Portal de Periódicos da CAPES

The Localized Geographies of Violence in the North Caucasus of Russia, 1999–2007

2010; American Association of Geographers; Volume: 101; Issue: 1 Linguagem: Inglês

10.1080/00045608.2010.534713

1467-8306

John Ο’Loughlin, Frank D. W. Witmer,

Russia and Soviet political economy

Abstract The second Chechen war, starting in the North Caucasus in August 1999, shows few signs of a ceasefire after eleven years, although the level of violence has declined from the peaks of the war's first two years. Initially framed by both sides as a war of separatists versus the federal center, the situation is now complicated by the installation of a Moscow ally into power in Chechnya and by the splintering of the opposition into groups with diverse aims and theaters of operation. The main rebel movement has declared the establishment of an Islamic caliphate in all the Muslim republics of the North Caucasus as its ultimate goal. Fears of regional destabilization of the entire North Caucasus of Russia are propelled by reports of increased militant activism in republics adjoining Chechnya due to possible contagion effects of violence in these poor areas. Temporal and spatial descriptive statistics of a large database of 14,177 violent events, geocoded by precise location, from August 1999 to August 2007, provide evidence of the conflict's diffusion into the republics bordering Chechnya. "Hot spots" of violence are identified using Kulldorff's SaTScan statistics. A geographically weighted regression predictive model of violence indicates that locations in Chechnya and forested areas have more violence, whereas areas with high Russian populations and communities geographically removed from the main federal highway through the region see less violence. La segunda guerra de Chechenia, que empezó en el Cáucaso del Norte en agosto de 1999, muestra pocas señas de un cese al fuego tras once años de lucha, si bien el nivel de violencia ha declinado desde el pico alcanzado en los primeros dos años de la guerra. Lo que inicialmente se encuadró por ambos bandos como una guerra de separatistas contra un centro federal, se ha complicado ahora con la imposición de un aliado de Moscú en el poder de Chechenia y con la fragmentación de la oposición en grupos con propósitos y teatros de operación diversos. El principal movimiento rebelde declaró como su meta única el establecimiento de un califato islámico para todas las repúblicas musulmanas del Cáucaso del Norte. Los temores de desestabilización regional en todo el Cáucaso del Norte ruso están siendo impulsados por los informes sobre incrementos del activismo militante en repúblicas vecinas a Chechenia, debido al posible contagio de los efectos de la violencia en estas áreas empobrecidas. Las estadísticas descriptivas temporales y espaciales de una gran base de datos que cubre 14.177 hechos violentos, geocodificados con localización precisa, de agosto de 1999 a agosto de 2007, dan evidencia de la difusión del conflicto hacia las repúblicas fronterizas con Chechenia. Los "puntos calientes" de violencia se identificaron mediante el uso de estadísticas SaTScan de Kulldorff. Un modelo predictivo de la violencia mediante regresión ponderada geográficamente indica que ciertas localidades de Chechenia y las áreas de bosques tienen más violencia, mientras que las áreas con alto contenido de población rusa y las comunidades que han sido removidas geográficamente, desde la principal carretera federal a través de la región, muestran menos violencia. Key Words: Chechnyacivil wardiffusionNorth Caucasusspatial-statistical analysis关键词: 车臣内战扩散北高加索空间统计分析Palabras clave: Checheniaguerra civildifusiónCáucaso del Norteanálisis espacio-estadístico Acknowledgments The authors thank the National Science Foundation's Human and Social Dynamics program (Grant No. 0433927) for the financial support that made possible both the cooperative field work in the North Caucasus in 2005 through 2007 and the events data collection. Gearóid Ó Tuathail and Vladimir Kolossov again proved to be inestimable colleagues in the larger project on war outcomes in Bosnia and the North Caucasus and continue to offer invigorating comradeship during the field excursions. Sitora Rashidova, Yana Raycheva, Mary Robinson, and Evan O'Loughlin painstakingly coded and geo-located the events data over many long months. Nancy Thorwardson of Computing and Research Services at the Institute of Behavioral Science, University of Colorado, prepared the figures for publication with her exemplary skill, punctuality, humor, and élan and cast her editorial cold eye over earlier versions of the article. The article also benefited from the comments and questions of colleagues at the 2007 Conference of Irish Geographers in Dublin, Dartmouth College's Geography Department, Kennan Institute at the Woodrow Wilson International Center for Scholars, and the April 2008 Boston meeting of the Association of American Geographers. Thanks also to Annals editor Audrey Kobayashi for her editorial work. Any remaining errors are ours. Notes 1. To assess whether the observations within each cylinder form a cluster, they are compared against the expected number of events with the given area and time period. Using the CitationKulldorff et al. (2005) method, for each spatial location, s, and each time unit, t, an expected number of violent events is calculated: where c st is the observed number of events cases at the given spatial location during t and C is the total number of observed violent events. From these individual calculations, the expected number of events within a given cylinder, A, can be calculated by the following summation: For each cylinder, the observed to expected ratio, ODE = c A /E[cA ], is calculated, and for each cylinder with an ODE > 1, the Poisson generalized likelihood ratio measure is where c A is the observed number of events in the cylinder. To evaluate the statistical significance of the cluster, a Monte Carlo method is used to generate 999 random permutations of the data for which the likelihood ratio statistic is calculated. A p value is then calculated by comparing the rank of the test statistic generated from the real data, R, against the 999 simulated values, p = R/(999 + 1).