This study examines determinants of leftist violence at the municipal level in Colombia from 2000 through 2010. A multilevel GLMM model with a negative binomial distribution is used to take advantage of the information available at the municipal and department level. Surprisingly, inequality was not a significant covariate of violence, and agricultural GDP tended to reduce, instead of increase, guerrilla violence. The main risk factors identified include physical characteristics such as rugged topography and prior violence, but also factors that are candidates for policy action, such as unemployment, incorporation of the poor into public services, repression, and the energy and mining sector. These findings suggest interventions to decrease risks of guerrilla violence beyond merely strengthening the state. While repression tends to escalate violence, targeted policies to provide health benefits to those currently underserved, and securing mining and oil operations can effectively reduce the risk of violence.

Abstract. Generalized linear models are often used to identify covariates of landscape processes and to model land-use change. Generalized linear models however, overlook the spatial component of land-use data, and its effects on statistical inference. Spatial autocorrelation may artificially reduce variance in observations, and inflate the effect size of covariates. To uncover the consequences of overlooking this spatial component, we tested both spatially explicit and non-spatial models of deforestation for Colombia. Parameter estimates, analyses of residual spatial autocorrelation, and Bayesian posterior predictive checks were used to compare model performance. Significant residual correlation showed that non-spatial models failed to adequately explain the spatial structure of the data. Posterior predictive checks revealed that spatially explicit models had strong predictive power for the entire range of the response variable and only failed to predict outliers, in contrast with non-spatial models, which lacked predictive power for all response values. The predictive power of non-spatial models was especially low in regions away from Colombia's center, where about half the observations were clustered. While all analyses consistently identified a core of important covariates of deforestation rates, predictive modeling requires parameter estimates informed by the spatial structure of the data. To inform increasingly important forest and carbon sequestration policy, land-use models must account for spatial autocorrelation.