GIS in Locating and Explaining Conflict Hotspots in Nepal Lila Kumar Khatiwada Notre Dame Initiative for Global Development 1
Outline Brief background Use of GIS in conflict study Data source Findings and recommendations 2
Background Nepal conflict (1996-2006) claimed over 15,000 lives and displaced thousands Originated in western hill in 1996, by 2004 most of Nepal districts were affected Conflict experts: Spatial-horizontal inequality in the districts best explains origin and expansion of conflict 3
Patterns of conflict High casualties in some districts: They were clustered together Low casualties in some districts: They were also clustered together Presence of spatial dependence 4
Data sources UNDP Nepal Human Development Report Central Bureau of Statistics, Nepal INSEC- human rights violation data UNSTAT- country boundary file 5
Use of GIS Visualizing, seeing the patterns Combining data: combined data derived from three different sources Mapping the casualties Testing autocorrelation, Morans I Building models, spatial error, lag models 6
Software use ArcMap Geoda Excel 7
Population killed in the districts 8
Why spatial dependence in conflict? Conflict is contagious: conflict of one place depends on its neighbor (autocorrelation) Geographically proximate units are similar to each other; conflict might spill over from one location to the next Population flows help in diffusion by facilitating the spread of arms, combatants, and ideologies conducive to conflict 9
Spatial dependence Underlying socio-economic process has led to clustered distribution of variable values Grouping processes grouping of similar people in localized areas Spatial interaction processes people near each other more likely to interact, share Diffusion processes Neighbors learn from each other Dispersal processes People move, but tend to be short distances, take their knowledge with them Spatial hierarchies Economic influences that bind people together 10
Moran s I A measure of autocorrelation similar in interpretation to the Pearson's correlation for independent samples, in that both statistics range between -1.0 and 1.0 Local Moran s I decomposes a global measure into each unit s contribution 11
Scatterplot 12
Moran s I results of high-high and low-low districts and their comparison Development indicators Low-low districts (averages) National Average High-high districts (averages) Life expectancy at birth, yrs. 64.89 61 47.34 Adult literacy rate 52.52 48 28.25 Percent population without safe drinking water 19.57 20.48 41.86 Female life expectancy, yrs 65.65 61.5 47.94 Female literacy rate 38.94 34.9 12.92 Percent women in administrative jobs 11.92 12.71 8.22 Infant mortality rate 50.96 68.51 104.49 Percent electrified households 41.52 31 6.37 Land inequality 0.52 0.54 0.38 Percent population with access to credit 23.59 19 3.24 Percent population in nonagriculture 37.25 31.33 18.88 13
Local Moran s I 14
Spatial lag model Spatial lag model deals with questions of how the interaction between socioeconomic agents can lead to emergent collective behavior and aggregate patterns It is used to estimate the effect of various socioeconomic factors present in the districts on conflict It includes the mean of the dependent variable in neighboring areas (i.e., spatial lag) as an extra explanatory variable 15
Dependent variable District level fatalities: both govt. army, rebel side and common people Death rate is standardized based on district population 16
Regression results Variables Coefficient Std. Error Constant 0.2030* 0.083 Per capita GDP in PPP$ -0.0000 0.000 Adult literacy rate -0.0018* 0.000 Women s share of income -0.3660** 0.121 Average landholding size -0.2129** 0.073 Electrified households -0.0015* 0.000 Infant mortality rate 0.0003 0.000 Percent in non-agriculture -0.0006 0.000 Percent household with radio 0.0031** 0.000 R 2 0.58 Log Likelihood 112.59 Rho 0.398*** 17
Findings Conflict is not independent of location and is not random but is a function of regional spatial effects A spatial lag model shows that districts with higher death rate had higher illiteracy, lower landholding size, poor infrastructure base, and unequal share of women s income 18
Findings It calls for addressing inequality issues in the districts Program and policy targeted for post reconstruction should be focused at the regional level the effect of any intervention may go beyond a district s boundary 19
Methodological contribution GIS can be used to locate the high conflict places Modeling with available spatial approach can answer many questions that are not possible from other methods 20
Thanks! 21