This project introduces a new framework for thinking about ethnic cleansing and tests this theory on the Lebanese civil war of 1975-1990 using an original quantitative dataset and about 60 interviews gathered during 14 months of fieldwork. Ethnic cleansing, or instances where one ethnic or sectarian group displaces members of another from a particular territory through violent means, can be either a political goal or a strategy in service of another end and this crucial aspect means that the practice of this strategy will vary widely across conflicts. Ethnic cleansing is often used as a strategy for dominating contested territory in nationalist conflicts which implies that some group is trying to maximize displacement using all available military forces. Yet my argument is that in civil wars fought at least in part across an ethnic divide, but where neither side has separatist ambitions, armed groups have strong incentives to minimize the amount of displacement. Ethnic identity is an easy and cheap heuristic device that can be used to identify potential enemies which enables armed groups to use collective violence such as ethnic cleansing. But with a lack of separatist ambition comes an understanding that the country will need to be governed after armed hostilities end. In addition, during the conflict process, armed groups have an incentive to retain non-coethnic resident civilians since all manners of in-conflict funding of armed groups are strictly increasing in the size of civilian population: taxation, predation, seizing and operating firms, and donations. For these reasons armed groups have an incentive to acquire more fine-grained information about resident non-coethnics to allow them to displace only those who are actively disloyal. This is easiest in areas where loyal and disloyal civilians are mixed, such as mixed residential neighbourhoods, which are therefore more likely to witness selective violence but less likely to experience full collective ethnic cleansing. To test this theory I rely on an original countrywide village-level data set on demographics, migration, and violence collected during 14 months of fieldwork based in Beirut and supplemented by about 60 interviews with political leaders and former militia fighters. The data stems from voter registration materials, government documents, NGO reports and other sources and allows for a rich set of quantitative tests and illustrative GIS mapping. The main finding is that ethnic cleansing was less extensive and more narrowly and competently targeted than is commonly imagined or appreciated.