The recovery of carnivore populations in North American has consequences for trophic interactions and population dynamics of prey. In addition to direct effects on prey populations through killing, predators can influence prey behavior by imposing the risk of predation. The mechanisms through which patterns of space use by predators are linked to behavioral response by prey and nonconsumptive effects on prey population dynamics are poorly understood. Our goal was to characterize population‐ and individual‐level patterns of resource selection by elk (Cervus canadensis) in response to risk of wolves (Canis lupus) and mountain lions (Puma concolor) and evaluate potential nonconsumptive effects of these behavioral patterns. We tested the hypothesis that individual elk risk‐avoidance behavior during summer would result in exposure to lower‐quality forage and reduced body fat and pregnancy rates. First, we evaluated individuals' second‐order and third‐order resource selection with a used‐available sampling design. At the population level, we found evidence for a positive relationship between second‐ and third‐order selection and forage, and an interaction between forage quality and mountain lion risk such that the relative probability of use at low mountain lion risk increased with forage quality but decreased at high risk at both orders of selection. We found no evidence of a population‐level trade‐off between forage quality and wolf risk. However, we found substantial among‐individual heterogeneity in resource selection patterns such that population‐level patterns were potentially misleading. We found no evidence that the diversity of individual resource selection patterns varied predictably with available resources, or that patterns of individual risk‐related resource selection translated into biologically meaningful changes in body fat or pregnancy rates. Our work highlights the importance of evaluating individual responses to predation risk and predator hunting technique when assessing responses to predators and suggests nonconsumptive effects are not operating at a population scale in this system.