The rate of population change through time (λ) is a fundamental element of a wildlife population's conservation status, yet estimating it with acceptable precision for bears is difficult. For studies that follow known (usually marked) bears, λ can be estimated during some defined time by applying either life-table or matrix projection methods to estimates of individual vital rates. Usually however, confidence intervals surrounding the estimate are broader than one would like. Using an estimator suggested by Doak et al. (2005), we explored the precision to be expected in λ from demographic analyses of typical grizzly (Ursus arctos) and American black (U. americanus) bear data sets. We also evaluated some trade-offs among vital rates in sampling strategies. Confidence intervals around λ were more sensitive to adding to the duration of a short (e.g., 3 yrs) than a long (e.g., 10 yrs) study, and more sensitive to adding additional bears to studies with small (e.g., 10 adult females/yr) than large (e.g., 30 adult females/yr) sample sizes. Confidence intervals of λ projected using process-only variance of vital rates were only slightly smaller than those projected using total variances of vital rates. Under sampling constraints typical of most bear studies, it may be more efficient to invest additional resources into monitoring recruitment and juvenile survival rates of females already a part of the study, than to simply increase the sample size of study females.