Managed forest ecosystems-agroforestry systems in which crops such as coffee and bananas are planted side-by-side with woody perennials-are being touted as a means of safeguarding forests along with the ecological services they provide. Yet we know little about the determinants of land cover in such systems, information needed to design effective forest conservation policies. This paper presents a first-ever spatial regression analysis of land cover in a managed forest ecosystem-a shade coffee region of coastal Mexico. Using high-resolution land cover data derived from aerial photographs, along with data on the institutional, geophysical, socioeconomic, and agronomic characteristics of the study area, we find that plots in close proximity to urban centers are less likely to be cleared, all other things equal. This finding contrasts sharply with the literature on natural forests. In addition, we find that membership in coffee marketing cooperatives, farm size, and certain soil types are associated with forest cover, while common property, proximity to small town centers, and the prevalence of indigenous peoples are associated with forest clearing.