Building and Using Datasets on Armed Conflicts (Nato Science for Peace and Security) (Nato Science for Peace and Security)

BUILDING AND USING DATASETS ON ARMED CONFLICTS NATO Science for Peace and Security Series This Series presents the res...

Author: Mayeul Kauffmann

146 downloads 909 Views 3MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

BUILDING AND USING DATASETS ON ARMED CONFLICTS

NATO Science for Peace and Security Series This Series presents the results of scientific meetings supported under the NATO Programme: Science for Peace and Security (SPS). The NATO SPS Programme supports meetings in the following Key Priority areas: (1) Defence Against Terrorism; (2) Countering other Threats to Security and (3) NATO, Partner and Mediterranean Dialogue Country Priorities. The types of meeting supported are generally “Advanced Study Institutes” and “Advanced Research Workshops”. The NATO SPS Series collects together the results of these meetings. The meetings are co-organized by scientists from NATO countries and scientists from NATO’s “Partner” or “Mediterranean Dialogue” countries. The observations and recommendations made at the meetings, as well as the contents of the volumes in the Series, reflect those of participants and contributors only; they should not necessarily be regarded as reflecting NATO views or policy. Advanced Study Institutes (ASI) are high-level tutorial courses to convey the latest developments in a subject to an advanced-level audience. Advanced Research Workshops (ARW) are expert meetings where an intense but informal exchange of views at the frontiers of a subject aims at identifying directions for future action. Following a transformation of the programme in 2006 the Series has been re-named and reorganised. Recent volumes on topics not related to security, which result from meetings supported under the programme earlier, may be found in the NATO Science Series. The Series is published by IOS Press, Amsterdam, and Springer Science and Business Media, Dordrecht, in conjunction with the NATO Public Diplomacy Division. Sub-Series A. B. C. D. E.

Chemistry and Biology Physics and Biophysics Environmental Security Information and Communication Security Human and Societal Dynamics

Springer Science and Business Media Springer Science and Business Media Springer Science and Business Media IOS Press IOS Press

http://www.nato.int/science http://www.springer.com http://www.iospress.nl

Sub-Series E: Human and Societal Dynamics – Vol. 36

ISSN 1874-6276

Building and Using Datasets on Armed Conflicts

Edited by

Mayeul Kauffmann École de la paix, Grenoble, France

Amsterdam • Berlin • Oxford • Tokyo • Washington, DC Published in cooperation with NATO Public Diplomacy Division

Proceedings of the NATO Advanced Research Workshop on Indicators and Databases for Risk Prevention Grenoble, France 21–23 June 2007

© 2008 IOS Press. All rights reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without prior written permission from the publisher. ISBN 978-1-58603-847-2 Library of Congress Control Number: 2008922611 Publisher IOS Press Nieuwe Hemweg 6B 1013 BG Amsterdam Netherlands fax: +31 20 687 0019 e-mail: [email protected] Distributor in the UK and Ireland Gazelle Books Services Ltd. White Cross Mills Hightown Lancaster LA1 4XS United Kingdom fax: +44 1524 63232 e-mail: [email protected]

Distributor in the USA and Canada IOS Press, Inc. 4502 Rachael Manor Drive Fairfax, VA 22032 USA fax: +1 703 323 3668 e-mail: [email protected]

LEGAL NOTICE The publisher is not responsible for the use which might be made of the following information. PRINTED IN THE NETHERLANDS

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

v

Preface Richard PÉTRIS Director, École de la paix, Grenoble The so-called “hard” or “exact” sciences, with their necessary emphasis on technology and on the technical, are hardly reputed for being very human, and, conversely, the socalled “human” sciences are often pronounced as “soft” because they cannot be based on the certainties associated with the former. The search for truth which is the essential dimension of the construction of a peaceful world therefore has to navigate between considerations of a philosophical nature and the concrete data of the hard sciences. If, ever since the humanism of the Renaissance period, we have been happy to lay claim to the wisdom of one of its great writers, Rabelais, who taught a moral lesson to the young Pantagruel with the neat formula “science without conscience is the ruin of the soul”, we nonetheless stand in awe before modern scientific advances and the extraordinary achievements that they have opened up. If everything is not permissible, at least everything seems possible! It was such an outlook which no doubt had something to do with our decision taken ten years ago here in Grenoble to give concrete shape to a utopian dream by creating in our city a “school for peace”. At that time we were reminded that a Brazilian bishop, Don Helder Camara, famous for the stands which he took in favour of human development and the promotion of peace through dialogue and a more just distribution of resources, had once remarked that Grenoble would be an ideal place to combine its advanced technological and computing research with the philosophical and humanist dimension to bring about the necessary synthesis for the promotion of the cause of peace in the world. And he had made this comment at the height of the cold war, as a way of confronting and fighting against the extremism that posed such a threat at that time. Since then, the Third World War has not happened and if one can pose the question of why the North Atlantic Treaty Organisation should be maintained, an organisation which without a doubt contributed to the prevention of such a war, we should not ignore the threats and challenges which remain, or are even greater today, in the general field of world security. This is why we were very gratified when we realised that an Alliance, which some felt was the product of a bygone age, was reorienting its scientific programme in order to “favour research on questions linked to security and by so doing to better reflect the new environment in which NATO finds itself”. And that in order to face the complex questions associated with security, there would be “a greater interaction between the exact sciences and the human and social sciences”. By submitting a project to hold an Advanced Research Workshop on the theme of “indicators and databases for risk prevention”, we gave ourselves the chance to deepen and prolong research which we were already engaged in, based on the idea that a climate or a culture of peace needed to be encouraged and that there were effective ways in which progress could be made in this domain. We believe that this area of study can be subjected to measurement and quanti-

vi

fication. We are now equipped with measuring instruments for natural phenomena, and in a world of increasing complexity, we are more and more concerned with questions of efficiency, of performance and evaluation in all domains, including those which are related to questions concerning the quality of life and the avoidance of social conflict. Comments on war are now inevitably accompanied by sets of “figures” concerning the loss of human lives and the “collateral damage”—all sorts of damage which come in its wake. To the sinister “body count” of open and full war, we must add the information concerning the horrors of all other forms of violence and ignored wars, if our intention is to take the real measure of human tragedies which they entail in order to awaken public conscience and encourage appropriate political action. But we must also today take into account the figures and statistics of new fields of research and action, concerning the environment and sustainable development, if there is going to be real progress on global prevention. The human development indicators of the United Nations Development Programme (UNDP) seem to us to show the way forward in this respect. Furthermore, we support the recommendation of the mathematician, historian and philosopher, Michel Serres, when he says “Since we control the whole world, we must learn to control our own control mechanisms” and we also share the belief of Martin Hirsch, president of the humanitarian organisation Emmaüs, who, just before taking up his French governmental appointment, declared on the subject of poverty indicators: “They are indispensable… To try to do without them would be like tackling road safety, without using devices for measuring and registering speed!” We were therefore clearly aware that all sorts of research is being undertaken in this vast domain and the time had come to bring it all together and submit it to further empirical evidence. All the actors concerned, the politicians, diplomats, the military, the NGOs, as well as the heads of enterprise, need a clearer vision of the state of a world which is more and more complex. The improvement of our information systems is therefore particularly useful, both for pure research and for its applications. The sharing of means and techniques of information, the development of “synthetic” coordinated indicators and even the identification of new data which will help with risk analysis and conflict prevention in tomorrow’s world, all these constituted the agenda for this Advanced Research Workshop. Our approach was to give as much freedom as possible to specialists in presenting their work, so that new ways forward and approaches could be identified and facilitated. The aim of this publication is to render account as faithfully as possible of the diversity of approaches taken by the 27 researchers and experts from 13 countries who participated in this Advanced Research Workshop. It will of course take time to evaluate the potential for progress in all this but these two figures point the way forward.

vii

Foreword Ranan D. KUPERMAN University of Haifa 1. Why Are Datasets Needed? The simplest reason why data is collected and stored, thereby creating a dataset, is to aggregate this information in order to measure performance over time [1–3]. A slightly more sophisticated use of a dataset is to try and explore the relationship between different measures, thereby discovering casual effects. Finally a third application of a dataset is using it as a source of reference. In this case, the user of the dataset seeks to identify specific cases for further analysis according to some criteria that have been classified by the creator of the database. The use of datasets in order to study the development of peace and conflict started in the 1940s with Richardson’s work on deadly quarrels [4] and Wright’s work on the causes of war [5]. In this case the tendency to create datasets could be attributed to the behavioral revolution in the social sciences during the 20th century, which encouraged large sample comparisons and quantitative forms of analysis. However, even after the demise of the behavioral movement, the tendency to collect and accumulate data seems to have only increased. This is most probably attributable to the expansion of low cost computerized systems. Thus, the creation of electronic datasets related to conflict and cooperation has become a relatively simple enterprise, and many organizations consistently engage in various forms of data collection related to conflicts.

2. Why Should Datasets Be Shared? Another advantage of the significant developments in computer technology is that the ability to transfer data has become relatively cheap and fast. Thus, there is a high demand to share datasets and thereby save time and resources duplicating previous work. However, sharing data has other objectives as well. It allows researchers to compare alternative collections of data [6,7], which in turn can indicate the reliability or validity of data collections. Large discrepancies between datasets coding similar events, indicates that at least one of them has been poorly coded. Frequently these variations are due to different interpretations of the raw data (in its original format before it was coded) or relying on different sources. In these cases, it is possible to fill in missing data and thus improve the validity of the dataset.

3. Improving the Validity of Datasets It seems that the demand for security related datasets is growing. This trend will most probably be accompanied with a demand to strengthen the quality of the data. In order to improve the validity of the datasets two procedures are suggested.

viii

The first relates to the quality of the raw material that is used in order to produce the data. Rather than relying solely on the reports of major newspapers or news agencies, data should be sought from sources that have better access to information, such as archived material and foreign language reports, and possibly even through direct contacts with local observers [8]. The second procedure relates to the integration of information from alternative sources. Standards regarding the validity of the sources will be required. Such a measure is especially necessary when there is a conflict between information from different sources. Many researchers circumvent this issue by coding both interpretations rather than making a decision in favor of one source. This allows the user of the dataset to decide which interpretation to accept. However, frequently datasets do not display information about the sources of the raw data and therefore the user has no criteria that can be used in order to decide which interpretation to prefer. It is thus suggested that datasets include full information about their sources. An alternative possibility is that the coders of the dataset determine themselves which source should be preferred. In this case, the coders must provide a detailed account of the criteria that was applied in order to accept or reject information (for an example see [9]).

4. Improving Our Ability to Share Datasets A second set of improvements is required in order to encourage the sharing of datasets. Essentially this requires the creators of datasets to think about establishing universal coding procedures. Because there are currently no agreed upon standards for classifying and categorizing data, comparing data between alternative datasets for the purpose of cross referencing or reducing missing data is not reliable [10]. Unless the coding procedures are identical, attempting to combine or compare data between alternatives sets may lead to considerable errors. This problem could partially be overcome if instead of coding for very specific types of variables, variables would be disaggregated into a number of subtypes. This approach was in fact adopted during the early coding of events datasets such as the Conflict and Peace Data Bank (COPDAB) or World Event/Interaction Survey (WEIS) and in a similar manner there have been attempts to combine the two datasets together [11]. However, because it is very difficult to maintain funding for generating general cooperation and conflict datasets, such a breakdown of variables into subcategories is not encouraged. Even when the group funding the dataset has a specific agenda, breaking down the variables into sub-categories should not consist of significant extra work and therefore should be encouraged. In addition, it is advisable to consider adding to datasets a copy of the original textual source. This can also be used in order to account for coding differences between alternative datasets. Besides dealing with coding procedures another important aspect of sharing data is accessibility. Many datasets cannot be easily downloaded from the Internet and the owners of the datasets are frequently reluctant to share their data. In recent years a number of scientific journals have started demanding as a precondition for publication that the authors make their data available. It is hereby suggested that also organizations providing grants for the creation of datasets adopt this policy. Yet even if datasets are available, researchers must know that the dataset exists and how to obtain it. Although there are a number of Internet sites that provide this information, such as at PRIO or ICPSR, there is preference to direct researchers to datasets that were compiled by aca-

ix

demics or governmental agencies. However, in reality there are many more datasets that are compiled by various NGOs and think tanks that can also be used. Thus, a more comprehensive index of alternative datasets is required. In addition to the description of what type of data is stored within each dataset, dataset indexes should also include information regarding the validity and reliability of these datasets, meaning what were the sources that were relied upon in order to code events, and what if at all were the coding procedures.

5. Establishing the Foundations of a Paradigm To a great extent the creation and maintenance of datasets can be viewed as a profession. Many organizations usually recruit a person who will be responsible for designing, coding, and maintaining their databases. Frequently this person is chosen on the basis of technical capabilities rather than research experience. In order to overcome this person’s lack of research skills a senior researcher is also assigned to oversee the maintenance of the dataset. However, it is very common that tensions start emerging between the managers of the dataset. Such a scenario will usually occur if the senior researcher fails to actually analyze and negotiate the raw data. Thus, persons compiling the data are confronted with many dilemmas regarding how to code the data because the classifications proposed by the senior researchers seem to be irrelevant to the actual information that is being reviewed. The coders might even fail to compile important data when the senior researcher is not even aware of its existence. Overcoming these problems requires the accumulation of experience. Yet it seems that these issues remain hidden and unrecorded. Another example where experience is not redistributed is the discovery of errors within existing datasets. This information remains out of sight and restricted to personal knowledge and therefore, many errors that researchers find in their datasets are not reported to other researchers who use the same datasets. Unfortunately it seems that each time a new research project that relies on the use of data analysis is launched the research team is destined to repeat the mistakes of previous researchers. In this respect, data analysts in the field of conflict and cooperation must pay more attention to the special and unique problems associated with the coding of datasets. There seems to be an overemphasis on developing statistical methodological issues, while very little concern about how the data that is used in these procedures is produced. Therefore, dataset managers should establish a forum for exchanging professional information with each other. Such a forum could serve as the basis for standardizing this field, and thereby a new generation of dataset managers will be able to build on their predecessors experiences and possibly devote more energy to solving problems and dilemmas that still remain.

References [1] Caplow, Theodore (1976). How to Run Any Organization. New York: Holt, Rinehart and Winston. [2] Rossi, Robert J. and Kevin J. Gilmartin (1980). The Handbook of Social Indicators: Sources, Characteristics, and Analysis. New York: Garland STPM Press. [3] Taylor, Charles Lewis [ed.] (1980), Indicator Systems for Political, Economic, and Social Analysis. Cambridge, MA: Oelgeschlager, Gunn & Hain.

x

[4] Richardson, Lewis F. (1948). “Variation of the Frequency of Fatal Quarrels with Magnitude”, Journal of the American Statistical Association, 244: 523-546. [5] Wright, Quincy (1944 [1942]), A Study of War. Chicago: University of Chicago Press. [6] Hoggard, Gary D. (1975). “An Analysis of ‘Real’ Data: Reflection on the Uses and Validity of International Interaction Data.” In Theory and Practice of Events Research: Studies on Inter-Nation Actions and Interactions. Edited by E.E. Azar, J.D. Ben-Dak and P.M. Burgess, pp. 19-27. New York: Gordon and Breach Science Publishers. [7] Reuveny, Rafael and Heejoon Kang (1996). “International Conflict and Cooperation: Splicing COPDAB and WEIS Series.” International Studies Quarterly, 40(2): 281-305. [8] SwissPeace, “Early Warning, About”. Available at http://www.swisspeace.ch/typo3/en/peace-conflictresearch/early-warning/about/index.html, accessed on 17 December 2007. [9] Kuperman, Ranan D. (2001). “Rules of Military Retaliation and their Practice by the State of Israel,” International Interactions, 25(3): 1-36. [10] McClelland, Charles A. (1983). “Let the User Beware.” International Studies Quarterly, 27(2): 169177. [11] Ward, Michael D. and Sheen Rajmaira (1992). “Reciprocity and Norms in U.S.-Soviet Foreign Policy”, The Journal of Conflict Resolution, 36(2): 342-368.

xi

Contents Preface Richard Pétris Foreword Ranan D. Kuperman Introduction Mayeul Kauffmann

v vii

1

The Purposes of Conflict Dataset Building The Creation and Use of Data: Scientific Requirements and Political Utility Wolf-Dieter Eberwein

13

An Overview and Typology of Conflict Data: The Advantages of Data Diversity Kristine Eck

29

Databases on Defense for Influencing State Policies Jean-Marie Collin

41

Concept Definition and Data Building Calling a Conflict a Conflict: Violence and Other Aspects of War Joakim Kreutz

49

Databases on Terrorism: Constructing a Database on Hostage-Barricade Terrorism and Abductions Galit M. Ben-Israel

63

Reviewing CIDCM Databases: Modeling and Indicator Development in Risk Assessments and Early Warnings Lyubov Mincheva

75

Indexes for Measures of Conflict and Cooperation Ranan D. Kuperman

97

Opening up Conflict Data Enhancing Openness and Reliability in Conflict Dataset Creation Mayeul Kauffmann

107

Merging and Meshing Data: Difficulties, Lessons, and Suggestions D. Scott Bennett

133

The ACLED Data Project: Purpose and Methods Clionadh Raleigh and Joachim Carlsen

161

xii

Visualization of Conflict Networks Ulrik Brandes and Jürgen Lerner

169

Author Index

189

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

1

! " # ! # ! $ # ## # # !

# %& % # '(% ) ) % ! (% ) ) # ' * ( * (# + , -&# . % / 0112! & % ! % & & + & ## & # + # % ! # & ! % # % ! # 3 + 5 -i 6 7 8,% . % & ### # ( ! 6 &#, (9 ,)! # ( !

# +%: ! #

# + ! & #% ;
# + ! # &7% % # # 3 + > 2 ?AB > EF111 . % > ! !G HEE 718I2JJEF! +GHEE718I2JFI0!$ G !&K +!#!

2

M. Kauffmann / Introduction

8! # ! ## # # ! " 7 ; # < ; % <8 %: % # % # !- % % !5 # # # #! ! ! " % # ! # # ! # 7 8#

% % : ! * % % ! & # & ! #( LM % ! 7 8% % % % + ! % % # 7 8 7 %

% # 8! 5 % % # !* # % ! ;< # ( % ( ( - N #!$ ; #<

% # G; % #%: O- & % # # %: # " ! #%: ! # + % # + G # # % # ! # % # P

!<7L0M LEM!Q221P # 8!- N # LIM % &;
% # ! ; <% % % ; <; < !( % # #

M. Kauffmann / Introduction

3

(((7( #( ( 8 ! A % # % % # " # !'

% % # ! 0 % ! - 7 % 8!' %

% : ! #$ $ # !$ ; #<%: # % & !% ; <;
% # P : # # # ! -) $% # % % # # !* $& # % % # # P # :# ! % -) $% !5 " # !* & %

% & !

+ / A # .' . ! + & % #!/& !5 # % % !. !N R ;
7 8 ! * + % # % # % % % % ! 6 % !* % 0*; : 51 # QS< %% % 7 # & 8% ## # ! : 7 : 85 1% 5 !

4

M. Kauffmann / Introduction

# %: # %

# & #!,)! % % # % ## # # % ! ! % % # # # # % ! ' " # )!*N R G # # % % # # % % # P # # # % !A , #/ A A6$))(: 7 A 6 $ )8 + G # # # ## # 7 8

## # # % % # % ! % &N /T# 6 # # + # # &% # % # % # + % & ## " % ! %&'' '' & ' & ) # # # V V &

# & ! # P % & # ! !# ## # # # %% % ! # # % ! + #% QW # % $& #LQM!QW QW ; < ! " &

& % !,# 01 & ;< &%; < # # ! : "

! # G: ; <; <; <; <; + <;
M. Kauffmann / Introduction

5

( % ! + # # # 7 # 8 ! # % #! % " !# QW 7 % + 8! ;111 < % % 111 111 # ! # % % 7 8! % % 7 # # % 8!' 01 0Q E07 8 011; < !!!* # % # # % ! - %

! !"!# $ % % ! # 7 + G O8 7 !#! # 8! ! % # % %

( % R # + EP ( % % !) 7# 8 % # % # # % ! !%!# & * % !# ! !%!"!# $ ' # # ! % % E;' % # # #

% # ( #

% !<7LJM!28!

6

M. Kauffmann / Introduction

& % & ! % %& #G % %% % + ! #% # % # & # %% % & ! !%!%!($ ' % + # ! ) # G % ! ; < ; <% ! # G # # # # & %

# # # % % & # ! * % # ;& #
P % % # # !Q # # # # " !( # % # % & # # % % % !* & % # ; <;% <; #
M. Kauffmann / Introduction

7

# 0 + %

7 %

8 7 # # 8! % ! # ! #7%8 % % NA ! % #78 % # % ! I # 7 # 8 % # G 1110111E111 & 7 # 1110111 & 00111E111 E % 8! E # 7 #% #8 % # GQ11111 & ! %

# !J

(' A "7 %8 #7 # %8 7"N A8 J # # +!' % # + 7 # 8

! # % # 0 7 8 # ,## + # # #% G ! " # ! $ %#&%'( ( ') )))&)))*)))+))) '),)) )))+)))

A # %

! # % # - ./% - . 27L!!!M'01- ./L!!!M8 # 0 !

8

I # 7 #X8

E # 7 #%X8

%

# 7X8 7X%8

M. Kauffmann / Introduction

(' # ) # # % % 7 ! % #8 % % 7 ! # #8G + "

5 # # 0 # ;%< % % #; #< % % % ! + % : # # # # % -) $% $& G % 7$&R # 8 % & #% 7$% R # & # " % 8!

! % % ## # % !, % % % % # ## # G + % # ## # # !$ % ! ! %% % + # !5 " ## # # !-

% % &# # # ! * % &!

: " % #

! # # % ! " # # % #! ! % # % #

M. Kauffmann / Introduction

9

% % % % % 7 # 8 ! % % % ! ) LM ( !* + WEI! % % WQW$* ( # , # ! L0M N #(!-!$* ) , W02! LEM N ,!. (!/!)!7 !8$, ( ( WW! LIM N : !A!;' A $
This page intentionally left blank

The Purposes of Conflict Dataset Building

This page intentionally left blank

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

13

! " " # $ % &"" '! " (#&( ( ') "#' ! " $% %') " ' % " "* %% # & ( +% % ( %# $ %,% ( # #

*% & % # " % & $ " # "* ( % & % & " % - * $# % , % % " * &, " * , "#, ," & # ", # & *. & # , % & " ( %# ( # * ", # ", ## *

&( ( # $ / % *** " ( & % 0* " # (%% * 2 #& + # % 345 67* % %% , +% ( % ( % , "&

% ( % * & &" # # & * & +(&% & (# (( ( & % ( % + # #3897* : ; ; # # , ( # %% * %% (% % # " "" & / 0*# ( & ( , % ( # & # % + " & *

14

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

% & # % % &*# ( " + ,% % " * % % (&( *) #% & % & % , # , " /0 < / &= #0 >4* " # &, " & # * %% " & ( # & # % &,& ( $*$ ( 5 ( % (% & " ," && & "* %%" # (& # * & & % % # % * #(" & % ( * (% % (& % # ( " $ " (*2 & & % ## * $% ," % # % %% * " & % * # & ( #( #( ( " " % & * 2 % " & *

/&# # 0 % # % & ( " $ " (*#( " (& % # & , ( #& +%, "# *

%" %# # % * % " % % , # % # ( , % ( - % * &" % * & %% (% % *?%% ( % , * & % # *% ( " $ " (*)+% % ; &; % # % $ " ( $ " ( +%*% % * 2 , # % % % % 5 , #%% ( # (%% 5 ( %% %% % *

4

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

15

"% * % % && # & *. +% &

) & " ( ( 3@7* ) & # # ( #% % , * &( & & % % ( % ( +%

&#&% *% % * % ( # % % , % % " " ( ( &= # * &= # " * (* % &= # # " (," & % & ( ( % % ( %& *% &= # % # ( " % (* &= # & % / (0* ( % " / (0* ( /A0 < / (0>" %

*)&& ( ( " /B %% $ " ( # & % +% 0 <3C7, %*64D>* . +% )&: +% &= #( #& % % & " % < ( % % % % % >* /A0 & * "#, $ (&" #* & # & % # % % * % " +% $* #( +%

# * ,% & " ( ( - % ( & % +% " % * , **% $, # %

%% ( " ( & *#" , % ( +% % " (% * " ,& " $ % #& % - * ! " ) E ( ( F (: ",/" && ,&

* &

"- 0<3D7,%*+ >*)- (" (#

% %% % # * % % % - " ( %% % ( * &= # % % %* % $ & % &&

16

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

" ( & % * & # # ( & ( E3G7 "( * . % & % ( & * % % , $" #%% , & #& % ( % % * # # *+% &E (*387 397 (

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

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

17

& #*%% % < ( , , ( - % "K> (# " #& &34K7*!( " ( % & ( # &*! #( #% "* %(& & % " & (% * % - % * ( % " % &* % " " & ( % %% #* . %% =," = , & % % % * & # # ( ( & " ( * , & (, ( &$ " # % *& % (< #> "" *# % " # %% , " * " # &" %& - $ (% $, , " L ( & # " (& #< (> % <& > * L, % % " = & * " ," % +% , & # & % # % * :( ( % / 0 # ( % % <( , ( #, (, *>* ?E 3497% ( %% , " & $ & & % 9* # / 0 % # # ( (

( % * % " " " ( # ( *2 , & % * ) , " &"% "" "! (34C7 " # " # #* ) " + % ( * " % ( # & (( <34D54G7>* &"" =$ <34G7,%%*868K> &% "" $ "$" % % ( #

%* & " # % * # " " %% # & 5 % " # & & ( % * K 2 % " = % ", 34675 "+% & (3487 #* 9 M % ,! ?N$34@7 ( # "% *

18

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

( % , "+ #( " % ' , " % % ( # % ' #( < ( >% * %# , #

*$ #& &# & 36H7*) # (# # &*)$% & % &

( *? ( # % #% ( % *L ( "% % & " % # & *) , (% % & ," & % & " % % * M # ( ( % + % & % * , "#,& *)M ( 3K7 ( #( ( # & % # # (, % %,&" % % # &* (& % (# & $ %+ ( *

" (, & % % % " & & , + # ( " + * %* % &= #, "#, % " % < , , % > 5 +% % &" ( " , " " ( ( ( # ( % " '+% & =& %* " & " & $(*, & ( # & " % % % # +% * ( %&&% & * % , ( , # # ( % * .% & ( % * "# # $ *# # ( " " % #( & * ( % - , (, $ , ( &" , I # % (@*, % % & % *

+%% #(* @ ) # ( % (, ** & , & & +%* %% , = (*

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

19

& #(& & % $& &" =$ 3647 && 3667*." % & & * # % # % % *" ( # % % )E 2 !&( .?. ! &(,& M* )L) =< > =<>," & (& +% % (*((( ( & "# &4* ( & " (4G9H6HHH% *.)E 2% # ( ( 8C " # %% .: 64 , : E.?.:66*) % # # .?. & % (

% & . * ( & & & % ( # + ( #( 9D*8 %, " & ( 8H*K % <&" )E 2

>* % % " & = % # % & & , , #% ,% < #*+>,"& % #% & <3647, %*4@>* &6*

. "

#

.<L > 4G9H4GGC

64

E.?.

4G9H4GGG

66

< %%>

4G9H4GGG

66

)E 2

4G9H4GGG

8C

3647*

% "

$ "

% " $ "

"

"

)E 2

4G9HGG 4K

4

48

9

48

K@

8H*K

.)E 2

4G9HGC 49

49

9

9

KH

8C*9

E.?.

4G9HGG 46

6

9

4

4H

8H

KH*H

)E 2E.?. 4G9HGG 4C

4

48

K

@

K4

K4*9

.

4G9HGC 48

4

D

4

K

6C

KD*4

.E.?.

4G9HGC 4K

8

8

K

6K

9D*8

3647

20

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility &*& % "

'# () '%*' + )( *E*<2$ > 4GD6 4GD6 4GD6 4GD6 (%< +> 4G@C 4G@C 4G@C 4G@C (%

4GC8 4GC8 4GC8 4GC8 % ( 4GGDGG 4GGD 4GGDGG 4GDHDD 4GDHDD 4GDHDD 4GDHDD ( 2(%<-> 4G9@ 4G9@ 4G9@ 4G9@

E E 4G9H98 4G9H98 4G9H98 4G9H98 + , A 4GCG 4GCG 4GCGDD 4GCG <)> 4G@6 4G@6 4G@6 4G@6@8 # ! 4G@G 4G@G 4G@G 4G@GCH 4G@9 4G@9 4G@9CH $ < E > 4G@9 + - . + $ 4GC4 4GC4 4GC4< + "*> 4GC4 *P*P 4GGK< > 4GGK< > 4GGK 4GGK< > # !( 4G9@ 4G9@ 4G9@< > 4G9@ $%<$ # > 4GCK 4GCK< > 4GCK< + "*> 4GCK + , )PJ 4GC4C6<* *> 4GC6< > 4GC6 4GC6<# **> $ 2 ? 4GD9D@<* *> 4GD9< > 4GD9 4GD9<# **> & A 4GC9CG 4GC9CC< > 4GC9 4GCCCD <)- > ( 4GG9< > 4GG9 4GG9 M<E" , > 4GGHG4 4GG4 4GGH<4>54GG4<6> 4GGHG4 $ < ( > 4GDK,4GD9<* 4GDK,4GDC,4GDG 4GDKDG 4GDKG4<# **> *> GH< *> ( 4GGG 4GGG 4GDDGG<# **> $ < > 4G@H@6<* *> 4G@6< > 4G@6 4G@H@G 4G@CC8<# **> (% 4G@GCH 4G@GCH< *> 4G@GCH ? 4G@8@9<* *> 4G@8@@< > 4G@8@@ 4G@8@@ ? )( < > 4G@8<* *> 4G@8< > 4G@8@K 4G@8<# **> 4GDG<* *> 4GDG< > 4GDG 4GDGGH<# **> 4G@9C9 4G@9C9 4G9CC9< +*"*> 4G@KC8 A A 2< -> 4G@4<* *> 4G@4 4G@4 4G@4@8<# **>

M 4GD8<* *> 4GD8< > 4GD8 4GD8<# **>

+* / - + )( # 4GCGDD ,% , < # > ", *4K4 E *4 L &<) - > 4GD@DC<* *> 4GDC A 4GDC 4GD@DD< *> <)- > 4GD4<* *> 4GD4 4GD4<4>,4GD4<6> % 4GCCCD 4GC9DK< +*"*> 4GC@CD $ < E> 4G@9<* *> 4G@9 4G@9GG<# **> ( 4GGD J <$> 4G@CCH<* *> 4G@C <M ! (> 4GC8CK<* *> 4GC8 4GD6 L 4GDCDD<* *> 4GD@DD< > 4GDCDD P )& 4G@G<* *> 4G@G # P( # < %*> 4GG4< , 4GG4 > % ?

4G9C9D<* *> 4G9C9D 4G9C9D<# **>

( - <E( > 4GC4<* *> 4GC4< + "*> 4GC6<# **>

( - 4GCDCG 4GCDCG 4GCDCG )&&# * 5 +*"* + "5 * * 5 # **# * 3647

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

21

&" =$ 3647 % <&8>*& # (*.&# % (" % # & ( %<>* %" % ( *%<&> #" (+ & "#,%<> ( % " < ** #* +>, % < > 5 %<> " "# *. + % - # ( ( " # $% " * % %<># "& " +"* % " % % < >, ** I * & 3667 % %& #( ( # # " <& K>* # " % & 4GGK ( 3687 " 898 , " (36K7 4GG@% + # C@K* . &# ( & # "# &<&9>*. ( # ( ( * <.6HHH, $ (34D7> # ( <.4GGK, ( 3687>, < ! ,6HH4,3697> + * && < " > # (# (&"H*D8 H*DD* (

H*D , " & (H*K@*

0 % # " % ( <4GGK>

1/ #%

23 #

@4

898

$ (<6HHH>

C8

K44

! <6HH4>

CH

K@H

L $ <4GG9>

9D

9H9

M *<6HH4>

D8

88H

2 L <6HH8>

CG

@@8

(<4GG@>

44@

C@K

& <6HHH>

4HH

CHH

& <6HHK>

4H6

@G4

. ( & *& /P " % # & (< $ % > + * # " *L &(<6HHH> + && # " & % % *0<3667,%*D8K>* &3667,%*D8K*

22

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

4 540 507 507 509 54& 54

4 548 4 576 548 56 544 566 57 H*DH 57

4 56 4 566 H*D9 56 H*D8

& <6HHK>

2 QL <6HH8>

4 546 565 577 577 579 568 H*D8

Q& <6HHH>

M *<6HH4>

4 56 54 565 547 577 H*DH 564 56&

Q& <6HHH,+*>

L $ <4GG9>

4 H*D8 564 507 565 577 565 578 567 560

(<4GG@>

Q! <6HH4>

4 H*G@ H*D6 560 50 578 575 565 578 567 560

L &(<6HHH>

.6HHH

.4GGK .6HHH Q! <6HH4> L $ <4GG9> M *<6HH4> 2 QL <6HH8> L &(<6HHH> (<4GG@> Q& <6HHH,+*> Q& <6HHH> & <6HHK>

.4GGK

4* (" ( # " ,4G@H4GG8

4 H*DD

4

3667,%*D86* & H*D5 H*@* & & :# <> " <84GD &# >*

2 &%&*) ( & & ( % " $ & +% % % * % , * . + & & < , >* & # &# ( & # * % && " " ( ( ' R ( & & # " ' &&R ) & #,% ( & * % +%

# % % * ( % ( *+% &= # , # % , % % ( &&* , % C* ,+%" % % % % & ( # (

+ *)( , % " "" * " # ( ( % ( # * %-- & # * ( " & % C # & # " ( M % % " % # & *

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

23

# ( * % % % % ( & " # & ( # , " & ( * " (= # % ' & % # (

% # # & % * % % ( & &+ , * ( % % ( " " "% % " *) , "#, ( %&( # % & & ( * &+% " " ( %% % & &* % % < % # >*., , #( $ & # % % & & * % , # # * ,% % ( " (" ( & % * & 1 ( / M # % ( # # %% & % # ( % #( &" * % " & ( & % ( $ ( % *M%$ ( " $ ( " " $ & % % *& # + "( " " % %# 36@5 6C7* ?$ ( / % % % " $ #0<36D7,%*949>*P % &" % % % (% * &# # % %% * ," (' " % " & % %%* & % * % % ( %# ,+

, &$ % *?$ & % " &# / $ % % # S (:& # ,& # " $ " ( "" " " $0<36D7,%*949>*$ ( ( (% " %% % % ' ?$ % % # ( % % 36D7* %%" " *

24

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

&% $ " (& %+ *E " ( $ " "

$ $ " # # # * % $ " ( & * $ " (& % ( % (

* . ( #, ( # ( # ( - & (& * & " # # #* # &, " $ ' % $ % " "# *) ( $ % ( ** " &

$ (& $ ( *) % ( $ * % " && % " # " % * %" ( - :" # "&

( ," % * % % ( ( * % % % $ ( ( - ( *% $ %

" *% % ""% % * ? ( # % # % %& % 6HHK* % ( & ( & ( +# +% , ** & # ( ( " (( * # & " & , # % , #*% ( ( % ( ( * % % #& ( & * %$ 4GG9& #& (* " # & &# * " &" " *& " # " % # "*/ " " # " ( $ % " ("*? % ,

& # % # $ # "& % (# ,% & , S %&: # & " #" , % ( ( "

0<3667,%*D99>* & (% & " +%* # % "$" % " % & ( & = & # % * " ( % " %% & # # & " & *

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

25

0 '

:

( % % (*. , " +% # < ( L &(>, " ( %<.> ( % ( % * % #, +% " % # # " * & $ ## & * #, % ( & & # % & #* ) ( (% ( % , " ( %" ( " * " + % " ( ( % , ** ( # #&& %*)& (/ & % &< " ( &> # ( # "03667* ( #( # ( % % %, % ( < ?$36D7> % " %& = #( " *.&# , & # & & & $ " ( ( % * , "#, (% # % & %* )( & # % , % ( " $ " (* % " & < , &( (, (>* % &" =$ 3647 " D* & %*+ % % ( ( # & * "& 3667*) & #, +% & * % ( (" +% *& % % * % # & % % (/" (0* ,+% " * & # % %%%

# ( : & * % # & % - * $ " (<& % & & % & (> % *)?$<36D7,%*94C>( , &#" & " $ " $ & % $* ? % % / # ! $ & $ (( & 0<36D7,%*94C>*

D & % " ( " " & /% 0*

26

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

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

$ & # , 4D4@4GGC0* % &, 6HH8, A * KC4, %%*KGCH* 34G7 =$ *,/) (" ' # ,4GK@6HH80*% &,6HH@,A *6H4* 36H7 $E*,/)( :M 2 ( ( (0, #

%%, %% (, 6HH9* % Y4, # & %II"""*%**I%& I Z%&I Z%%4*% * 3647 &" * =$ *,/ ;) % # 0,[ %H48HK, %&6HH4* )# &%II$*"- & * I% I6HH4I%H48HK*% < 4H&6HHC>* 3667 & *,/ # "' % % %+ % 0* .$"$' ,6HHK,A *KD@,%%*D4KD9D* 3687 (J** ?*,/ % = # " ,4D4@4GG60, 4GGK, % , YGGH9,))& ,? # 3 & 7* 36K7 (*,/ % # 0* .$"$' ,4GG@, A *KH4,%%*88@89G* 3697 * ! )*,/M M # # 0,6HH4, %6899, ( ** $* 36@7 &" * *,/# # ()A ( " . '0*% &,6HH@,A *6H4,%%*46K*

W.-D. Eberwein / The Creation and Use of Data: Scientiﬁc Requirements and Political Utility

27

36C7 *< *>,< "$' =/ %& !,L (,6HHC*6 #* * 36D7 ?$)*,/ # ) 0* .$ ' , 6HH6,A *8G9,%%*949969*

This page intentionally left blank

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

29

! " # ### #$% # & '

#

( ' ) ! # ! & # # #

'

# # # # # * # ! % # & ' !

# # + '% #

, , # # ,

) #

#

( ! ) ! #

! &

30

K. Eck / An Overview and Typology of Conﬂict Data: The Advantages of Data Diversity

-. /012#3 4' + ' 4' ' &

5/6 7 )8 ' # /09-5-6)8 : ; # /09. ) ; 8 ' 516 596 #; +/0<. # ) (")$ = 4 ( >4 #

5?6 5<6) # ! # ! > , ">$ ) /@/<+-../ # # ) ( # ' )8 #A 3 "A3$ +/0@.526B

-? )8 /#...# A3 ( )A3 + ' /00.# + 4' 3 ; " !"

#$ $ # -../ ( '( 3 ; # "3;$# ' /09< A3 #

3;+A +# + # A3 ( ' /0@. # ( > ">$ A > 5@6>8 (> ;'">;$ ; * # & > B 3("B$# ' )) 4 % 3 '% >

K. Eck / An Overview and Typology of Conﬂict Data: The Advantages of Data Diversity

31

* #4 * ) "A%$506 '+4 +> "4>$ 5/.6# ( )#A3 > A%8 * B C#4> # (( + %! #4>10-..9"// $ A3 1 4

! = % : 8 -..1 % 5//6 % : 8 ! D 4 8 5/-6 D*# # ' ' # ( ! # ' ! 3 ! ( '8E F5/16 !# )

A3# " ' A38 + $ !# ,/ !

#> B,' (5/?6C (8 & ># # & #+ ( H4 # & & (I )# A3# % I ! + ' ; & ! ! B ) A3 +

% #B,' 4( + # ( #

/

% <. #'#-..?5/96

32

K. Eck / An Overview and Typology of Conﬂict Data: The Advantages of Data Diversity

&'(' ) " # / , A38 , # # ' # '

+ # " # $ # !

( A3 + # # ' B + E F # # ' # '

# # '#

)

# #A3

* +# '

# # # )8 ) + % # # + ' ) ' ! # # ' J! % : # ' # + , > ( # % # ( '!

, "A3$ "

# $" %

"

# $" %

+4K

D+4

+4K

K. Eck / An Overview and Typology of Conﬂict Data: The Advantages of Data Diversity

33

# '! 2#?.. B AD 4 = /00? A3 , # + ) ! " + $ % : ' "; ' 4' 4 $# ! / * # & , A3 ' , # # ! %; A + # > ( " !$ # # + ! ( &'&' + ," - ( 4 #' A3# A L /0@0+

& #

' H ) ( 8 H " & ' $H># H>' # % # D* & & ! + -..#... B ) ' # ' 5/<6# &

'

34

K. Eck / An Overview and Typology of Conﬂict Data: The Advantages of Data Diversity

' L # # & # # ' # ! 5/26> ( # 5/@6% ' % ! # ; + -..1 /00@+-..-# ; 11 ! 5/06 I /<M///M I 11 # # 'L '

% # ( "A3! # /09<+/0@0 A +3; $ - # ( (A3& -? + " # #+ #+ $# /#... + )& /#... ! " $# /#... L , % : & /.. /#... &'.' / % 0 + + *1* # # ## % / > # #+#

: * 5-.6#

/09<+-..9

K. Eck / An Overview and Typology of Conﬂict Data: The Advantages of Data Diversity

35

& 3

# + > +3 #! # ,

(" 3 , $ B + A3# :' # &# + # ) HA # ! > 3 # ! # #E# # N #

! NF5-.6 A4 E # # F 5-/6 # # ! & & ' # # 4// , O8 D # ! # 8 -..- " $ A4 , A4+O 8 %# , C # # L

(

36

K. Eck / An Overview and Typology of Conﬂict Data: The Advantages of Data Diversity

3

#+# A38 + # # ' 5-16B + * + " (#$+ + + # + + B + -? A38 # + & # ! ' " # # # $ ' -..

L # 4 # # + * #

'! ; /009#'?..#...+@..#... C ! C

' # A3 , & "/#... $# "; ; 3 %$ ' %# A3 ; /009 ' A38 +# &

% # 4 ! # ! # ) 1 # % : # # #!

/02?" #)4 # D $ ' # ! & # # 1 ) D ! #) " $ & /#... " $

K. Eck / An Overview and Typology of Conﬂict Data: The Advantages of Data Diversity

37

;###

# ,# # # #

# D ' # # # # # & # & ! '( ) *

& & & 8 # ' ! ( ' # & " ( $ & ; #! # & # #> ; '

5-96 & # ' ' ! ' # !# D 4 # # *3 # #

% ' ! #8 #

# ! 4 + 4 8 ' > , B , & # '

38

K. Eck / An Overview and Typology of Conﬂict Data: The Advantages of Data Diversity

, B ' ! + 4 ! ' ' 4 # *3 4 # ( +/00. # ( ! I # # ( # # # ( ( ")# A3#>#$

(8 ( 5-?6 # ') ,# , !B ' C ( ' K

, + # ('

# 3 #

# 4 ( B) A3 4 P + ' ' # &

K. Eck / An Overview and Typology of Conﬂict Data: The Advantages of Data Diversity

39

' ' , , + # !

# # # # # #

# #

# % # , + + , ' , 5/6 5-6 516 596 5?6 5<6 526 5@6 506 5/.6 5//6 5/-6 5/16 5/96 5/?6 5/<6 5/26 5/@6

34'#

5

% # * A :#:#/012 7 )# #A 3#/09- : ; # #4 43 #:#/0<. : ; # 6#4 43 #:#/0<. = 4 >4 # % %#) 3 #D P'#/02- = 4 >4 # #4 3 #B C#/0@- A 3 "A3$#JJ # -..2+.2+./ > ">$#JJ #

-..2+.2+./ "A%$# JJ , + J J J' J# -..2+.2+./ 4> C ; #JJ '# -..2+.2+./ = % : # # # # %#02"-..1$#2?+0. D 4 # ) H ! #$ #9@"-..9$#@/9+@?@ 4( '#,% % (789*&::.# #%AQB#-..? '#8 #A33 S/# -..? > B,' # # !"

# &::;#!A 3#-..2 > T > 4#

# 5

#> % %U#B #C D"-..1$#-2+?. C , * #3 C # B ; # ' I # %#02"-..1$#/@0+-.- ; :4 #) ! #: #K#)>4#/00<

40

K. Eck / An Overview and Typology of Conﬂict Data: The Advantages of Data Diversity

5/06 :;# 3 D # ># : # : > ,# > W # > #> ; #+ ; #-..1 5-.6 B : D3*#> #+ $ #-/"-..?$#/9?+/<< 5-/6 > 3 ">3 $#JJ '# -..2+.2+./ 5--6 A4 4 #JJ # -..2+.2+./ 5-16 ' : C #+ #$ 99"-..2$#-11+-9< 5-96 > ;#) ' # $ # 9/"-..9$#112+1?< 5-?6 > '#< #!A 3#-..<

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

41

!"" "" # $ % & $' " "$' ( $ # )$ * " " "$( & !" # $ # ( # # # $' %&

! +,-.( " " $ #" # # "& " $"" # $ "" " & / #

" " "$

0& +,-1# #$ " 2 " 3 *

463# " 3

7# $ "2 "" 4& 0 ( $" " / # # $ #%#" $

" ( & )"

" "

$ ( 8 • • •

( # $ $ # " 0 &

" # & $ " "" $ + "" 9 3# +-: " ; # =,..> # ( & ?8 $$$&"&&

42

J.-M. Collin / Databases on Defense for Inﬂuencing State Policies

(

$ $ # ( ) " & * $

$

( ) " & ( $ " $ & +,-. ,. ( $ " ! # % # !& " $ $ $ / & ( *" # ( " $ '

" # # " /" & ? "$'

#$ ( "" & "$ $8 • • • • •

$ 0"0 # ) @+A 3 B'6' " 3 7@CA# ! $ " # 3 # ( !

# " #&# " $" "" * & D#E #E # *" & "" * $ " " $$ "# " # #&

"" " $ # # / $ $ & ( *" $ % $ " " "

# $ / " " ( "F & +,,>3 6 "7 ( " @G:A&

$ " #" # " $ ( #$ # "" & " / ( " %& (*" # " $ ' $$! ( $ &3 $ $# C " "

$ # *" ! C ' " "

$ H38 8II$$$&"&I"I J"I C...I+JJ &"

J.-M. Collin / Databases on Defense for Inﬂuencing State Policies

43

$ * " "( & " # " ( ) " & ) ( ) $ # " $ " # """ " "" & $+,=- ? # ( 0 $ & " ( ) # $ ##' " " $ # / $ ( ) * " & $' " "" $ "&$ " " $& ( # $ $ $ ' & * K " " 9 # 6" 7K " K " K " " &!

" $ ( ) & ! " C. #3 & $ 0 ( )" & E/

" $ " ( ) " # $ / &(#$ /(

$$' $'#" L $ &E*" $' " " 8 •

•

+,,:# $' "#

*" #2 ? #4 2" " #

4 "$' " $ /" #$$ !" & " "

"

" $ # $ 6 $$$& "&7& $ " F " ( ) " " # L " " &?"

44

J.-M. Collin / Databases on Defense for Inﬂuencing State Policies

L# " " @-#,A& "L" & " # $ # " &

$ " ( & " *" C...# $ $ & " ( ) $ $ & (

$ / / # 0

/" !" 6+,,,7#

$ 6C..C7#

/" ! 6( C..:7&

&

$ " " # $ " ) 0 ( & 0" $ " &? ) " $ & "

# 2 "

" &4 ( D (@+.A# " " 0 ! C...( & )

$" ( H M "# $ E $ &("#$' "

"" & ( *" " $ * "D ( %# # $$

$ " ! C..G# % # "" & D " C. # & % " # ( #" " $ & " $$ & C...# " *& / # # & F$# $

" &

J.-M. Collin / Databases on Defense for Inﬂuencing State Policies

45

@+A @CA @GA

!#$' "0 # #HM8 ) " %& "#$#%

!# #' "8' " 3 & D D

#D '"E "# & '#N

8E&% 8 )"" #+,,>#+C-& @1A D D

#D '"E "6&7# ( ') **+#+,,=# 8 )"" #C=1& @>A D#D '"E "6&7# ( ') **,#+,,:# 8 )"" #+11& @=A D#D '"E "6&7# ( ') **-#+,,-# 8 )"" #CC1& @:A D#D '"E "6&7# ( ') ***.///#C...# 8 )"" #CC1& @-A

#0 1 1 21 1 1 #3#$ F #C..G#=1& @,A D D

# & 31 1 41 5$ 1 ) 11 1 )55 1 1 1 & 8 )"" # $ F !# C..>#1C& @+.A

# ' 51 ' 3412

5 1 # )" ; O# Q++# 8 " ; OI3#C..C&

This page intentionally left blank

Concept Definition and Data Building

This page intentionally left blank

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

49

Calling a Conflict a Conflict: Violence and Other Aspects of War Joakim KREUTZ Uppsala Conflict Data Program, Uppsala University, Sweden

Abstract. Arguably the most important aspect for researchers and policymakers is to have access to reliable information, especially when focusing on armed conflict. In order to know when there is greater risk for the outbreak of war, whether containment has been effective, or when post-conflict measures are appropriate, there is a need to clarify what an armed conflict consists of. The most commonly employed definitions of armed conflict are being measured through battle-related fatalities tied to the political goals of the warring sides. This definition excludes several other indicators of human suffering, such as violent crime, genocide, starvation, and forced migration. This chapter reviews some of the problems with expanding the concept of conflict and argues that it is beneficial to use a narrow conflict concept in order to study the relationship between war and other phenomena such as criminality and genocide. Keywords. Crime, violence, indirect deaths, war deaths, genocide.

Introduction The study of war has a long tradition, particularly among historians and military strategists, but also within the social sciences. Carl von Clausewitz’s dictum that “war is the continuation of politics by other means” emphasises the use of violence as the ultimate means in the quest for power or control over a territory. Following the behavioural revolution in the social sciences after World War II, it was argued that the occurrence of war could be explained as a repetitive pattern originating in basic and observable conditions, i.e. each event is not unique. These views formed the theoretical framework for the first conflict data projects. One such project was the Correlates of War (COW), which collected information on conflicts and also other key variables that could explain (or “correlate with”) the outbreak of war including iron and steel production, demographic data, diplomatic linkages, etc. [1]. The original intent of conflict data researchers was to explore why wars begin. The scope soon widened to also focus on why, and how, wars end. These two main questions guided the development of the academic definitions of war, as it was necessary to distinguish the concept as separate phenomenon with identifiable start and end-dates. In practice, war became defined as a contest between organized groups, at least one of which was the government of a state, with political goals, and where violent acts between the groups reached a certain severity threshold. The COW definition of war requiring at least 1,000 battle-deaths in a year became so widely accepted that when other projects started coding organized violence with fewer fatalities these were not referred to as alternative definitions of war, but as “minor

50

J. Kreutz / Calling a Conﬂict a Conﬂict: Violence and Other Aspects of War

armed conflicts” (Uppsala Conflict Data Program, UCDP), “disputes” (Militarized Interstate Disputes, MID), or “crises” (International Crisis Behaviour, ICB). Information on events during conflicts was less in demand and remained largely the topic for studies on military strategy or the anthropology of violence. An outlook that changed in the 1990s when policymakers, activists, and political scientists became increasingly concerned with an alleged era of “new wars” [2]. These, it was argued, were intrastate rather than interstate, more violent, fought for economic rather than political gain, and not fought in battles but consisted of attacks on defence-less civilians. The “new wars” argument was often accompanied by a growing scepticism, mainly from the policy community, towards the established conflict data projects as it suggested that the conflict data approach to defining war misrepresented the severity of armed conflict [3]. Since the end of the Cold War and the changes in the international system that followed, it is clear that there is a need to review and contemplate the definitions of war that are employed in conflict data research. The latest war between states registered by the UCDP ended on 26 November 2003 when India and Pakistan agreed to a ceasefire, but there was still some 32 intrastate conflicts active in 2006. This paper will not attempt to comment on the “new wars” argumentation as this has been done elsewhere [4], but instead to consider some of the suggestions for changing the definition of war. In order to study changes in the phenomenon of political violence, it is of great importance to resist the temptation of manipulating the definition of war. Expanding the definition to include other events will limit the analytical clarity of the concept and lead to less useful research findings to guide policymakers. As conflict data projects become ever better at providing detailed information on different aspects of war, the need for clear and consistent definitions increases. Without stringent definitions, it will not be possible to increase our knowledge of the causal relationships between: direct and indirect conflict deaths, conflict and genocide, or conflict and crime. Thus the challenge is not primarily in changing our understanding of war, but of identifying and systematically study other phenomenon that often, although not exclusively, can be found simultaneously as an armed conflict.

1. Non-lethal Measures of the Severity of Conflict In order to understand trends in warfare across time and space, it is necessary to develop some kind of measure for the severity of a conflict. Traditionally, conflict data projects have focused on the number of fatalities reported in a year but this may not be the most appropriate measure. If the aim is to fully understand the impact a conflict has on a society, there are many indicators that arguably are more useful, such as the number of violent incidents, number of injured, or number of refugees. These measures could be used as indicators of the immediate impact as well as the long-term consequences of the severity of conflict. 1.1. Incidents as a Measure of the Severity of Conflict Attacks on infrastructure or economic centres can have more impact on a society than the use of armed force that lead to some casualties. Furthermore, non-lethal attacks also send a strong message that can be used to challenge or undermine the authority of a

J. Kreutz / Calling a Conﬂict a Conﬂict: Violence and Other Aspects of War

51

government. The Colombian anti-government group M-19 initially focused on highpublicity operations, so-called “golpe revolucionario publicitario”, to create civil disorder. Their first operation consisted of stealing the sword of the national hero Simon Bolivár from its exhibition in Bogotá in 1974, an attack that did not lead to any casualties but had significant political effects in Colombia [5]. Another important effect of a large number of incidents, even if these are non-lethal, is the ability to interrupt the normality and create fear among the population. This is a concept that has been embraced in the growing research agenda on terrorism, where systematic data is collected on incidents rather than fatalities. Arguably the main challenge for this field of study is the lack of a coherent universal definition of what terrorism constitutes, even though the emphasis has become the intent of actors to produce fear for sociopolitical purposes [6]. The focus on incidents is an interesting approach to measuring the severity of conflict, but it is still not operational for use in the creation of global comparable data. There remains some work with regards to defining exactly what an “incident” is and the potential number of global incidents that need to be collected will be a great challenge. There will also be difficulty finding reliable information on such a measure globally. To illustrate the potential scope of such a project, it can be noted that according to statistics from Iraq there were on average 175 attacks per day by insurgents and militias in December 2006 [7]. Simply counting the number of incidents may limit our understanding of and actually misrepresent the severity of a phenomenon. It is necessary to remain focused on the scale of different incidents, as the impact of the attacks on the World Trade Center in New York on 11 September 2001 has been significantly larger than another terrorist incident that occurred just over a month later near Susanville in California. The attack on the twin towers led to 2,749 deaths, while the attack by the Earth Liberation Front resulted in a government owned barn being set on fire. 1.2. Injuries as a Measure of the Severity of Conflict The immediate effect of disrupting economic and social life and spreading fear can be even greater if some casualties occur in these attacks. According to the original military definition, the term “casualty” referred to all who were lost to active military service, meaning that not only the dead but also those severely wounded, captured, deserted and missing were taken into account. Indeed, it is not an uncommon view among warring parties that wounding an opposition soldier is better than killing the soldier, as the immediate effect is that other opposition soldiers will stop fighting to take care of their colleague. Injured and wounded victims in a conflict will also require resources through hospital treatment and help with reintegrating into society. Conflicts that lead to many injuries will also have long-term impacts as the affected society will have to continue to accommodate victims even after the fighting has ended. It would therefore make sense to measure the severity of conflicts by counting the number of injuries resulting from the fighting. Such an approach would however have to deal with similar challenges faced by attempts to measure severity through the number of incidents. It would be necessary to provide a more detailed definition of “injury” and it will be very difficult to collect global information about all injuries.

52

J. Kreutz / Calling a Conﬂict a Conﬂict: Violence and Other Aspects of War

1.3. Refugees as a Measure of the Severity of Conflict A third potential effect of war that has a great impact on societies are the numbers of refugees created by conflicts. This aspect could also be employed as a means of measuring severity both immediately and over time. Immediate effects are that refugees in the vicinity of conflict locations will need resources in the form of food, health care, and shelter. At the same time the country will be less able to produce these resources as the refugees have left their jobs, houses and lives to avoid the conflict. Furthermore, after the conflict has ended, many refugees will not return as they are worried about the security situation in their old home and may have created a new life somewhere else. The result will be that when a society needs to be rebuilt after a conflict, the most able members may not be able or willing to participate. Even though the creation of refugees sometimes can indicate the severity of conflict and there is accessible global information available to prove this point, it is a measure with limited use. Forced migration is not consistent across all conflicts, instead it depends on several other factors such as the location of the fighting and the type of violence employed. Interestingly, the number of fatalities in the conflict is not linked with increased number of refugees [8]. Refugee flows are not only caused by conflict, but also for a number of other reasons such as the risk of human rights abuses, lack of democracy, and environmental disasters. These different reasons for why people flee their home countries can be difficult to separate. According to figures from UNHCR, there were over 340,000 refugees from Vietnam in the world on 1 January 2005, despite the country not being involved in a conflict since 1988. Rather than using forced migration as an indicator of the severity of conflict, it is a phenomenon that needs to be studied on its own in order to provide more information about the relationship between refugees and war.

2. Non-violent Deaths in Conflict The basic definition of war for most conflict data projects generally consists of the following three criteria: identifiable organized actors, identifiable political objectives, and a given threshold of deaths caused by the fighting between these actors. One of the most disputed aspects of the definition is with regards to the focus on deaths caused by fighting. It has been suggested that it would be better if the data provided information about war-related deaths, including the so-called indirect deaths caused by, for example, disease and malnutrition. 2.1. Direct and Indirect Deaths Following epidemiological surveys in the eastern part of the Democratic Republic of Congo, the International Rescue Committee (IRC) reported that the conflict had led to some 3,8 million deaths, the large majority were due to treatable and preventable diseases [9]. This estimate was significantly higher than those provided by conflict data projects, mainly because it included both direct and indirect war deaths. The former are fatalities that are directly caused by violence, while the latter are caused by “such phenomena as illness, disease, or starvation that would not have occurred in the absence of conflict” [10]. The suggestion that indirect conflict deaths should be

J. Kreutz / Calling a Conﬂict a Conﬂict: Violence and Other Aspects of War

53

included in severity estimates on conflict is often expressed by case study experts who argue that not including all aspects of a given conflict is morally wrong as the suffering of the population in a particular conflict location should not be ignored. In a study about the consequences of a conflict in a particular country such an approach is clearly advantageous, but it produces limitations for systematic studies trying to identify structural factors that contribute to human suffering. It is worth bearing in mind the lessons of Kalyvas (2006) and Durkheim, who noted that social scientists cannot be accused of wishing to condone crime or be devoid of all moral sense just because they study a phenomenon through a “cold, dry, analysis” [11]. Indirect deaths from starvation, disease, and the like are significant problems that often do not receive enough attention by the policy community. However, the solution is not to incorporate these figures into the conflict severity estimates. Any such attempt would only make our understanding of conflict less precise, it would be harder to identify the end of conflicts, and there is a risk that the indirect deaths by malnutrition or disease would receive even less attention than they currently do. Taking the example of the Democratic Republic of Congo, the argument that the severity of the conflict is better represented by the mortality figure of 3.8 million victims would suggest a similar approach to other countries in the region. Thus, the short-lived conflict in the neighbouring Republic of Congo 2002 where around 100 people were reportedly killed should be reported as an estimated 45,000 casualties [12]. Furthermore, the violence between the government and rebel groups in the Democratic Republic of Congo ceased in late 2001 and a peace agreement in December 2002 solved the political aspects of the conflict. Even though the conflict behaviour had ended, the IRC surveys included the mortality rate in the post-conflict phase. In order for us to be able to analyze the phenomenon of conflict, there is a need to identify when the conflict has ended even though the long-term effects of the conflict often remain. According to Ghobarah et al [13], many victims of indirect deaths may occur after the fighting has ended. They find that “civil wars greatly raise the subsequent risk of death and disability from many infectious diseases”, and also there is a risk in post-conflict societies for “increases in homicide, transportation accidents, other injuries, and cervical cancer” [13]. Finally, problems concerning malnutrition and disease are not only caused by conflicts, but are common among refugee populations and many poor countries. The occurrence of a conflict in the same location may worsen the situation as it may be more difficult to administer humanitarian assistance. In the end, without more studies clarifying the links between the direct and indirect deaths in conflict regions it will continue to be difficult to design policies in response to future situations. 2.2. Deaths of Combatants Outside of Combat One of the most commonly reported type of fatality statistics from conflicts consist of so-called combatant deaths. Information about the number of combatants killed has long been used by military strategists in analyzing the capabilities, tactics, or for comparing militaries’ cultural or organisational sophistication [14]. In the end, however, the forces of the conflict actors consist of individuals who during conflict often become exposed to extreme situations. On 3 June 2006, a Chinese military airplane crashed in the province of Anhui, some 200 km west of Shanghai, which led to 40 people being killed. It was just one of several reported incidents of deaths during military training in China that year. There have not been any suggestions that the victims of the accident should be considered as conflict-related and there is no conflict

54

J. Kreutz / Calling a Conﬂict a Conﬂict: Violence and Other Aspects of War

in China registered by the most recent updated UCDP dataset [15]. However, if the incident had occurred in Iraq and involved an Iraqi government airplane, a dataset focusing on combatant deaths would have included the victims of the accident as part of conflict severity. Similar to the arguments for why injuries could be a useful measure of the severity of a conflict, the casualties in the form of accidents and suicides among military personnel can be politically important. Although, these types of incidents are less useful for explaining the amount of fighting in a conflict. Additionally, relying solely on combatant deaths will create difficulties in regard to the comparability of conflict and non-conflict locations and for identifying the start and end-dates of conflicts.

3. The Difference between Violence and Conflict A common mistake is to assume that all violence that takes place in a country where a conflict is active must be part of that situation. Such a statement is common even though the logic behind it is rarely investigated since it would suggest that either (a) there is no violence in countries without conflict, or (b) that all other types of violence ceases when a war starts in a country. This can be disputed, as arguably the opportunities for criminal violent behaviour increases rather than end during a conflict. It is possible, with the help of a careful examination of the definitions used for identifying armed conflict, to distinguish between conflict violence and other types of violence. In particular, the following aspects of violence need to be assessed: 1. Who is the perpetrator of the violent act? 2. Who is the target of the violent act? 3. Why is the perpetrator employing violence against the target? A basic overview of the first two aspects can be seen in Table 1, while the third aspect is qualitative and needs more deliberation. The methods used for collecting information concerning these factors differ in different data projects and have evolved over time. Sometimes, projects provide an estimate for the yearly activity of conflicts based on an evaluation of the existing material, while other projects claim to have detailed information for each incident. It can be argued that focusing on each incident and then aggregating the information into a yearly conflict estimate provides a higher level of validity with regards to the aspect that is coded. There is, however, a risk of less reliability as the number of “unclear” incidents increases. One potential approach to handling such problems is the use of several estimates (for example the UCDP uses a low/best/high yearly estimate system) and then in the presentation to continue to focus mainly on the “best” estimate given. For the conflict in Iraq, for example, UCDP estimate 2,802 to 6,622 deaths in incidents when both the perpetrators and the victims could be identified as part of the conflict. There were, however, numerous other incidents where the attack clearly was targeting the Iraqi government forces, but also where there was no evidence or statement about whether the perpetrators of the attack were an organized group. Such incidents, which nevertheless were deemed to be likely to be battle-related, led to an additional 2,758 to 2,894 deaths. These figures do not include any attacks on civilians, fighting between different organized groups or the numerous victims that were found killed for unclear reasons throughout the year.

J. Kreutz / Calling a Conﬂict a Conﬂict: Violence and Other Aspects of War

55

Table 1. Perpetrators and victims of violence in conflict countries Who (perpetrator) State

Organized group

Individual

Who (target) State/soldier

interstate?

Intrastate?

crime

Organized group

Intrastate?

non-state

crime

Non-combatant

one-sided?

one-sided?

crime

3.1. The Perpetrators of Violence One of the key factors for the pioneers in the systematic study of war was to explain the fighting between different states. The consideration of military capabilities or domestic opinion are but two examples of the many studies carried out in this area. Following the development of comprehensive conflict data projects, it soon became clear that intrastate conflicts–or civil wars–were at least as common as fighting between states. According to the UCDP, the use of force by groups seeking to secede from a country or in the contest for political power has been more predominant every year since 1946 than warfare between states. The terminology in table 1 was chosen from the different types of data collected on organized violence that was carried out by the UCDP and presented annually in the Journal of Peace Research and the Human Security Report. States are the traditional key members of the international system and their use of violence can manifest itself in both inter and intra-state conflict. In intra-state conflicts, the representatives of the state will be the target of attacks from organized groups, but such groups often also use violence against each other. There is, however, an additional dimension that often has been overlooked in discussions about armed conflict and strangely has also been neglected by most conventional criminologists, namely individuals’ increased use of violence during, but not exclusively within, wartime. One of the first projects focusing on the systematic study of war theorized that the phenomenon occurred as a consequence of the breakdown of social and cultural relationships [16]. During the circumstances of war, it was argued by contemporary scholars, all the factors that lead to crime are driven upwards: “family life is ripped apart, children are neglected, destitution spreads, while scarcity of goods generates theft and begets illicit markets. Crime is also caused by a general demoralization, and violent behaviour increases as a mimetic outcome of the spectacle of ‘killing, maiming and terrible destruction’. [...] The dark figure of crime is assumed to go up, due to the weakening of institutional agencies such as the police and the judiciary” [17]. Individuals can sometimes attack and kill representatives of the government, such as police and members of organized groups. Indeed, individuals may even participate in large-scale violent demonstrations that force the government out of office without it being a case of armed conflict. Following a sudden fuel price hike in Yemen 2005, anti-government demonstrations turned violent as some individuals in the protest fired on police. The demonstrators were not part of an organized group but ended up in gun battles with local shop-owners who had armed themselves to prevent looting.

56

J. Kreutz / Calling a Conﬂict a Conﬂict: Violence and Other Aspects of War

The example from Yemen illustrates another effect where instability, or the outbreak of armed conflict, can make neighbourhoods, families, or friends choose to arm themselves as self-defence forces or potential vigilante mobs. Such developments are often more difficult to overcome than trying to settle the conflict as the concept may become part of the official or unofficial business community and the groups are less likely to demobilize voluntarily. One of the countries in the world where such structures are most visible is Colombia, where at times some 80 per cent of the casualties of violence are not caused by fighting in the conflict [18]. Indeed, the government has recently been forced to negotiate a demobilization agreement including some political concessions with the paramilitary groups who originated as private armies protecting large landowners and drug lords. These types of arrangements are not only found in conflict countries as, according to Guatemalan human rights organizations, vigilante death squads attacked and killed 57 suspected criminals in February 2006 alone because of the alleged poor performance of the local police. 3.2. The Targets of Violence The second important information necessary to classify a situation as a conflict is to focus on the victims, or rather the targets, of the violence. Most conflict datasets focus primarily on victims of the fighting, which includes either the combatants from the fighting parties or also the so-called collateral, or civilians caught up in the violence. There is an important distinction between these two approaches as the legal guidelines for warfare, for example the Fourth Geneva Convention, makes a distinction between combatants and legitimate military targets on one side and civilian and civilian objects on the other side. Despite the advent of so-called “smart bombs”, it is still very common that civilians are hurt in modern warfare, especially in urban areas such as the conflict in Iraq. However, it would be illogical to count only the combatant deaths in a battle where some civilians are caught in the crossfire, hence the use of battle-related deaths as a useful indicator for conflict severity. This measure consists of all people, soldiers as well as civilians, who are killed in military operations during a war and is useful to describe the scale, scope and nature of the fighting that has taken place. Thus, providing that the other criteria for what constitutes an armed conflict are fulfilled, all people killed in warfare are registered and aggregated to identify the severity of conflicts. For 2006, there were 15,994 (low estimate) to 30,865 (high estimate) people killed in conflicts according to UCDP, a figure which represented a global increase of some 37 per cent1 compared to the year before. In the Human Security Report 2005 [19], it was suggested that the traditional focus on conflicts between states, or between states and rebels, was incapable of providing the full extent of global battle deaths. It introduced an additional dataset of fighting between organized groups, with or without political motivations, called the non-state conflict data. There is so far only systematic data on non-state conflicts since 2002, but the phenomenon has been found in countries where the government is challenged and elsewhere. One common factor that seems to lead to the outbreak of non-state conflict is that different rebel groups are not only fighting against the government, but also among themselves. However, there are also cases observed when the government encourages such behaviour. After the rebellious Communist Party of Burma in 1989 was dissolved as several of its military units mutinied due to dissatisfaction with the 1 This increase is calculated using the “best” estimates.

J. Kreutz / Calling a Conﬂict a Conﬂict: Violence and Other Aspects of War

57

leadership, the strongest military force that emerged was the United Wa State Army (UWSA). The new group wanted to secure the political rights of the ethnic Wa group in north-eastern Burma. Government representatives quickly opened up talks with the UWSA leaders and offered a ceasefire with some local autonomy for the Wa providing the group used its military strength against rebels in its vicinity. UWSA agreed in principle to the agreement, but soon declared it was not willing to fight its former allies in the Kachin Independence Organization. It did not have such qualms about other antigovernment groups in the area and has ever since been involved in battles against forces such as the Mong Tai Army and the Shan State Army-south Command. Fighting between organized groups influences societies and can, as the example illustrates, also influence the dynamics of civil wars. It deserves to be treated as a separate unit of analysis that sometimes, but not always, can influence conflicts. Based on the limited knowledge we currently have of this phenomenon, it appears that non-state conflicts are more common in countries that recently have ended a civil war. This in turn may lead to several potential findings concerning, for example, the quality of governance and the availability of arms in such countries. The battle, whatever form it may take, should be understood as a phenomenon that consists of two combatant sides with potential civilian side-effects. This is a distinctively different concept than what UCDP refers to as one-sided violence, even though this phenomenon sometimes has been used during conflicts or even as a strategy against the opposition. One-sided violence is defined as events when an organized group, either the government of a state or any other organized entity, deliberately targets and kills civilians. The definition incorporates some incidents that elsewhere can be defined as genocide, politicide, terrorism, mass murder, ethnic cleansing, or massacres, but at the same time clearly delineates a certain empirically distinctive type of behaviour. Some commentators have argued that events such as the deaths caused by the Pol Pot government in Cambodia in the 1970s, or by the Rwandan government and its’ affiliated militias in 1994 should be included as part of the severity of these conflicts [3]. Such an approach would not necessarily improve the information provided about these conflicts, or outbursts of severe one-sided violence, without a closer investigation of the events. After the Khmer Rouge managed to take control over “Democratic Kampuchea” in 1975, there followed two years of peace from the civil war. The government ended up fighting interstate conflicts against Thailand and Vietnam during this time and there were some reports of local commanders in the southern and eastern districts of the country that mutinied against the government. Unfortunately it has been difficult to access reliable information about these events and it nevertheless did not constitute a large-scale war. The violence unleashed by the Pol Pot government during these years may have been fuelled by suspicion about potential rebellions, although there has not been any evidence presented to support the claim that the main reason for the forced evacuation of the cities and the mass executions was the conflict. In Rwanda, the conflict was in a process of de-escalating following the establishment of a multiparty government in 1992 and the Arusha peace accords in 1993. How this decrease of conflict intensity could have influenced the Hutu extremists into overthrowing the government in April 1994 and initiating genocide is a question that needs to be further studied. Nevertheless, it is a matter that should not be confused with the conflict itself.

58

J. Kreutz / Calling a Conﬂict a Conﬂict: Violence and Other Aspects of War

3.3. Why the Perpetrators Attack the Target Violence can take many forms, and only when registered in the form of fighting between organized entities, of which at least one is the government of a state, can it be considered a conflict. There is, however, another important aspect of violence that needs to be discussed. Violence, even between the types of parties that have been described in this chapter, does not necessarily constitute a conflict. At the same time, a conflict must not necessarily be violent. According to Wallensteen [20], a conflict is defined as a social situation where at least two parties strive to acquire, at the same time, the same limited resources. According to this basic definition, these “parties” could be individuals, organizations, or states and the resources that are competed for could be almost anything. As a consequence, it can thus be argued that there exist numerous active conflicts in any given country and at any given time. It is when these conflicts are concerned with the core aspects of state sovereignty, or “who is to make authoritative decisions over a number of people in a certain area” [21], and the actors choose to employ violent acts against each other that it constitutes an armed conflict. Indeed, some conflict data projects do not include the notion of violence at all in their definition of conflict, such as the project at the University of Heidelberg. Others focus on conflicts that include violence or the threat of violence, which is the case for the MID and the ICB projects. As was the case in the discussion on battle-related deaths when it is the intent of the perpetrator that classifies a situation as part of an armed conflict, it is the intent of the state(s) and/or the organized group that creates the conflict. Measuring intent is extremely difficult and has to be empirically based, either through analysing the actions (if rebel group X fires on government soldiers, it can be assumed that their target is the soldiers) or by basing the coding on statements made by the actors. Since the criteria for identifying a conflict consists of having good information about both the actions and the statements, there will always be a degree of uncertainty surrounding some cases. Hence, UCDP maintains an updated list of unclear cases which, when more information becomes available, may eventually become listed as active conflicts. One example of this is the conflict in Pakistan (Baluchistan) which in 2005 was reported as unclear, but in the 2006 update of the dataset has been included as active in every year from 2004 to 2006. To also include the political aspect, the incompatible statements made by states or organized groups into an overview would remove some of the uncertainty with regard to which situations constitute armed conflicts (see Table 2).

Table 2. Classification of violent acts with or without political incompatibilities Perpetrator and Target State targeting other state State targeting org. group State targeting civilians Org. group targeting State Org. group targeting other org. group Org. group targeting civilians Individual targeting state Individual targeting org. group Individual targeting other individuals

With political incompatibilities? Yes No interstate conflict ? intrastate conflict ? one-sided one-sided intrastate conflict ? non-state non-state one-sided one-sided crime crime crime crime crime crime

J. Kreutz / Calling a Conﬂict a Conﬂict: Violence and Other Aspects of War

59

In the analysis of armed conflict, it is of great importance to retain the understanding of the political nature of the phenomenon. There is an important factor that separates armed conflict from other phenomenon which is apparent in Clausewitz’s acknowledgement that violence is a means to reach a political end. UCDP have chosen to focus on statements by the actors in a conflict that spells out their incompatible positions over a political issue concerning government or territory. These situations are different compared to occasional outbursts of violence between states, as there are other means available for reaching the political objectives such as negotiations or sanctions. Soldiers from neighbouring countries that kill each other may not necessarily constitute an armed conflict, if there is no stated incompatibility. For example, soldiers from Thailand and Myanmar fired at each other for about 30 minutes in August 2006 after a misunderstanding about the exact location of the border. There were no casualties in this incident, but there has been in similar situations in previous years. Similar incidents can occur between government representatives, such as police and members of organized groups, even though they have not stated any political objectives against the administration of the state. Almost war-like fighting has taken place with organized criminal groups who sometimes have launched large-scale campaigns against government targets, such as the Colombian group “Los Extraditables”, led by Pablo Escobar, in the early 1990s. In 2006, fighting between the Brazilian police and the group PCC (Primer Comando de la Capital) led to over 100 casualties, which was more than many observed armed conflicts. The PCC did not, however, have any proclaimed political ambitions apart from their criticism of the conditions in Brazilian prisons. Without clear political goals, it is not possible to identify potential compromises or solutions that can stop the violence even though it obviously had a great impact on the life of citizens in Sao Paolo. To develop systematic definitions and start collecting information on these types of incidents is an important future development for the study of organized violence, but not necessarily for scholars focusing on the phenomenon of war.

4. Conclusion Armed conflict, or war, is a phenomenon of the greatest severity, and the effects for a society experiencing it can hardly be underestimated and may be felt for decades after the actual fighting ended. In order to improve our understanding about what causes war, what makes it more deadly, what can be done to limit it, or how it can be stopped, the research community faces substantial challenges. Using advanced statistical methods, it is possible to identify structural factors such as poverty or the management of income from national resources and this knowledge can help policy makers design long-term plans to limit the risk of having to deal with the outbreak of conflict. These studies must, however, be accompanied with a detailed investigation of wars to make it possible for us to act quickly, using the correct information and the right response, and improve our ability to stop the violence before it has spread across a country. In order to meet these two challenges, it is of utmost importance that the research as well as policy community have access to reliable and comparable information from countries around the world – in conflict as well as in peace. The best information only becomes available through the combination of different methodologies, as mortality studies inform us of the amount of destruction in a war-torn country while media based data can tell us more details about who, how, and when the deaths occur.

60

J. Kreutz / Calling a Conﬂict a Conﬂict: Violence and Other Aspects of War

In order to improve the usefulness of, and to promote the collection of data on organized violence, whether it be political or spontaneous, it is important to use clear and transparent definitions. A multitude of different data collection projects is not a problem, as there are numerous important factors of violent and non-violent phenomenon that may occur in a conflict country that currently is not systematically collected. Some of the examples mentioned in this chapter are the indirect deaths of conflict or the criminal statistics for conflict countries, which are not yet available on a global scale. Some projects coming out of the COW data effort such as the MID datasets have pioneered studies of the micro-foundations of conflict, and recent additions of the UCDP such as the one-sided and non-state conflict datasets have followed suit. Much remains to be done and the ever present challenge for data collection efforts is not the lack of ideas, but the hunt for additional funding. As policy makers around the world increasingly emphasize the need for evidence based policies, it must be acknowledged that there is a value in keeping existing data projects updated and revised. These are the building blocks for scientific progress and it is only with detailed information for long term series that the research community will be able to advance our knowledge of armed conflict. Finally, it is of great importance that researchers as well as data users within the policy community and journalists refrain from simplifying the complexity of conflict. Datasets could, and should, always be scrutinized in order to identify problematic areas and the need for improvement, but they should not be criticized for adhering to its definitions. One of the most policy-relevant research projects during the last decade has been conducted by Barbara Harff with the ambition of designing a model to assess risks of future genocides and politicides. One of her findings is that the risk of genocide/politicide increases when there is an intrastate conflict [22]. However, Harff would never have been able to even test that argument unless she had access to systematic data which treated conflict as something different from the attacks on civilians that constitute genocide/politicide. The importance of more disaggregated data for different phenomena, even if they occur at the same time in the same country, could not be clearer, especially if Professor Harff’s recommendations succeed in preventing genocide somewhere else. One of the important future developments for conflict data collection and research projects is to improve our understanding about different aspects of organized violence and how these interact. The increased use of the term human security is aimed at promoting research focusing on the threats against individuals, and the best approach to fulfilling that ambition is to investigate the diversity of threats rather than referring to them all as a single phenomenon.

References [1] [2] [3] [4]

[5] [6]

Singer, J. David. 1972. “The Correlates of War Project: Interim Report and Rationale.” World Politics 24: 243-70. Kaldor, Mary. 2002. New and Old Wars: Organized Violence in a Global Era. Cambridge: Polity Press. Leitenberg, Milton. 2006. “Deaths in Wars and Conflicts in the 20th Century, 3rd Ed.” Cornell University Peace Studies Program Occasional Paper No 29. Melander, Erik, Magnus Öberg and Jonathan Hall. 2006.”The New Wars Debate Revisited: An Empirical Evaluation of the Atrociousness of New Wars.” Uppsala Peace Research Papers No 9. Uppsala: Dept. of Peace and Conflict Research. Kirk, Robin. 2003. More Terrible than Death: Massacres, Drugs and America’s War in Colombia. New York: Public Affairs. Goodin, Robert E. 2006. What’s wrong with terrorism? Cambridge: Polity Press.

J. Kreutz / Calling a Conﬂict a Conﬂict: Violence and Other Aspects of War [7] [8] [9]

[10] [11] [12]

[13] [14] [15] [16] [17]

[18] [19] [20] [21]

[22]

61

O’Hanlon, Michael E. and Jason H. Campbell. 2007. Iraq Index: July 5, 2007. (www.brook.edu/fp/ saban/iraq/index.pdf). Accessed 9 July 2007. Melander, Erik and Magnus Öberg. 2007. “The Threat of Violence and Forced Migration: Geographical Scope Trumps Intensity of Fighting.” Civil Wars 9(2): 156-173. Coghlan, Benjamin; Richard J Brennan, Pascal Ngoy, David Dofara, Brad Otto, Mark Clements, and Tony Stewart. 2006. “Mortality in the Democratic Republic of Congo: A Nationwide Survey.” The Lancet 367: 44-51 Wille, Christina and Keith Krause. 2005. “Behind the numbers: Small Arms and Conflict Deaths.” Small Arms Survey 2005. Geneva: Small Arms Survey. Kalyvas, Stahis N. 2006. The Logic of Violence in Civil War. New York: Cambridge University Press. Mathers CD, Bernard C, Iburg K, Inoue M, Ma Fat D, Shibuya K, Stein C, Tomijima, N 2003. “The Global Burden of Disease in 2002: data sources, methods and results.” GPE Discussion Paper No. 54. (www.who.int/evidence). Accessed 9 July 2007. Ghobarah, Hazem; Paul Huth, and Bruce Russett. 2003. “Civil Wars Kill and Maim People: Long After the Shooting Stops.” American Political Science Review 97(2): 189-202. Lacina, Bethany and Nils Petter Gleditsch. 2005. “Monitoring Trends in Global Combat: A New Dataset of Battle Deaths.” European Journal of Population 21(2–3): 145–165. Harbom, Lotta and Peter Wallensteen. 2007. “Armed Conflict, 1989-2006.” Journal of Peace Research 44(5): 621-632. Sorokin, Pitirim A. 1937. Social and Cultural Dynamics, 4 vols. New York: American Book. Bonger, Willem Adriaan. (1936) An introduction to criminology; translated from Dutch by van Loo, Emil. London: Methuen. cited by Ruggiero, Vincenzo. 2005. “Criminalizing War: Criminology as Ceasefire.” Social and Legal Studies 14(2): 239-257. Bejarano, Jesús Antonio. 2003. “Violence, Security and Economic Growth in Colombia, 1985-1995.” Colombian Economic Journal 1(1). Mack, Andrew, Ed. 2005. Human Security Report 2005: War and Peace in the 21st Century. New York: Oxford University Press. Wallensteen, Peter. 1994. Från krig till fred (From War to Peace.) Stockholm: Almqvist & Wiksell. Heldt, Birger. 1993. “Armed Conflicts over Government and Territory 1989-91”, in Heldt, Birger, Ed. States in Armed Conflict 1990-91, 2nd Ed. Uppsala: Department of Peace and Conflict Research Report No. 35. 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.

This page intentionally left blank

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

63

!" # $$$ #% & ' ( )# # ) * + , # $ "-./"--0( , #*$$ ) #*% $$ 1 & '( ( + , ((2# 3$ *

4 5*67("-./(#+$8 %#! %+ &+8!+' %# 9: * , ;<; (*( 1 # )## # )#* # () # # =% 1):() # >"?*# 9: # (8#)#% $*# %# $ $ # $ =# # % # % $ + )$ *%# &>6?($77-'

# 1 $ * &# ' ) * # # 1 &)#1 #)# ') $$%# ) ".+ $ (@) # # # A(% 9 =# %#: $$$ 8 #* # *$$$ % #%>7? ( */("-;6($ %%# #) % # % C ! ) # 9: * 0 % # @: $ A , #! $4 #%) **-( * (*:&(#@ A'($% $ ()# # 1 *

"

D%

64

G.M. Ben-Israel / Databases on Terrorism: Constructing a Database on HBT and Abductions

) ## #

# % : %#$ (: )%## 9: $ ##)4 $ $ )#>0? 8 # # #"-;6# 9: %# % # % &!' $(3 $ * # ## #@ ) # A$ $( : ) #@# $ *A6### $ $# 1 ) %% 1 $ *( # $ *$* %% &>6?( $76.'

# $$$ #%% % ()# # ) * %% + , #$ % "-./"--0( , # $ %#1 #$ =$ # 8#$$%# ( @ % *# %#( * # *%#$$# %# 1 (# #%1 ( $#%% # # #*) #) :(% % #* :>? # 1 #%1 * 1 # 1% $ ( * )*#) # 1 = $ *>?A>E? >6?% $ %$ #: # * $ % # & $ ( # 1 %% ( 1(

(': ) &$ ( *(#($ 1 ( $ ( ( # $( ' $ &"' $$ , @A* &6'= % 1 * # : #### % * # * %#% $ ( %$$ (%$ % # *( * % * * # # #: %#: 1 (@ * 11# ,% ,$%#( ) ( *( F *( $ % $ 9 % 1 ( #( # = 1 ) 1( % $$A&>6?($77<' >6? # * : $ $ % % 1 #: =$# 1

* : % # : $( * ( * )# 1 * $# # ) )1(# #: # # $ 1 % %% ( $ ( 1 *)%% G & G' ## #

6 ## % % ) ##: $$ # (*F :* * # )*#% :*%% # %*

G.M. Ben-Israel / Databases on Terrorism: Constructing a Database on HBT and Abductions

65

3 (## # )$ # *$1 *# %&>6?($776' 8# # $ *# %

# 7/ # 1%%1 )%# (67)% $$$ % $#*H%#("E) % %) # # %##8 #"E("6% #*% :%#$$%: # 7) % #%)

% % %#) " 6

# % #% %#H

# ) # # # ( $ * # 1# &( # $( ' $ : ) # # #

7 # ## $H 0 * %#+ I , H E 1 H . ) H7 ; #) $$## 1 # &%% $ %# %#()# #1# 'H / # 1 ( % ( ) # # ( )# # ( # % % + $ % $ )#( % $ )# ## *% *$ %##H - #1 *$ ( ##%# % * $ % H # % % # + I $ # % ) # 1 % #H "< $"-./"--0

8) # 1 %# % % %# # % #% 1 ( ## 9: % % #* %# 8 %# ! %+ &"-./' # 1 #$ (# # %# (## C= (*#, J &"--0' $ %# "

7 3 * * % ) (*# # $ )# #) 1 * ,

66

G.M. Ben-Israel / Databases on Terrorism: Constructing a Database on HBT and Abductions

% & ' ()# #) *+ , #$ "-./"--0( ,

567K7"(2./ */-("-;6

# ) ) # # =% 1 ):() # ( =# %#: $$ $ $ ( ".+ $ (@) # # # A(% # 9

#)

##:

#+8!+ %# 9: * , #% 10<* ;<; ( "6<< $ 9 *( 1 # =# %# : $$$ )#* # $ %%# ##: #) # % #% C ! % 7"; )# 9:*0 $ % #:$ 9 =# % < , #: $$ #! $( $ ( # $ 6 $ %# $ : $ $

*"7"0("-;0

#: #3 ## # #% )# #$$ ( # % $ )## *#8!+

1 "/"-("-;0

# % # !%% $ 8!+ $ ( %# # # * %# # #%( : 6 1 :6 # =# %#%"0 1 # # $ ) $ # $ % #$ : $$ *$

#7("-;E

#+ # # % *# = # % ) ) ## ## , 6 * # "< ## # $ ; ): ") 1 1#( * $ ) # 0 # # # $1 $HE#) # #) % ;): H7 (#*$# $ #): ): H "" 1* # # *(# #*) # - %# * :# : ## )) #

##:3 67 7#$ $ % $ (#) # $$ ) :

1

* % ) ##

4 *-(#*%# # 9: * ( ($% $ # # ): $ #

# % : % #$ "< ( : 6%## 9: $ ##6 4 $ ( 1$ )#

#J (# E0E$ ( %# + ( * * ( $ 8 # *7<<$ * ## :): # "0( ) # ( ( ) #66$$ #8!+ " %F% (6;1 ):

;<))

# #

*$$ F##

#*%# ) ( %# * $ 8 # $ #6 #:): #( ) # #7 (" 1 1 ))

G.M. Ben-Israel / Databases on Terrorism: Constructing a Database on HBT and Abductions

5 "E("-;E

% * ) ) ## ##(: #7 )%# %# # " &)# )#%#%# #3% *') : = 6 #))

##: 8 % # # ! 8 %60$ ( :## % *# J%L1( % $ # #%( :#% * #

5 6;5*0("-;.

8 8 #"7-: #$ ) #% %)* %%% # (( ## # #%+ (8 4 3 ) ( "67<(# # ## 9: %

%$ # (#% #) %# ( # 9:( 1 :$ * #, (! *%; 5)()##* # : %# # # #, ( )# # 1 )% "% $*) ## #)) :(# : + $ %% <7"E# = ## 9:

(# % # %0<+ # "7# $ G # # 9:)/+ $ J *(8 ( ! 4 , ) , ( &+!4' 6 * #% & 8'

#*) $$ *$$*# G %$ + +

# 1 :

* %# # 41# % 8$( %# 1 %$ *#

# % #% # # *7

# #: %.# 9:( "#)##$ # %( 0E G # #) %) #*%# % # 1

#""("-;/

67

!%% # ) # # $ %# 8* %"<

& "# ) ' *, %E%# % 9 # # #%

1 1( #*: # $1 $#$# ) #$ ()# # ):# % & ')#) : $#$#%# ! $ # * # # 9: ## % # ## #)* ) $ # 1(# #(#$ # #

:$) # * $ % #%# - (6 )$ 1(7E ): "<< ) )

68

G.M. Ben-Israel / Databases on Terrorism: Constructing a Database on HBT and Abductions

$ 6667("-;-

) $%0 % # #+ ! 1 *#6 8 &+!8' 1# )# ) %# * 1 1# # #%% ##(# ): ! * #* : $ )# * # # $ (: $# # # #) ## ()##* ! ( $ 6$4 #% $: #$ + $

%# % * 9 =# %# %$ % #3

# 1

#*:6/* * # ) # # 0*#( %: # #(#+!8

#) ##

$ .;("-/<

8 1 % # ! 8 : 1 J ,( #:" 1 ) : ( ; % " 1 ): # # 3 *

$ "6"7("-/0

# # 8+ ##: # : $$ % %

$# 9: % E<<+!4 %$ $ % 9 1 1#: ( ) #0"$ % 1# , $ #($$= *"E : #%, *(#% * $ ) * * +

!%% # $ %# # %E< %# % 9 %

4 ### $) $ #+!8 ## J ( %# ##%#( *( # : % %# # ) # # #) =(J # 3# ## : ##: ) ##%# % * $ % ##)# # ) ## # #("$ 6 ): #6# ) $*#

% % # ) ) ## %# * ## : #E # # # $ # )# #"# " * ): H E % (0 # "" $()# #% ) 9 5%) ( 8 (( # 4 * % # 6 ): ;#$ )) #$

)# ) #*%% % # *1

G.M. Ben-Israel / Databases on Terrorism: Constructing a Database on HBT and Abductions

#;("-//

#+!4 , # #% 1# #% # #*$ % +!4 #) 9 :$ *)# 1 # # (#*

# * # # # )* #*# , # # :-$ #(/%)# ) ) 8 *(# )$$*# % 8 # : ) #@# A

% #L&3 ) ) ## 1 #* ') ## : 7 ( $1 6 # $ % :

"7("--6

$ ( #$

(): $$ #% ! #)*% # # :## # $%): L % $ #+ M $%% 1% # ,, M F & ' : %#

# + 31 #% 1#%# : $$ # #)@ ) ##

5*"("--7

69

# % !%% $ )

&,, %# M ': # $$$ $ %# 1 # %# :L # % 1L31 (# :# # %5 ( J 4 ( #% "<<+ #1 #* " $ H $ ) %0"; 1#$ )## %#$ $! * # 1%#: 31 # # % 31 # # 6%# # 9:$ 1 1 *) 5 J#

3 *)% #5 #5 *

# :8 % ( $ # # # ## #

4-"0("--0

# % #$ # 1 * $ ) *#

&,, ## 1 ## )@ %# * M ': ) ## ( # $ $$8 (# # :#L (# )) ##( # N= (% 9 (6<< &!' + $ ( *) ##+ $ +,( # *A ! 7: $$) $(()# # : H "" ) )) % % # $ 9 $ 1)# # () # N : ! #) % >7?

70

G.M. Ben-Israel / Databases on Terrorism: Constructing a Database on HBT and Abductions

#% % ") % 1 * % $ * # $ ( $ % ( $( $ $

( $ % #* :( $# $# $ 1 ) ) # #$ %# : $* (## +(% + ( # (% %% (L (% *%# + 34%% # +3 %%% ( #13 O1 (% % 8 ()#:$ # # 9: : $&"-;6'( )+ 3# %1 1 # $ "-;.( (% *% #+ 3%% # PL ,#:# 3 %%% L # ( % % # + 3 %% L ,#:# 3 %%% # 1 )): F % #$$% $ # #$# F % # 1 )) %# $$ # # # # # = & ' * 1 = ) # )# *#& # %% G (@# A $ *( * $'H $$ % # &' @4#A &)Q'H $ &@ # A(@ 1 A'H %5) # $ *$$ %$ *+ % *&@) %% A(@) 3: ) # A'H1 )%5) ## *#*%1 , &#(# $3 Q'H1%@# A $ % % ) ##@A "#

#% )1 # ")% : # # : $$ 1 1 % )1 ( *I$ $ ) * # %## #% $ ( #$% $ ( ) % #)# * #)#$ % ## )# # ) ) #1 % + : $$ # % &"-;6'(1*&"-;E'(J%L1&"-;E'( 1 &"-/<' # #3 &"-//'( # 1 % % @) A $$ $ ## #3%% &"-;0'( %% &"-;.' : $$ %7<< &"-/0'(#)% ) #1 % # : $$F *(% )#)# # $$

*$ ( % *#$ ) # # #3 %% &"-;0'(#$ & %% (# +' * :& (%#

( % ' # ( $ %#) #%# ( #* *$ %#: #1 %#@% A:&"-;/'( # *#1 &"-;-'( )#)

G.M. Ben-Israel / Databases on Terrorism: Constructing a Database on HBT and Abductions

71

# *% : $$# %#1 ## # ## *## &*# ## )##* # # % ') ## *%)# ## 1 # # #1 % %##$ )#)$ # )1() *%% & #)# 1#$ ! )#)#$'$1 #

# # *# %% & # % # : $$ % 5 J#' F1 )#% ) )#) #)#1 (# %#: $$): # ) ## ! )# # ) # # *# # & #1 %) )# 1%#1 '# #1 % ) ##$ *% % 1 % $( # : $$ % # % #( # , 1&"-.-'

##1 ##) % #* ) # # %#+ , #"--< % #(#%% %#: $$ %# #3 "-// ) # 1 %#+!4I8# , (% *#+ , 1 I "-// )*$% 1 * 1*#+ )$ 1 # )#: $$ % 3 &"-/-'( ## &"--0' # 8 : &"--0' # # # *$( )# # *(## *)#: $$ %# = 1 *#&6<<7'%= $# 1 ) $ # % #%%#"--< 1 # #

#: $$ %+ &"--6' ## N= &"--0') )1(#: $$ %# , 5 J# &"--7' )# # $ ) # ) % 4 1 #): $$ # * :# 1 )# # 1# 11#9 *) $$ #) % = $H # % )# $ : $$ : : ) $

#*%#: $$ %5 J#&"--7' *)* # % # # &"-;-' # * #% % # #( # #1 % # *% # # ) # ## ##): ( $*)###*): *#% %#: $$

# % # # $

# % % * % K * * $ ( #$()*%$ 1 # 1*#1 ) %$

* : ) #: $$( %# % # )# # #%% 1 %% %#1

72

G.M. Ben-Israel / Databases on Terrorism: Constructing a Database on HBT and Abductions

$ % & & 4 # # 1( $) # # )

#*I* % $$ 1 4 $$* # * # # $ * : *I 1 $ % 1 ( $ ( $ * $ *

: # #* 1$ # # % * # % ( = $ ### ) #* $ :) $ # $() $ :) ##$ # % %# $ ( $ ( #$ #

%)# ## $ ) =$ # ) % *( : 9% *($ ) '( ) & 4 % # 9 $ ) # ) 1#)( % * & '(%% $ # )$$( #) # % % % 4 #)# % # =*#) * ): $$8= $ %# 9: $ $ ) )* #)$ # #% * # ) # $ % # * % ( 1* )% ) # % 1 #%3 3# ##&"-;0'(%= $(#) %% % #1*8= $ # *%% ) >.? )#".$$ : #) #

* $ N : $ >;?# % $$ )66 # * % *>/?($ # "-;.(# % $$ )$6" 4 $$ % 1 3 # * 1 %* *$ #$$ * ###)# %% *

# ) $ ) # : ) # 8 = $( # = &>6?($70"' #J * # $ "-;0 1 N# % $ ""("-;0(# %#+$ 8 %#! %+

# % ! #* $ #) %J *#

# #( # # $ #( #* 1* #* )# @ F A # >-? $ # "/ # ) : )# )$$#$ % $ "/ >"<""? #$% #) 1 %# 1 % $ ):

G.M. Ben-Israel / Databases on Terrorism: Constructing a Database on HBT and Abductions

73

& >"?

P# ( (: , # +( 1 1("-/.("E>)? >6? ( @ N # : 1 1 A( (C(E(&"-/-' >7? ( ! " ! "# ( "-./"--0(# %#%%+# $#*( %G 1 *( %(6<<E( 6"<6".>)? >0? O ($ (J(5 (6<<<(>)? >E? @ A( + J ) ( G ! #$II))):I

*9$R$S

#>6E*6<<;? >.? @3 A( #4%% N % 8(G !#$II))) % I>6E *6<<;? >;? @3 A( N : $ ( G ! #$II ) : $ I) : IT6;U > 6E*6<<;? >/? O# %% &'( % (O ( ( (("-;.( 777>)? >-? @J * # A( N : $ ( G ! #$II ) : $ I) : IJ *U# U >6E*6<<;? >"""? ($ "6("-;0("

This page intentionally left blank

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

75

!"

" # $ # %# # #

% % # # $ #% % % # % # $ & # %%% #$'% %# ( )& (% )$ %% % ## $+ % % # # % , # % ## $'%# % % # & % # % #$'% %%# %%$ - & & & % & # &./&% $

0 ## .$

#$ -% # $'% %%% # $ - % 1 2 $ # & & % %# $ # % % 3% $'% #%% % & % $ &%& # #4(

) ( ) % $ % % % %5 '% % % % % # 4 % #$ '% % % # % ## $ &% ##

76

L. Mincheva / Reviewing CIDCM Databases

% ## 67.89$ ## ##%# $'%

&# & 4 $

% $- : #% 7$'%%- ; :<-;0&%

%// $$/#/=& % >./0 : <%//$$$ //=& % :< & %//$$$/ =$'% :% &%%

%%% 4 %:%& %%$'% %-;0 $ % % % & %# $ 2 % >./0 :&%%% $!&% ## & % $'% 2 #4 $ & % :& %% % >./0: % & 2 # % $ %% $ % %- ; :& %0 ' ;<0 ';= % >./0:% % #> 0 > ,$

! "#$%&$" '%0 ';/-;0<0 ' ;/- ; :=% # # #$ 7??@ % 4%,- >$'% : . #% # # $ # # %%# $ %% 2 : % $ !"! '% %0 ';/-;:%# % $!%( # )&%% %:1 &#5'% %4 %# ## $- # % % % & # #& % % % # $; # % % #$ '% & %/ & & $ '% 3 % # 7 % # # % ,# & 0 $ % 3 &%% %& # $ % &% #A% $& % %& $

L. Mincheva / Reviewing CIDCM Databases

77

& # & 4 $ +%#&% 0 ';/-;0% %# % $'% %#7?88%%BCCD$ 3 %E. 1 % # &%: #BFG ## $ '% 7@7& # #% % # $'%7@7 % #% & % # $ ; % % # < =HB &<=G@%/ &<=@C & <= 777 $ %# % $ '%## % # %# $'% % %# % # % & # $ ;7 B #% $;7% $ 3 #7?88%%BCCD$ % % % % 7?HC 7?GC& # #BCCC$'%%% # # 7&CCC %$;B ###%$'%#%% .- % # % $ D@$FI#% $

-% #%BB$GI# $ 2

% %7D$8I %7B$FI$'%# ## %7C$HI %;-,%8$GI$

$ # &7?88.BCCD$-6B9&$DG$

78

L. Mincheva / Reviewing CIDCM Databases

$ '# &7?88.BCCD$-6B9&$DF$< ##7 B #%,-10 <#- ; =' ;A% % %3 $=

' # ( # 2$! 50%# # # % $+# % $ #

$!& % 3 $'%0 ';/-;0 % GC . % $ % #% &

% $ '% # $ %

% 7&CCC$ %# & & % & #

$ & 2#H7G # %. % $ '% % :1 $ !& 4 % G8 ### $ '. % & B@ & DC $ ' 3&%0 ';/-;0 # $'%# % #& %#% 1 $ '% # # $ %# ## 2 % $ & %# %&%%

# $

L. Mincheva / Reviewing CIDCM Databases

79

) ( & $ *$ &( ' #%0 ';/-;04 $'% % ## # & %% # 3 % # # % $ %% G&8CC. % 7?88.BCCD &7@7#%:$'%% BI#% #<%% % # # =$ ! 3 5'% %

$ 2 %# #% $'% %# # # # &$$ # # % # % . $ '% % #:$ 2 %. # # $'% %# % # % # %% $'. %.4 %%% # % $'% #D7#% G8 % # %

%$' 7% $ % % % 2 % 1 % %# %$% 3$ 4 #% 2 $ & # % %% # $<; &679&$D7$='%# 2 %# %$

# % # !

(

" !

#2/ 2

0

>/>0

2

/% 2

# #2 2 0#% % 0# % >/ % J%

% #>0 ' <<K 2=/>0=

0 # 3

-

>0

J : %

%% #

<# / = <# / #=

% #

##

0 %2 2%

#/ '% # +'%L$C8$-679$

80

L. Mincheva / Reviewing CIDCM Databases

'%# % #% #% # $ &% $%%# %%% < = 1 &<=# & <= $ % % % #

$ % ' ;& % # % %6B9$ '%:1 # $%%# < 2 # >0= # # $%5M2 % % % % % # $ 2&# <# # % # # = % # %%# # $'%& # . # $ #%0

%% %# &% 3 #%%212#&

% # %# 2$B ;%&#%% # % #% %$ # # % # # 4 ##$ ( )# % % # & & %&$ ' & ! ) *% " + '% % 0 ';/-;0 : # $ # %% % 2 #%% % $'% $ # & % ##%# # $!&% % #$! #####%%4# $'%2 ## %#%# # # $ '% >./ 0: % # % $ 7??F% # %#% &%: 0 1 2

% #& % $ '% : :$ '% # : $ #%0 ';/-;0 & 4 # ./$ '% # # % 2 # % % $ '% : $ & B '% $ :&% & % % & 3 % % $ '% $ : %//$$$/:/: 5N7F

L. Mincheva / Reviewing CIDCM Databases

81

# ./ # % # % $'%&% ./ % % $ D $ % #%>./ $'% %#% $ ' ,-" '% % >./ : # . &%:1 $+ ! ##& #% >./ :& % # # ./ % (&2& /# % O&% # &%#% %O%

&% & & 3% )<6D9&$8F=$'%#./ 3 % % % %%#%# # % $$'#./% #$M%% & %&$$& %%3# # % $M%%% &% &$$& % % $> ##%:1 $ % # ./ % %

%$,(. ) %#

&%#./ 2 # % % # 7?88 BCC7$ '% % DG # ./& % $ 2 # % # % 2 %:' B$ '' ,-" . '% % # # $ # % 2% % % $ % % &% ## & O #% # # $

# '> <2= , - &8/FF.7/?7 +&7?FF - &?/F?.7/?C + &8/?B.77/?8 +&7C/?D.8/?@ &@/[email protected]/?@ - &7B/?F.G/??

-

0% > 0 > > > 0%

( 78&CCC.8C&CCC 8&CCC.BC&CCC 7D&CCC.DC&CCC BB8&CCC 8C&CCC 8CC&CCC.7&CCC&CCC 7C&CCC

-6D9&$HC$'%##%7?88.BCC7$ D

&%#%. #%% # $

82

L. Mincheva / Reviewing CIDCM Databases

'% # 2 $ '% % & & & % & & % 2$ ' D %

# % # $ % %

%GC 3%0 ';/-;0$ & % $ % % % 1 5 % $ & # # $ (0 % ) % &# & . &&$'% # (% # % )$'% % % $ % &%%

# $ '% % # % % $ % # % % $ '% % $ % %% % % 7?88$

# .'%>/ /

7$ # %

$

B$0

&%%% % % 7?88$

D$0

2# 1 $

#$ 2%C7C #%0:1 $

@$%

#%% % #% # :< M=$

#% % E.1%E.1 # :$

#%% # % $

8$

# $

H$ 2/

'% #4 <2 K=/7CC2>0$ '%1% 3 % # $

-6D9&$HB.H8$

L. Mincheva / Reviewing CIDCM Databases

83

$( % &

&# $% ./ % 7?88 7??H% ?C # # <6D9& $HB=$ ;%&% ./% % 2 / ## % # $ !& % % < =%2# % # & <= # & <= $ 2 %0 $'% % C 7C # # $ / &'% % % %&% %$% % %%/&% %% # #3 $ '% #%% % #% #% :$'% # %3#%& # % #7FBC%%%7??C$ 2& ## % # %$ % ##

% #%$ '% % %(.)%(.) # $'%& % # # % % #%$ &3%& %% # # #% #$ #%% .% $ 0 ( &('%# & $ ./#% 8C % # $M%#%' ;% % 2# &%% % # %# # % 7 7&CCC%$ 1" ( '% 2 $ #% % #% $'%# &%# &%%% %. % #% $, % % 3 % % $ '% >./ 0: % / # $ % % &% $0 1% . 3 $# #>0$

84

L. Mincheva / Reviewing CIDCM Databases

') ( & . ,-" &( # # % # ./$ M # % # $ ' % %#% # $%%(. ) %# & %% % % 8CC&CCC&% 2 G&8CC $!&D8 % % %:$'% # # $ ' #% # &%:% % 3 % 0 ';/-;0$ '% 2 # ##%1 # $'% # % # % (.#) $ %& %% # % # % 1 $'%1 % %2#%0 ';/-;0$'% # % $ 4#$ '% %### $ & %#2 $'% ' @$'

2 % % %# &% 2 $ 2 % # ## % #$7C % % # ## %#$C8$'% &%% # (%2# % )&(%#./)&

( )$ %# % ##$ !&% ## $ # # % %% % #./$?C&%?8I# # $HH $?F % # % # $ # % < =%% #%$% %%

$B8 &G@I# #<6D9&$HG=$

# 0; /

/

1

*

0 % 2

!%

7$GC

$C8

0

0.7?88

D$D?

$C?

#

2

B$88

$CG

D$8C

$CD

%% #

%

B$8H

$C?

'

B$8F

L$C7

-6D9&$HH$2 '%#D$D? #% 2 <.7?88=% #$'% #
L. Mincheva / Reviewing CIDCM Databases

85

. & 2 ! " + 34 '% 0 ';/-;0 % ## # & $ '%: # $-% # %0 ';/-;$!% 0 ';/-; % $'%# 4 ## # # $ ' # %& % ./& # $ -

%#% %4## % $ <'% # # % 4 # % % %% : % 2 $= P%#% <7= # 3 # $'% % %1 A
&8%%# & @% <6@9&$88D=$ '% % ### & %%%7?FC 7??C% % % # % #$ % 7?FH ' > 4 /4 % BCCB % & % : % # # % BF8 $2 HC#%#

% 37?FC 7??H&%% # $ '% # % % % %4 #% $- # %' 8$ #%%%%3 &%3%& % ##$' 8 # ## %$;2 % #3 %% % 2$

86

L. Mincheva / Reviewing CIDCM Databases

# 5 #%7?FC 7??C<2#=

>

"# /6 J # #

J P 7?FF.

>./

3 : 7?FF.?@

-

7??7.?H

>

% 3

7??7.?D

+

7??B.?8

+

7??B.?8

+

-

7??B.?8

%% 7??@.?H

-6@9&$88@$

)' 6/ * 3 . %5% # %#% # # 5'% # 4 % $ '% % # # #% & # $'%#1% $ % % #%. :##% : $ '%& # $ % #% #$" %%% 0%

$@ % > ! ##1 # % # <6H9& % 8=$ # % % : $ % %% # :$ # # $ % #%. $M %# $'%#

3# &%# 2#%3 # $'% %%# 2$,'M-' %2#

$ %7??C&# <0M-?@=& < -?@=&

@ < %//$$$/ / $ = # %#% % $ 2 %% %##% #% :$ %:1% $

L. Mincheva / Reviewing CIDCM Databases

87

< ,-?@=$ # % %2%% %%. : & $ > # %2 $ # # 3 $'% #% #

# # 3 $'% % # % 4 # $ > # $ '% # % % # <'! ;QQ=$ %2%% ##

# 7$" ## #%. /

<,'M-'=$>1% # . > # 2$%% %%% %

%% # #$

[email protected]??8< -?@=& #. 2$ 0 <0M-?@=& . 2 # %%%### $ < ,-?@=$ . 2$ > #

<'! ;QQ=$>## &-7-&%% 7??C %2& &&#& $

> #

% -< -'+= #%# : % %#%%%$'% % #%0

$ 8- # #7??H&## @ : % 7??D.?H D : % 7?FH.?B B : % [email protected] 7 : % 7?8H.GD C : % 7?8H

3 < M+?=$>3 % 7??C %2% %# 3 %##% $

-#P>

$'% % 8P & & &[email protected]??8$ @P [email protected]??8$ DP% . # [email protected]??8$ BP %. [email protected]??8$ 7P% 4[email protected]??8$ C % #%$ -6@9&$8G8.8GH$

88

L. Mincheva / Reviewing CIDCM Databases

#%% &#&# % # $'% 3 < M+?=$ % #$'%

% # 3 % 2 # %$ ># % $ % # % %1 & $'2 #% $ % : % % # % < -'+=$ % % #%$'%%% & % %:$ '% % # % % $ '% % % # %

% $ '%#&# & &

%% $% 3$ '% 2 7$'% 2%# 3$ %% F$ 2 H$F&%% % #%- $% %2D$GF& %% % # 7??8$ %% % % # # # 5'%% % %%% % %% $'% # 2 @$C<6@9&$8HB.8HD=$

89

L. Mincheva / Reviewing CIDCM Databases # 8 #% 7??C<2#= !% 0 #

> #

>M #

2

3 : R

K

!%

H$8

- R

!%

!%

8$H

%% R

!%

!%

@$?

P

!%

K

8$F

+ & & 3%

K

K

D$8

K

K

D$@

;

B$?

B$F

'% %E1 4 .&%%% # 7??8$> :7??8$ < # 0%

#8%%=$'%#( ) ( ) %# # %# $'%% %

C.C$@NA[email protected]$GNKAC$G7.7$CNA7$C7.7$@NKA7$@N%%$-6@9&$[email protected]$

0 & (9 ! ) % % % # %% ## $ '% % # % % # $ '% 0 ';/-;0 $ '% >./ # #% # #%# # $ &%4 #% %# #% 7??C$'%#%0 ';/-;0:$'% % # % %% # % # $ '% >./ : % & %%

$'%2$ %% # #< = # &<= <=$ M % % & % # % 4 # % $ % .2 # 5 >./#4# & % $ % DD#D8./ % 7?88 7??H # % $-.#%&BG %& # & < ##%=;%&8# %# &H #&77 #% # & 7D#%# <6@9 $88D=$ !&% #./ % #$!&% & $ #

90

L. Mincheva / Reviewing CIDCM Databases

& 3 & %% % 2 # $ % %## $'% % %#%%#% (%%# # %%## #)<689&$GB=$ % $ 3 % 1 & 2$ 9 ,-"

' %& % % % #./5 ' F 3 % ./ $ '% $'% %# & % & # % $ : % % % 1 % # # % $ % % # &## 3 % & %#%1 3 & # 2$ % %%# &% & % & % %%% # %# ./ $ '% %% # % % $'% ## $ '% % 1 2 & # % % & % %& # % # & ## # #% %$

# :- #./ /

-%# %

#%

-%#% ## 3 % %#%1 3 %# 2

1 % 2 ## # &#%#%# $ ; /%% - % %# % %% # % %% ## # -6@9&$8H7.8HB$

L. Mincheva / Reviewing CIDCM Databases

91

9'

/& ,-"

'%# % $'% #% $'% #% # $-% # # # & % % % $ '% % 3#% ##$'% % $ '% % S, & 4 %$'% # %&%$ '%% . $'% . $. 4 %& #% % 3#% ##$

# ;> # M# % ##< = 7$7$ % 7$B$'% #% 7$D$ %& 7$@$ 7$8$0% # 7$H$0% :& 7$G$ ##<# = ' 2 #

& # % ## B$7$- &%& # B$B$- &%&& % B$D$ %# . B$@$' ##. B$8$' ## % .'% #2 & # # %% # &%% D$7$'% # D$B$'% # # D$D$'% # 0 3&#% @$7$# % @$B$# # @$D$-# : -6@9&$8GH.8GG$

5 8$7$ % 8$B$'% #% 8$D$ %& 8$@$ 8$8$0% # 8$H$0% :A& 70% % <=& H$7$2% # #% % H$B$0% % 8 % G$7$'% #% #

G$B$ G$D$ :% G$@$ G$8$2 # G$H$ G$G$ #3% :# . F$7$#% F$B$ % : F$D$ F$@$; F$8$' F$H$ 2# F$G$ & 2&

92

L. Mincheva / Reviewing CIDCM Databases

' ? 2 $ '% % % #2 ##$ % % % 2 # %$ 9) $ '% ./ % # $'% $'%# + & %% % % 2 A % + > & %% % $'%# # &%%# $'%,# %%> -<>-=&%% % #$'% : # & %% ; +- % 2/ 2 ,$ #$'%# % # $ 0 # % $ # % &#

%%%$;D%

% # 1 %$ '%%%. % %%% % % # + $ &#%% 2% #+$

% % % #%# % % # $ '% #

# %%% % # # <7= % &$$%2 # &
% # & $'% %#%2 ##$

5, '% % % # % # 4 $ # % & % ( ) (

L. Mincheva / Reviewing CIDCM Databases

$ .# #% % #689&$G8.GG$

93

94

L. Mincheva / Reviewing CIDCM Databases

)$ ;%& # ## # # $ '%#%#%# & # # % & 4 # & % ./ % $ '% % # % # % # % $'% % %# # # $ % % 3 # # 1 # $' 77 # $ % # % & % % $ %

#% $ '% # % &# & $'% # ./ % & & & & % $ # & % & $;% % #$'%& & # $ & % %% % % $ '% 2 # & # & ##$ # - # % # %

'%>./

'#

&

2 # A M#A ##$

'%

-4

># A > # A ># $

'%>./

-

2# % A M#./A A A % A

$

'%- ;

-

M A

# A $

L. Mincheva / Reviewing CIDCM Databases

95

679 $ $&>T$&>'$$&! ##+$&-0$'$&,$ $ %$-$&('%- ; 0: %#,$-$;00 )& T$$ >'$ $<=& && 4 % # ( / $ * $(&7??F& $BG.DF$ 6B9 >T$&+ $!$&>'$$& $& % $&,#T$ $&(> ; # 0 )&0 % #% 0 - & - 7.@& BCC8& %& $ %// $$/#/0 '; $#< 7DBCCG=$ 6D9 ! ##+$&( #%! 5#> 0 - 7?88)$ ( && BCCD& ; & "$ ?G7& $ 8G.GD$ %//$$$///! ##0-BCCD$#$ 6@9 ! ##+$ >'$$&(- #! )$:

&7??F&-&"$D88&$887.8G?$ 689 ! ##+$&( #! -4 %# )& T$$ >'$$& && 4 % # ( / $ * $(&7??F&$GC.GF$ 6H9 ! ##+$ >'$$&/ * & 0&BCC@$

This page intentionally left blank

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

97

! " ! ## $ $ #! ! % ! # $ # # # &# $$ % %&$$% #!% !% !%

! " ! #! ! %# # $ $ $ #% # $ ! ! $ & $ #&$ ! # # $ $

#! ! % ! # $ $ $ # # & !' %" & () * # , # # & # $ % $ ) * & ! # #- . $ " $$ $ % #/01 # # # ! # $ # $ 2 3/41% 052067 # $ # #! . /81 #& ! $ $ "

98

R.D. Kuperman / Indexes for Measures of Conﬂict and Cooperation

# $ $ ! #! % ! $ % ! $ !

# " & %# & % &

! # # ) 9 $*% & !9: 3:7/;1% #! # #- " 04 $ #!% & $ $ &$3 #07< ! & 9 ! ! & ! ! ! !' # $ &$ # ! & &$$& & #! ! = & % & />?@1 9 9$ : 9$%& # &$ & $ $$ $ $ # % # # ! ! # $$ $ $ # #% % ! % % % /A261B ) #-* # # $ &$= & %&$ ) * " &$ %# ! $ &$ ! &$ $ # : /51 #4

" 9 9$ # C ' : = F D B I/;1

! : B #! B

! $ D & E

#! C !G !

$% 04 04 04 04 04 @ @ @ @ @ @ ; 0HH

R.D. Kuperman / Indexes for Measures of Conﬂict and Cooperation " : J&$ & "!

' 448 ! 9? ? 20HH 400B, 264 444 !

K- ! 25A 440 - !

258 054 #, % % !? !# 2A@ 06>C 9 2AH 0A8 & 2AH 0A; ? &$ 2@6 0A4 &$ ! 2>5 2>@ 068 ? K 050 ! %& 9 2>4 4H0E ! 2>H 4H4 $ , $ 2;6 0>H %% 2;6 0A0 & $ 2;; 404

37 2;; 064 !? 2;0 004 ? ? & 2;H 000 & ? - % % 2;H 06;= $ 285 044 ?$ ? # 28; 0@H:& $ 28H 084 24; 040 $? ,?# ? 244 060 244 080 9 3 7 206 H@8: ! 200 0;4! # !% % 200 0;0! 2H6 H48 2H4 0H4 $ $$ ! 2H0 H40 2H0 H6;" ?

2H0 H4> !? HH H609 H0 H00B

%! % #

H@ H04< ? ? H@ H80&? 0H H6> ? ? ?#$ 04 0H0E 0> H@0 $ ? $, 05 H84L?$ 06 H@@ K ! 06 H08 & $ $? $,% 4H H@4: 4> H>; ? 45 H88? 45 H@>B ? ?

46 H54$ % % $ 8H H649 ! 8; H689 8; H;0 % % % 8; H;4 M ! ?$ # 8@ H>8 ;> H>0 & ! ;> H>4 >4 H@;: $? $ ? >; HA8: @> H50 9 # $ @> HA0 ?$%# !%% %#

& A; HA4 ! 58 B I/51%8A@

HH HA H> H@ 04 08 00 0; 06 0; 40 0A 0; 0A 0> 48 04 0> 0> H6 00 08 H6 08 0> H@ 4> 0H 08 H> 0> H@ 0H HH H; A4 @@ H6 05 06 0> 4; 4A 44 4A 44 8H 8@ 4> 00 4; 40 05 0@ 0A 0> 0; 06 0; 0H H6

99

100

R.D. Kuperman / Indexes for Measures of Conﬂict and Cooperation

E $ $ # ! #! $ &$#! $

$ 9$ = & % ! &$ " #: # !$ B !%&B J /61 # 0A ! ' ! # ! # ! % $ & &$ & # 9 $ & 9 ( D9& 9

#! $& # 9 ( $ #! ,/0H1 $ # # % $ #-! " - # ,J !$ #- 9$ %

9$ # # % #! $ # ! & " #& ! 39 & C7 & #8%& & & 2C #&0; #& ! 0 !

$ $ 2 % 2C & 08 & $ % 3 $ & ! $7 ,J ) & * # % # ! # #& 2C & 2C )B & ' *=% 2C % & ! &#

" ( ,J0N C #& ! 0 C ' # E B, B' C0 ' E .

0 H C H H 44 H H H H H H ' 06 4; >> 04 H # >; H 4A H 0A H E H H8 0> H5 88 H> H B, 8> 4H H H 8A 05 0H H B' >> 06 ;6 H; 8@ 88 44 >A H C0 H 0; H6 H H8 H H H 0H H H H 8; 8H >4 H 8H H A8 80 H ' ;0 05 ;H H4 46 46 0; 8@ >0 H >0 H E H H H H 00 4H >H H >H H H 00 H 0 2C H H .

0H $$ H4 HA H5 H 0%& #&& #3 HA H5 H6 H5 H H H0 H0 # ! # 2 $ $ ! & $ 7 2C 0% $ B I/51 #! #&& ! %& H%& 9 #&& ! % " H@0%& '$ 0

R.D. Kuperman / Indexes for Measures of Conﬂict and Cooperation

101

$ #! ! $ /0H1

# % & %$ & # % % !# #- $ &'$ 0 % '$ 0% &$ $ ! #! ,$ 0 0HH%& 0 0HH

) ,J

102

R.D. Kuperman / Indexes for Measures of Conﬂict and Cooperation

& - $& ,J #

# # # # $ , $! , #- &$ = & % $

$!# $ ! #-

! # # $ ! ! # $ #! #

$ # ! /00?041 # #! $ %! ! $ # % ! $ $ % $ &#

$ $ $&! # #& $ $ # # $ - $ ,$ $ # # 2 9 $ # $B% # ! ! #! 2 9 9 % $# # (" *

! 9$ # % & $ $ $$ $$ ,J%# ! ' % # $ # # $ $ &# " #&$ % # ? & % & # % ! ! 2 9 9 ! % # $ & # # $ $ $ " 9$ &$ ' % ## & 2 9 2 9 %& ! &2 # O % # # ! ! $ # #!

% # & !. $ $#&(' % $ #& 9$ ! $( = & $ ( $ & (

R.D. Kuperman / Indexes for Measures of Conﬂict and Cooperation

103

$ $ !$& ! #

# # $ #&& & ! ! #$ ! ! ! %& ! $ ! $2 % ! % # ! & ! $ ,

# $ $ # # - $ = & % # # & # # # ! % ! # # & #! #

$ # ! $ , 9$ %

$# # % !9! # #! $ $ /001 # $ #& $ ! ' % % $#

= #

+ /01 L %).!B !B #- * %4HH4% L >5%882;> /41 F : F%

%&< 9I: B %065H /81 : % ).$$ $$ $ *% $ G4HH@ G !B !B % %$ %4;24@ % 4HH@ # IKK&&& $ 9KK"4HH@K & K0>P /;1 9B : 3B7%) !9: *0@ !4HHA # IKK&&& $ K:P $! />1 # C%= ! L9 % ) 9% ' 9 9* %066@%L >4I@%452;@ /@1 D B %)$ & !9 $*% =, B B 3BB7% 4HH4 # IKK&&& 2 - K& 9$ K /A1 , %) # 9 C 93EC7*%! % ! =%065H /51 : FB%) 2 B .B * % 0664%L 8@I4%8@6285> /61 B B%) $! I& 2D9 B * %4HH;%L 8H%0H620;0 /0H1 ,Q% !" #$#%&#'(%%C %.& %0654 /001 % ) * ! +, -+ ., / %D %ID$ C 9%4HH> /041 %)B $ ! . B D# *+ %4HH@%L 4%024H

This page intentionally left blank

Opening up Conflict Data

This page intentionally left blank

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

107

! "" # " $ % & % "'( ' ( "%

') % ' $ " $ $ $ $ $ *"$" $ '

$ %"" " " $# " $'( % ' % + ," - ' % ," " % '( " ' ( %

"%'( " % ' ( % !$ '(

"

% " ' % # % . ' /' 0' 1' 2' 3'

"% $ % $ ( % ! $ % %$ !% % $ 4 , '

5 ! 7 8 (9 ;< 7 0?@@@ A 7 ' ('. B00 +@-18330?' R code reproduced here is covered by the General Public License Version 2, June 1991.

108

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

D %" %

" ' ! % % . $ & $; '( % %' " 'D % # $ ! # ' # '( , "'$ & " ' E& ! %& * $ % !

$ ' &F ""% +GH/I-

! '(

% # & ' ) $% * ! "# ') $ % % $ % ' % % % 'J" $ &% % ," % % ' K" % 'D $ $ % ""' ( %% $ $ "

% $ %

' & %

% ' % ') " % " $; ' !" # %&%

'/( " " '( # % $ '( , &% /) ' '

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

109

"% ' L% M '4%% $* %'( $ " % &')

$ $ %$ ') " & $ ' &

') $ % 'E$ +" -

% ' $ "% % % % ') $ &' ; % &% ' , " ' $% % " $ $% 4$ % $ + $E $ - O "+ " -' ! % % % &

H $ %" $ $L( ML( JM$ % . ( '4 !+- % $ &% O + G0I$ " $ /-' ( % ' $ $% ' $% L M L M+! " % -$ L M " .%L M $ L M $% $ LE M ' " % " %

% $ &" "% ' % " &, $ %% " ' $%& ' ( )*+ ( +% - " ')"" $ '( * , " ' % "') , % % " '

110

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

" $ $ $%% " +''

%& ."%& " -' % + % # - " $ &

%" ')"$% %"$ % ' $%%

$ ! ') $ %"&% '(& " % $ ; ; 'P%"$ % % + -$ ,

'4 $ & " 0$ , +% %# -' # % *$ E '( $ " ' $ % Q& ( % +%%%' ' -$ & LDM LR &M E' J! % $ #% $% $ ' $ "% $ % $% $''$* $* $$$!+& -$ ,+! -$ $ $ '(L M % & " + L% M- ' ( L M % L M L% M+'' -'(% 'P%"$ ! +! -' ( % $ " % % % " ' (

$ $ " " $ ' ,& -. ( ) ' (%% " + % &$ -' 0( &% " %% %&'

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

111

( % + %%-% ; % + &! &! -' ( % ; & " " ')% " " $ " % O

" " ' (" "

'P%"$ ," " ' $ ! ' $ ) ) '1D ! % % $ " .% " % ' $ ! % " " % "H

'D $! ), % ' !* ' ( % % %

'( % &2/@@1'

" & $ " ' /@@2 /@@3' " /@@3 % '(

S/@@8' ( (! & . % % " $ ! &%+ " & -$ + " % -' " $ ! " % '$ $

L M ') $ '

1)"%$" % && % '

112

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation D %

!

"### "$### # %

#

/@@1;@0;2

/@@1;@0;2

/8

/8

/8

/@@1;@2;@

/@@1;@2;0

3@

8@

32

/@@1;@3;@

/@@1;/;0

/@

1@

0@

/@@2;@;@

/@@2;@3;0@

0@@

3@@

2@@

/@@2;@8;@

/@@2;/;0

@@@

/@@

2@

/@@3;@;@

/@@3;@3;0@

@@

2@@

0@@

/@@3;@8;@

/@@3;;2

/@@

/2@@

/0@@

/@@8;@;@0

/@@8;@;@1

/

/

/

Q

&

!# #

/@@1

///

/@@2

/32@

/@@3

03@@

/@@8

/

' +#-

( +#J-

&

)

&

)

/@@1

+L M-

/@@1

@+L%M-

/@@2

/+L; %M-

/@@2

+L%M-

/@@3

/+L; %M-

/@@3

+L%M-

/@@8

@+L%M-

/@@8

@+L%M-

L M %@@ 1TT " ' L; %M 2@@ "'

* +#F

!

/@@1

/@@3

L%M @@@ "'

/+L; %M-

@@ "')

$ 2@@ $ L; %M$% L M'

+ +#JF

! /@@2

/@@3

L%M @@@ " '

, % #

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

113

( 3 % " % " '(% % . $ ($ (/'P%"$" (/ %('( (( / . " % % ' $ (1(3 . (3

(1' (" ' ) $ ( " $ % $ #+ $ #-'(

% #% %! ' !!"/ 0 ' ) (" % $ " " ' $ % 03$LA S M LM'4 $ %" + S S0@- " LM'( $

'( , * " ' ! "$ %

% ),J *' (8 % !%

"; " ),' )J% )J + $ L*@1M(8$ .OO%%%', 'O O O*@1-' (? ! % /'% +(8-% " $ L M" " + ! -' $!@0? "+/@/@@00 /@@0-2 ' (/1@ " +3 $ /1@O2U3-'( ! (?' ( ?0" %% %

' !L% M %'(% % " % # $ %% % 0' ( $ %;# % % % " % % # '( % % '

114

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation - ),$ " +!-

.

!

/ "## / "$## !@@ /@@0;@;@ /@@0;@;@ !@@/ /@@0;@;@3 /@@0;@;@3 / / !@@0 /@@0;@/;@ /@@0;@/;@ / / !@@1 /@@0;@0;@/ /@@0;@0;@/ 3 3 !@@2 /@@0;@0;@2 /@@0;@0;@2 0 0 !@@3 /@@0;@0;2 /@@0;@0;2 !@@8 /@@0;@0;/@ /@@0;@0;/@ !@0? /@@0;@0;/@ /@@0;@1;@0 //3 /1@ *@/@ /@@0;@0;/@ /@@0;@1;@0 // // *@1 /@@0;@0;/@ /@@0;@1;@1 1 3@ % G1I' ( " !+% 0@@@ -' %# ),J *'

0" ),$ +L M $!-

1" ),$ %&+L M $!-

2" ),$ +L M $!-

/@@0;@;@ /@@0;@;@3 /@@0;@/;@ /@@0;@0;@/ /@@0;@0;@2 /@@0;@0;2 /@@0;@0;/@ /@@0;@0;/ /@@0;@0;// /@@0;@0;/0 /@@0;@0;/1 /@@0;@0;/2

3

4 % '@ / /'@ 0 @ 1 @ 2 @ 3 @ 8 /'@ ? @ T 3'@ @ 0'@ '@ / 00T'0

" % S/@@0 /@@0 /@@0 /@@0 /@@0 S/@@0 S/@@0 /@@0 D /@@0 /@@0 " /@@0 /@@0

% '@ /'@ /'@ 3'@ 0'@ '@ /38'1 /3T'T 128'/ 011'8 /88'8 /8'8

" ),$, +L M $! /@@0K /@@0K/ /@@0K0 /@@0K1 /@@1K

% 0T8?'3 123@' T?@'1 1T'/ /0'1

"

),$ +L M & /@@0 /@@1 /@@2 /@@3 /@@8

% /@@'0 @2T'8 110'@ /82T'/ /02@T'?

% 0'@ /'@ 0T80'3 0101'1 212' 2?@'3 31?'T 8?2'T 212'3 1??'? 183'8 2/2'8

'Q""),$ " & ) 5 /@@0 /+L2@@@TTTT M/@@1 /+L2@@@TTTT M/@@2 /+L2@@@TTTT M/@@3 0+L/@@@@ M/@@8 0+L/@@@@ MV

.E"@ $G$2-$ /G2$/@-$0/@ +'0'-

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

115

6% % % 6)

6.

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

116

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

, ' " "), ( % % # ' (% G1I% /'"& & % % & &+''LS'ML SM$ L/@@0ML/@@0M-'( % +0'0'- " # + &" -' $ +0'- % S/@@0 0 /@@2$ %2$@@@ /@$@@@$% S /@@30 /@@8$ %/@$@@@ @@$@@@'

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

117

!$* " -+ " # ," ' ) $ " $ "

% 'J % " ," " " + -$ ' ; "' " S 0 '( " S 'J" %" ' %

% L" M' $ "/; , $ %

.L %"" @@@ "/; M' E % %; %@@ TTT; $% " @@@ ; '(%+ - +(1-" ')! $ ;"% J %;%; + S/@@2 /@@3-'P%"$ "" @@

' L" M% "" ' D & ' $ $ AR+ -% "1)') ! L AR M$ ! L AR ! " M'( % ! "') * L" M$ % $ %$ '

( % %& ! <

=# % =# %

S/@@27S/@@2

@

S/@@27 /@@2 @ S/@@37S/@@3

3@@

S/@@37 /@@3 3@@ S/@@87 /@@8 @

@ 3@@ 3@@ @ @

118

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

'4% !% / '( ; ,

"' )

$ "" "

% ' ,$ " ," ' D $ " "$ " " ! '( " " # !' $ . " "! $," '" & %" '( $ $% &%!%" "+ % " %"" -$ $ $%! ' (% " ' $ "

$ % F $ F E'

% + -$ " " % " &%!% '( % &%% % & % " ' " " % ' ( % &% " " % $ , ' " %" '2( '%;#

J;

G2I$% ' " " % " 'P%"$ % %'" $ $ " % '( " %+ $ &$'-'3 & ' ( " H %%

'P ! $ .( % +

"% $ * &-'( % 2( L /@@1' "M+ !-$% LX M% " % L/@@1'M+" -' 3 4F % G3I$ &F LD &/@@1''

M+% $% " -

L /@@1' "M% " 'J $ ;" $ + L /@@1M %-'( & 4D /@@1'; &'

4D /@@1'' "

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

119

+ " " -' ( % & + $ % "$

" -' ( , " + L"M L" M-. "'@'@$ "'@'$"'@'/$"''@$"'/'@$"/'@'@' $

"" $ &$ % ' $ " % L" '@'7@S/@@8M+@O@ % $ -'( "+ - $ " ' $

" '8 $ %

, " ' $ " " % ' % ! %

% 2@ 'D $ " %% " '( %$% % %% " $ &%%! ' ( % "% " $ 'D$ + %- %% $ LM + 4 % -L M

+ E!%4 % %7& E! -H $ % % 'P%"$ ! + -H $" + - " % $ '?( & " %

$ + -' $!* + . (, " & " % +'' - " ' " " % & " # " ' ( " "

' %" ! 'EF S R1 R1 'P %

& & '( & + DKE- 1'

8( % $ % %' $L8'@M% /@@8' " %% ' ?( % %L M G1I+),J -$% $ L)J M$ , 'D

$ % '

120

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

:3B%3%BD03 25%65"2)535E%5600! ,25%5E%5600! F25%5E%5600! F%3G%5E%5600! - 25%5E%5600! 25%5E%5600! H:2B:$I3$ 35)253($2"B :3B%3%BD03 02I3 B0%3:%BD03 :EHE065 B0%3:%BD03 BE06525%>2:"% B0%3:%BD03 B253G B0%3:%BD03 BE065 >2:"% B0%3:%BD03 BE065 JB:&B:@CC>2:"% B0%3:%BD03 BE065 %23"%BH>2:"% :3B%3%:2))3: E5 >E:3B&:EKD3)25 25"3:%25%E "3% (6"3:! ! (L6HB%3L! "

!# # (E0! $ "

,(E0 ,! F(E0 F! F(E0 F! - (E0- ! (E0 35 :3B%3%:2))3: %& 'E5 >E:3B&:EKD3)25 25"3:%25%E "3% (6"3:! (L303%3L! (E0! ,(E0 ,! F(E0 F! F(E0 F! - (E0- ! (E0 35 :3B%3%:2))3: ()*E5 >E:3B&:EKD3)25 25"3:%25%E "3% (6"3:! (L25"3:%L! (53K! $ "

,(53K ,! F(53K F! F(53K F! - (53K- ! (53K 35 , (P% 4;R1 +%DKE2'@'@%-

121

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation *D #

/ ( # / (

/@@8;;0@ *C @.@@.@@ T/'3?'@'

)D(

/

/

D

?@@

/@@8

/@@8;;0@ *C @.@@.2 T/'3?'@'

)D(

/ 032

032

?@@

/@@8

/@@8;;1 *C 2.@0.@@ T/'3?'@'/

R( / 032

032

?@@

/@@8

/@@8;/;1 *C 2.@0.@@ T/'3?'@'/

)D(

0 032

031

D" T// D

TT

/@@8;/;1 *C 2.@0.@@ T/'3?'@'/

)D(

1 032

032

/@@8

TT/

S % 4 'P +"- " 0@'(%%& $S % DD " 'S &

'T! (3+! -'( D + %, / 032- % ( 2' ( ' 4" % $ %& + "-$% % ' (2$ ,

$ &%% %$% % + )R

-' " %

L M $ ! %% % ' ! " R' (R TT8 ' ! $ "% $ +

-$ %$ $ " % " "@'R &" +G8I$ '/-' ( " " ! " "') "

'(&! (2$" 0@" /@@8+ "- "'@'@ " 1 /@@8+"- "''@' TS S '%$JD%& ", % & % . 25"3:%25%E JB063"/!@!@!L6"L!/M??!@??N 25"3:%25%E JB063"@!AOC!AOC!L:L!/M??!@??N 6HB%3 "3% (/P@/K&3:32(@ 25"3:%25%E JB063"A!AOC!AOF!L6 " +" : L!/P@@!/PP/ 25"3:%25%E JB063"F!AOC!AOC!L:L!/PP@!@??N @( " $ .OO' 'O !' L (QM'& LM

LEM '

122

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

$$ * + * + . $ " !

$ %& & '"$%

%&$% "" $% % " " $ "%$ 'P " " " ' $$ ) )') ' 4 $%

" "' ( ' " ' $$!2 *( . * ' ( % ! & $" ' " % " , ' $$$) ( ) % ( % $ $ %'P $" %%

% + . "* "J$ " $ -' $$,0 ) ) ( %" % " 'D $ $ & % 'A"F

" " + $ "" -$ &

'( %$ " $% %" ' $$3) (' 2 ( ' $ " $ "') ' $ " ( &

% '( % " ! +% # -'

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

123

% " " $% " 'D $ "" %"$ H

% %& %& ' " % " &% % $ $

! LM % $ $') %&$ L M + % $ % ,#-$ ') $ $ '(& ') $ " " ') $"% $% % " + $ " % -' (!% % #/# ( "% ' " $

'( . " %' ,

0 4(

& / % '( " '( " 'DJF +G?I$ 1' 2'0-'P

" ' 4 $ % ! %+ -$ &' 4 ! $ ! % % ') " " $ ! %

$ E$" +$$ & " -' (

$ ' )D " %')$%)D 033; " $ " ' , % %

'GTI ! %, $ " '

/ L& M! '

124

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

4 " $ ,' ! $RF +$ED$ E'''-'R " %'0 (% $ " " % + -' ( "" '("

% +$$ -$ ' D$ @@Y % $ , ; + DKE *% .DD % DD)/@@@ DD/@@@-' $"

& $ % "+ -2 %; $ %; % 7 4 R &' 1 ! +%" - %(3' ,!+ 4( $ &% "

' + $ - JD+ D -' JD DKE +D K E- , 'P%"$ DKE" H % ' $% LM L M" JD+ R DKE$ DKED"/@@@ J/- " . " + L " &% M-'( $% LM L M % (3'

+!! % 4 R1 / 0

/ 032 032

/ ( / 032 031

# / ( D D"D

?@@ ?@@ T//

/@@8 T/ TT

1

032

032

TT/

/@@8

0D" .OO' 'OE ' Z U@?' +E 8 S/@@?-' 1' D JF + G?I$ " $ 2'/- L 1 1 1 ,(AOFAOCM & R1 4 $ & 4 R1 '4;% $ ! ;% 'R L M "% % L MH %"R " % R %" 4 'D $ " % ! " ' "$ DKE; +JD-" % %;; $ %;; '

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

125

& % % " &% " .OO%%%' ' O O ,[ " [% [&O')% JD$ % + -' ) JD " &% $ % & + %#% %JD % %-' ,$* ( D "

'4 $ $ $; $ # ; ' ( + G@I # -' ) $ + -$ %"

% + ),J *$GI /-' ,, ( "% $ % + $ % "-' ( D ) )+D))- /? $ %T2 2'( $ +

% $! - % '( % $ ' ) D)) $ D)) '( " $ " D)) ' $D)) & " ! % " $ $ ' ) $ &;?;3 $$ %%

' /3D))%; + -'(

; $ , !%&' 3 + $ $ '''- . %& $'( %' D)) %+ %%&"

-$ /3

$ % ' $& $L ()X )M L\ ])"M' 2( + %L M-.QR4#SLT)&2XI05EHY:"%6JKG$*8Z9[1\ -]. +,;'7^=_ 3) $ %

, '( %DKELE)M , + " & -'

126

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

( &% H % % % $' ,35) ( "+ %- 'D $ % " ' $ L M' % ;$% # +L@'2M L'2M-' 4 $ ;;" , "$ " +'2 /H 1'2 1$ '/2 '/-'( +% "- '( ;;" # )3@22T' ,6 ( $ +% %1 -$ $ )D ?3@ '( ! . •

•

( LQQQQM $%%0S/@@? /@@?@0 + % % $ # -' " + -' (! LQQQQ;;M %

+/@@?;@ S/@@?-'

&/@@?OO0 , %& ' &;; QQQQ$;;QQ$;;QQ$QQQQ" " ,' 4 " % + -$ % $ #8' ")D ?3@; $ ! +! # " -' ,7% 85 *+ 9 ( % " L"M$ "' ,$ " + $LQ MLM-$"L"M % 8 $ /% % /@@?;@;3(2@820^$''3S /@@?$2.@8 20 $A(G "% & I'()D ?3@ ./@@?;@;3([email protected]^' )D ?3@ " .OO'%& 'O%&O)D [?3@ + @S/@@?-'

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

127

' JD %" % & ' % " $ $ $#$;$ "L;33M' (! & R)X $% "%;@+- B@+ -' X % " 'P%"$

" $ ;??+L M-$;88+L M- ;33+L M-$% H + $ '''-%" ;@B@ " $% %;??$;88;33 '( " # " "

" ') $

% $ " L"M% & %'" % L"M + ,""$ $/ 0$$J $ -' ,:0 0 ) $" " " ') $% & " ! %' ,:") + L M-$ ! !$

% ?' " ' $" " , $ ' L "M& + -$ ! ' ) &$ $ !

! ! +''L $ @@M L M$LX$ ! /23 M L !M-')

$ % % + $DKE;

L"+/23-M % ! -' ,:!5 ;) ; $ % $ '(! & E+_E-% % ' QE % & " _E$ +% -'

?( ! , $ ! , %! '

128

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

( A %+ G?I$ " $ 2'/- ! ; 'P%"$% # $

" ! 'D QE $

# % % " " + LQE M% " "

-' *>. < # . ?/ !$"" " % " $ ; '( & + AR -'P$ " , ' $ %%& $ % ') " %$ %& ; % %

'(% & % %' $ % $ '( ; " , " $ % ' D $ '( % ' ) ' ( $ " , ' 4& %, $ " $ %+E D %$E DD-" " " ";$ %'%& E D%$ ) D A $"% " E D %G/I'4 %% " $ ! ' 3 0 "( *)( ( ( $ &'

'D % %% $

" " ' % % ' ( E D % " # ') " $% % %' $ " ," !$" +

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

129

$ G0I-'D" % " $% % ;;& ' 4 %$ L&!M &% %! T' $ % ! * % ' $ $ %$

' % $ "! %! % " ' $ %/% %" /'2' /'3''D % $ % $%% % H $% % " . " 8@" 7 & TT87"H *" + /'2' /'3'- "/@') $ , % / "% + $ % $ %-$ % ! ' D " !

"+ $ G0I$ '3?/-'4 % ') % " 'Q "" " ! ') !$ % " % " %'4E D %$% ')

$ %$ % " +%" -$" ' " 'E D % " % ", % $ * ' 3! ' < = ) ( E D % '( " % % ' E D % " $ %A % 'D " %" # $ " ' P$ " E D % " " . " % % % ' T % %$ % ' /@ 8@" $ .OO' ' ' OOO O O'4 % " $ .OO' ' ' OOOO% % O O O/'2'O

130

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

)

$ % " " /'D % " $ + %" - $ % E D % '( " ""

%$ %$ " %' " % " ; %' E D %

*"# %% + ;& -'; % % &! $ !

" ''' % ! " % ' $ +% - "08 +% % G1I-$% '+%

J -$ 8@ + %& 0@ G2I-' '$ $% % ' ( % " ' ( ; " $ + - $% ' $

H " % " 'D $ % " '4% $% % "% Z(, , ! % "' ( $ % '( " %& ' $ " "E D % '( "

$ %. & % '( % % ' 4 % ' + # " % $ % &' ( & " % ! ') $ % $%" ' +% - % ' /% + A -% ; %, $ * ""E D %+ " -'( " " "'

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

131

& " ') " % '( % & $ & % %'( ! +% -$ ' ( & % % % + &-% &' + $

$ $ %$ $ '''- " " &'( & "" %" ' R % %

% 'P%"$R % // ' ,$ & " + $ 4 -'P%"$ 4 $ " 4

/0H% $ & R % ' " D)D $% )D ; # +)D O) /30@@./@@3-' )

$ $ F % 'D"

'/1 - ( %!% % , ') $ % $ %

' $ % " L M $ 7 7 %

') $ $

$%" " %% '( " % %. % ; $% % * 'R " H%

" ' //R+ %LRM- %% R %+ .OO%%%'

'O O

-' ( ' %R' /0) $ ! ' /1) '$ $A $ )JE D $ ' ! ;

+ -)D O)/30@@./@@3'

132

M. Kauffmann / Enhancing Openness and Reliability in Conﬂict Dataset Creation

GI &'$LJF A . M$ "

$

R $ /@@2' RR W$" .OO%%%' '' O

OR[ OR[ '

' G/I & '$ L ""% ( . " " M$ '+ '-$%) <>% /) = * $ $/@@?' G0I J;) A''$L ( . P ;J ( M$ '+ '-$%) <>% /) = * $ $/@@?' G1I ),J *$L)J;) M'/@@?'" .OO%%%', 'O O

% O; ' % 3S/@@?' G2I

''$L)" J D M$' ?7, $% Q&.;P $T81' G3I R D%$ L ( A%. 4; ( D R ; A M$ R$ S /@@@$ 8 ' " .OO' 'O '

+ 1S/@@?-' G8I J'D'$L . $E $ D M$ ' + '-$ %) <> % /) = * $ $ /@@?' G?I A 'D' 4 ''$L) D ." E ) D ?3M' - - $ TTT$ X' /2$

' 0T0;10'X 0'/ $

/@@3$ // S /@@8' " .OO "%%%' !''&O `& O ;1' ' GTI A ! A $ L( D #A M$/@@8$T1 '" .OO ''O O O [ ' ' G@I ), J *$ L M' " .OO%%%', 'OO O' DL0''D M'E 8S/@@?' GI 4& ED%$LD%O D.) D

ZM$ /@@@$ 1/ ' " .OO''O' X '/ +%& -' D L " %M$

';0 .OO''O O " [ [ [ ' ' G/I ' 'R'$L D D%M''>' ' $/@@$X'01.0$

'3?;3?8' G0I $ L( /@@8 # " M' /@@8'" .OO' ' O; O OP@/08@@00' !'E 1S/@@?' G1I "%&+ '%&-$LE# M'@S /@@?'" .OO%&' " ' 'O %&O !' ZUE a U3@@/3' E 1 S /@@?' E " " .OO%&' " ' 'O %&OE ' G2I "%&+ '%&-$LE# M'@S /@@?'" .OO%&' " ' 'O %&O !' ZUE a U3@@/3' E 1 S /@@?' E " " .OO%&' " ' 'O %&OE '

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

133

! " # $ #

#

% & '$# ( * %

$ + * ,

# #

# - " #

! *

# * "# "

* . *

*

** ## !#

* #/ $

#

# 0

*

. #0 " #

$# * 0&

"

" 1 &! " " # 0 "

0 1 &! " # # - " . $ #+ 2 ## $ % ! #- ! 3 "%

#*

4

5 4 *$1 &

* * # 3 * * #

2 6*" ' ( * +

/

7

134

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

* 1 & # 6888 #! 9*"*

*

* . * # ' " (:*" $ # % " %

#* % " "

*# # * -

"

* *

# #! *

* #

*

# /

#

" 0 *

" .;! ';( # *$ # - . #

* . " " " "

# * *

+ * $#

* ## " # + " $ '1 &( '1 &2 ( * * " " " + " " ..;

* #

5 45 " #! " "

"1 & + " #*

* 0!3 8<<8 06 6 .* ' ( ' ( #

# . " # * # % " ! * " " #

$ * " ! 1 &- *" * !+

" ## ' (

#

# = ; # 0 # . # " #! ##

* * 0 # : * 0 - /# 0

# .

* * "

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

135

# > * # % * - # " " 4

1 &1 & # - # "# " " * # ! ## /$ "

# # "

#= # " 1 & #" #

#

! * ! '+ # ( # " " *

#** 4 688?** * "4688?1 & 688? @** * 688?1 & <88 4= 688?1 &* = 688? *" # $#" * 0 * #

4 0 '4 <88( * $# " # " #!

#/# <88 #' <88('688A( . 0

* $#" * # " $#" * $#"

'( ' ( 6B8 3>: # !* ?6C .> # !

#'(! + #/ / /"

= # # *

# 6 !: # $$ B

4

0

<
B?

0

<
B<

9

<
CD

9

<
DD

136

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions 4 #

!"

0

<
AD

0

<
A8

9

<
68

9

<
68

B888 C888 B888 6D88

# # !

!"

0

<
B?

AD

B888

0

<
B<

A8

C888

9

<
CD

68

B888

9

<
DD

68

6D88

# # * * $# " *

*

" # 0"#

" # #/# ##E #/# # # ' (

#

#

#

# * ; - * ! "" * #! 5 4 - * 5 .-+0 # * * +

* C!+

D! ' "( * * *

* !

$ #! *

%

5

5

5;

8

5;

8

5F

5F

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

137

& # *!

5

5;

8<

5

5;

5

5F

68

5

5F

6

' #

%

%

( %

5

5;

8<

8

8

5

5;

8

8

5

5F

68

5

5F

6

:

* *

"

'

( # #6 . * '

(

*

# *

#6 *" # #

#* # # #6& "* " E ' #( ' #6(#"

# # 1"

# " * *

5 #

$$A 3# # *$ * , " **

+

" * #

.

# 1. > " * * # * # + #

# 4 * * ** * E*

* *

** *# ,

0 3 :, & : , ! ** $ : * # * -

138

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

+ 3 -

" #* * #* 3 -

* # * * " * # "* " * * # B )) * * ) " + " " "#

/ * -

# * * # ,

* * * - " / - #* # ! 9*" #

## - 2 # 4046 * 5 1 & #6888 ! 0 1 & * #'# (

1 & " 0 1 &" *0

# ! #

# E"## "

B +

" * " 1.*/ # *$ " #" *

@

1. #

# 0 # #! " #" # *

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

139

+ 5 G1 & ! ;51 & !

+ #5 G1 & ;51 & D8G

4B # * D8G " '

(.

* *

0

# *

- # # $* + #4B* # $ 469*" 4B*

#%

# :#H+

"

46 * " "##*1 & *

0 * $ " " . # 4 B

140

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

##* : #

#"

*

# - # # # - - "# 0

##

#- "

! * # " " " +0 # # * * '# ( ! - " #

" !- #" " '8888 D888 (! "

* .

# "

##

#- "#

"

' %( "#

# - ! # "#

#1> * 1> I - # 0 # " J! : " / # %

# ## >" -

#

# +

# # * ,

# - "# " # " # 0 /

#

" "

* " * "

# # # - " " # 4- # " 0 #

# $

$ "

# " " + ## #

- " # #

# * * # # ""

* $#$

IJC. # C $# " - = 0 # #*I6J # * #

IBJ $#

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

141

" # $

# " #

5 #

# # #* $ # / * 3 #

* #/# # /# ! " 4 C D" !

* " - " # #

* $ $

5 *

"# " #

* # #

*

# " $* -

. #

# #

$ * " *

22*** *

+ $ * * *!

142

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

+ & * * * * *!

* $ # ; $ ; $ +;!&, #%

' #/ " ( *

D88888 , "

# ; $ " *# # , +; * %

/ / * $# " # # # " * * * #

-# " "

>

+;

-# " " * D# "

#

#

" * * " #

%

# > * * # @ #% # * # @ #% @ #% " #

* " / $ * ! * # ' ( 6D8#

* # !

#

= * #%

# * $ " * +;, * 5" ## = 3 . " 3 *

22***

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

143

*

" * / 2 # * *

" #>, & ##

.; #

#

0

K# K!+

#*

# : * #

D9

0 0 # # E # " " *

#

"

! # " * /

"

! *$ " *

*

$ $ ( ( )% )) - * *

* * 0

1" / " '# ( # 0

0

# 9*" " # * $# -

"

#

*

4 0 '&

(*

# " !!

"# ! + " # $# " $#" 4

" +

' #( $# " . # $# " * # #

** " 4<<8 4<< ! #

1> " $ 1#*" $#" * D = *$ 0 : / * ##: * 5.L/$ #

144

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

* * : $#" " " #

# "

$#" ! " "

* 9*" # " # # $#" 0" '( ' ( " # #

'K ('K ( "'+/KM (" * " : # *

%* - * "

%

" ! # # # # $ #

9*"

40 3 : 3>:!, ' # ( "
+3>: # * #

& #, +;, &;.> # !. >% .>! .>

BAA/'3 " ( * 6B8 3>: # *?6C .> !.>

" * % #3>:5

'. : (" * '4 . &

( + " 3>:

40* #"

" 3>:, 'BAD( ;

5; ;

5; ::..

& # , A 'BAD( 5; 'BAC(;

<66 <<: #! # 3>: BAD <
( # BAC<

* * <
# * <
: #

EBAC BAD E * #

& # , # "

A

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

145

$%

& !!

+ ""

$' #(! % "# *# 4 0 3>: , # #

? D88888 # 5 # 4

#!

#

* * # #

+ # *! ## #% '3>:

#

( 9*" " *

# * # * " $# "

ICJ!+

! 0

*4

5

* -#

! 5

'. : ( .>

BAA/

N3 " ( " # * * " + *

% *

40 # '&

( # * #" '&

("#

" # #+ * * -#

* # " % '

(! #

#% .

#% " 0

. 0 " #

* * "

0 # '( ! .;# " / #" " " 9*" " " ##/.;, 3>:, #

" "

" " # * # > ## #- # # "# #. / - ! * #!

?

. ' #

(' #

( #

146

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

* # 0

4

# * . * * # * * "# *" #

#+ " * : # * " **# * 4

* - # ! "

0 " $

- #! " $ "

- #

" " ! 5 " * # * " 4

'*( * " '

$( " # * - # "#"# ! "

$ "

*"

* " *. " # " " *" # *

"

* 3 #'$ (*

* 1.. 4 % ?!

" * * # " * * "#"#

" '1 &(** # . '#* (/** $ " "
"

0 * O

" * # # * #*

* " * "# " O

+ * $ # - #1 &*

* # - #1 & * # C )* + )* + -

- #

" P!I )* +,!E )* +!J

#" " * *

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

147

. * #

/

$ $ $ $ $ $

B? CD D? AB A6 AD

?B / ? / AD /

( +0 / / / / A

13 *

$ $ $ $ $ $

B? CD D? AB A6 AD

/ 2 ) 3 ?B ?B ? ? AD AD

( +0 2"3 6 B C D A

+ # '" ( /

# * '" (

#

* !4 " * ##

# - !* * " * # 6888 S# 6888 *

6888 # * - #

" # 6888 S# 6888* - 0

*" 4

" #

#

#- !. $ # * "

* *%* ""

% 6888 688 - #!+ # * # # " * #.

" " ; * "

#

"

+ 0

" "# " # * " $* " *!# * $* " "# E 0 "D688B S#6?688D * # # ** * # " *

# ! # >

"# #

148

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

* / *" 4

** * * # * * $ "

*

" " D688B B 688B@ " S#688C B688C@

" S#688D S#6?688D * * . * * " "

" - " "

" /

## # /"#" IDJ! #

$ "# " # " 9*"

* * # *

*

' ( - # " # $ "

"# !+ #*$* !

* * "# # # #/ # # / #!.;* " # $# #/ " # " " #

#

" # # $ - #* " # * # * %

"

# !* # " * ###% / # -! S

# # * *

4

##

D888 68888 D888 D88884# 9*"

# - "

# #% * $

" ! - # " "4 * #

# "

+ "# " " E

0

+ ' " ## ( * # * 0

&;.> . &; .

> ! # "

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

149

.( !

# "#

: : : * * #*$

0 "

# *# % 5 * # ##

* * 9*" % $ " % "

"

40

" # * $4 0

= ! 9*" *

# "

" - &: !* "

" ** & #,

* & #"C! /0 . " " #

*

*# - " " " " * " *" 0

. # * *" . * "#

!" . - % #

$ " 4# / "

#

"#

$*" * * # "

# 1(

2 ! % $ " '

( / 0

+3.. * #

*# 5 # 0 *#

"* 0 "

" # #

! + " " &

K +

40&! 4- # #

*# &+ " # #

150

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

## # * <5

* #

*

& ! * #!

* " *

" 40 23#* * O8 3# "# - 3 2* !

( ' # ( # " 4 0

" # 0 # # # " # ' #/#(. ," # * ' #/- (. * * " * #!

# * *

# 1. 2. 3. 4. 5. 6. 7.

#/# " ##@ #/- " #- @ * " *@ */# " * # *

$ " !@ #/* " # * * *# # !@ . " % % # *

IAJ!@ # /# " #

#* * *"# @

# " *!

: # * #+ * #! #- ! " #

"

!. *"

* * #

* # /# * #/# #

0# ! *

* # 4 0 * * *

# *" * * . ' 5 *(

' * * 5 " 5(' * * 5 .( / ! " #" # < /

* * #0 " % #

* #

1. * " . "

22

2;2

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

151

* -

* " # - . # ! * / ! # ' ( #* #*" * * # *# %

# 0

* "' (* # ' #( ' #6(' (" E # ## " " " # # # . # ' #( ' #6(" # % # # 8

# *

** # # $" ' #(!

'( '6( 6!

" # # " " .0

* * # 9*"* # # *# E # E -

# :

#

# *! * / # + # # # * # # #

# # # # " ""

$ " # "

" * 5 "

! 0 **

* " # 0"! $ * #* # *" " # " # * " 4 0 $ # , # " *" '"", #( * 5

@ *

$ # 0 # #* 5 #

# * 0* * + * # * #

* " " "# = 0 #

* , # +

# # +

# , * */

$ * $ # 8 #

'*(" # E #

# *

# "

152

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

#1" " * * # ! #! " " *$ * # # # # "" 1" *

# # . * # # * #$ # # + * * * " $ *! # + "# # " '*(! * * * * : #" # # 1"B

B # & * * * * .

C

A #

"8 # D

D"A

*3 ## # # # #

# # # #

4( 05%

& +

#" " * /

/

! # #

" /

#

@ 0 # " @ " # 1> ! - # $#

@ # 0

#

" @ - # * *

" . #

"

*

!.; # #/# # /## . * - * # 6 B

/#@ "

# " 2*

!@ / # /# " 0

#* #* #!@ / # # /# *

# " " " ! 0

#

4#

6 &78$ '/ # (. ' ( # 5/5;

5;/5

""' # ( # * $#

# * * -" *

#* * #*/" /**

. /!

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

153

* " #

* #

# ! . " ## #"

#" #% #

# . 1> /" % ! .1> / " % ! * #

* * # .1>

# * .1>/# #!* # * -

# # .1>/.1>.1>/

#.1>/ +> * !1>

# # *

" "

* #

#

1> * .1>

* # 1> # * ' 0(

$# ###

-

. # "

# ##1>1># # *

% / >

" # #

1>! #

* #

# 3 *

# $ * # * # 64

640 * # # #"

' #( ' #( #"

'

( '

6( #

" ** " #

* * ' #(

1> ! # 6 * # # E

/1>

/.1>.1>/ + # * $# #

/ #

* $ $#

, -

* * # #

* * ' (

4

* # * *

"" #

#

=

* *

"#

! # # #

"

154

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

& - .9!0 *

>*# " # " *

. % * * .>

*

'

(' ( #

* # . *" $# / % * : $ '

( " * #

" .

"" , # $# *#

+

/ 0

" / +

2 #* * # # *$

# $ 4

# " *

#

* + "# $ # " . 0- # " * $ * - # . * " * " * : ""

*#

# #

0 ;

*

#*

%

#

* *

,

"

# # # ! >

" * # $ *$ * #/ 0.* -

* " * *

+ 0

2 *# *" !

*

*" /

. * * #

* " .

: " " $#

# / / 9*"

" * /

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

155

#

*

. #

" # %

*$ * #

" #

-

// *"

#

/ #$* #" #*# .$ *

0 # " * 2 +"# ' (61" # " ! $*

4 0 "# "

# * $* 0 / 0 !: '$*( ** ## " * # " " * # 2 $*

"$ 0 0 $# * " # " 0 # @ * * " "$ -

" *# *

-

0 :

2 # > * $2 $* #

" 0 * > #*# * %

" * " * # *$ /, ** *! * * -# $# # 51 # !+

# * * 4

#

*

* * "$

. * 6 > '

/ "#

(I?J

156

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

$#

# "

0 / 0 " # 3>:! .>!> * -

*

" * #

* "$ 4# * " = "0

.;

"-

*

# #/# # /# /# ' #( # # !

** $** # .>3>: # *

!4 - # $#" #/# ! ** $**" # * " ' +( 'T+;( 'T;( # * @ $# # *" "

" ! + -

** * * - # # " / ** $**" * #

"" !. $*" *

#"

#

"

$ !4# ** $**# " # # "# #

! * *

" #

9*" "

"# -$ # # *$ " " *"

#!# *$ @ *

/

/

" ' ( . #

# *

% *

= 9*" % * #

% "#

* = * "

+ 2 * " "# $# * # * 9"

* " * " " *

+ *

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

157

! " #$%&' ( ') *+,-.(/ % " 0 /12345.1674* / //

/8-*9*: /8- - - -; - :6442* -/$ $ /$ $76 -

*$ $

*$ $76 -

$ $

$ $76 - $$ $ $76 - "$$/ $ $766 - "-$$/ $ $76 - "-$$/ $ $76 * *

* '3 . "( " #$3 % + * IOJ# " #" :$

* $ # # ' ( #* $ # 0 !

" " = $! $** #

*E $#* * #" !4 # * 0 " " # #

51 *

" *# $#* // # * * 2 !" *

" # " * $0

$ " *

0 ! " $#* / $# " * / " * " !B:

# 2 4 # * * * " B

4 * #

158

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

: -

" * * # .' !!

"

" * $# " ' # ( '#(. "

* # # -"

# # .

"# * 3>: .> #

* : * * # +

* #" * #+ -

* # # - "

# * $ - # " " ## *8 6 # ## * * # % # 5! + " * $ *

& # * * BAC BAD;

" 5* 3>:, # BAD! # $#

# 0

. *

/ 4:14:2 * * # BAC -" 3>: BAD : % #% /

+

$* "

* " " "

* $

*

/

" - *

#

9*"

* # $ * " " .( 5! 0 . - # #!

'# ( "

/

.1>1>* # ! # # / "

#* .

*$* - #

! #

- # 0 - #. * # -

.1>

D.S. Bennett / Merging and Meshing Data: Difﬁculties, Lessons, and Suggestions

159

# #* .

*

* * # "

# "# #

$$#*"

' 4

$

,

#$ "

* +

# #

# 0 * #

0

* " # $%

.#

"

*#* # *

% **$* * * * # " ,) IJ

+'51+3 (9

9

6888U6A ?
5( ! 9

9!0* 9"T5" #&

<<6 IBJ

+; <- :++5" #&

688C ICJ 1 F : '.

#

+;"

. OA(9

9

<<<U6DB
688C IAJ 1 4&1 '. B <
-! 688CU6BB/DC I?J :;(!

5 * (!

*T$1*/ 9
. (9

*>*688 UCDBC8
This page intentionally left blank

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

161

!" # $ % & ' ( &(

&)* )* *+ , - ( % ##

&(

+ ) %*

*% ( %+ * % # # ## # *

+ & (

. # % !* . & . % +

*. / 01 ) & #

2% + # - #%2% . % . ! () /334# ( % 5 6 + ## % %&

*

& )% )+7 & 8 & !* $8 !'0/1. 9 % 0:1. ! , 6 , $!,,'0;1 % &+(& . % * % % * )(

.

# % + &

* #& %& . " < = .

* * (

#*

)(- % ( 6 % ( # #+ % & %& % ( *' ## # )* &) . ## # >/'*%%&

*. )) .&. )

* #> :' %% - & . #. )

* ;'*%%& # ) * * %%. #*. ) +

#

. % !*%&2

. # 4."7?3@A7. " ( *> ? B ++ 6 ! % # ( # $,# '+ / & ) CC(((+ +C(C +2# )6

*

01+

162

C. Raleigh and J. Carlsen / The ACLED Data Project: Purpose and Methods

# % # % ) &) )& (*&(

) 6.(*

( 6 +

# *% ) % =

( *+ %

% %(

( (

)* % +,# & * # %% =) !

) +#% *

.%##

6

#. ) #+ %? & && *

% * -

6 % & (

+ . % - .) #& ) #. #? * & #

% #

%&(

+# ## # % (#?

% %

* # * %&(

.) +

#%&) D@D?/33:$((&% D@3' !

DD?/333+&*#) # $ & * ' *+

C. Raleigh and J. Carlsen / The ACLED Data Project: Purpose and Methods

163

9= *.&(

& & %% % & . ) ) % % 0E.A1+!& 6*%- &&(

) %#) &

.(

'*% ? . )*# # *&(

* #

% *>/'*% ? . )*# # * #% &

)

) &>:'* %*&

) .% - . < $ '() %#& * ) & 04.A1>;'*# .#? )(&

)> E'*# .#? )( &

). A'*& %* % . .* . ## *(

# + % % .

- % & (

. *% (

.% * - +&(

)# ).% . #. % 0@1

*0D1+ # ## &(

% ) 031+2 .* %(% &

) % % ( & - &(

+ ! 2 )& % ## # & #(&(

+

*% %&) (()%% # ## # *+ % *#* %% &&6# %% 01. 0/1. #

0:?E1.

& ) % % F0A?D1+!.

%&(

. & # & )* #& 0/31 ) % . ) )% %

0E.A1+ )% & .)%(

# * * % +

# *# % &

&. # ( . & * = *. * .

# F 0/?/A.E1+ = = # & # ( -

)+, . *% &. %%- 0/41. )F 0/@1.) & % 0/D1 %- %

0:31

) # * ? >)#

* )(* + *. # C) )*. %#%(

9" $9"' 6#

* % % % * %%

)*%&# +

# % &

*# % ) & % %

164

C. Raleigh and J. Carlsen / The ACLED Data Project: Purpose and Methods

01 9" 6 0:1 =%%&%-

#&(

?%& +

*% = &&(

% G## *& % )#H( #& %% &

% ( *% *$&& F'

%# %

0:/.::1+ % . * % )

)**% #& +%*

*) #& ( # * & ) # F .(%% # % %

%-#% $ # )*.%#. %# %(. )# % F '+* % % ## * % % # %% $

#. . '0:;. @. D1+ F(% * & #() ##-.- + . #) -% ## # &%

*

&

&(

& ( .

#& + % % % ) % *

)60.:E1+) %#& + & %

) + /+ :+ ;+

$- * '. $# % & #.( *'. $ ).#& . C & '. $*% &*# #'+

#)(-*%& ?&&(

# %+J&') (

*./') (

*()* ).:') (#& #

*. ;'& # & .)* #& )# +" &&E' ) % )) =

. A' ) + % & % % . *

* . %) & &. * & %

) ) #& &*+

6 ( #%% ## # 6 +, & % ) # (

) ) % & *+ ! ## * = ! #6(#

C. Raleigh and J. Carlsen / The ACLED Data Project: Purpose and Methods

165

% # =.( * + .#

% #& % .) = ( ) * *( ## # # % + "#$ $ $

)*% ((%% % #&

#+ #) = # % *.

* 6

*

& + ( *) - ## # & %

( * * #+

)( ) % -# ) $+#+7?. # ! &* $ !'.&% % ,

*2

#* '+ )*%%% % (( ) # # )0:A1+:7) &

% +2 !% )

## $# '. % ) & + ) $ *. .*

' $ '+ & )!% % &# - * *

# % % 6 & % .*

% .% #. ?## 0:41.& . % + "% # 9F# ## # * % =

* $-&' !

+, .

&#.)

%% + !20D. :@1 ( & ) ( & )# % ## # % ! +!

( % ) +, ! %(

$

= '+;(! # # + ## # % %F$)*6 *'+

) ( ( % ## # =

)* = ) % (+ 6# * - . ) ( ( % )* 6 =

( ## # %# % =

+ % # % * # ) * . # %

(. ( * % *(% * )# =

6 + ) . ( #) &

)+ ( ( % ## # #

#=

) &

+%( & #( & (# )## *

%%+*%% & .(( *. ( &

) ( % !K $! K * # #'= *+ %*& )##

( : 6!5 "7&

2 6% &

)( 6 # % # + ; ! = % ))

) )+

166

C. Raleigh and J. Carlsen / The ACLED Data Project: Purpose and Methods

:3333+ (

* = * 7 $7? * #' ) %%* +

& )-.)( ( + - )% # .) #)# + ## # *%&(

.(& = ) -&%(

& 6+ & #%

*# & * * ) - + 9 C 7 & /33A @;/) & & #%% D;E?/33E+ (& /E333) & & # )*%

% % 5 6 + % . * #* F 6(# *( ## # + ""& ! !$ * * &

( * % ( %##

*F . # %##+ ## # #? &

* &

&(

(( *(% *?& + # ) )% ( *.) & ) & ( ? )* %%

)( + ##% ## # . C %% & #&%+- %

% % (

&* - * ) % &)*%% +, % & & **

. ) %& %% &# +5# ) # #% ) ) = *% +*

%0D1& %

.) % ( )* *# % +

) ) % )

*+&

*## ) % . %( C #+ 2 ) %&

%%#%% .()

% ( %%# & * % %#L7 & % ( ( & &+ ) # ( + ? ## # + ###

% ) & . # ( # % &#(.) # ## # & % *#&

). * 6 F# ) +! *) # )( & % + - &# # ) )* # ?&&(

& *)

. % -

# 9 C 7 *?*

(

## # %% % *+ #& % ( # )* #

?( ## # ( * ) ? ## # ( *+

C. Raleigh and J. Carlsen / The ACLED Data Project: Purpose and Methods

167

01 0/1 0:1 0;1 0E1 0A1 041 0@1 0D1 031 01 0/1 0:1 0;1 0E1 0A1 041 0@1 0D1 0/31 0/1 0//1 0/:1 0/;1 0/E1 0/A1

#. MN&

# ./334+P # % & +< % ## # # !* % & 2

J.! #+ ! . +./33A+P8 & !* 5# < ( !* % , # % +< ! #.! , .#DDA+ ." . 2 .6 6.

#

!) #M &

! . /33/+P %D;A?/33"( <.'(! :D$E'AE?A:4+ *. + +.DD@+P! ,

*2

#

% 9!, # * #<.+ &M)

..()$ * + .+( M%+ . . .N&

# .6%% .

* ?K M" ! ) . /33:++* , ( +7-% 7-% 9& * + ,

. +M & + ./33:+P*. #*. &2

<.#(

!D4$'4EQD3+ . M6%% ./33;+ P # & &(

<. -. /( EA$;'EA:?EDE+ *.++M2 % &

./33+!$ +7-% >"(R 67-% 9& * +-.:;3+ . M)

.DDA+P& #M*<. M)

.. , ) #+7-% ."++ % (

*> $?::'+ #. M 69 . /334+ P #. & #

%+<(0 /A$A'A4;QAD;+ 2 .".N&

# M #./33A+P#! % & 2

<+ # % ! . #:?! ) :. ."+ 2 .5

)

,+./33;+P %5#%L- #

&2

<.' (! ;$:'+ .2 .DDE+ + ) #0# 1>"(R 6 ) #9& * +-.//D+ . !. /33/+ P # % % )<. ' !;A$'Q:3+ 2. *+./334+ $) + ) #>"( R 6 ) #9& * +--.;3/+ 8 *& .! "+./33A+ , + ) # > "( R 6 ) # 9& * +-&.;@E+ 5 #. &

M 8+ S. /33A+ P % % & 2

. D43Q/33<. (0 +/E$:'.:E?::E+ #. ./334+ , ) # 9& *% + # .N&

M #./334+P F. &(

+< ( 0 9 &(+ .2+./333+P % !)%& <. '!;;$'34?/4+ .(

"+M+!#.D@4+P= * #*<., #(

!@$/';/E?;E/+ 8. &M % ! #!+.DD@+, +7-% >"(R 67-% 9& * % % ! # !+@@+ %%.

6 +. /33+ 0$ ) +>"(R 6T56> )9!-&*)* # &+-./D:+ , . !(

. /33/+ P F = "# % & < K F ).9& *%7-% +CC + + #CC=C=(C=(@+ . ./33:+P = * %L< ' E$;':D4?;/+ U)*. ./33A+P F = . & &%&% EE &# +< K F ) . 7-% 9& * 7-% + & ) CC(((+ +-+ +6C)C( 6# /@+%$ 3 ) /334'+

168

C. Raleigh and J. Carlsen / The ACLED Data Project: Purpose and Methods

0/41 .8 !6 ./334+P %&2

<.'(! ;;$:'/D:?:3D+ 0/@1

.) M5

)

%%./33:+ / +5 .+2&( + 0/D1 8 .+./33A+ + ." 9& * + 0:31 8 . . /334+ ) , 0$ %* # ! !!"+ CC + C VD@3@/3 0:1 . ./33A+1(* 0+ ( ?

#

& &!2 6.

7.# /3?/+ 0:/1 5. . /33:+ ( # ) + ) #.98>"(R 6 ) #9& * +-.;3:+ 0::1 . !+.D@E+ , ) + .2+9& *% 2 +D4+ 0:;1 .) 2+.DA4+P # *%& .J . <., '!$/'E:?A4+ 0:E1 #. .N&

# .

M 7 #./33A+P % & $ ' )6. J ?/33A< 7 % !* % & 2

.

+CC(((+ +C(C CW)6+%+ 0:A1 5. !. /334+ % % ) + - ? "7&

2 6. ). , + $ & +' 0:41 .

? 6. 8 S M " 2 . /33A+ 02! / 0) + 72 2 6. 7. 3Q ,)

*+ CC(((+ +C(C# (%C WW2 W3A+%+ 0:@1 # .N&

. U)*M #./334+/ /) # $+ ;@ #% ! . #.+

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

169

Visualization of Conﬂict Networks Ulrik BRANDES a , Jürgen LERNER a a University of Konstanz Abstract. Visualization is a powerful tool to derive insights from massive, noisy, and possibly inconsistent datasets. We propose a method for the visualization of conﬂict networks that show a set of actors together with hostile or conﬂictive relations on the systemic level. Our method highlights the most involved actors, reveals the opposing groups, provides a graphic overview of the conﬂict structure, and allows for smooth animation of the dynamics of a conﬂict. The visualization technique can deal with potentially complex network structures and distinguishes visually between bilateral and multilateral conﬂicts. Keywords. Conﬂict data, network visualization, time-dependent visualization.

Introduction Event data describes “who did (when/where) what to whom” and are among the most widely used indicators in quantitative international relations research [1]. From a high level view, event data are used for two types of purposes: ﬁrst, the assessment of current or past political situations, second; for statistical validation of theories about the likelihood and outcome of conﬂict and cooperation. Currently there exist many event databases (some of which are described in this book) that differ largely in scope, granularity, and in whether they are hand-coded or automatically extracted from, e. g., news sources. Due to the typical size of these datasets, it is hard to derive insights from them without automatic support. In this chapter we present a method to visualize event data given as a set of pairwise conﬂictive or hostile interactions. Well-designed images of conﬂict data provide support for at least three different purposes. First, they give the analyst a graphic overview of the data, which may reveal expected or surprising patterns, and thereby can lead to hypotheses that may be validated or rejected later. Second, visualizing data is a powerful tool for error detection and data cleaning. Last but not least, images are very convenient to present and communicate insights to others. A straightforward way to visualize conﬂicts is to make use of geographic maps and, e. g., highlight countries involved in conﬂicts. While such drawings have the advantage that they are very common and most people are familiar with them, a different visualization strategy can give additional insights into the data. As a matter of fact, conﬂicts among large sets of political actors do not normally happen at random but often reveal a grouping of actors into two or more blocks, characterized as follows: conﬂicts within a block are rare or weak and conﬂicts between actors from different blocks frequent and more serious. Groups need not be deﬁned by ofﬁcial treaties or alliances but are rather determined by the interactions themselves. Naturally, this grouping of actors does not

170

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

necessarily coincide with geographic closeness or distance so that geographically determined positions do not reveal the group membership of actors. A distinguishing feature of our method is that we show the conﬂict network where actors’ positions are determined by the relations themselves, thereby revealing the network’s structure. Given a list of events, our method visualizes the resulting conﬂict network in such a way that actors are far from each other if there is a strong negative (i. e., conﬂictive) relationship between them and close to each other if they share the same opponent(s). An example of visualization of a conﬂict network is shown in Fig. 1. If events have timestamps attached to them, the static visualization can be turned into an animated scatterplot showing the dynamics of major conﬂicts over time. From such a video, an analyst can recognize or discover the major actors engaged in conﬂict during certain periods of time, see how they are grouped together, and which are their main opponents. The observer is also enabled to detect time-points where the conﬂict structure changes signiﬁcantly. Since our animation is smooth by design, it can be recognized easily which actors enter or leave a conﬂict during transitions. In contrast to a pure dyadic analysis, networks give additional information about indirect ties (e. g. enemies of enemies), density, complexity, and structure of the actors’ network environment.

Figure 1. Visualization of the conﬂict network constructed from events related to the War in Bosnia. The nodes represent political actors, the edges represent conﬂictive relations which are often military engagements. Three groups that are in conﬂict with each other are revealed. The dominant members of these groups are {BOSSER (Serbs in Bosnia), SER (Serbia)}, {UNO, NAT (NATO)}, and {BOS (Bosnia)}.

Related Work Many event databases exist that differ largely in scope (e. g., which actors or events are included; which time-period is being considered) and granularity (e. g., aggregation of actors ranging from countries over ethnic groups and organizations to individual persons;

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

171

granularity of time-stamps ranging from years to days). Hand-coded datasets (such as [2, 3]) are typically coarser-grained than machine-coded event sets (such as [4,5]). King and Lowe [4] report similar performance for an automated extraction tool as for human coders. Nevertheless, it can be expected that machine-coded data is more likely to contain events that are obviously incorrect for a human coder (cf. Sect. 4.2 below). Although our method is applicable to human-coded as well as to machine-coded event data, it is especially appropriate for the visualization of large, ﬁne-grained, and potentially noisy datasets. For the examples in this paper we use automatically coded data from the KEDS (Kansas Event Data System ) project [5]. One typical use of event data is to analyze the outcome and likelihood of conﬂict. For instance, Schneider and Troeger [6] examine the inﬂuence that conﬂicts and cooperative events in war regions have on ﬁnancial markets. They demonstrate that this impact not only depends on the severity of conﬂicts but also on the degree to which economic agents could anticipate events. Guidolin and La Ferrara [7] analyze the effects that the onset of violent conﬂict has on asset markets. Schneider [8] reversed this line of research by examining how political events can be foreseen by using data from ﬁnancial markets. Obviously, studies as in the three last-mentioned papers rely on the validity of conﬂict data, i. e., to what extent does the dataset represent the true level of conﬂict or cooperation at a given point in time. Thus, a possible usage scenario of our visualization technique is to detect coding errors and clean the data before doing the analysis. While studies of conﬂict often focus on the dyadic level (i. e., the relationship between only two actors), there is increasing interest in applying network analysis techniques to understand world politics at the systemic level. Maoz et al. [9] provide an overview of the potential use of network analysis in international relations research. In many studies, network structures such as alliance and ethnic (e. g., linguistic or religious) afﬁnity networks, as well as trade relations, are considered as independent variables from which it is sought to predict the level of conﬂict (cf., e. g., [10,11]). Harary [12] and Maoz et al. [13] analyze the balance (i. e., does it hold that two enemies never have common friends; is the enemy of an enemy a friend) of the network of friendship and enmity in world politics. Other papers aim to understand the effect of several structural characteristics, including reciprocity, triangularity, polarity and bipolarity, on conﬂict [14,15,16]. Note that the bipolarity index that we introduce in Sect. 2.1 is different from those considered by Esteban and Ray [14] since our index is deﬁned as a function of the structure of conﬂictive relations. A paper closely related to ours is Hämmerli et al. [17], who applied network analysis and visualization techniques to conﬂictive and cooperative relationships. The main difference is that in their paper actors that are strongly in conﬂict are drawn closely together, whereas our visualization technique separates strong enemies (cf. Fig. 1), thereby revealing opposing groups. The method presented in our paper is also a tool to dynamically visualize news sources—a topic that has received considerable attention (cf. e. g., [18,19,20]). Wong et al. [21] proposed a method to generate animated scatterplots from data streams, such as sequences of news articles. (Scatterplots are widely used in statistical graphics, see, e. g., [22,23].) However, the scatterplots in [21] show similarities between documents and not hostile relationships between political actors as will be done here. The basic version of our visualization method [24], which is restricted to display a single bipolar conﬂict is augmented in the current paper to deal with several and potentially multipolar conﬂicts.

172

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

Outline of this paper. In Sect. 1 we provide background information on the type of data that is being analyzed. Our method for visualizing the conﬂict structure embodied in a set of events is introduced in Sect. 2 and extended to smooth animation of event series in Sect. 3. The utility of our method is illustrated on event data from the Balkans in Sect. 4. We conclude with a discussion of open problems and future work.

1. Event Data Our method is applicable to event data given as a series of pairwise interactions. Although it is independent of the data format, we will focus on a particular coding scheme to make the exposition more concrete. The Kansas Event Data System (KEDS) [5] is a software tool that automatically extracts events from text such as news reports. In Sect. 4 we will use KEDS data for the Balkans region. Formally, an event series is a sequence a1 , . . . , ak of tuples ai = (ti , si , oi , ci ), where • • • •

ti is the time-stamp (date, given by the day), si is the subject (source actor), oi is the object (target actor), and ci is the code (event type)

of event ai . We say that actors si and oi are involved in event ai . Events are classiﬁed using the World Event/Interaction Survey (WEIS) codes [25]. Each event is assigned Goldstein weights −10 ≤ ω(ai ) = ω(ci ) ≤ 8.3, which are psychometrically determined scores depending only on the type of event (see [26]). A positive weight indicates the degree of cooperation of the corresponding type of event, whereas a negative weight measures hostility. Examples for Goldstein weights associated with event types are the following. 072 054 160 173 223

EXTEND MIL AID ASSURE WARN SPECIF THREAT MIL ENGAGEMENT

8.3 2.8 -3.0 -7.0 -10.0

Apparently, extending military aid is a highly cooperative action, whereas warnings are mildly hostiles and military engagement is extremely hostile. To analyze conﬂict, we will only make use of negatively weighted events, i. e. hostile actions. The following excerpt indicates the coding of actors in the Balkans data. NATO_OFFICIAL [NAT] NATO-LED_STABILIZATION_FORCE_IN_BOSNIA [NAT] SERBS_IN_BOSNIA [BOSSER] RATKO_MLADIC [BOSSER] MILOSEVIC [SERGOV 890101-971230] [FRYGOV 971231-001005] [SERSM >001006]

Several tokens in the news may be interpreted as referring to the same aggregated actor. In the above excerpt, NATO (NAT) is represented by (among others) potentially unnamed ofﬁcials and SFOR.1 Similarly, the actor BOSSER is represented by (among others) the general term “Serbs in Bosnia,” as well by speciﬁc persons like Ratko Mladi´c. On the other hand, the same token may represent different actors at different times. For 1 The

(Stabilisation Force) was a NATO-led multinational force in Bosnia and Herzegovina.

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

173

instance, Slobodan Miloševi´c represents the Serbian government (SERGOV) until December 1997, the government of the Federal Republic of Yugoslavia (FRYGOV) until October 2000, and after being replaced by opposition-list leader Vojislav Koštunica only himself (SERSM). Given an actor coding, textual statements are parsed into events like the following example which took place on 10 July 1995. 950710 NAT

BOSSER

173

(SPECIF THREAT)

POSSIBLE AIR STRIKES

This event is an action initiated by the NATO (active) and directed at the Serbs in Bosnia (passive). In addition to the event code (173), a textual description of the type of event (in this case a “speciﬁed threat”) is given in parentheses. The rest of the line is the stemmed form of the text fragment that has been turned by the KEDS parser to the corresponding event. Often, this text gives valuable additional information, in this case information about the nature of NATO’s threat. Datasets derived from serious conﬂicts can be quite large. For instance, the KEDS dataset encoding the Balkans conﬂict consists of more than 78, 000 events. To detect emergent patterns and utilize indirect relations, we transform the data into a network. Any set {a1 , . . . , ak } of events gives rise to a directed and weighted interaction graph G = (V, E, ω) that we call a conﬂict network. This graph G is made of a set V of vertices, a set E of edges and a set ω of weights, in the following way. The network’s actor set V is the set of actors involved in any event as the source or the target, i. e., k V = i=1 {si , oi }. There is a directed edge e = (u, v) ∈ E if there is an event with source u and target v, and we assign a weight ω(e) that is minus the sum of all negative weights on events initiated by u and directed to v (i. e., edge weights are positive and indicate the degree of hostility; cooperative events are disregarded).

Figure 2. Force-directed drawing of hostile interaction in the Balkans from 1991 until 1997. The darkness of the edges is proportional to cumulative hostility weights. This kind of graph visualization is inappropriate for conﬂict networks as it does not distinguish between important and non-important actors nor does it reveal the structure of the network. In this chapter we present a method to draw conﬂict networks in a concise and easily understandable way (see Fig. 1).

Figure 2 shows an example of a conﬂict graph drawn by standard force-directed layout techniques [27]. The complexity of Fig. 2 already indicates the insufﬁciency of

174

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

general-purpose graph drawing techniques and the need for other analysis and visualization methods that are more appropriate for this application. In Sect. 2 we develop a method that extracts the dominant conﬂict structure, ﬁlters out minor actors, and produces a less complex image that is easy to understand. It is unlikely that a focused data set yields an interaction graph with more than one signiﬁcant non-trivial connected component. However, since connected components can be analyzed separately, we may safely assume that all interaction graphs are anyway connected.

2. Visualizing Conﬂict Structures In this section we focus on extracting the structure of conﬂicts from static event data, i. e., we ignore time-stamps and consider the data to be given as a set rather than a sequence. The actors’ positions are determined in a way where actors that are strongly in conﬂict with each other are far apart in the drawing and actors that are not connected by a conﬂictive edge, but have conﬂicts with the same other actors, are drawn closely together. Thus, the drawings facilitate the recognition of groups of actors that ﬁght the same enemy. We start in Sect. 2.1 with the assumption that the network contains only one major bipolar conﬂict. This rather restrictive assumption is generalized in Sect. 2.2 to multipolar conﬂicts and in Sect. 2.3 to several parallel conﬂicts that overlap in the data set. The static methods that are developed in this section will be augmented to include dynamics in Sect. 3. During the computation of the conﬂict network’s group structure we will ignore edge directions. The rationale behind this is that if there is a strong negative (i. e., hostile) edge between actors u and v, then u and v should be in different groups—independent of whether the edge is directed from u to v or vice versa. However, edge directions will be taken into account when determining whether an actor is more active or more passive and highly asymmetric edges will also be shown as such. 2.1. Single Bipolar Conﬂict A ﬁrst attempt to determine the two opponent groups of a bilateral conﬂict would be to try to divide the actor set V into two disjoint subsets U and W , such that all edges go from U to W or vice versa and, hence, no edge connects two actors of the same group. See Fig. 3 for a ﬁctitious conﬂict network of selected Balkan actors (the real network of these actors is much more complex) and the matrix P of derived group-membership values. However, the discrete assignment of actors to the two groups of a bipartite conﬂict is completely impractical for empirical data. Firstly, the requirement that all conﬂictive relationships must be between the groups, and hence none of them within any group, is typically not supported by the data. Secondly, the attempt to determine a partition of V 2 such that the sum of edge weights between the two groups is maximized is impractical as well: the problem is computationally intractable, highly sensitive to noise, requires actors to be purely in one group or the other, and reveals no prominence of actors. 2 The groups U and W form a partition of V if their union equals V (V = U ∪ W ) and their intersection is empty (U ∩ W =).

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

175

⎤ BOS NAT CRO MOS SER BOSSER ⎦ P =⎣ 1 1 1 1 0 0 0 0 0 1 1 0 ⎡

Figure 3. Visualization of a ﬁctitious network that constitutes a bipartite conﬂict structure. The group-membership values are shown in matrix P . Actors are either entirely (value of one) or not at all (value of zero) in a given group.

We relax the idea of a strict bipartition by employing the recently introduced framework of structural projections and the closely related structural similarities [28]. This will lead us to a method that poses no algorithmic problems, is robust to noise, can handle actors that are members of both groups, and ﬁlters out unimportant actors on the ﬂy. Instead of mapping actors to one class or the other, structural projections yield real-valued degrees of membership to classes. For a relaxed bipartition, actors that are strong members of one group have major conﬂicts with actors that are strong members of the other group but only minor conﬂicts with actors in their own group. An example of visualization of such a real-valued assignment is shown in Fig. 4; the associated group-membership values of the most involved actors are in matrix P (1).

Figure 4. Bipolar visualization of Balkan conﬂict 1989–2003. Dominant actors include those set out in Fig. 3. Actors are members of the ﬁrst or second group to the extent that they are mapped close to the left or right coordinate axis, respectively. (See the membership values of the most important actors in matrix P (1).) The angle (left vs. right) encodes the ratio between the two group’s membership values. Involvement of actors is proportional to the distance from the origin. The aspect ratio (shape) of an actor encodes the ratio between activeness (height) and passiveness (width).

⎡

⎤ BOS NAT CRO MOS UNO SER BOSSER . . . P = ⎣ .7 .5 .3 .2 .5 0 0 . . .⎦ 0 0 0 0 .2 .8 .7 ...

(1)

176

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

Note that the degree of membership assigned to actors varies. E. g., BOS is a much stronger member of the ﬁrst group (value of 0.7) than, e. g., CRO (value of 0.3). On the other hand UNO, though closer to the ﬁrst group (degree of membership is 0.5), is also a member of the second group (degree of membership is 0.2), because conﬂicts with other actors in the ﬁrst group (e. g., with Bosnia) are reported. Many of the unimportant actors close to the origin are ﬁltered out because their level of hostility is not sufﬁcient to place them prominently in one group or the other. Thus, our method not only determines a relaxed bipartition, but also indicates which actors are most responsible for the division. The determination of the optimal membership values so that the weight of edges between the groups is maximized is derived in [24]. (Also see [28] for the general framework of this method.) Here we reproduce the results only. Group-membership values. Given a conﬂict network G = (V, E, ω) on n = |V | actors, let A be the symmetric adjacency matrix of G, deﬁned as the n × n matrix whose rows and columns are indexed by the actors of G and where the entries are deﬁned by Auv = Avu = ω(u, v) + ω(v, u). 1. Compute maximum and minimum eigenvalues λmax and λmin of A together with associated normalized eigenvectors vmax and vmin . √ 2. Let P be the 2 √ × n matrix with x = (vmax + vmin )/ 2 in the ﬁrst and y = (vmax − vmin )/ 2 in the second row. 3. The membership values of actor v are the two real values in the column of P that is associated to v (see (1) for an example). 4. The involvement of actor v is deﬁned to be the norm of its membership values, 2 + P2 . i. e., the involvement is P1v 2v Any eigenvector algorithm for real symmetric matrices can be used in Step 1 (see, e. g., [29]), and there are many readily available software packages. Activeness or passiveness of actors. Activeness is deﬁned as the net weight of the events in which an actor is involved as the subject initiating the event, i. e., activeness of actor v is the value u∈V ω(v, u). Symmetrically, passiveness adds weights of events received, i. e., passiveness of actor v is the value u∈V ω(u, v). Indicator for the ﬁt of the bipolar conﬂict model. The bipolarity (or ﬁt to the bipolar model) is deﬁned as the ratio between the minimal and maximal eigenvalue, i. e., λmin . β(G) = λmax The index β(G) measures to what extent are conﬂicts only between the two groups and ranges between zero and one. It is one if and only if the graph is bipartite (i. e. if the model ﬁts perfectly) and it is zero if and only if there are as many conﬂicts within the groups as there are in-between (i. e., if the model does not ﬁt at all). 2.1.1. Graphing Bipolar Conﬂict Space The graphical attributes of our visualization are determined as follows. The actors’ position in the two-dimensional drawing indicate their group membership and involvement: Actors are mapped in direction of the left or right coordinate axis to the extent that they are members of the ﬁrst or second group, respectively. We propose a coordinate system

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

177

where the x-axis points to the upper left corner and the y-axis to the upper right corner. This coordinate system seems to be preferable to the more usual one (one axis vertical, one horizontal) since it prevents the misguided interpretation of superiority of one group over the other. The angle (left vs. right) encodes the ratio between the two group membership values. The ratio between activeness and passiveness determines the aspect ratio (height vs. width) of a node, so that actors who initiate conﬂictive interactions, but are not the subject of retaliation are high and narrow. Involvement is proportional to the distance from the origin and emphasized in the size of an actor. Finally, we indicate the ﬁt of the bipolar model using a bipolarity gauge on the right-hand side of the images. Figure 4 shows the bipolar visualization of the network derived from the Balkan Conﬂict from 1989 to 2003. The circles around the origin link points of identical involvement. They help see that the most involved actor during the whole period of time is Serbia (SER), closely followed by the Serbs in Bosnia (BOSSER) and Bosnia (BOS). The bipolarity of this network is rather low (only around 0.42), indicating many conﬂicts within groups. Despite the low level of model ﬁt, our method still yields two reasonable opponent groups: Serbia and the Serbs in Bosnia opposed to Bosnia and Croatia (CRO). The NATO (NAT) is opposed to SER and BOSSER, due to the massive air strikes in 1994 and 1995. Since NATO initiated more events than it receives (i. e., is more active than passive), it is displayed as a high and narrow actor. 2.2. Multipolar Conﬂict The low ﬁt of the bipolar conﬂict model to the complete Balkan data set (Fig. 4) indicates that many conﬂicts occur within groups and hence the assumption of only two opposing groups is not satisfactory. We call a conﬂict structure where k ≥ 2 groups are mutually in conﬂict a k-lateral conﬂict (for k = 2 we get a bilateral conﬂict). Here we extend the method developed so far to deal with k-lateral conﬂicts. Doing this is straightforward from the analysis point of view, although the visualization has limitations if k gets larger than three for two-dimensional visualizations. The reason for these limitations is that it is not possible to draw four or more points in a two-dimensional space such that all pairs are at the same distance. In the following we derive a method to draw conﬂict networks in a two-dimensional image that reveals more general than just bilateral conﬂict structures. The position of a particular actor in the drawing should express which other actors it confronts. If two actors u and v are connected by a hostile edge of large weight, then we want to draw u and v on opposite sides of the image. The difﬁculty lies in the fact that we have to draw not only two authors but the whole network such that all confronting pairs are simultaneously as far from each other as possible. This objective, which contrasts to most objective functions for graph drawing that traditionally want to keep edge lengths short [27], is of course due to the fact that edges encode negative relations. The good news is that this problem is efﬁciently solvable, as will be derived next. Let G = (V, E, ω) be the conﬂict network with actor set V of cardinality n = |V | and let A be its symmetric adjacency matrix (as deﬁned in Sect. 2.1). Since we want to draw the conﬂict network in two-dimensional space, the positions of all n actors are represented by two vectors x, y ∈ Rn . If for two actors u and v the entry auv in the adjacency matrix is large (i. e., if they are strongly in conﬂict), then they are wellrepresented by the coordinate vector x if the entry xu is (say) strongly negative and the entry xv strongly positive. Then, the value xu auv xv is negative and has a quite large

178

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

absolute value. Summing this up over all pairs of actors, x is determined to minimize the objective function ΦA (x) =

xu auv xv = xT Ax ,

u,v∈V

under the condition that x must have unit length to keep the drawing to the screen size. It follows from an alternative description of the eigenvalues of a matrix that this term is minimized if and only if x is equal to the eigenvector of A associated to the smallest eigenvalue λmin (see, e. g., [29]). The second coordinate vector y is chosen to minimize ΦA (y) under the condition that y is normalized and orthogonal to x. This is solved by taking for y the eigenvector of A associated to the second smallest eigenvalue λmin . This method can reveal network structures beyond bilateral conﬂicts. For instance, Fig. 5 shows the visualization of the Balkan Conﬂict network (1989–2003). Three groups can be detected whose dominant members are {BOS} (ﬁrst group), {BOSSER, SER} (second group), and {UNO, NAT} (third group). However, actors are not strictly assigned to one or the other group but their membership can also be intermediate. For instance, conﬂicts that are reported between UNO and BOSSER (Serbs in Bosnia) are stronger than conﬂicts between UNO and SER, which leads to the drawing in Fig. 5 where BOSSER is more distant to UNO than SER to UNO. Actor CRO (Croatia) is much less involved and therefore drawn smaller and closer to the origin in the center of the drawing than the aforementioned ﬁve actors. CRO has strong conﬂicts with the second group (most notably with SER) but no strong conﬂicts are reported between CRO and either members of the ﬁrst or third group. Therefore, CRO is drawn exactly opposite to SER and between the ﬁrst and third group. Note that the complete Balkan data set indeed yields three groups that are mutually in conﬂict. In particular, it is not possible to assign the Balkan actors to only two groups without having strong conﬂicts within a group, leading to the low ﬁt of the bipolar model (Fig. 4). The visualization of the Gulf Conﬂict network (1979–1999) (see Fig. 6) does not reveal three groups. Instead we can see a bilateral conﬂict (ISR vs. LEB) overlaying a trilateral conﬂict that is mostly formed by USA, IRQ, and IRN. In Sect. 2.3 we extend our method to handle such situations. Interpolation between bipolar and tripolar conﬂicts. One further detail has to be taken care of: a conﬂict is not necessarily either bilateral or trilateral but can show an intermediate structure. For instance, Fig. 7 shows the conﬂict structure among a selected subset of Gulf actors involved in conﬂicts in Iran and Iraq. The dominant structure is a triangle formed by USA, IRQ, and IRN. However, conﬂicts reported between USA and IRN are weaker than those reported between IRQ on one side and USA or IRN on the other side. The question arises whether this network is best represented by a balanced tripolar structure similar to Fig. 5 (which would ignore that two of the three actors are closer to each other than to the third), or by a bipolar structure placing IRQ on one side and USA along with IRN on the other (which would ignore the hostile edge between USA and IRN). We argue that the best way to represent such a conﬂict structure lies in the middle, i. e., to show it as an intermediate between a bipolar and a balanced tri-polar conﬂict. Figure 7 shows a triangle which has three unequal sides, but where USA and IRN are closer to each other than to IRQ. Such a representation can be computed by appropriately scaling

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

179

Figure 5. The graph on the left-hand side visualizes the conﬂict network arising from the Balkan Conﬂict 1989–2003. This network matches well the model of a trilateral conﬂict shown on the righthand side. Three groups that are mutual in conﬂict are revealed whose dominant members are {BOSSER, SER}, {UNO, NAT}, and {BOS}. The black-white gradient of the edges indicates the main direction. For instance, the edge from NAT directed to SER has a higher weight than the reverse edge from SER to NAT, meaning that more actions were directed from NATO to the Serbs than the reverse.

Figure 6. Visualization of Gulf Conﬂict 1979–1999: Two overlapping dominant conﬂicts can be detected. The trilateral conﬂict model shown in Fig. 5 (right) does not ﬁt well to this network. A remedy for this fact is to separate almost independent conﬂicts ﬁrst, as will be described in Sect. 2.3.

the eigenvectors we project on (see the algorithm description in Sect. 2.4). If the weight of the edge between USA and IRN got smaller and smaller, then these two actors would move towards each other until the pattern of a purely bipolar conﬂict is reached. 2.3. Parallel Conﬂicts The complete data set for the Gulf Conﬂict (1979–1999) did not match well a bilateral nor a trilateral conﬂict model (see Fig. 6), since it consists of two almost independent conﬂicts: that between ISR and LEB on one hand and the mutual conﬂicts between USA,

180

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

Figure 7. The dominant cluster in the Gulf Conﬂict data set (1979–1999) contains USA, IRQ, and IRN and shows a tri-polar structure, although a higher conﬂict intensity is reported between the pairs (IRQ,IRN) and (IRQ,USA) than between (USA,IRN).

IRQ, and IRN on the other. A way to obtain a better visual representation of this data set is to ﬁrst separate almost independent conﬂicts into different sub-networks and then visualize these independently as described before. The separation of independent conﬂicts can be done either by a network clustering algorithm that computes dense clusters (corresponding to subsets of actors that are strongly in conﬂict) or interactively by the analyst who chooses a subset of actors he or she is interested in. The Gulf Conﬂict data set contains three signiﬁcant clusters that have been obtained by a slight adaptation of a spectral clustering technique [30]. The strongest one (already shown in Fig. 7) has a mostly tri-polar structure. The two other conﬂict clusters (that have a very trivial bipolar structure) can be seen in Fig. 8. The formerly overlaying conﬂicts (see Fig. 6) are now displayed separately, each according to its structure.

Figure 8. The remaining two of the three major clusters in the Gulf Conﬂict data set (1979–1999) (the cluster containing USA, IRQ, and IRN has been shown already in Fig. 7) have a trivial bipolar structure and are shown in the above two diagrams.

2.4. Visualization Algorithm We summarize the method outlined in this section in the following algorithm for the twodimensional visualization of conﬂict networks. The algorithm takes as input a directed graph G = (V, E, ω) where the weight ω(u, v) of an edge (u, v) is determined by the conﬂictive actions targeted from actor u to actor v. 1. Divide the actor set into dense clusters C1 , . . . , Cp , either by a clustering algorithm (such as [30]), or manually during data analysis. 2. For all clusters C, compute A as the symmetric adjacency matrix of the subgraph deﬁned by C and perform the following steps to visualize C (let n = |C| denote the number of actors in C). (a) Compute the two minimal (negative) eigenvalues λmin and λmin of A together with (orthogonal and normalized) eigenvectors x and y.

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

181

(b) Construct the 2 × n matrix P whose ﬁrst row is equal to x and whose second λ row is equal to y · λmin (i. e., y scaled with the ratio of the next-to-minimal min eigenvalue divided by the minimal eigenvalue). (c) Draw actor v as an ellipse at the position deﬁned by the v’th column of P . The ratio height/width of the ellipse is proportional to the outdegree/indegree (i. e., activeness/passiveness) of v. The product of height with width (proportional to the area of the ellipse) is proportional to the Euclidean length of the v’th column of P (encoding v’s involvement). (d) Draw the strongest edges of the subgraph deﬁned by C (the number of edges drawn is a free parameter). The width of an edge (u, v) is proportional to the symmetric weight ω(u, v) + ω(v, u). A dark-grey to light-grey color gradient is directed from u to v if ω(u, v) > ω(v, u) and directed from v to u if ω(u, v) < ω(v, u). The darker side of this gradient is a ﬁxed grey-value (close to black). The lighter side is this grey value scaled with ω(v, u)/ω(u, v) if ω(u, v) > ω(v, u), so that the gradient becomes more pronounced if the ratio gets larger. An example of a strongly skewed edge is that from NAT directed to SER in Fig. 5. Note that the side closer to NAT is darker than the side closer to SER.

3. Animating Conﬂict Dynamics The images generated as described in Sect. 2 already reveal the actors and conﬂicts that are dominant over the whole period of time. However, due to varying conﬂict intensity and changing oppositions and alliances these images might not represent well the structure at speciﬁc time-points. Likewise, conﬂicts of short duration might be ﬁltered out. To obtain a more detailed insight into the evolution of conﬂicts, we will introduce a technique for smooth animation of the above type of scatterplots for limited periods of time. The event graph G is used to generate a sequence of graphs Gt , each of which represents the view on the set of events at the speciﬁc time t. A graph Gt yields one frame of the ﬁnal video and this frame shows a detailed image of the situation at time t. How the events are viewed at a certain time-point is determined by a scaling function η : R → R≥0 , which models how events move into the data when time increases and how they fade out. Examples of possible scaling functions are triangular shaped scaling functions with time radius r, as deﬁned and illustrated in Fig. 9. The function ηr does not ⎧ ⎨ (t + r)/r if |t| ≤ r and t < 0 ηr (t) = −(t − r)/r if |t| ≤ r and t ≥ 0 ⎩ 0 if |t| > r . Figure 9. Left: Deﬁnition of the triangular shaped scaling function ηr with time radius r. Right: Illustration of ηr for r = 2.

consider events with a time-stamp more then r away from the current time-point. Events

182

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

move into Gt linearly until t is larger than their time-stamp. Then, they fade out linearly until they have zero weight. For a ﬁxed scaling function η and time point t, the graph Gt = (V, E, ωt ) is deﬁned as follows. The actor set V and the edge or event set E are the same as for the input graph G. The weight ωt (e) of an event e at time t is deﬁned to be ωt (e) = ω(e) · η(te − t), that is, the weight of e at time t is its absolute weight ω(e) times a scaling factor which is dependent on the difference between the time-stamp te of the event and the current time t. The graph Gt may be reduced by removing events with zero weight, as well as isolated actors, since these do not inﬂuence the analysis and would be invisible in the ﬁnal video. Given a graph G, representing a list of events, the movie is generated by the following steps. 1. Select a sequence of time-points t1 < · · · < tN in a given time interval. 2. For each i from 1 to N Compute the visualization of the graph Gti . 3. The images for all time-points yield the frames of the video. In order to maintain the overall appearance of the frames one further detail has to be taken into consideration. If v is an eigenvector of A associated to eigenvalue λ, then so is −v. Thus the eigensolver algorithm could return either v or −v as a solution to the eigenvalue problem. To prevent that this assignment switches from one frame to another (which would result in interchanging the axes of the coordinate system from one frame to another) we have to ensure that the eigenvectors we use point in a well-deﬁned direction. For the bipolar conﬂict projection the canonical direction for the eigenvector vmax associated to the largest eigenvalue is simply the direction in which each entry of vmax is positive. (It is standard knowledge in algebraic graph theory that all entries of this eigenvector have the same sign.) We deﬁne the canonical direction for the eigenvector vmin for time-step t recursively by the direction of this eigenvector for time-step t − 1. The direction of vmin is chosen such that the angle between vmin at time t and vmin at time t − 1 is smaller than 90 degrees. Thus, only the direction of vmin for the very ﬁrst time-step is arbitrary. This translates to the fact that there is no absolute meaning attached to the two opponent groups. A second computation of the movie could interchange the ﬁrst and second group, but then it has to reverse the assignment for all actors and at all time points, which results in the same opponents. For multipolar conﬂicts the computed coordinates are only unique up to rotation or reﬂexion of the two-dimensional image space. To ensure that the images are not rotated or reﬂected from one image to the next we determine the two-dimensional orthogonal transformation that minimizes the distance between images at time t and time t − 1 and apply it to the drawing at time t. More precisely, let t be a time-point that is not the ﬁrst and let Pt−1 be the 2 × n matrix holding the two-dimensional coordinates of all n actors at time t − 1. Further, let Pt be the 2 × n matrix as returned by the eigensolver at time t. Set X = Pt−1 PtT and compute the Singular Value Decomposition (SVD) X = U SV T of X [29]. The optimal coordinates at time t are given in the matrix Pt = U V T Pt . Of course, animation can also be applied to sub-networks representing independent conﬂict clusters. In Fig. 10 we show three selected time points in the Gulf Conﬂict cluster containing USA, IRQ, and IRN. In the smooth animation, the radical change of the relative positions of USA and IRQ can be easily followed.

183

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

Figure 10. Dynamic visualization of Gulf Conﬂict (cluster containing USA, IRQ, and IRN) (left: June 1989, middle: March 1990, right: August 1990). Note that especially the relative positions of USA and IRQ changed completely during this period.

4. Application Example We apply our method to visualize a data set from the Kansas Event Data System (KEDS) [5] (see Balkan data set [5]) in a prototypical implementation. The animations3 are realized in SVG format (Scalable Vector Graphics, see W3C Recommendation at http://www.w3.org/TR/SVG/), thus they can be viewed on any web browser with an appropriate plug-in. 4.1. Dynamic Bipolar Visualization

1

1 Volatility 0.9

0.9 0.8 0.7

0.8

0.6

0.7

0.5 0.4

0.6 0.3 0.2

0.5 0.1 0 1988

0.4 1990

1992

1994

1996

1998

2000

2002

2004

1990

1992

1994

1996

1998

2000

2002

Figure 11. Left: Volatility proﬁle of the bipolar visualization of the Balkan conﬂict. Right: Bipolarity (ﬁt of the bipolar model) of the Balkan Conﬂict.

The varying degree of polarization can be inferred from the model ﬁt indicator curve in Fig. 11 (right). Although there is great variation in the magnitude of the model ﬁt, it is often close to one and at all time points considerably distant from zero. Thus, the simplistic assumption of bipartite conﬂicts already ﬁts the data sufﬁciently well—if we restrict the analysis to relatively short time intervals. (As it has been noted before, summarizing the whole data set over 14 years in a bipolar model provides suboptimal results, e.g. see Figs. 4 and 5.) The volatility indicator [24] measures the sensitivity to noise of 3 The animation of the bipolar visualization are available from http://www.inf.uni-konstanz.de/algo/research/conflict/

184

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

our drawing technique. The volatility proﬁle for the Balkan conﬂict is shown in Fig. 11 (left). The fact that this indicator is small during most time-steps gives guarantees the stability of our method. The peaks in this plot, where the volatility reaches a value of one, can be recognized in the animation by a sudden movement of the actors. The following ﬁgures show selected time-points of the Balkans conﬂict.

Figure 12. War in Bosnia, ﬁrst semester 1995 (only edges incident to UNO are shown). Left: Two opposing groups and UNO trying to mediate. Right: BOSSER’s troops in conﬂict with the UN. The heavy edge pushes UNO to the group on the right-hand side.

Figure 13. Left: NATO bombing in Bosnia, second semester 1995. Note that BOSSER changed from high and narrow (being the source of events) in Fig. 12 (right) to broad and ﬂat (being a target). Right: Dayton peace talks.

Figure 14. Left: Conﬂict between Turkey and the Kurds before the change from Reuters North America to Reuters Business Brieﬁng (June 1997). Right: During the change: TUR and KUR are still visible to the left and to the right of the origin and are rapidly moving towards it.

Important changes in the conﬂict structure took place in 1995 and 1996. Figure 12 (left) shows the war in Bosnia, where Serbia and the Serbs in Bosnia (BOSSER) are opposed to Bosniaks and Croats. The UNO, which is trying to install peace in Bosnia, has conﬂicts of similar strength to all of them. This changes when troops of the Bosnian Serbs captured weapons from UN peace keepers and declined to return them (Fig. 12right). After the Bosnian Serbs did not respond to an ultimatum, the NATO started air strikes

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

185

under the order of the UN (Fig. 13left). The opposing parties ﬁnally participated in peace talks which took place in Dayton, Ohio and were signed in December 1995 (Dayton Peace Agreement, Fig. 13right). After this, events in the Balkans calmed down and the media focused on the conﬂict between Turkey and the Kurds. The conﬂict between Turkey and the Kurds also exempliﬁes a problem with the data that we were not aware of before seeing the animation. In July 1997, there is an abrupt change in media coverage in the sense that reports on hostilities between Turks and Kurds are suddenly missing. Figure 14 shows the conﬂict structure in the Balkans with only a few days in between. The change is also visible in a signiﬁcant drop in the bipolarity curve (Fig. 11right), where the highly bipolar situation rapidly changes into a more complex one, and in a peak in the volatility curve (Fig. 11left). That this change is indeed supported by the data can be veriﬁed by printing the events involving TUR and KUR. During a period of one month from May 10’th 1997 to June 10’th 1997 many hostile events between these two actors are reported (see Fig. 15). 970514 970514 970514 970520 970521 970522 970522 970522 970522 970524 970526 970527 970602 970604 970604 970605 970607 970609 970610

TUR TUR TUR TUR TUR KUR TUR KUR KUR KUR TUR TUR TUR TUR TUR TUR TUR TUR KUR

KUR KUR KUR KUR KUR TUR KUR TUR TUR TUR KUR KUR KUR KUR KUR KUR KUR KUR TUR

(MIL ENGAGEMENT) KILLED (MIL ENGAGEMENT) TROOPS CLASHED (MIL ENGAGEMENT) TURKISH PUSHED AGAINST KURDISH (MILITARY DEMO) HUNTING DOWN (MILITARY DEMO) BUILDING UP FORCES (MIL ENGAGEMENT) ATTACKED KILLING (MIL ENGAGEMENT) TURKISH PUSHED AGAINST KURDISH (NONMIL DEMO) STAGED PROTEST (DENIGRATE) CONDEMNATION (MIL ENGAGEMENT) ATTACKS ARMY (MIL ENGAGEMENT) TROOPS CLASHED (MIL ENGAGEMENT) BOMBED (MIL ENGAGEMENT) KURDISH KILLED IN TURKISH (MIL ENGAGEMENT) KURDISH KILLED IN TURKEY (ARREST PERSON) JAIL (MIL ENGAGEMENT) KILLED (DENY) DENIED (DENY) DENIED (DEMAND) DEMANDING

Figure 15. Hostile events between TUR and KUR from 05/10/97 to 06/10/97.

In contrast, during a period of one month from 11 June 1997 until 11 July 1997 there is no hostile event reported between TUR and KUR. There are no prominent historic events explaining this sudden “peace”. However, turning to the data description gives the information that this is precisely the time when KEDS sources change from Reuters North America to Reuters Business Brieﬁng, with the latter apparently not covering the conﬂict (or TUR and KUR being ﬁltered out during preparation of the data for the KEDS parser). 4.2. Multipolar Visualization of Independent Conﬂict-Clusters The conﬂict network arising from the Balkan data set contains several interesting subnetworks corresponding to separate conﬂicts. Figure 16 shows six selected conﬂict clusters that have been identiﬁed by a spectral clustering technique [30]. Note that some of these sub-networks have a trilateral structure (such as the war in Bosnia shown at the top-left). Other clusters have a purely bipolar structure, as e. g., Slobodan Miloševi´c (SERSM) versus the UN War Crime Tribunal (UNWCT) shown at the bottom-right. Yet other sub-conﬂicts have an intermediate structure, as e. g., conﬂicts on Cyprus shown at the top-right. Naturally, it could be argued that some of these conﬂicts should not be part of the Balkan data set. Thus, the visualization presented here may also serve as a support for data cleaning.

186

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

Figure 16. Strong conﬂict clusters in the Balkan data set.

While the six conﬂict sub-networks in Fig. 16 seem to be reasonable and correspond to known historic events, we detected another conﬂict cluster in the Balkan data set that surprised us. According to the network shown in Fig. 17, the USA-Government (USAGOV) would be a serious opponent of the USA.

Figure 17. One conﬂict cluster found in the data set seems hard to believe. The hostile edge between USAGOV and USA is probably due to some systematic errors of the KEDS parser (see text for a more detailed description).

To ﬁnd out the reason for this strange conﬁguration we printed all major hostile events with weight < 5.0 between these two actors in Fig. 18. Apparently the KEDS parser repeatedly interpreted certain recurrent statements in the news as military demonstrations of the US-Government targeted against the USA. (Military demonstration is a serious hostile event with weight= −7.6; the most hostile event has weight −10.0.) Note that although, this error seems to be repeated quite often, the resulting conﬂictive edge would vanish against the hostilities between (say) BOSSER and BOS. In particular, it would not be visible without the prior clustering. The story around Fig. 17 is a good

U. Brandes and J. Lerner / Visualization of Conﬂict Networks

930224 930714 930729 940210 940525 950104 950803 950929 951008 951119 951119 951120 951125 951202 951203 951213 960113 960113 960203 960711 960917 961006 970624 990501 990504 010616

USAGOV USA USAGOV USA USAGOV USA USA USAGOV USAGOV USA USA USAGOV USAGOV USA USAGOV USA USAGOV USA USAGOV USA USAGOV USA USAGOV USA USAGOV USA USAGOV USA USAGOV USA USAGOV USA USAGOV USA USAGOV USA USA USAGOV USAGOV USA USAGOV USA USAGOV USA USAGOV USA USAGOV USA USA USAGOV USA USAGOV

187

(ARREST PERSON) ROUNDED UP (MILITARY DEMO) SENDING TROOPS (NONMIL THREAT) CONSIDERING STRIKES (MIL ENGAGEMENT) FIRED AT PRESIDENT CLINTON’S (MILITARY DEMO) SEND TROOPS (CUT AID) EMBARGO (CUT AID) EMBARGO (MILITARY DEMO) SEND TROOPS (MILITARY DEMO) SEND TROOPS (MILITARY DEMO) SEND TROOPS (MILITARY DEMO) SEND TROOPS (MILITARY DEMO) SENDING TROOPS (MILITARY DEMO) SEND TROOPS (MILITARY DEMO) SEND TROOPS (MILITARY DEMO) SEND TROOPS (CUT AID) VOTED TO CUTTING OFF FUNDS (MILITARY DEMO) INSPECTED TROOPS (MILITARY DEMO) INSPECTED TROOPS (ARREST PERSON) HOLDS (MIL ENGAGEMENT) ASSAULT (MILITARY DEMO) ORDERED TROOPS (MILITARY DEMO) SENT TROOPS (CUT AID) VOTED TO CUT OFF FUNDS (SEIZE POSSESSION) EXPANDED (MIL ENGAGEMENT) KILLED (NONMIL DEMO) DEMONSTRATED

Figure 18. Hostile events between USA and USAGOV. Military demonstration is a serious event (weight= −7.6) that repeatedly has been interpreted by the KEDS parser as being targeted against the USA.

example to illustrate the utility of visualization for data cleaning and/or improvement of automatic event parsers.

5. Discussion In this chapter we presented a general method for the visualization of conﬂict networks. We focused on the description of the visualization technique and brieﬂy demonstrated its usefulness. The images produced give deep insight into the conﬂict structure and, as illustrated in the examples, may lead to the detection of coding errors. A future ready-to-use conﬂict visualizer would certainly beneﬁt from interaction possibilities allowing the analyst to trace back the events responsible for conﬂict edges and ﬁnally trace back the original news reports [24]. Another issue for future work is to augment the method to simultaneously visualize other relations (such as membership to ofﬁcial alliances, or geographic closeness) and supplementary attributes of the actors (such as ethnic composition, distinction between state and non-state actors etc.). Visualizing attribute data can be conveniently done by color or texture of the nodes representing actors.

References [1] Kazuo Yamaguchi. Event History Analysis. Sage, 1991. [2] Nils Petter Gleditsch, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, and Håvard Strand. Armed conﬂict 1946-2001: A new dataset. Journal of Peace Research, 39(5):615–637, 2002. [3] David Singer. The correlates of war project: Interim report and rationale. World Politics, 24(2):243–270, 1972. [4] Gary King and Will Lowe. An automated information extraction tool for international conﬂict data with performance as good as human coders: A rare event evaluation design. International Organization, 57:617–642, 2003. [5] Phillip A. Schrodt, Shannon G. Davis, and Judith L. Weddle. Political science: KEDS-a program for the machine coding of event data. Social Science Computer Review, 12(3):561–588, 1994. (Data sets available at http://www.ku.edu/∼keds/data.dir/balk.html).

188 [6] [7] [8]

[9]

[10] [11] [12] [13]

[14] [15]

[16] [17] [18] [19] [20]

[21]

[22] [23] [24]

[25] [26] [27] [28] [29] [30]

U. Brandes and J. Lerner / Visualization of Conﬂict Networks Gerald Schneider and Vera E. Troeger. War and the world economy: Stock market reactions to international conﬂicts, 1990–2000. Journal of Conﬂict Resolution, 50(5):623–645, 2006. Massimo Guidolin and Eliana La Ferrara. The economic effects of violent conﬂict: Evidence from asset market reactions. Technical report, Federal Reserve Bank of St. Louis, 2005. Gerald Schneider. Banking on the broker? Forecasting conﬂict in the levant with ﬁnancial data. In Frank Wayman, Paul Williamson, and Bruce Bueno de Mesquita, editors, Prediction: Breakthroughs in Science, Markets, and Politics. University of Michigan Press, to appear. Zeev Maoz, Ranan D. Kuperman, Lesley G. Terris, and Ilan Talmud. International relations: A network approach. In A. Mintz and B. Russett, editors, New Directions for International Relations. Lexington Books, 2004. Zeev Maoz. Network polarization, network interdependence, and international conﬂict, 1816–2002. Journal of Peace Research, 43(4):391–411, 2006. Zeev Maoz, Ranan D. Kuperman, Lesley G. Terris, and Ilan Talmud. Structural equivalence and international conﬂict. Journal of Conﬂict Resolution, 50(5):664–689, 2006. Frank Harary. A structural analysis of the situation in the Middle East in 1956. Journal of Conﬂict Resolution, 5(2):167–178, 1961. Zeev Maoz, Lesley G. Terris, Ranan D. Kuperman, and Ilan Talmud. What is the enemy of my enemy? Causes and consequences of imbalanced international relations, 1816–2001. Journal of Politics, 69(1):100–115, 2007. Joan Esteban and Debraj Ray. A comparison of polarization measures. Technical report, UAB and CSIC, 2007. (available at http://ideas.repec.org/p/aub/autbar/700.07.html). Joshua S. Goldstein, Jon C. Pevehouse, Deborah J. Gerner, and Shibley Telhami. Reciprocity, triangularity, and cooperation in the Middle East, 1979-97. Journal of Conﬂict Resolution, 45(5):594–620, 2001. R. N. Rosecrance. Bipolarity, multipolarity, and the future. Journal of Conﬂict Resolution, 10(3):314– 327, 1966. August Hämmerli, Regula Gattiker, and Reto Weyermann. Conﬂict and cooperation in an actors’ network of Chechnya based on event data. Journal of Conﬂict Resolution, 50(2):159–175, 2006. James Allan, Rahul Gupta, and Vikas Khandelwal. Temporal summaries of news topics. In Proc. ACMSIGIR’01, pages 10–18, 2001. James Allan, Ron Papka, and Victor Lavrenko. On-line new event detection and tracking. In Proc. ACM-SIGIR’98, pages 37–45, 1998. Clive Best, Erik Van der Groot, and Monica de Paola. Thematic indicators derived from world news reports. In Proc. International Conference on Intelligence and Security Informatics (ISI 2005), pages 436–447, 2005. Pak Chung Wong, Harlan Foote, Dan Adams, Wendy Cowley, and Jim Thomas. Dynamic visualization of transient data streams. In Proc. IEEE Symp. Information Visualization (InfoVis ’03), pages 97–104, 2003. John M. Chambers, William S. Cleveland, Beat Kleiner, and Paul A. Tukey. Graphical Methods for Data Analysis. Wadsworth, 1983. William S. Cleveland and Robert McGill. The many faces of a scatterplot. Journal of the American Statistical Association, 79(388):807–822, 1984. Ulrik Brandes, Daniel Fleischer, and Jürgen Lerner. Summarizing dynamic bipolar conﬂict structures. IEEE Transactions on Visualization and Computer Graphics, special issue on Visual Analytics, 12(6):1486–1499, 2006. Charles A. McClelland. World event/interaction survey codebook (icpsr 5211), 1976. Joshua S. Goldstein. A conﬂict-cooperation scale for WEIS international events data. Journal of Conﬂict Resolution, 36(2):369–385, 1992. Michael Kaufmann and Dorothea Wagner, editors. Drawing Graphs. Springer Verlag, 2001. Ulrik Brandes and Jürgen Lerner. Structural similarity in graphs. In Proc. 15th Intl. Symp. Algorithms and Computation (ISAAC’04), pages 184–195, 2004. Gene H. Golub and Charles F. van Loan. Matrix Computations. John Hopkins University Press, 1996. Ravi Kannan, Santosh Vempala, and Adrian Vetta. On clusterings: Good, bad and spectral. Journal of the ACM, 51(3):497–515, 2004.

189

Building and Using Datasets on Armed Conflicts M. Kauffmann (Ed.) IOS Press, 2008 © 2008 IOS Press. All rights reserved.

Author Index Ben-Israel, G.M. Bennett, D.S. Brandes, U. Carlsen, J. Collin, J.-M. Eberwein, W.-D. Eck, K.

63 133 169 161 41 13 29

Kauffmann, M. Kreutz, J. Kuperman, R.D. Lerner, J. Mincheva, L. Pétris, R. Raleigh, C.

1, 107 49 vii, 97 169 75 v 161

This page intentionally left blank

This page intentionally left blank

This page intentionally left blank

Complexity and Security (Nato Science for Peace and Series) (Nato Science for Peace and Series)

Read more

Mini-Micro Fuel Cells: Fundamentals and Applications (NATO Science for Peace and Security Series C: Environmental Security) (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Energy Options Impact on Regional Security (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Energy and Environmental Challenges to Security (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Environmental Security and Public Safety (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Smart Materials for Energy Communications and Security NATO Science for Peace and Security Seri

Read more

Stem Cells and Their Potential for Clinical Application (NATO Science for Peace and Security Series NATO Science for Peace and Security Series A: Chemistry and Biology)

Read more

Nuclear Power and Energy Security (NATO Science for Peace and Security Series B: Physics and Biophysics)

Read more

Hypermedia Seduction for Terrorist Recruiting - Volume 25 NATO Science for Peace and Security Series: Human and Societal Dynamics (Nato Science for Peace and Security)

Read more

Problems of High Altitude Medicine and Biology (NATO Science for Peace and Security Series NATO Science for Peace and Security Series A: Chemistry and Biology)

Read more

Geographic Uncertainty in Environmental Security (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Nuclear Doctrines and Strategies (Nato Sciences for Peace and Security)

Read more

Nanomaterials: Risks and Benefits (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Virus Diseases and Crop Biosecurity (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Nanotechnology - Toxicological Issues and Environmental Safety (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Nanomaterials: Risks and Benefits (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Boolean Functions in Cryptology and Information Security (Nato Science for Peace and Security)

Read more

Quantum Communication and Security (NATO Science for Peace and Security Series: Information and Communication Security) (NATO Security Through Science Series. D: Information and Com)

Read more

Imaging for Detection and Identification (NATO Science for Peace and Security Series B: Physics and Biophysics)

Read more

Harbour Protection Through Data Fusion Technologies (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Green Chemical Reactions (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Countering Nuclear and Radiological Terrorism (NATO Science for Peace and Security Series B: Physics and Biophysics)

Read more

Hamiltonian Dynamical Systems and Applications (NATO Science for Peace and Security Series B: Physics and Biophysics)

Read more

Optical Waveguide Sensing and Imaging (NATO Science for Peace and Security Series, B: Physics and Biophysics)

Read more

Hamiltonian Dynamical Systems and Applications (NATO Science for Peace and Security Series B: Physics and Biophysics)

Read more

Hamiltonian Dynamical Systems and Applications (NATO Science for Peace and Security Series B: Physics and Biophysics)

Read more

Security and Reliability of Damaged Structures and Defective Materials (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Remote Sensing of the Atmosphere for Environmental Security (NATO Science for Peace and Security Series C: Environmental Security)

Read more

Metastable Systems under Pressure (NATO Science for Peace and Security Series A: Chemistry and Biology)

Read more

Extreme Photonics & Applications (NATO Science for Peace and Security Series B: Physics and Biophysics)

Read more

Recommend Documents

Complexity and Security (Nato Science for Peace and Series) (Nato Science for Peace and Series)

COMPLEXITY AND SECURITY NATO Science for Peace and Security Series This Series presents the results of scientific meet...

Mini-Micro Fuel Cells: Fundamentals and Applications (NATO Science for Peace and Security Series C: Environmental Security) (NATO Science for Peace and Security Series C: Environmental Security)

Mini-Micro Fuel Cells NATO Science for Peace and Security Series This Series presents the results of scientific meeti...

Energy Options Impact on Regional Security (NATO Science for Peace and Security Series C: Environmental Security)

Energy Options Impact on Regional Security NATO Science for Peace and Security Series This Series presents the result...

Energy and Environmental Challenges to Security (NATO Science for Peace and Security Series C: Environmental Security)

Energy and Environmental Challenges to Security NATO Science for Peace and Security Series This Series presents the r...

Environmental Security and Public Safety (NATO Science for Peace and Security Series C: Environmental Security)

Environmental Security and Public Safety NATO Security through Science Series This Series presents the results of sci...

Smart Materials for Energy Communications and Security NATO Science for Peace and Security Seri

Smart Materials for Energy, Communications and Security NATO Science for Peace and Security Series This Series presen...

Stem Cells and Their Potential for Clinical Application (NATO Science for Peace and Security Series NATO Science for Peace and Security Series A: Chemistry and Biology)

Stem Cells and their Potential for Clinical Application NATO Science for Peace and Security Series This Series presen...

Nuclear Power and Energy Security (NATO Science for Peace and Security Series B: Physics and Biophysics)

Hypermedia Seduction for Terrorist Recruiting - Volume 25 NATO Science for Peace and Security Series: Human and Societal Dynamics (Nato Science for Peace and Security)

HYPERMEDIA SEDUCTION FOR TERRORIST RECRUITING NATO Science for Peace and Security Series This Series presents the resu...

Problems of High Altitude Medicine and Biology (NATO Science for Peace and Security Series NATO Science for Peace and Security Series A: Chemistry and Biology)

Problems of High Altitude Medicine and Biology NATO Science for Peace and Security Series This Series presents the res...