ABSTRACT We analyzed 53 years of banding and band recovery data along with estimates of harvest and population size to assess the role of harvest and density dependence in survival patterns and population dynamics of black brant (Branta bernicla nigricans) over the period 1950–2003. The black brant population has declined steadily since complete annual surveys began in 1960, so the role of harvest in the dynamics of this population is of considerable interest. We used Brownie models implemented in Program MARK to analyze banding data. In some models, we incorporated estimated sport harvest to test hypotheses about the role of harvest in survival. We also examined the hypothesis of density‐dependent regulation of mortality by incorporating estimates of population size as a covariate into models of survival. For a shorter period (1985–2003), we also assessed hypotheses about the role of subsistence harvest and predation as sources of mortality. The best supported model of variation in survival and band recovery allowed survival rates to vary among 2 age classes (juv, second‐yr plus ad brant) and the 2 sexes. We constrained survival probabilities to be constant within decades but allowed them to vary among decades. We also constrained band recovery rates to be constant within decades and to vary in parallel among age and sex classes. We were limited to decade‐specific estimates of survival and band recovery rates because some years before 1984 lacked any banding, and banding in some other years was sparse. A competitive model constrained survival estimates to be the same for males and females. No model containing harvest or population size was competitive with models lacking these covariates (relative quasi‐Akaike's Information Criterion adjusted for small sample size [βQAICc] > 13). In the best supported model, band recovery rates declined from 0.038 ± 0.0028 (F) and 0.040 ± 0.0031 (M) to 0.007 ± 0.0007 (F) and 0.007 ± 0.0007 (M) between the 1950s and 2000s, a clear indication that harvest rates declined over this period. Survival rates increased from 0.70 ± 0.02 and 0.71 ± 0.02 for adult males and females, respectively, in the 1950s to 0.88 ± 0.009 and 0.88 ± 0.01 for males and females, respectively, in the 1990s. Survival rates in the 1990s were among the highest estimated for brant and did not increase in the 2000s with additional reductions in sport harvest. For the shorter data set from 1985 to 2003, models containing covariates for either sport or subsistence harvest were less competitive than models lacking these terms (βQAICc > 3). For the best model containing subsistence harvest, the estimate of β linking subsistence harvest to survival, although imprecisely estimated, was near zero (β = −0.04 ± 0.30), consistent with the hypothesis that subsistence harvest had little impact on survival during this period. We conclude that while harvest likely influenced survival and population dynamics in earlier decades, it is most likely that continued population decline at least since 1990 is a result of low recruitment.