Seeing the Forest for the Trees: Does Physical Geography Affect a State's Conflict Risk?

Abstract

This article considers why and how physical geography may influence conflict patterns within African states. It juxtaposes arguments relating to four features of the physical environment— distance, resources, terrain, and size—to those purporting rebels base insurgency tactics on the strategic value of locations. Using GIS and spatial econometrics, a geographically disaggregated dataset of population, distance to capitals, borders, resources, terrain, and road densities is tested against conflict data from ACLED (Armed Conflict Location and Events Dataset). ACLED disaggregates data on internal conflict into georeferenced individual events. The analysis covers six of the most conflict prone states in Central Africa. The analysis confirms that an area's physical attributes do not have a uniform effect on the likelihood of experiencing a conflict event. Areas of high strategic value, including densely populated areas and military zones, have a higher risk of conflict than rural, peripheral areas.

Keywords:

• Africa
• civil war
• military and political geography
• physical geography

This research was funded by the Irish Research Council for Humanities and Social Science.

Notes

¹For more information on ACLED or to download the publically available beta version, please go to www.acleddata.com

²No conflict in this study is deemed “separatist.”

³Data are available in either raster, shapefile or point form. Geographic data, including road network files, can be accessed on the UNEP Geodata Portal (UNEP 2004, 2006). All the data values for this project originated as shapefiles/vector or raster files. These two common spatial data file types relay local information about locations. Shapefiles store information about polygons (for example, county attributes), lines (for example, roads), or points (for example, conflict events). Raster files store information as pixels or images (for example, slope). Using GIS techniques, spatial data can be added and manipulated to a grid set structure.

⁴None of the states in this sample has a federalist system of government. In such systems, the heightened power of regional governments in comparison to central government underlies a different type of political geography.

⁵Terrain data are available from http://geodata.grid.unep.ch The image used in this analysis (DEM 5 Minutes- Africa_etopo5.tif) was developed by the National Geophysical Data Centre in 1990. The Continuous Fields Tree Cover-Africa_tree_umd.tif) details the percent of tree cover across space in 1 km increments.

⁶Diamond data is available from PRIO: http://www.prio.no/CSCW/Datasets/Geographical-and-Resource/Diamond-Resources/.

⁷Smaller, unincorporated mines are not included in MRSD.

⁸Information on Central African airports is available from www.pilotsfriend.com.

⁹Information on population is available from http://www.glowa-volta.de/cd_v3.1/landuse/populat/part2.htm.

¹⁰“Distance to capital” is highly correlated to “distance to major city.” This variable is replaced with a “capital” control dummy for full models.