Accounting for the dynamics of one-sided violence

Abstract

This article presents the Konstanz One-Sided Violence Event Dataset (KOSVED) which allows researchers to study the dynamics of civilian abuse in 17 civil wars. The dataset provides, based on a multitude of sources, detailed information on the number of civilians killed or harmed by government or rebel troops. Where information is available, KOSVED also documents the dates of these events as well as the identities of the perpetrators and the means used in terrorizing the civilian population. The authors argue that the content analysis of news reports offers relatively accurate figures on those events that the perpetrators cannot hide from the public and that receive prominent media attention. Presumably, such information motivates potential short-term retaliatory acts by the group that has been the target of one-sided violence. The analysis suggests that, over the course of a conflict, almost all actors attack unarmed citizens, although to radically different degrees and relying on different means.

Keywords

event data mass killings massacres one-sided violence political violence time series

Introduction

This article summarizes efforts to gather for 17 armed conflicts the number of civilians whom organized actors killed or harmed in episodes of so-called one-sided violence. This form of political violence encompasses a wide range of cruel acts, ranging from the massive harming and killing of non-combatants to genocides but also including violent events on a smaller scale. According to our definition, instances of one-sided violence are lethal or harmful acts that an organized group, which can be either a rebel organization or government actor, directs against unarmed individuals. The aggression results in the immediate physical harming or death of more than one person.¹

To assess the human suffering for specific civil wars, we have created the Konstanz One-Sided Violence Event Dataset (KOSVED). We acknowledge that trying to measure the carnage among civilians for single events is an ambitious endeavour. Previous studies already disagree about the number of civilians or unarmed soldiers who have, during or outside of civil wars, fallen victim to mass killings. Wayman & Tago (2010: 7) show, for example, that Rummel’s (1998) estimate of the yearly number of victims in massive episodes of one-sided violence after World War II often exceeds the corresponding figure by Harff (2003) by a factor ranging between ten and two.

There are several reasons, both excusable and non-excusable, for the widespread tendency to remain vague about the number of killed and harmed civilians. First, the statistics of one-sided violence are most often controversial and, as a consequence, highly politicized. Both the planners and the executioners of these acts share the desire to deflate the figures. If unable to cover up their part in the murdering and maiming, the perpetrators portray their actions as self-defence. This tendency to deny and hide the responsibility for the violence goes hand in hand with the attempt to inflate the suffering of the own group. Second, many acts of one-sided violence take place in remote areas; no population census or register accurately accounts for the people who live in these ‘killing zones’. A third reason for the widespread reluctance to pin down the exact number of victims stems from the monstrosity of these acts that many people believe to be unimaginable. A cynical illustration of this originates from interwar writer Kurt Tucholsky (1931[1925]: 148, own translation) who lets a fictitious French diplomat cynically say: ‘War? I cannot find this to be so atrocious. The death of one human being: this is a catastrophe. One hundred thousand dead: this is a statistic.’² Slovic (2007) develops in this context the notion of ‘psychic numbing’ and maintains that experiencing many sufferers arouses less empathy than being confronted with a single or few victims.

Fortunately, the goal of accounting for the dynamics of one-sided violence is not as far-fetched as it was some years ago since our cumulative knowledge on the origins of civilian abuse in and outside of war is growing at a fast pace (e.g. see Valentino, 2004; Kalyvas, 2006; Downes, 2006; Schneider, Banholzer & Haer, 2011 for an overview). Especially Eck & Hultman (2007) provided the research community with an invaluable dataset that allows us to analyze the onset of this form of political violence cross-sectionally. This resource on one-sided violence documents 573,000 fatalities from 1989 to 2008. Half a million of these killings are attributed to the government-sponsored massacres in Rwanda in 1994 and revenge acts committed by the rebel troops. The Uppsala Conflict Data Program Georeferenced Events Dataset (UCDP GED) that provides disaggregated data on conflict events including casualty figures of battle-related and one-sided violence for Africa from 1989 to 2010 (Sundberg, Lindgren & Padskocimaite, 2010) is a recent follow-up to Eck & Hultman (2007).

In this article, we add to these efforts a data resource that details one-sided violence by the individual event. KOSVED, which we introduce in this article, distinguishes between specific acts of civilian abuse and the perpetrators and targets of this form of political violence. The data can be aggregated temporarily and, if the information on the location of the events is comprehensive enough, also spatially. We present the dataset in detail and discuss in the light of recent controversies over war fatality estimates (c.f. Burnham et al., 2006; Johnson et al., 2008; Obermeyer, Murray & Gakidou, 2008; Spagat et al., 2009) the reliability and validity of our figures. To this end, we compare our estimates with those reported in Eck & Hultman (2007) and UCDP GED (Sundberg, Lindgren & Padskocimaite, 2010). The article concludes with some descriptive evidence and a discussion of the relevance of dynamic event datasets.

The empirical basis of the study of one-sided violence

The occurrence of one-sided violence during war, and the startling number of civilian casualties that results from it, are puzzling for several reasons. Most explanations resort to an explicit or implicit rational choice model. According to this adage, the perpetrators conduct at least the massive acts intentionally. One-sided violence is therefore either a consequence of the strategic calculus of the military leaders or of the organizational structure of the perpetrating armed units. There is considerable empirical support for both explanations which complement rather than contradict each other (Schneider, Banholzer & Haer, 2011). Valentino, Huth & Balch-Lindsay (2004) show that government forces are more likely to target civilians when military tactics are inefficient against a guerilla movement that relies on a hit and run strategy. In other conflicts, some armed units rely on this strategy after a victorious battle with the aim to annihilate future opponents (Schneider, Bussmann & Ruhe, 2012). Kalyvas (2006) argues that the government or the rebels use ‘selective’ violence as a deterrent against possible defectors in territories under partial control and ‘indiscriminate’ violence in a territory largely controlled by the opponent. Humphreys & Weinstein (2006) demonstrate that military units in which the soldiers act largely on their own present a greater danger to the civilian population than more disciplined organizations. Weinstein (2007) similarly argues that insurgencies that attract fighters because of the exploitation of a resource become more violent than conflicts in which the soldiers are primarily recruited for non-material motives. In the Democratic Republic of Congo, some armed units have relied on the hierarchy of armed units to coerce soldiers into massive acts of violence against civilians (Schneider, Banholzer & Albarracin, forthcoming).

The quantitative evidence that is presented in support of the theoretical advances in the literature on one-sided violence is largely cross-sectional. While Valentino and his co-authors compare mass killings with a magnitude of at least 50,000 dead non-combatants, Humphreys & Weinstein (2006) rely on interviews. Other researchers employ single country evidence to support their claims. A prominent example of the case study approach to one-sided violence is Kalyvas’s (2006) examination of the Greek civil war.

If we want to study how large scale massacres but also less deadly instances of one-sided violence evolve over time and across a number of armed conflicts, we can primarily resort to the UCDP One-Sided Violence dataset, which provides yearly counts of the number of unarmed victims killed by government and rebel troops (Eck & Hultman, 2007). Although the coding relies on single events, the dataset was presented until the release of UCDP GED at the yearly level of aggregation. This resource includes events with more than 25 civilian victims and clearly identifies the actor who is responsible for the killings.³

If we believe that not all instances of one-sided violence are as centrally planned as the Holocaust or other genocides, we should analyze the individual occurrences of this form of political violence separately. In other words, KOSVED considers an act of one-sided violence as an event if there is sufficient evidence for it to be based on an autonomous decision by a perpetrator to kill or harm civilians. Temporal or spatial aggregation of the data is one way to alleviate the statistical conundrum that the events included in the dataset are not completely independent of each other. This problem does not, however, require the researcher to move to the highly aggregated units of analysis typically used in macro-quantitative research.

Obviously, the ambition to offer more conflict-specific information has the disadvantage of limited comparability. It allows the researcher nevertheless to study more precisely how the civilian victimization has evolved over time and space within individual conflicts. KOSVED, one in the growing number of event datasets in conflict research (Schrodt, 2012), currently offers this service for 17 conflicts. It is a partial complement to the Armed Conflict Location and Events Dataset (ACLED) (Raleigh et al., 2010), which especially adds battlefield information to the analysis of those conflicts covered by both datasets. Combining both sources in a study on the Bosnian civil war, we were able to show how military battles and civilian victimization evolve and co-determine each other (e.g. Schneider, Bussmann & Ruhe, 2012).

KOSVED relies in a similar way to ACLED and UCDP GED on the content analysis of a variety of press reports to estimate the number of victims, the identity of the perpetrators and targets, the location of an event and the means used to kill and harm civilians. Although news reports are often the only information available for certain conflicts (see Bocquier & Maupeu, 2005), the usage of journalistic sources is not innocuous. Davenport & Ball (2002: 440–441) analyze the differences in the number of killed people in the Guatemalan civil war by comparing journalistic reports, documentary evidence and interview-based information: ‘In the number of killings reported in each year, not only is the scale of reported violations different by a factor of nearly 50, the pattern reported in the press is completely different from the patterns reported in NGO and interview sources’.

Surveys might at first sight be a better strategy to estimate the number of people killed and harmed in war. As the controversy over the estimates of casualties for the war in Iraq shows, this alternative to written sources does, however, not necessarily yield more accurate numbers. Johnson et al. (2008) demonstrate that the usage of cluster sampling is not appropriate in Iraqi cities like Baghdad and that the initial estimates by Burnham et al. (2006) were inflated. Similar controversies have emerged over the survey-based epidemiological estimate by Coghlan et al. (2006) that up to four million people perished as a direct or indirect consequence of the civil war in the Democratic Republic of Congo or the early calculations for the Kosovo conflict (Spiegel & Salama, 2000). The mortality figures provided by Degomme & Guha-Sapir (2010) for the conflict in Darfur strongly suggest that a multitude of surveys are necessary to gauge exact fatality estimates. Furthermore, Obermeyer, Murray & Gakidou (2008) criticize the fatality estimates of Lacina & Gleditsch (2005) who resort like KOSVED to archival evidence and news reports. However, Spagat et al. (2009) show in return that Obermeyer, Murray & Gakidou (2008) use questionable aggregation rules to create a survey based assessment of the global number of war fatalities with which they try to counteract claims of a declining fatality rate in recent years.

We would like to stress that the strategic analysis of one-sided violence justifies the reliance on press reports as the main source of information. The models require data that are available at short notice and inform about decisions on one-sided violence. By relying on news reports we are sure to include events that the leadership of the conflict parties and the international community clearly notices.

Dataset construction

KOSVED defines, as indicated, one-sided violence as violent acts that an organized group directs against non-combatants and that result in the immediate physical harming or death of more than one individual. Although this conceptualization closely follows Eck & Hultman (2007), we lower the threshold of fatalities to more than one person killed or harmed. Furthermore, unlike Eck & Hultman, we did not integrate intent into our definition because many reports will not allow us to draw conclusions on whether the perpetrators attacked civilians intentionally. However, we differentiate between non-combatants as primary or secondary targets whenever such information was available. The latter category refers to acts during which the perpetrators directed the violence against a representative of the opponent’s side and accepted the possibility that these acts led to the killing or harming of civilians. Concentrating on whether one-sided violence takes place within internal wars, our dataset only includes instances of one-sided violence that were perpetrated in the two years before the onset of a UCDP/PRIO civil war (Gleditsch et al., 2002), during the armed conflict or in the two years after its termination, but does not assess violence taking place outside of civil war situations. Although one-sided violence does not necessarily occur exclusively in the context of an armed conflict, Eck & Hultman (2007) have shown that only 1% of the total fatalities of one-sided violence took place in countries that did not experience a war at that time.

The 17 civil wars covered in KOSVED (see Table I) are not a representative sample. We combined several criteria in the selection of these armed conflicts, most notably whether the conflict was, at the start of our research project, also planned to be included in ACLED which offers complementary information on the conflicts. A further benchmark for case selection was the presence of an internal or an internationalized armed conflict as defined by the UCDP/PRIO dataset on armed conflict (Gleditsch et al., 2002). As Africa experienced many conflicts after the end of the Cold War, in the main period of investigation for the KOSVED project, a majority of the wars included occurred in the sub-Saharan part of this continent. However, we also specifically refer to conflicts from different geographic regions, and we tried to vary the magnitude in which the included wars experienced one-sided violence.

Coders relied on sources like African Contemporary Report or Africa Confidential to become acquainted with the African wars and checked NGO reports for further information. To identify and code instances of one-sided violence, we selected several sources included in LexisNexis. The coders consulted for each conflict at least the New York Times and BBC Monitoring. For African conflicts, we relied additionally on Africa News and for all other conflicts on Associated Press Worldstream. Coders also screened several other sources, such as Global News Wire, Agence France Press and United Press International for some conflicts. Besides LexisNexis, we consulted Keesing’s Contemporary Archives for selected conflicts.⁴ Taking into account the varying reliability of news reports, KOSVED provides the minimum, median and maximum number of victims in a reported manifestation of one-sided violence. What distinguished KOSVED from other data projects is that we not only coded the numbers of civilians that were killed but, if the information was available, also the number of civilians that were harmed during the acts of one-sided violence. Unlike similar datasets, KOSVED furthermore distinguishes among the methods of killing (primitive weapons, shooting, bombing, suicide attack, tank attacks, aircraft attacks or weapons of mass destruction) whenever the respective information was available. However, for the large majority of events, the type of weapons used was not reported and thus had to be coded in the category ‘unclear’. For example in the conflict in Chad, we have reports that ‘primitive weapons’ killed 113 people, that 128 were shot, and that 5 died during air raids, while 1,684 fell under the residual category. Similarly, we gathered information on characteristics of the targeted civilians (e.g. age, gender) and the perpetrators but these variables also largely contain missing observations as most press reports do not provide detailed information. Note that the lack of detail also characterizes other data gathering efforts such as the Iraq Body Count (Hicks et al., 2010).

We report for each event whether the incidence or the number of victims is contested and whether the report is based on a primary or secondary source of information. As a matter of lucidity we also post on our website a file with just the most basic information on the minimum number of civilians killed for each conflict per week or per month by each conflict actor. The time series listed on the project homepage refer to the most conservative estimate; we provide this information at the weekly and

Table I.

The civil wars covered by the Konstanz One-Sided Violence Event Dataset and the main perpetrators

Civil war (starting and ending month)	Number of events (weeks)	Aggregate min. (max.) number of killed victims	Min. killed per event (per week)	Most extreme killing (min./max.)	Min. number of wounded aggregate (Ratio harmed/killed)	Main perpetrators (attribute. killings)^a	Perpetration ratio most violent groups
Angola (2.1994–12.2003)	205 (513)	4,323 (4,976)	21.0 (8.4)	200/260	1,744 (0.4)	Unita^b (3,062) Government^b (199)	15.4
Angola–Cabinda (5.1996–7.1999)	10 (165)	139 (139)	13.9 (0.8)	57/57	17 (0.1)	FFA (74) Rebels^b(8)	9,3
Azerbaijan (1.1990–3.1995)	436 (272)	3,286 (4,566)	7.5 (12.1)	200/1,000	4,236 (1.3)	Armenian (819) Azerbaijani (596)	1.4
Bosnia-Herzegovina (3.1992–10.1995)	421 (186)	12,765 (18,949)	64.3 (146.6)	2,000/6,000	9093 (0.7)	Serbs (12,056) Bosniaks^c (2,092)	5.8
Chad (1.2004–1.2008)	60 (214)	1,580 (1,659)	26.3 (7.4)	300/300	678 (0.4)	Janjaweed^b (1,038) Government (33)	31.5
Chechnya^f (6.1993–8.2004)	126 (586)	2,680 (4,901)	21.3 (4.6)	384/700	3,355 (1.3)	Government (1,667) Rebels^b (346)	4.8
Congo-Brazzaville (5.1995–5.1999)	59 (213)	1,186 (1,230)	20.1 (5.6)	250/250	79 (0.07)	Ninja (252) Cobra (40)	6.4
Congo-Kinsasha (4.1995–12.2002)	461 (403)	54,946 (66,959)	119.2 (136.3)	13,000/13,000	36,244 (0.7)	Government (3,348) AFDL (2,550)	–^e
Côte d’Ivoire (9.2004–11.2004)	73 (114)	1,058 (3361)	14.5 (9.3)	133/1,000	1,705 (1.6)	Government^b (450) FANCI (124)	3.6
Croatia (9.1990–10.1995)	63 (266)	1,052 (1,168)	16.7 (4.0)	120/120	485 (0.5)	Serbs (709) Croats (324)^d	2.2
Liberia I (12.1989–12.1996)	93 (364)	3,732 (7,483)	40.1 (10.3)	1,000/3,000	9,352 (2.5)	AFL (1,191) NPFL (279) / LPC (403)	–^e
Liberia II (6.2001–12.2003)	38 (129)	786 (1,963)	20.7 (6,1)	200/1000	1006 (1.3)	MODEL (104) LURD (43)	–^e
Macedonia (3.2001–8.2001)	22 (23)	106 (117)	4.8 (4.6)	60/60	348 (3.3)	State (101) Macedonian AF (98)	1.0
Niger (5.1990–7.1998)	16 (429)	142 (165)	8.9 (0.3)	63/63	37 (0.3)	Government (83) Rebels^b (56)	1.5
Sudan (6.1985–4.2008)	169 (1,196)	6,149 (6,276)	36.4 (5.1)	1,000/1,000	862 (0.1)	SPLA (2,662) LRA (1,652)	1.6
Sudan-Darfur (6.2003–10.2008)	158 (277)	3,069 (3,342)	19.4 (11.1)	200/200	1,657 (0.5)	Government (1,798) Janjaweed (562)	–^e
Uganda (01.2000–12.2006)	454 (363)	3,383 (4,088)	7.5 (11.3)	200/337	1,395 (0.4)	LRA (1,934) Government (992)	1.9

Estimates always correspond to the most conservative count.

^a The sum may not correspond to the number in the previous column due to events that are not attributable to a particular group.

^b The numbers also refer to killings perpetrated by coalition partners.

^c Modal numbers of Serb killings for Sebrenica massacre (6,000) instead of minimal number (2,000). The Croats are left out; 1,210 killings are attributed to this side.

^d Including acts perpetrated by Croatian Muslims.

^e Not reported because of large number of violent actors and shifting coalitions.

^fFor Chechen territory.

monthly level of aggregation and document which actors were part of a particular coalition. In this aggregation, for example, all events that are not clearly attributed to a specific perpetrator are eliminated, together with a variety of other information that is deleted from the aggregated files. Researchers are advised to consult the more comprehensive data files. Further details on the coding criteria are described in the web appendix and the KOSVED codebook.

We compared the KOSVED data to other data sources, in particular to the UCDP one-sided violence data (Eck & Hultman, 2007) and for the conflicts in Africa to the UCDP GED data version 1.1 (Sundberg, Lindgren & Padskocimaite, 2010). It turned out that a systematic comparison is difficult for various reasons starting with different time periods covered but also partly different actors that were coded. In the online appendix we present the correlations between KOSVED and UCDP GED for the total number of civilian deaths per week or per month, together with a discussion of more specific discrepancies between the datasets.

Whereas for most conflicts there is quite a high analogy in the weeks without casualties, the agreements in the number of civilians killed per week are sometimes quite low during periods with one-sided violence. These discrepancies alert the user of KOSVED and similar research endeavours to carefully evaluate the sources and to compare in case that there are double gathering efforts. Such caution is especially warranted for cases in which the coding rules were similar, but the correspondence in the events covered and the number of reported victims is low. In any case we recommend that researchers consult the original datasets before they conduct any statistical study and try to compare KOSVED with another dataset if possible.

The usage of one-sided violence across 17 civil wars

The KOSVED dataset includes 2,850 incidents for 17 conflicts. Of the events covered, the number of people killed according to the minimal (maximal) estimation is 103,916 (130,178) individuals or 35 (44) civilians in the average event. This underlines the importance of accounting for civilian and not only battle-related deaths in studies of armed conflict.⁵ Extremely violent instances of one-sided violence with more than 1,000 dead took place twice in Bosnia and four times in the Democratic Republic of Congo. Such events are rare; the median number of killed individuals is five (six) in the minimal (maximal) estimation. The number of civilians harmed is not systematically smaller than the number of killed civilians reported. In Chad, for example, we have a minimum of 1,580 killed civilians and 678 reported injuries, whereas in Ivory Coast, there are a minimum of 1,058 killed civilians but 1,705 harmed civilians. We assume though that the reports on killings are more reliable and less severely underreported than reports on injuries and thus concentrate on the reported dead civilians in the following. To some extent, the fatality numbers depend on the ‘technology’ that the perpetrator relied on in their attacks. Figure 1 summarizes this information.

The percentage of cases where we could not attribute the killings to a particular type of weapon differs by conflict; for Bosnia, this ratio amounts to 75%, and for DR Congo it is 95%, where a considerable number of killings were perpetrated with primitive weapons, accounting for the high average number of victims for this type of event.

Of all KOSVED events, 70% of the incidents are reported in one source only, 15% in two sources and the rest of the events in three or more. While the average number of reports per event amounts to 1.9, the corresponding figure is extremely low for Chad for which 59 of the 60 reports refer to one source only. Obviously, the paucity of information for such a conflict partly reflects the one-sidedness of the media industry in the developed world, which often does not report regularly on conflicts in what it considers peripheral. It is reasonable to suspect that the government side mainly profits from the underreporting of certain events as it often controls the access of international media to the conflict zone.

The level of precision also varies with regard to whether or not the civilians were the primary target of an incident. According to KOSVED, this was the case for 1,281 events (45%), while unarmed civilians were what we call ‘secondary targets’ in 360 cases (13%); we do not have sufficiently precise information for the other events. Information on the identity of the perpetrators is available for 83% of the events; the Lord’s Resistance Army – with 277 events – is the organization which relied most frequently on one-sided violence. The dataset includes 52 cases involving sexual violence.

Table I provides an overview of one-sided violence in the different conflicts covered by KOSVED. Obviously,

Figure 1.

Number of events by type of weapon

our estimates are often lower than those that rely on additional sources and that were assembled after a war. A case in point is the Bosnian civil war for which the Sarajevo-based Research and Documentation Centre (RDC) established a death toll of over 97,000 direct casualties including civilian and military victims because of the war (Nettelfield, 2010: 165).⁶

Table I strongly suggests that the frequency and intensity of one-sided violence differ greatly among the conflicts covered. While the rebel forces are the main perpetrators in many conflicts, the Russian government seems overall to be the main culprit in the war in Chechnya. The ratio between the number of civilians killed by the most violent group and the corresponding figure for the second most violent group differs widely; it ranges between 31.5 (Chad) and 1.0 (Macedonia). However, these figures crucially depend on the number of actors involved, the coalitions that are formed and that are broken, and the level of precision about the individual acts of violence. We have for these reasons not calculated this reciprocation ratio for a conflict like the multilateral war in the DR Congo with its shifting coalitions.

The wide range of the reciprocation ratio also suggests that dynamics of one-sided violence vary between the conflicts. We depict in Figures 2 and 3 the ups and downs in the total number of killed people for the entire civil war in Azerbaijan and with a differentiation for the main actors for Bosnia.

The conflict in Azerbaijan saw widespread human rights violations. It is, however, quite typical for this conflict that a few extreme mass killings overshadow a large number of much smaller events. Figure 2 also shows that time periods of intensive one-side violence are followed by less violent phases. This suggests, in line with the growing literature cited partly above, that the civilian suffering is largely a consequence of the strategic calculations by the military and political leaderships. Figure 3 reveals that extreme events like the massacre of Srebrenica are relatively rare and that many low-level instances of civilian victimization are much more typical than the large-scale massacres that make it into the international headlines. It is, in light of this evidence, not surprising that the distribution in the intensity of the killings in the Bosnian civil war followed a power law (Scharpf et al., 2013).

Conclusion

The emerging literature on one-sided violence indicates that at least the more severe acts are planned and that this form of political violence often has a strategic nature. If we want to test such arguments, we either need to resort to archival evidence, as Kalyvas (2006) does, or to dynamic data. This article has introduced a data resource which follows the second logic and allows the researcher to test time series and longitudinal arguments on the ups and downs of one-sided violence within the context of a civil war. It will help us address questions of why some conflict parties resort to violence against civilians quite extensively at a certain point in time while they refrain from using this instrument in other periods.

The Konstanz One-Sided Violence Events Dataset (KOSVED) covers 17 civil wars so far. We are aware of the limitations that media sources have for the precise estimation of the civilian suffering in conflict zones, but would like to reiterate that such information is often the only type that is available in the short term. Even if other data sources, for instance based on surveys or census figures, are made available in a timely fashion, event data projects like KOSVED supplement such alternative information-gathering methods. In our view, the policy community should have a strong interest in dynamic information on human suffering as it can potentially be used to forecast the risk of a particularly severe act of one-sided violence. If decisionmakers resort, as we hope, to real-time forecasting resources to predict and prevent the worst from occurring, we would have moved far beyond the interwar cynicism traced back to Tucholsky’s (1931[1925]) fictitious diplomat who equated the killing of a hundred thousand with a mere statistic.

Figure 2.

One-sided violence in Azerbaijan (minimal number of reported victims)

Figure 3.

Bosnia with actor groups (mode of reported victims)

Footnotes

Replication data

The datasets described in this article can be found at http://www.prio.no/jpr/datasets and on the KOSVED webpage: .

Funding

We would like to thank the German Foundation for Peace Research, the European Commission, and the University of Konstanz for financial support.

Acknowledgments

We thank Jule Krüger, the JPR editor and three referees for helpful comments.

Notes

References

Bocquier

Phillippe

Maupeu

Hervé

(2005) Analysing low intensity conflict in Africa using press reports. European Journal of Population 21 (2–3): 321–345.

Burnham

Gilbert

Lafta

Riyadh

Doocy

Shannon

Roberts

Les

(2006) Mortality after the 2003 invasion of Iraq: A cross-sectional cluster sample survey. Lancet 368(9545): 1421–1428.

Coghlan

Benjamin

Brennan

Richard J

Ngoy

Pascal

Dofara

David

Otto

Brad

Clements

Mark

Stewart

Tony

(2006) Mortality in the Democratic Republic of Congo: A nationwide survey. Lancet 367(9504): 44–51.

Davenport

Christian

Ball

Patrick

(2002) Views to a kill: Exploring the implications of source selection in the case of Guatemalan state terror, 1977–1996. Journal of Conflict Resolution 46(3): 427–450.

Degomme

Olivier

Guha-Sapir

Debrati

(2010) Patterns of mortality rates in Darfur conflict. Lancet 375(9711): 294–300.

Downes

Alexander B

(2006) Desperate times, desperate measures: The causes of civilian victimization in war. International Security 30(4): 152–195.

Eck

Kristine

Hultman

Lisa

(2007) One-sided violence against civilians in war: Insights from new fatality data. Journal of Peace Research 44(2): 233–246.

Gleditsch

Nils Petter

Wallensteen

Peter

Eriksson

Mikael

Sollenberg

Margareta

Strand

Håvard

(2002) Armed Conflict 1946–2001: A new dataset. Journal of Peace Research 39(5): 615–637.

Greenhill

Kelly M

(2010) Counting the cost: The politics of numbers in armed conflict. In: Andreas

Peter

Greenhill

Kelly M

(eds) Sex, Drugs, and Body Counts: The Politics of Numbers in Global Crime and Conflict. Ithaca, NY & London: Cornell University Press, 127–158.

10.

Harff

Barbara

(2003) No lessons learned from the Holocaust? Assessing risks of genocide and political mass murder since 1955. American Political Science Review 97(1): 57–73.

11.

Hicks

Madelyn Hsiao-Rei

Dardagan

Hamit

Serdán

Gabriela Guerrero

Bagnall

Peter M

Sloboda

John A

Spagat

Michael

(2011) Violent death of Iraqi civilians, 2003–2008: Analysis by perpetrator, weapon, time, and location. PLoS Medecine 8(2): e1000415.

12.

Humphreys

Macartan

Weinstein

Jeremy M

(2006) Handling and manhandling civilians in civil wars: Determinants of the strategies of warring factions. American Political Science Review 100(3): 429–447.

13.

Johnson

Neil F

Spagat

Michael

Gourley

Sean

Onnela

Jukka-Pekka

Reinert

Gesine

(2008) Bias in epidemiological studies of conflict mortality. Journal of Peace Research 45(5): 653–663.

14.

Kalyvas

Stathis N

(2006) The Logic of Violence in Civil Wars. Cambridge: Cambridge University Press.

15.

Lacina

Bethany

Gleditsch

Nils Petter

(2005) Monitoring trends in global combat: A new dataset of battle deaths. European Journal of Population 21 (2–3): 145–165.

16.

Nettelfield

Lara J

(2010) Research and repercussions of death tolls: The case of the Bosnian Book of the Dead. In: Andreas

Peter

Greenhill

Kelly M

(eds) Sex, Drugs, and Body Counts: The Politics of Numbers in Global Crime and Conflict. Ithaca, NY & London: Cornell University Press, 159–187.

17.

Obermeyer

Ziasd

Murray

Christopher J

Gakidou

Emmanuela

(2008) Fifty years of violent war deaths from Vietnam to Bosnia: Analysis of data from the World Health Survey Programme. British Medical Journal 336: 1482–1486.

18.

Raleigh

Clionadh

Linke

Andrew

Hegre

Håvard

Carlsen

Joachim

(2010) Introducing ACLED: An Armed Conflict Location and Event Dataset. Journal of Peace Research 47(5): 651–660.

19.

Rummel

Rudolph J

(1998) Statistics of Democide: Genocide and Mass Murder since 1900. Münster: Lit.

20.

Scharpf

Adam

Schneider

Gerald

Nöh

Anna

Clauset

Aaron

(2013) Die Blutspur des Vetos: Ex Ante-Prognose zur Gefahr von extremen Massakern in Syrien [The blood trail of the veto: Forecasting the risk of extreme massacres in Syria). Zeitschrift für Friedens- und Konfliktforschung 2(1): 6–31.

21.

Schneider

Gerald

Banholzer

Lilli

Albarracin

Laura

(forthcoming) Ordered rape: A principal agent analysis of wartime sexual violence in the DR Congo. Violence against Women.

22.

Schneider

Gerald

Banholzer

Lilli

Haer

Roos

(2011) Cain’s choice: Causes of one-sided violence against civilians. In: Jakobsen

Tor Georg

(ed.) War: An Introduction to Theories and Research on Warfare and Collective Violence. New York: Nova Science, 57–82.

23.

Schneider

Gerald

Bussmann

Margit

Ruhe

Constantin

(2012) The dynamics of mass killings: Testing time-series models of one-sided violence in Bosnia. International Interactions 38(4): 443–461.

24.

Schrodt

Philip A

(2012) Precedents, progress and prospects in political event data. International Interactions 38(4): 546–569.

25.

Slovic

Paul

(2007) If I look at the mass I will never act: Psychic numbing and genocide. Judgment and Decision Making 2(2): 79–95.

26.

Spagat

Michael

Mack

Andrew

Cooper

Tara

Kreutz

Joakim

(2009). Estimating war deaths: An arena of contestation. Journal of Conflict Resolution 53(6): 934–950.

27.

Spiegel

Paul B

Salama

Peter

(2000) War and mortality in Kosovo, 1998–99: An epidemiological testimony. Lancet 355(9222): 2204–2209.

28.

Sundberg

Ralph

Lindgren

Mathilda

Padskocimaite

Ausra

(2010) UCDP GED codebook version 1.0-2011. Department of Peace and Conflict Research, Uppsala University.

29.

Tucholsky

Kurt

(1931) Lerne lachen ohne zu weinen. Berlin: Ernst Rowohlt [Learn to Laugh without Crying. Originally published on 23 August 1925, under the pseudonym ‘Peter Panter’ as ‘Französische Witze (I)’ in Vossische Zeitung].

30.

Valentino

Benjamin A

(2004) Final Solutions: Mass Killing and Genocide in the 20th Century. Ithaca, NY: Cornell University Press.

31.

Valentino

Benjamin

Huth

Paul

Balch-Lindsay

Dylan

(2004) Draining the sea: Mass killing and guerrilla warfare. International Organization 58(2): 375–407.

32.

Wayman

Frank W

Tago

Atsushi

(2010) Explaining the onset of mass killing, 1949–87. Journal of Peace Research 47(1): 3–13.

33.

Weinstein

Jeremy M

(2007) Inside Rebellion: The Politics of Insurgent Violence. Cambridge: Cambridge University Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB