Population modeling exercises can lead to both expected and unexpected results useful for wildlife research and management, even though inferences must often be qualitative, given underlying assumptions. Our main objective was to use empirical data on wolf (Canis lupus) kill rates and growth of the Western Arctic caribou (Rangifer tarandus) herd (WAH) of Alaska, USA, to assess the potential for predator regulation. We used available data and published literature to construct a deterministic density-dependent population model fitted to trends of the WAH from 1976 through 2003. By increasing wolf densities in the baseline model, we failed to reject the hypothesis that wolves at a density of 6.5 wolves per 1,000 km² could regulate a caribou herd at a density of 0.4 caribou per km². In addition, our model may be conservative by underestimating the regulatory potential of wolves. We suggest that this relatively simple predator–prey system shows signs of a predation–food 2-state model. Elasticities from matrix models may be deceiving. Although herd growth is most sensitive to changes in adult female survival, survival of younger cohorts may be more easily influenced by natural conditions or management action. Management of the WAH near maximum sustained yield may not be attainable if desired, but modeling exercises such as this elucidate options. In conducting this research, we also discovered by Monte Carlo simulation that survival and productivity data from radiocollared females and calves were negatively biased and failed to predict herd growth. Thus, researchers should consider potential effects of neck collars on vital rates of female tundra caribou and concomitant offspring when using sample data to model population dynamics or test hypotheses.