Applied ecologists routinely use demographic models to predict population trajectories. Survival rates throughout the life cycle, which are required for these models, are often difficult to obtain, especially for long-lived or mobile species. Detailed information for pre-adult age classes in particular is often lacking. Using a 20-year dataset from several hundred individuals, we used Markov chain Monte Carlo methods to fit hierarchical models that describe survival rates for both adult and sub-adult Hawaiian stilts Himantopus mexicanus knudseni, an endangered island endemic. We constructed the complete-data likelihood and used data augmentation to estimate missing values and incorporate data that were not collected during formal sampling. Survival estimates were lower and more uncertain during the first 2 months of life compared with the remainder of the first year. The probability of first-year survival averaged 0.55 (95% credibility interval: 0.07–0.90), but varied considerably among cohorts from different years and islands. Probability of adult annual survival differed little between females (0.79; 0.71–0.86) and males (0.80; 0.72–0.87), but increased as birds aged from 1 to 20 years (0.77–0.85). Our analysis confirms that earlier work, despite being based on few data, provided good point estimates for survival rates. Our new analysis, however, provides the first comprehensive assessment of uncertainty in survival rates and detailed information on the nature of variation in first-year and adult survival. This information will help inform new demographic models and can be used to guide management actions.