Most Old World monkeys show male-biased dispersal. We present the first systematic data on male dispersal in a provisioned multilevel group of Rhinopithecus roxellana, based on 4.5 years of field observations in Shennongjia National Nature Reserve, China. We evaluated both ultimate (inbreeding avoidance and male mating competition) and proximate (food availability and predation risk) factors influencing male dispersal. The focal group contained 34-53 individuals, in 3-4 one-male units (OMUs) and 1 all-male unit (AMU). We observed 37 dispersal events involving 10 of 11 adults, 7 of 8 subadults, and 7 of 15 juveniles. Most interunit transfers within the focal group occurred around the months of mating season. Adult males competed for the leader positions of OMUs mainly through aggressive takeovers, and young males transferred from the OMUs to the AMU at the median age of 41 months, forced out by leader males. No young males older than 4 years remained in natal or non-natal OMUs. The male mating competition hypothesis was supported. The young males emigrated voluntarily from the focal group at the average age of 58.6 months, and no young emigrating male was observed to return, suggesting inbreeding avoidance also played a role in the dispersal of young males. Most emigration/immigration events were parallel dispersal and occurred during intergroup encounters, suggesting increased predation risk during the dispersal period. Males were more likely to emigrate/immigrate during the months when preferred foods were most available. We compared the dispersal patterns in R. roxellana with those in gelada baboons and hamadryas baboons, both living in multilevel societies. Similar to R. roxellana, young male geladas disperse at puberty, but they may return and breed in their natal groups. Males in hamadryas also disperse, but much less commonly than in R. roxellana. Provisioning may have influenced results, and confirming studies on unprovisioned groups would be valuable.