A semi-parametric spatial regression approach to post-war human security: Cambodia, 2002-2004

Human security in post-war societies depends on incentives to forego violence in local interactions. The government of Cambodia monitors domestic violence, land conflicts and serious crime in over 13,000 villages and urban neighbourhoods. We use three annual data collections to estimate the response of these conflicts to the legacy of the war, poverty and resource competition, urbanity as well as governance quality. Bayesian spatial regressions help identify socio-economic thresholds beyond which conflicts expand or contract significantly. We find numerous non-linearities in the propensities for violence. Notably, predicted rates decrease in response to quality of governance only at a high level of service provision, which, realistically, most communities may not soon achieve. This may justify dedicated programmes addressing particular types of conflicts. We propose alternative analytic approaches, including some that would make the problem of endogeneity more tractable once updated poverty estimates become available.