Political Cleavages and the Propensity for Victimization

Do victims have particular demographic characteristics that differentiate them from non-victimized persons? Does ethnicity delineate between the victims and the non-victims of violence during armed conflicts, as the media often claims? Have the political cleavages dividing the communities increased the severity of violence? This study addresses these issues by analyzing data from “the People on War project” (the PoW project) through latent class analysis (LCA). Data were collected by the International Committee of the Red Cross (the ICRC) at the 50th anniversary of the signature of the Geneva Conventions in 1999 throughout twelve armed conflict zones in the world. [Supplemental materials are available for this article. Go to the publisher's online edition of Terrorism and Political Violence for the following free supplemetal resources: Table 2: Exposed Risk of Violence During Armed Conflict and Demographic Factors (With Regional Values), and Table 3: Risk of Violence During Armed Conflict and Demographic Factors (% Having Taken Sides)]. The study was funded by Grant-in-Aid for Scientific Research © of the Japan Society for the Promotion of Science (No. 21530159). The author appreciates the support. Notes Note. Table 1 shows the distributions associated with the models in Table 2. The first column of total denotes the weighted percentages of each latent class in a country, while other columns are the conditional probabilities of an indicator given each latent class. “Humil” means to having felt humiliated, “relatives” means to lose contact with relatives, “family” means to have their family members killed, “food” means to have their food taken away by combatants, while “rapes” means that somebody you had known had been raped. The Israeli data asks the question if somebody you had known had been sexually assaulted. Mahmood Mamdani, When Victims Become Killers: Colonialism, Nativism, and the Genocide in Rwanda (Princeton, NJ: Princeton University Press, 2001). A large number of works on ethnic conflicts have relied upon the rhetoric that distinguishes “good sides” and “bad ones.” Robert J. Donia and John V. A. Fine, Jr., Bosnia and Herzegovina: A Tradition Betrayed (London: Hurst and Company, 1994). Greenberg, Inc., The People on War Report: ICRC Worldwide Consultation on the Rules of War (Geneva: International Committee of the Red Cross, 1999), iii. Reports are obtained from http://www.icrc.org/web/eng/siteeng0.nsf/htmlall/greenberg_reports?opendocument (accessed April 17, 2010.) Greenberg Research Inc. is a U.S. private research company which commissioned the PoW project from the ICRC. Datasets are currently distributed by the Swiss Foundation for Research in Social Sciences (the FORS) of Lausanne University at http://forsdata.unil.ch/projects/pow/default.asp?lang=e (accessed October 18, 2011.). Barbara Harff, “No Lessons Learned from the Holocaust? Assessing Risks of Genocide and Political Mass Murder Since 1955,” American Political Science Review 97, no. 1 (February 2003): 57–73; Matthew Krain, “State-Sponsored Mass Murder: The Onset and Severity of Genocides and Politicides,” Journal of Conflict Resolution 41, no. 3 (June 1997): 331–360; R. J. Rummel, “Democracy, Power, Genocide, and Mass Murder,” Journal of Conflict Resolution 39, no. 1 (February 1995): 3–26; Will H. Moore and Stephen M. Shellman, “Fear of Persecution: Forced Migration, 1952–1995,” Journal of Conflict Resolution 48, no. 5 (October 2004): 723–745; Havard Hegre, Gudrum Ostby, and Clionadh Raleigh, “Poverty and Civil War Events: A Disaggregated Study of Liberia,” Journal of Conflict Resolution 53, no. 4 (August 2009): 598–623; Anjali Thomas Bohlken and Ernest John Sergenti, “Economic Growth and Ethnic Violence: An Empirical Investigation of Hindu-Muslim Riots in India,” Journal of Peace Research 47, no. 5 (September 2010): 589–600; Quy-Toan Do and Lakshmi Iyer, “Geography, Poverty and Conflict in Nepal,” Journal of Peace Research 47, no. 6 (November 2010): 735–746; Mathias Czaika and Krisztina Kis-Katos, “Civil Conflict and Displacement: Village-Level Determinants of Forced Migration in Aceh,” Journal of Peace Research 46, no. 3 (May 2009): 399–418; Stathis N. Kalyvas, The Logic of Violence in Civil War (Cambridge: Cambridge University Press, 2006); Stathis Kalyvas, “Wanton and Senseless? Logic of Massacres in Algeria,” Rationality and Society 11, no. 3 (July 1999): 243–285; Stathis N. Kalyvas and Matthew Adam Kocher, “The Dynamics of Violence in Vietnam: An Analysis of the Hamlet Evaluation System (HES),” Journal of Peace Research 46, no. 3 (June 2009): 335–355. Inger Skjelsbaek, “Sexual Violence and War: Mapping Out a Complex Relationship,” European Journal of International Relations 7, no. 2 (June 2001): 211–237; Elizabeth Jean Wood, “Variations in Sexual Violence During War,” Politics & Society 34, no. 3 (September 2006): 307–341; Elizabeth Jean Wood, “Armed Groups and Sexual Violence: When Is Wartime Rape Rare?” Politics & Society 37, no. 1 (March 2009): 131–162; Manus I. Midlarsky, “The Demographics of Genocide: Refugees and Territorial Loss in the Mass Murder of European Jewry,” Journal of Peace Research 42, no. 4 (July 2005): 375–391. It should be noted that some of these small-N case studies attempt to reduce the selection bias by selecting the cases of the different outcomes. See Wood, “Variations in Sexual Violence” (see note 4 above); Midlarsky (see note 4 above). Philip Verwimp, “Machetes and Firearms: The Organization of Massacres in Rwanda,” Journal of Peace Research 43, no. 1 (January 2006): 5–22; Philip Verwimp, “An Economic Profile of Peasant Perpetrators of Genocide Micro-Level Evidence from Rwanda,” Journal of Development Economics 77 (2005): 297–323; Philip Verwimp, “Testing the Double-Genocide Thesis for Central and Southern Rwanda,” Journal of Conflict Resolution 47, no. 4 (August 2003): 423–442; Aderoju Oyefusi, “Oil and the Probability of Rebel Participation Among Youths in the Niger Delta of Nigeria,” Journal of Peace Research 45, no. 4 (July 2008): 539–555; Aldo A. Benini and Lawrence H. Moulton, “Civilian Victims in an Asymmetrical Conflict: Operation Enduring Freedom, Afghanistan,” Journal of Peace Research 41, no. 4 (July 2004): 403–422; Tom Bundervoet, “Livestock, Land and Political Power: The 1993 Killings in Burundi,” Journal of Peace Research 46, no. 3 (May 2009): 357–376; Ana Maria Ibanez and Andrea Velasquez, “Identifying Victims of Civil Conflicts: An Evaluation of Forced Displaced Households in Colombia,” Journal of Peace Research 46, no. 3 (May 2009): 431–451; Machiael Mousseau, “Urban Poverty and Support for Islamist Terror: Survey Results of Muslims in Fourteen Countries,” Journal of Peace Research 48, no. 1 (January 2011): 35–47; Macartan Humphreys and Jeremy M. Weinstein, “Handling and Manhandling Civilians in Civil War,” American Political Science Review 100, no. 3 (August 2006): 429–447; Hazem Adam Ghobarah, Paul Huth, and Bruce Russett, “Civil Wars Kill and Maim People-Long After the Shooting Stops,” American Political Science Review 97, no. 2 (May 2003): 189–202. Ibanez and Velasquez (see note 6 above). Benini and Moulton (see note 6 above), 417–421. Bundervoet (see note 6 above), 373–375. Verwimp, “Testing the Double-Genocide Thesis” (see note 6 above), 439–442, Tables 9 and 10. Kalyvas, The Logic of Violence in Civil War (see note 3 above), 173–328; Kalyvas and Kocher (see note 3 above), 343–352. Bundervoet (see note 6 above), 367–368, Table IV. Ibid., 369–370. Benini and Moulton (see note 6 above), 416. Among the cases of the PoW project, the Colombian case has not been included, because its questionnaires differ significantly from those of other cases. Paul Collier, Anke Hoeffler, and Nicholas Sambanis, “The Collier-Hoeffler Model of Civil War Onset and the Case Study Project Research Design,” in Paul Collier and Nicholas Sambanis, eds., Understanding Civil War: Evidence and Analysis, Volume 1: Africa (Washington, DC: The World Bank, 2005), 2, 26, footnote 6. See Common Article 3 of the Geneva Conventions, available at http://www.icrc.org/ihl.nsf/FULL/365?OpenDocument (accessed April 17, 2011.). Evan R. Gottesman, Cambodia After the Khmer Rouge: Inside the Politics of Nation Building (New Haven: Yale University Press, 2003); Timothy D. Sisk, Democratization in South Africa: The Elusive Social Contract (Princeton: Princeton University Press, 1994). This definition is broader than that of other conventional studies. Kalyvas, The Logic of Violence in Civil War (see note 3 above), 19–20. Verwimp, “Testing the Double-Genocide Thesis for Central and Southern Rwanda” (see note 6 above), 428–442; Verwimp, “Machetes and Firearms” (see note 6 above), 12–20; Bundervoet (see note 6 above), 361–369. Ibid., 357, 367–368, Table IV; Ghobarah, Huth, and Russett (see note 6 above), 196, 198–199, Tables 1, 2, and 3; Wood, “Armed Groups and Sexual Violence” (see note 4 above), 134; Verwimp, “Testing the Double-Genocide Thesis for Central and Southern Rwanda” (see note 6 above), 439–440, Table 9; Verwimp, “Machetes and Firearms” (see note 6 above), 17–18, Table VII; Ibanez and Velasquez (see note 6 above), 446–449, Table III. Verwimp, “Testing the Double-Genocide Thesis for Central and Southern Rwanda” (see note 6 above), 439–440; Verwimp, “Machetes and Firearms” (see note 6 above), 16; Bundervoet (see note 6 above), 369, Table IV; Wood, “Variations in Sexual Violence During War” (see note 4 above), 307–309; Hegre, Ostby, and Raleigh (see note 3 above), 619. Craig Etcheson, After the Killing Fields: Lessons from the Cambodian Genocide (Texas: Texas Tech University Press, 2005); Cynthia McClintock, “Why Peasants Rebel: The Case of Peru's Sendero Luminoso,” World Politics 37, no. 1 (October 1984): 81. One study suggested the correlation of −0.14 between educational levels and incomes in the fourteen Muslim countries. Mousseau (see note 6 above), 46, Appendix. Kalyvas, “Wanton and Senseless?” (see note 3 above), 243–285; Kalyvas, The Logic of Violence in Civil War (see note 3 above); Wood, “Variations in Sexual Violence During War” (see note 4 above), 327, 332. Kalyvas, “Wanton and Senseless?” (see note 3 above), 263–265; Kalyvas, The Logic of Violence in Civil War (see note 3 above), Chapters 6, 7, and 11; Kalyvas and Kocher (see note 3 above), 343–352. Benini and Moulton (see note 6 above), 414–420. Bundervoet (see note 6 above), 357, 371–373. Verwimp, “Testing the Double-Genocide Thesis for Central and Southern Rwanda” (see note 3 above), 440–441. The study uses latent class cluster models. The analyses of this study are conducted by Latent GOLD 4.5. Latent GOLD 4.5 uses the EM and the Newton-Raphson algorithm to estimate the models. Although the EM algorithm is extremely stable, a well-known issue in latent class analysis is the occurrence of local solutions rather than global ones with some models estimated on some data. In order to avoid this problem, this article has followed the guidance of the software technical guide. See Jeroen K. Vermunt and Jay Magidson, “Technical Guide for Latent GOLD 4.0: Basic and Advanced” (Belmont, MA: Statistical Innovation Inc., 2005): 49–54, available at http://www.statisticalinnovations.com/products/LGtechnical.pdf (accessed July 4, 2010.) In general, the article takes the solutions with the best log likelihood values, which have been replicated in several runs. Although the examples are those of M-plus, for a good guidance of local solutions with real data, see Tony Jung and K.A. Wickrama, “An Introduction to Latent Class Growth Analysis and Growth Mixture Modeling,” Social and Personality Psychology Compass 2, no. 1 (January 2008): 305, 313–314. Allan McCutcheon, Latent Class Analysis (Newbury Park, CA: Sage, 1987), 6, 11; Allan McCutcheon and Colin Mills, “Categorical Data Analysis: Log-Linear and Latent Class Models,” in Elinear Scarbrough and Eric Tanenbaum, eds., Research Strategies in the Social Sciences: A Guide to New Approaches (New York: Oxford University Press, 1998), 71–94; Bengt Muthen, “Latent Variable Analysis: Growth Mixture Modeling and Related Techniques for Longitudinal Data,” in David Kaplan, ed., The Sage Handbook of Quantitative Methodology for the Social Sciences (Thousand Oaks: Sage, 2004), 345–370. One example of latent class analysis in the political science field includes the following article. Lisa Blaydes and Drew A. Linzer, “The Political Economy of Women's Support for Fundamentalist Islam,” World Politics 60, no. 1 (July 2008): 576–609. McCutcheon (see note 31 above), 17. Carl-Erik Sarndal, Bengt Swensson, and Jan Wretman, Model-Assisted Survey Sampling (New York: Springer-Verlag, 1991), 87–99. The Nigerian survey uses the population estimates of 1997 based on the census of 1991. The Philippines' survey uses the population estimates of 1998 based on the census of 1995. The Cambodian survey uses the general population survey of 1998, the Colombian one the adjusted census (no date is recorded, but it seems to be the census of 1995), the El Salvadoran one the adjusted population of 1995 based on the census of 1992, the Israeli one the adjusted population estimation of 1999 based on the census of 1995, and the Lebanese one the most recent census (no date), respectively. The South African survey also has the records of population statistics of 1998. “Technical Appendix, Quantitative Research,” the PoW Georgian data file, no page; Hans Anker, “Abkhazia: Revised Draft Sample Design (May 11, 1999),” PoW Georgian data file, no page. “(Draft) Methodology Somalia People on War,” no date, PoW Somali data file, p. 3. “Sampling Methodology for Federation of Bosnia and Herzegovina,” PoW Bosnian data file, no page. Public Opinion and Marketing Agency Medium-Index, Ltd., “Sampling Plan for the ICRC project in Republika Srpska, December 1998 (survey realized by “Medium”),” PoW Bosnian data file, no page, footnote 1. No record remains as to upon what basis the allocations to sub-units had been undertaken. “Sampling Afghanistan People on War,” Afg38, 16 June 1999, PoW Afghan data file. Weights are the inverse of the selection probability. Standardized weights are applied for the PoW project. Leslie Kish, Sampling Survey (New York: Wiley, 1965). “Weighting,” PoW Georgian data file, no page. “(Draft) Methodology Somalia People on War,” PoW Somali data file, p. 3. For the Afghan survey no record is deposited in the data file on what basis samples are distributed. For the Afghan case, the codes of 99, 97, and 96 are recoded as missing values, as each of them includes only 1 case. Yet, the code of 98 includes 275 cases. Also, education variable of the Afghan data does not include 0 years level of education. Considering the interrupted modernization under the civil war in Afghanistan, naturally these cases are recoded as 0 years of education. When non-politicized ethnic groups are included in the models, their coefficients have turned out to be insignificant. In the Filipino case, only religious variables were included. Afghanistan is divided into 39 administrative units. Yet, due to the considerations of parsimony, the analyses of regional effects have been based upon the four regional units. In the Israeli case, regional variable was divided into Israel proper and the occupied territories by cross-tabulation. Kalyvas, “Wanton and Senseless?” (see note 3 above), 263–270; Kalyvas, The Logic of Violence in Civil War (see note 3 above), 173–328; Kalyvas and Kocher (see note 3 above), 343–352; Verwimp, “Machetes and Firearms” (see note 3 above), 18; Benini and Moulton (see note 6 above), 414–420. Those of the eastern region are 0.42, while those of the southern one 0.40. The percentage of those who had taken sides in the northern region is higher than any other region. These results conform to the conventional wisdom which points out the intensive military and political struggles in northern Afghanistan. Benini and Moulton (see note 6 above), 417–421. Here, when calculating the weighted percentages of political cleavages, the finest administrative units available are used. Log likelihood BIC statistics of the Lebanese case is 8533.49 for three classes, while 8501.29 for five classes. Those of the Nigerian case are 5150.53 for three classes, while 5128.88 for four classes. Those of the Filipino case are 6001.44 for three classes, while 5988.22 for four classes. Those of the Somali case are 7892.80 for three classes, while 7824.13 for four classes. Other than global model fit statistics, such as BIC, the examinations of local fit statistics, including the local independence assumption, are recommended in several articles. For instance, large values of bivariate residuals suggest local dependence. Jay Magidson and Jeroem K. Vermunt suggest three possible methods to tackle this problem. One of them is to delete several indicators. (Jay Magidson and Jeroen K. Vermunt, “Latent Class Models,” in David Kaplan, ed., The Sage Handbook of Quantitative Methodology for the Social Sciences (New York: Sage, 2004), 183.) Yet, each indicator's question of the PoW project asks about a different aspect of war-time violence. Even if several of them are deleted, similar outcomes appear. The second suggestion is to add more latent classes (Ibid.). Yet, I do not take this method, because the three-latent class models are most interpretable and comparable for the research questions of this article. As Beth A. Reboussin, Edward H. Ip, and March Wolfson warn, “sometimes a statistical analysis yields models that are not immediately interpretable to experts in the field, even though the models that are selected by LCA may be optimal according to criteria that are based on goodness-of-fit statistics.” Beth A. Reboussin, Edward H. Ip, and Mark Wolfson, “Locally Dependent Latent Class Models with Covariates: An Application to Under-Age Drinking in the USA,” Journal of Royal Statistics Society, series A, Statistics in Society 171, no. 4 (October 2008): 878. This study considers relaxing the local independence assumption by including direct effects between indicators with large values of bivariate residuals for the purposes of robustness checks (Vermunt and Magidson (see note 30 above), 24–25; Magidson and Vermunt (see note 51 above), 183–184), yet does not place much emphasis upon these results, due to the consideration of interpretability and parsimony. The inclusions of direct effects yielded similar results in terms of the overall significance of covariates, yet slightly better global model fit. Although the coefficient of a regional variable of the Cambodian case has lost statistical significance at any level, the one of political cleavage variable in Cambodia which had been significant only at the 10% level becomes significant at the five percent level after including several direct effects. Also, notice that not all the published articles include direct effects stringently for all the pairs with above certain values of bivariate residuals. (L. A. Ghandour, S. S. Martins, and H. D. Chailcoat, “Understanding the Patterns and Distribution of Opioid Analgesic Dependence Symptoms Using a Latent Empirical Approach,” International Journal of Methods in Psychiatric Research 17, no. 2 (2008), 101; Naomi Breslau, Beth A. Reboussin, James C. Anthony, and Carla L. Storr, “The Structure of Posttraumatic Stress Disorder,” General Psychiatry 62, no. 12 (December 2005): 1345; Blaydes and Linzer (see note 31 above).). “Taking sides” simply means the respondent's support for either side of the armed or the political conflicts. The PoW surveys have never asked the respondents which side they had supported, perhaps because these questions had remained too risky in the post-conflict settings. Kalyvas, “Wanton and Senseless?” (see note 3 above), 259–265. The PoW project includes the ethnic categories of Serb, Croat, Bosniak or Muslim, mixed, and others. So, those who identified themselves as Bosniaks simply had Muslim identities rather than those of the supporters for a multi-ethnic Bosnian state. Carol McQueen, Humanitarian Intervention and Safety Zones: Iraq, Bosnia and Rwanda (London: Palgrave Macmillan, 2005), Chapter 3. Russell R. Ross, ed., Cambodia: A Country Study, 3rd ed. (Washington, DC: Library of 1987), Craig Etcheson, The and of Press, Elizabeth Jean Wood, and Civil War in El (Cambridge: Cambridge University Press, and the International in the and Central and the Journal of Social and Political no. Killings and a in (New York: (see note above), Maria H. Society, and the in to the International available at (accessed July (see note above), Here, also three-latent class models are for all the due to the consideration of are to the coefficients at at the 10% In the El Salvadoran case, the of the coefficient for the class of the to violence is not because its coefficient out to be The is for the of the coefficient for the class of the to violence in the Somali case. Although the samples of are the results are all adjusted by the effects of the variables into Stephen of in and eds., New New Building the (Cambridge: Cambridge University Press, to in (London: The of and European University of 1991), (see note above), in The Regional of the in ed., and in and the New of 2 (New York: from to in the 2005), available at (accessed July In Table 2 of this study shows that the of the coefficients of the to violence in these are which means that the respondents not to be to violence. In has one of the in South This is to the coefficients at at the 10% Tom study on the is an case, as the author had obtained statistics on the support for Bundervoet (see note 6 above), Table on is of

Criminology · Political economy · Political science · Political violence · Politics · Sociology · Law · Psychology · South Asian Studies and Conflicts