State capacity has attracted renewed interest over the last years, notably in the study of violent conflict. Yet, this concept is conceived differently depending on where the interest lies. In this article, we focus on bureaucratic autonomy as a distinct concept and discuss its connection to state capacity in detail. Using panel data over 1990–2010 and a novel indicator of autonomy, we estimate the separate effect of state capacity and bureaucratic autonomy on child mortality and tuberculosis prevalence. The evidence suggests that bureaucratic autonomy has a stronger impact than commonly used measures of state capacity or traditional macroeconomic variables.