This paper investigates the role of state capacity for political violence. Most previous studies have suffered from inadequacies of country-level data, questionable validity of indicators or theoretical shortcomings. This paper aims at overcoming some of these challenges. We focus on one specific aspect of state capacity: the role of governmental manpower. We argue that its subnational effect on political violence follows a non-linear, inverted-U shape. We investigate this hypothesis in the context of southern Sudan, covering the period from 2006 to 2010. We use unique data on the geographical distribution of public personnel across 75 southern Sudanese counties. The data are matched with geocoded data on violent events as well as various socio-economic indicators. Our fixed-effects estimations indicate that particularly low or high levels of state capacity are associated with low levels of violence. Counties with intermediate numbers of state personnel experience the highest numbers of violent events.