Most variation in relative murder risk comes from space-time regional clustering. Population density effect on murder risk is insignificant. Time trend contributes minimally to murder risk variation. Northern regions show consistently low murder risk. Karas and Khomas regions face long-term high murder risk increases. Relative risk map reveals clear north-south increasing gradient. Space-time structured heterogeneity has highest standard deviation at 0.683. Unstructured heterogeneity standard deviation is 0.089. Model fit improves significantly only with structured heterogeneity addition. Karas and Khomas classified as high-risk consistently over 2002-2006. Omaheke and Hardap high-risk in 2003, 2004, 2006. Northern regions account for 62% population but low murder risk. Urban migration to Khomas and Karas boosts crime without jobs. Model robust to prior and starting value changes. Intrinsic and Laplace CAR priors both viable for heterogeneity. Namibia ranks high globally in murder rate at 0.168 per 1000 in 2006. Regional clustering explains over 55% of space-time variation.