SUMMARYLocal government extractive capacity, as measured by the amount of graduated personal tax (GPT) collected relative to district wealth and population and, more qualitatively, as reflected in the nature of enforcement, varies considerably in Uganda. This article explores the reasons for this variation, first by investigating aggregate data at the cross-district level, using data on taxation as well as survey data from the Afrobarometer, second, by a focussed comparison of two districts, one with high-and one with lowextractive capacity. I find that generalised trust can explains some of the variation in extractive capacity across districts. The case studies trace the differences in trust and extractive capacity back to pre-colonial rule; the better performing district having had centralised rule, a tradition of tax-payment and a higher degree of social cohesion, while the poorer performing district had a more fragmented and less cohesive history of governance. The article thus points to the importance of understanding institutional path dependencies when assessing the feasibility of reform.