ABSTRACT Habitat loss and anthropogenic mortality are recognized as threats to populations of large carnivores worldwide, yet their relative importance to extinction risk has rarely been quantified. We used population viability analysis (PVA) to estimate extinction probability of an isolated population of black bears (Ursus americanus) on the Bruce Peninsula, Ontario, Canada under different management scenarios. We used random-effects analysis of variance to estimate components of variance in extinction risk explained by 4 management actions: 1) preventing habitat destruction, 2) reducing or eliminating incidental non-natural mortality, 3) reducing or eliminating harvest, and 4) reducing the fraction of reproducing females in the harvest. Habitat area reductions had the greatest effect on risk despite uncertainty in bear density. Incidental non-natural mortality had a greater effect than the rate or age and sex distribution of harvest. Quantifying the variation in outputs of PVA models associated with different management options is an improvement over qualitative comparisons of relative risk and enhances the applicability of PVA to management. This study highlights the importance of protecting habitats on adjacent private lands when reserves are too small to support populations of bears, and of protecting reproducing females from non-natural mortality—results that could aid managers of other large carnivores in focusing management efforts to ensure persistence of populations.