Universal Patterns Underlying Ongoing Wars and Terrorism

Professor Jorge A. RestrepoDepartment of Economics

Universidad Javeriana

Professor Michael SpagatDepartment of Economics

Royal Holloway, University of London

1 Department of Physics, University of Oxford, Oxford, U.K.2 Department of Economics, Royal Holloway College, University of London, Egham, U.K.3 Department of Economics, Universidad Javeriana, Bogotá, Colombia4 Department of Industrial Engineering, Universidad de los Andes, Bogotá, Colombia5 Department of Economics, Universidad Nacional de Colombia, Bogotá, Colombia6 Department of Economics, Universidad de Los Andes, Bogotá, Colombia7 CERAC, Conflict Analysis Resource Center, Bogotá, Colombia

Neil Johnson 1,7 Michael Spagat 2,7 Jorge A. Restrepo 3,7 Juan Camilo Bohorquez 4

Nicolás Suárez 5,7 Elvira Ma. Restrepo 6,7Roberto Zarama 4Oscar Becerra 5,7

• Is there a common structure of modern warfare that remains more or less constant across diverse wars?

• Aside from the inherent interest of this question, the answer can have important implications for the practical conduct of war, including medical and insurance planning.

• What is the relationship between terrorism and modern warfare?

• The distinction between the two is often blurred, e.g., in the concept of the “war on global terrorism”. But there does appear to be a real relationship there and knowledge of its nature will be very welcome.

Introduction

• There are remarkable regularities and similarities in the size distribution of violent events in Colombia, Iraq, Afghanistan, Indonesia, Israel-Palestine and Northern Ireland

• We have similar results, with interesting variation for Casamance (Senegal) global terrorism, the US Civil War and the Spanish Civil War.

• We can understand the findings for modern wars in terms of a model of the coalescence and fragmentation of insurgent groups.

• One can learn about the nature of an insurgency from studying the size distribution of casualties that it throws up.

Introduction

• Hammes (2004) argues that typical wars these days are irregular, guerrilla or terrorist-type affairs. Pentagon planners are trapped in a time warp, preparing for increasingly irrelevant conventional wars in his view.

• Lind et. al. (1989) wrote earlier of “fourth generation war” marked by greater battlefield dispersion, importance of small, manoeuvrable groups, diminished need for centralized logistics, blurred distinctions between military and civilian targets, parallels with terrorism and emphasis on breaking the will of the enemy.

• A key idea here is that there are important commonalities that apply across a range of modern wars.

Introduction

• Our team maintains a database on the Colombian conflict, 1988-June 2005.

• This dataset has more than 21,000 events.

• We include only clear conflict events, requiring each event to have a military effect and to reflect a group action.

The Colombian Conflict

Ana Cristina Restrepo

Simón Mesa

Wilmer Marín

Sandra P. Oviedo

Katharina Grosso

Gustavo Perdomo

CERAC Data-coding team

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for Colombia

• – the probability that a particular event will be of size x.

•

• Such a distribution is called a “power law”.

• Take logs of both sides and the plot is a straight line.

•

Mathematics

xp

minxxallforCxxp ,

xCxp logloglog

• Fat Tails

• Scale invariance

Power Laws

Log-Log plots of 1 – the Cumulative Distribution Function for the Power Law, Lognormal and Normal distributions

The three curves have the same mean and standard deviation.

• CERAC Integrated Iraq Dataset (CIID)

• Takes events from the Iraq Body Count Project, icasualties (for coalition troops, Iraqi security personnel and defence contractors), ITERATE, a terrorism database.

• Eliminates some IBC events, includes coalition casualties, gets additional events and matches events across sources to build confidence and avoid double counting.

Iraq Conflict

Venn Diagram for Iraq (CIID) (After May 1st 2003)

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for Iraq (CIID)

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for Iraq (CIID) (After May 1st 2003)

• Afghanistan, based on the work of Marc Herold, an economics professor at the University of New Hampshire, plus ITERATE and icasualites

• Data on separatist conflict on Indonesia provided to us by Professor Ali Kutan of the Southern Illinois University.

• Data on Casamance in Senegal provided to us by Professor Macartan Humphreys of the Colombia University.

More Modern Conflict Data

More Modern Conflict Data

• Israel-Palestine, based on a count from a chronology in a book by Anthony Cordesman of CSIS.

• Northern Ireland, based on the work of Malcolm Sutton, a journalist.

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for Afghanistan

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for Indonesia (separatist)

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for Senegal

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for Israel-Palestine Conflict

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for Northern Ireland

• Professor Ron Francisco of the University of Kansas political science department has done two wonderful datasets posted on his web site.

• They include killings and injuries and are based on a number of history books.

• He has done the US and Spanish Civil Wars and has the Russian Civil War on the way.

Old Wars: US Civil War and Spanish Civil War

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for US civil war

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for Spanish civil war

• Previous work by Clauset and Young (2005) used MIPT database and found power laws in terrorism.

• They got coefficients of 1.7 for G7 targets and 2.5 for non-G7 targets.

• We used the ITERATE database that records major international terrorist attacks, 1968-2004 and got similar results although we think that the data still needs more work.

• Thus, non-G7 terrorism looks much like a new war and G7 terrorism looks much like an old war.

Terrorism

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for non G7 Terrorism

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for G7 Terrorism

Log-Log plots of 1 – the Theoretical Cumulative Distribution Function for Severity of Events for old wars, new wars and terrorism

Log-Log plots of 1 – the Theoretical Cumulative Distribution Function for Severity of Events for old wars and new wars (without Iterate)

Power Law Coefficient: Variation Through Time for Colombia

Power Law Coefficient: Variation Through Time for Iraq (CIID)

Feb2004 Jun2004 Oct2004 Feb2005 Jun2005 Oct2005

1

1.5

2

2.5

3

3.5

4

4.5

Non-G7 Terrorism

G7 Terrorism

1990 1992 1994 1996 1998 2000 2002 2004

1

1.5

2

2.5

3

3.5

4

4.5

Colombia

Iraq

Power Law Coefficient: Variation Through Time

• We find extraordinary similarities in the size distribution of violent events in Iraq and Colombia.

• Iraq and Colombia differ strongly in a number of highly visible ways.

– Colombia has “rough terrain”, i.e., extensive mountains and jungles, completely contrary to Iraqi geography.

– Iraq has strong ethnic/religious cleavages, completely unlike Colombia.

– The ideologies of Colombia’s insurgent groups, more or less Marxism, differ strongly from the ideologies of the insurgent groups in Iraq.

• But a common underlying logic renders both conflicts structurally almost identical along a key characteristic dimension.

Conclusion

• Moreover, we get strikingly similar results for Afghanistan, Indonesia, Israel-Palestine, Nothern Ireland and Senegal (Casamance), i.e., there seems to be a very reliable pattern to modern insurgency.

• The Spanish and US Civil Wars, the older wars, seem rather different.

– They are much bloodier overall.– Event probabilities decline much less rapidly with size in the old

wars compared to the modern wars.– They do not fit power laws as nicely as modern wars do.

Conclusion

• So “Modern War” seems to be a valid category of analysis.

• Geography, ethnicity, religion and ideology are not unimportant - various wars and global terrorism do differ and good analysis must make reference to local specifics.

• But there is an underlying logic to insurgency that should be central to the study of modern warfare.

• There is a potentially high payoff to in-depth, micro-level studies of individual wars combined with comparative work ranging across these wars.

Conclusion

Supplementary Material

Communal Violence

• We use a different part of the Indonesia data provided by Ali Kutan of SIU.

• We use data on the Rwandan Genocide provided by Professor Christian Davenport of the University of Maryland.

• These are different from modern insurgencies. The Rwanda case is really mass slaughter while the Indonesia case is different communities killing each other.

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for Indonesia (communal)

Log-Log plots of 1 – the Cumulative Distribution Function for Severity of Events for Rwanda-Genocide