This paper contributes to explaining variation in violence in post-independence East Africa. By focussing on Tanzania and Uganda, our comparative 'most-similar-cases' research design ensures significant across- and within-case variation in violence against relatively similar background conditions. As a conceptual starting point the World Bank's Collier/Hoeffler model and theory are applied to both countries. It is argued that while the model's fit is relatively good, Collier/Hoeffler's main theoretical proposition that civil war onset is best predicted by the existence of opportunities (or financing availability through resource extortion) for rebellion does not correspond as well with the East African context. We propose a modified rational-choice framework focusing on micro-level economic motivations for state capture, which is argued to help explain cycles of violence in Uganda until the 1980s, as well as the 'paradox of peace', i.e. the puzzle why East Africa has experienced sustained periods of absolute (Tanzania) and relative (Uganda nationally after 1986, Zanzibar since the 1970s) peace despite identity fragmentation. From this perspective, peace followed from successful elite strategies to co-opt opponents through a mixture of rent-sharing and non-violent authority, leading to identity-encompassing, hierarchical one-party structures able to overcome collective action problems of state capture and defence.