Because ungulates can be dangerous prey, predators generally kill vulnerable ones (e.g. juveniles, individuals in poor body condition). Adult female ungulates usually have a high constant survival rate until they become frail at older ages. We tested the hypothesis that relative to other mortality, selective wolf predation on old female elk is the principle driver of actuarial senescence in northern Yellowstone National Park, USA. We monitored radio-collared elk (n = 196) from 2000-2008 and 2011-2016 and determined the cause of mortality for 77 individuals. We used a multi-state survival model in a competing risks framework to estimate age-specific wolf-caused mortality compared to all other causes of mortality. Elk mortality was low (< 5%) for 2-13 year-olds. After age thirteen, the probability of wolf-caused mortality increased sharply. The probability of wolf predation was 0.10 (95% CI =0.08, 0.14) for 15-year-olds and 0.42 (95% CI = 0.32, 0.54) for 20-year-olds. In contrast, the probability of non-wolf caused mortality was 0.08 (95% CI =0.06, 0.11) for 15-year-olds, and 0.15 (95% CI = 0.11, 0.26) for 20-year-olds. Older elk were more likely to be killed by wolves during severe winters (20-year-old mortality = 0.68, 95% CI = 0.41, 0.85) relative to mild winters (0.32, 95% CI = 0.16, 0.60). These results suggest that prey senescence can be shaped by selective predation and that the population-level impact of predation depends on prey age structure. Additive predation effects are more likely in a population with an older age structure, and among old individuals.