Researchers have been struggling to capture the learning outcomes of role-play simulations of political decision-making, which shows in inconsistencies in findings. In this systematic review study we argue that research designs should not ignore the contextual features of these simulations. This review aims: (1) to comprehensively map variation in learning environment components, and (2) to increase insights into their relationship with learning outcomes. A systematic search in SSCI and ERIC databases yielded 36 studies that were eligible. The following learning environment components were comprehensively mapped: simulation features (structure and agency), student characteristics, the broader learning context, and learning outcomes. Findings reveal that more than half of the studies investigate learning outcomes without taking any other learning environment component into account. Learning outcomes have never been studied in relation to the simulation structure or the broader learning context. Findings are discussed with regard to avenues for future research.