis is extensively comprehensive with the historical events of each case, it does not
4
The present draft only contains the theory on genocide occurrence in civil wars, descriptive statistics
will be presented in the poster session.
5
The epistemology of the Word genocide is: ‘geno’ Greek term for family, tribe or race, ‘cide’ Latin term
for killing. For further clarification, please see Lemkin 1943.
6
For a detailed discussion on definitions of genocide by different disciplines and a criticism on the groups
excluded from the UN CPPCG, please see Jones 2011.
Page 3 of 12
help to understand the occurrence of genocide as an event that takes places outside a
singular explanation.
Recent comparative studies try to overcome this case-based understanding of
genocide occurrence and examine a number of factors believed to increase the risk of its
onset, including political upheavals, power concentration and civil war (Harff 2003,
Valentino and Huth 2004, Gurr 2000, Krain 1997), state repression (Rummel 1995), and
ethnic fractionalization (Bhavnani 2003). However, these studies have focused on
unconditional effects of a particular explanatory factor, saying little about the
interaction between parties. The complexity of this event taking place cannot only be
explained be the presence of a set of variables that are necessary but not sufficient
conditions for the onset of genocide. This implies that there is a specific context under
which these factors not only are needed conditions but also triggers of mass killing.
The identification of cases of genocide is not exempt of complications (Gurr and Harff
1992). Most of the analysis is post factum, taking a look at the consequences to evaluate
whether a particular event of violence directed to innocent civilians was genocide or
not. Taking into consideration that the information on the numbers of victims can be
underreported, the number of victims of genocide during civil wars in the last fifty five
years rises to the staggering figure of 18 million deaths (Harff 2003)7. This not only
shows the high cost of civilians’ life that genocide has caused during this period, but
also the worrying fact that this type of event takes place more often than what it is
normally conceived (one onset of genocide ever year and a half), bringing about
devastating consequences in the country and region where it takes place. It is for these
reasons that a better understanding of genocide onset will allow us to assess the risk of
different countries to experience genocide, prevent its occurrence, deal with on-going
cases, and create policies to overcome the consequences of it. It is the aim of this paper
to present a cross-case explanation of genocide onset in on-going civil wars.
Building upon the existent literature on genocide within civil wars (Harff 2003,
Valentino and Huth 2004, Gurr 2000 and Krain 1997), I develop a model of genocide
occurrence in contexts of on-going war.8 Particularly relevant is that genocide in these
cases is explained only under structural models. However, they do not take into
consideration the endogeneity of their model with the causes of civil war. In this paper,
I aim to explain what is distinctive for the onset of genocide within civil wars.
7
The estimation of the number of civilians’ deaths is, most of the time, under reported. The lack of
precise information on this issue is reflected on Harff and Gurr (1996) article available at Jongman (Ed.)
(1996). Also, there is a debate on the span of time that should be considered in order to identify an event
as genocide and the type of violence inflicted on the civilian population (direct or direct means of
violence). For the later discussion, see Fein (1993) and Balcells (2011)
8
Genocide has also occurred in periods of peace (understood as the lack of armed confrontation between
two or more parties), where the government is not in a stable regime and lacks the support of part of the
population (twelve out of forty-four cases occurred in such environment). It is part of my PhD Project to
account for these cases.
Page 4 of 12
3. Theory (incomplete – references missed)
Genocide in civil wars
In order to understand genocide occurrence within civil wars, it is essential to fully
comprehend the nature of the event per se. In a context of civil conflict, there is an
armed dispute between two or more organised parties, one of them being the
government. The causes of this type of asymmetrical warfare vary across cases;
however, what they have in common is that once the armed confrontation has started,
each party will maximise its possibilities to defeat its rival. On this dynamic of
continuous fight, the civilian population is caught in the middle, being utilised by the
confronted parties to either gain territory (logistic advantage) or reduce support to the
rival. However, the consequences of this war dynamics are translated into human
atrocities, where the intention is to end the life of a group of the population or to
seriously damage its legacy towards the greater aim of winning the civil war. In such
context, the probabilities of genocide increases.
The relation between war and genocide is implicit. A context of internal armed
confrontation accustoms a society to violence, eroding the boundaries of ethical
behaviour and justifying the use of violent actions against the opponents and the civilian
population. This behaviour is reinforced by the security dilemma of the context, which
facilitates a discourse within groups to enhance rivalry towards the other groups
(intercommunal enmity) (Jones 2006). In combination, this generates a framework that
favours the decision to carry out extreme intentional killing.
Analysing genocide, it is possible to identify two components on the dynamic of the
violence: the perpetrator of the violence and the group towards which the violence is
directed.
Perpetrator – The perpetrator can be any of the organised parties confronted in the
civil war. However, the government has hold this position in the majority of cases of
genocide during wars; thirty out of forty-four from 1955 to 1997 (Valentino, et al.,
2004). Governments have the strategic advantage to access the state’s structure, being
easier to deploy the required policies and infrastructure to carry on violence towards a
selected group. As Valentino (2004) argues, if the government feels threatened by a
particular group, genocide will be the tool to reinforce its position in power. In this
paper, I consider cases of genocide where the government is the perpetrator.
In a context of civil war, the position of the government not necessarily is as strong as
expected; hence a weak infrastructure of the state can make the opposition more likely
to fight tighter for the control of the country. Here, the dominance of strategic positions
favours the advantage of a particular group, and therefore such specific disputes
accentuate the intensity and dynamics of the battles. Since the control of those areas
prevail as a logistic target, the control of the population related with the area is crucial.
The ways of doing it vary, but taking this into consideration the tight dispute and the
trapped-status of the civilian population, the possibility of genocide increases.
Page 5 of 12
Victimized group – They are groups, most of the time minorities,9 which are
identifiable due to its national, ethnical, religious, racial and political position or
actions. The victims, independently of the type of group they belong to, are considered
by the perpetrator as a threat either to the power position the later holds or to the
stability of the country. This can be due to historical reason or latent structural factors as
discussed below.
Signalling the final solution
Once identified the elements that compose the concept of genocide, it is possible to
explain how a situation of tension between perpetrators and victims turns into a onesided massacre. Political violent events are always preceded by a period of increasing
hostilities between the parties; when the violence is one-sided that tension can be staged
or perceived unilaterally. Hence, to account for genocide it is important to explain the
pre-genocidal situation.
In a broader understanding, genocide occurrence is composed by two stages: the pregenocidal period and the actual event of genocide. The former (t1) is when the
conditions under which drastic measures to kill part of the population are intentionally
designed and carried out. It is a dynamic process where the interplay of factors
generates tipping points and increases the odds of genocide onset. After a phase of
cumulative pressure and intentional violent strategies – especially during on-going wars
–, the deterioration of the political context achieves a critical point that triggers the
occurrence of genocide (t2).
Figure 1. General model of genocide onset*
Pre-genocide stage
(t1)
Genocide
(t2)
1. Structural factors
(Transmission and return of
information)
2.
Destabilising events
4. Critical point
3. Deterioration
*This causal mechanism captures the general model for genocide onset. Microfundations of this (power, interest
and beliefs of the actors) are specified below on the cases to support the theory.
9
Minorities group at risk are groups of civilians that have previously suffered policies of inequality,
direct discrimination and terror from the government.
Page 6 of 12
Here, I focus on the pre-genocidal situation (t1) and I argue that a better understanding
of this stage provides a model of the occurrence of genocide across cases. In the pregenocidal period two different conditions are met: (i) structural conditions that favour
the deterioration of the political context towards genocide and (ii) specific destabilising
events ––such as crucial battles or death of a leader – that submerge the already hostile
context into a spiral of increasing animosities from the perpetrator to the victimised
group.
a. Structural conditions (risk factors)
Structural conditions are historical and cultural factors that, if present, make a society
more likely to experience confrontation between groups. These structural factors can be
applied to the occurrence of war as well as one-sided violence (Eck and Hultman,
2007). Literature on genocide has been consistent with the identification and empirical
analysis of the impact of such structural factors on the likelihood of genocide
occurrence (Harff 2003, Valentino and Huth 2004, Gurr 2000 and Krain 1997).
Particularly relevant is the study of Harff (2003) where she identifies three strong
predictors of onset of genocide in cases of state failure; political upheaval, autocracy,
and trade. Data on these particular variables is available from 15 year prior to the
conflict.
The structural factors, already addressed by the literature, reinforce the predisposition
for genocide occurrence (prior genocide, ethnic/ideological fractionalization, regime
type, trade openness, discrimination by the state and instability). Even though they are
not sufficient to explain the event taking place, they are the bases that predispose the
occurrence of genocide. However, there is the necessity to explain why certain conflicts
have genocide and others not. In other words, how do these structural conditions
become not only necessary but sufficient conditions for the occurrence of genocide?10
In an environment of war, the context situation is already violent and renders to keep
deteriorating. The struggle between different parties within the country brings into
discredit the power position of the government. Under these circumstances, the threat
perceived by the perpetrator increases with every confrontation, pushing them to be
more severe on the decision to end the threat. Hence, genocide comes about as a tactic
to end the support to the opposition (Valentino, 2004). Following this, victims can be
associated with the armed group that is actively confronting the government by their
shared ethnicity, nationality, religion or ideologies (i.e. Guatemala 1968 – 1996).
However, there is a distinction to be made: the group of the civilian population
victimized can have a passive role (no–cooperation with the challengers) or an active
one (providing the fighters with support such as information, food, etc.). In this case, if
10
Eck and Hultman (2007: 244) proved that the variables that predicted the incidence of genocide (based
on Harff 2003) are also explaining the magnitude of one-sided violence. Therefore, the mechanism to
explain the incidence of one-sided violence has to be different from the one on genocide.
Page 7 of 12
there is a historical precedent of active support from the civilians to forces against the
government, the government will not hesitate to be dramatic on its measures to win the
war.
H1: previous war-like confrontations between the government and communal groups
located in a strategic area for the on-going conflict increases the probabilities of the
government to carry out intentional mass killings.
H2: the probability of genocide increases when the government is waging war against a
guerrilla group which is related with an ethnical or national civil group..
b. Destabilising events (in progress)
Destabilising events are particular episodes that threat to shift the distribution of power
between the contestant groups of the civil war. Also identified as shocks, the presence
of these events trigger cultural and historical animosities, dragging the context of ongoing conflict into a spiral of deterioration11. The origin of the shocks can be either
endogenous or provoked by one of the parties in the conflict (i.e. death of a leader,
winning a crucial battle, internal displacement of the population), or exogenous to the
situation (i.e. scarcity of resources). The main differences with the structural factor are
(i) that the destabilising events occur once civil war is on-going and (ii) they are s
turning point that makes genocide more likely to happen. I argue that this is exactly
what makes a difference between cases of civil war with genocide from those that do
not experience such one-sided violence.
Accordingly with the causal mechanism explained earlier, conditions that exacerbate
scarcity and competition among groups can increase the risk of onset of genocide.
Hence, the likelihood of genocide depends on the specific combination of factors that
are inter-related, where (i) the effect of each variable may depend on the presence of
other variables, (ii) particular risk factors may serve as substitutes for others and (iii)
combination of factors may exacerbate the effects of the others. Bearing in mind the
connection of structural factors as necessary but not sufficient conditions for genocide
onset, I explain the destabilising events that start a deterioration of the context.
One of the most destabilising factors is the death of a leader. This can occur in both
sides; however the consequences are different. The death of the rebel leader is seen by
the government as an opportunity to reinforce its position in power. Therefore, if the
rebel group lacks popular or civilian support, there is no need for mass killing.
However, if the death of the leader takes place in the government side,
11
As catalysers of mass killing, time between the shock and the occurrence of genocide is important.
Page 8 of 12
H3: the probability of genocide increases when the government’s leader is killed and
the contestant party is related with an ethnical or national civil group.
Since the dispute to control the country is decided in different stages of the on-going
conflict, specific battles can grant the victory to either the government or the contestant.
This is particularly relevant when the fight has been active for a long period of time and
there is no difference of power between the government and the challenger. Once the
battle is won, the government will opt to reaffirm its position, ending with the rivals’
sources of support. Hence, groups related with the contestant by either ideology or
communal characteristics are under risk of being attacked by the government.
H4: the probability of genocide increases when the government’s position is
comparable to the rebels and the rebels are supported actively by the civilian
population.
4. Research Design (incomplete)
In order to account for genocide occurrence I apply rare-logit (King, G. 2000). Given
that the number of cases of genocide in on-going conflict is thirty-one and the purpose
of this research is to find whether the interplay between structural factors and shocks to
the system explain genocide occurrence, it is crucial to take into consideration all the
civil wars between the period 1946-2010, and observe what makes a difference in those
cases that ended in genocide. If I prove that in a context of war the latent structural
factors are triggered by destabilising events, then the theory will offer a more complete
approach to explain genocide occurrence.
Data
The number of armed conflicts comes from UCDP Armed Confrontations 19462010, I take into consideration those cases with civil conflict where the government is
identified as one of the parties (n = 267). UCDP has also data on One-sided violence
(Eck and Hultman, 2007). However, I do not consider using this data at this stage for
three main reasons; (i) its span of time is 1989–2010, (ii) it also codes terrorist attacks
as part of one-sided violence, and (iii) captures low intensity conflicts perpetuated by
rebel groups.
The dependent variable, that is cases of genocide, is from Harff (2003) – 30 cases
during civil conflicts. This dataset is based on COW and MAR, and represents a
systematic effort to gather information on the structural factors related with genocide
occurrence. Updated until 2003, it has been complemented by yearly Risk Assessment
documents until the year 2011 (this document is already published and can be coded
Page 9 of 12
into the dataset). The unit of analysis is country-year, with a total of 42 cases of
genocide between 1955 and 2003. The structural factors are taken from this dataset.
It also has a 1–5 scale on the civilians’ deaths. The lower threshold is 300 and the
higher is +250.000, in a minimum period of six months. The figures do not play a role
on the definition of genocide of politicide for this data.
The shocks to the systems (independent variables) are taken from Keesing's World
News Archives, Banks 2005 and Political Terror Scale 2010. (Selection of data in
progress)
Operationalization
The operationalization of genocide is an on-going debate within the academic
community. In its broader sense genocide refers to the acts that “collectively endanger
the physical life of group members” (Harff, 2003:58). There have been different
attempts to create a precise definition of genocide. On the one hand, stressing its aspect
of eradication or substantial elimination of part of a population, genocide has been
defined in relations to large number of victims such as a minimum of 1 million or 50
thousands deaths (Rummel, 1995; Valentino, 2004, respectively). However, the
victimized group’s size changes in each case, making this definition difficult to apply
when the victimised group is small. On the other hand, definitions focused on the
identifiable characteristic that belongs to the victimised group (Harff, 2003). This
attempt considers cases with a minimum of 300 deaths, which not necessarily captures
the essence of genocide.
A different approach is needed to account for the two main characteristics of
genocide; the intentionality to eradicate a victimized group and the element that
identifies a group as such. To accurately achieve this, I take into consideration the
proportion of the victimized group in relation to the population of a country, and the
proportion of civilians from such group killed. If 25% of the population (one fourth of
the total victimised group) is killed by the government, then the case should be
considered as genocide. (In progress)
Methods (incomplete)
Rare Logit
Conditional frailty event history analysis
5. Discussion
6. Conclusion
Page 10 of 12
Appendix A.
Cases of genocide in civil wars between 1955 and 2010
Country
Estimation of Deaths
Start Year
End Year
Guatemala
1978
1990
46,800 – 96,000
El Salvador
1980
1989
32,600 – 66,000
Chile
1973
1976
12,000 – 24,000
Argentina
1976
1980
8,000 – 16,000
Serbia (Yugoslavia)
1998
1999
9,000 – 18,000
Bosnia-Herzegovina
1992
1995
152,000 – 298,000
Nigeria
1967
1970
72,000 – 644,000
Democratic Republic of Congo
(Zaire)
Uganda
1977
1979
9,300 –19,000
1971
1979
144,000 – 288,000
Uganda
1980
1986
160,000 – 320,000
Burundi
1965
1973
86,200 – 173,000
Rwanda
1994
-
Somalia
1988
1991
19,300 – 39,000
Ethiopia
1976
1979
9,000 – 18,000
Angola
1975
1994
364,000 – 728,000
Angola
1998
2002
80,000 – 160,000
Sudan
1956
1972
294,000 – 588,000
Sudan
1983
2002
1,112,000 - + 2,480,000
Sudan
2003
2010
249,000 – 498,000
Iran
1981
1992
8,100 – 19,300
Iraq
1963
1975
27,600 – 56,900
Iraq
1988
1991
224,000 – 448,000
Syria
1981
1982
24,000 – 48,000
Afghanistan
1978
1992
784,000 – 1,568,000
China
1959
-
32.000 – 64,000
Pakistan
1971
-
+ 256,000
Pakistan
1973
1977
4,600 – 10,000
Myanmar (Burma)
1978
-
4,000 – 8,000
Sri Lanka
1989
1990
17,000 – 33,000
Cambodia
1975
1979
200,000 - +1,000,000
Indonesia
1975
1992
132,200 – 264,000
Page 11 of 12
+256,000
References (incomplete)
Balcells, L (2011). “Continuation of Politics by Two Means: Direct and Indirect
Violence in Civil War.” Journal of Conflict Resolution 55(3): 397‐422.
Balcells, L (2010). “Rivalry and Revenge.Violence against Civilians in Conventional
Civil Wars.” International Studies Quarterly 54 (2): 291‐313.
Eck, K & Hultman, L (2007), “One-Sided Violence Against Civilians in War: Insight
from New Fatality Data” Journal of Peace Research, 44: 233.
Diehl, P F. (2006), “Just a Phase? Integrating Conflict Dynamics Over Time” Conflict
Management and Peace Science, February 1, 201128:41-64
Fearon, J. & Laitin, D. (2002), “Ethnicity, Insurgency and Civil War.” American
Political Science Review. 97(1): 75-90
Gurr, T. R., Harff, B., and Marshall, M. G (2003) State Failure Task Force Report:
Phase
III
Findings.4
August
2003.
Available
at
<
http://www.cidcm.umd.edu/publications/publication.asp?pubType=paper&id=9>
Harff, B (2003) ‘No Lessons Learned from the Holocaust? Assessing Risks of Genocide
and Political Mass Murder since 1955’ American Political Science Review, 97 (1): 5773
Jones, A. (2006), Genocide: A comprehensive Introduction. New York: Routledge.
Krain, M. (1997). ‘State-Sponsored Mass Murder: The Onset and Severity of Genocides
and Politicides’ Journal of Conflict Resolution, 41: 331-359 pp.
Krain, M. (2005) International Intervention and the Severity of Genocides and
Politicides. International Studies Quarterly, 49: 363–387.
Rummel, R. J. (1995) ‘Democracy, Power, Genocide, and Mass Murder’ The Journal of
Conflict Resolution 39 (1): 3-26.
Rummel, R. J. (1994) Death by Government. New Jersey: Transaction Publishers.
Valentino, B. (2004) Mass killing: the final solution. London: Cornel University press
Valentino, B., and Huth, P. (2004) ‘Draining the Sea: Mass Killing and Guerrilla
Warfare’
Page 12 of 12