Reliable estimates of survival and dispersal are crucial to understanding population dynamics, but for seabirds, in which some individuals spend years away from land, mortality and emigration are often confounded. Multistate mark–recapture methods reduce bias by incorporating movement into the process of estimating survival. We used a multistate model to provide unbiased age-specific survival and movement probabilities for the Endangered African Penguin Spheniscus demersus based on 5281 nestlings and 31 049 adults flipper-banded and resighted in the Western Cape, South Africa, between 1994 and 2012. Adult survival was initially high (≥ 0.74) but declined after 2003–2004 coincident with a reduction in the availability of Sardine Sardinops sagax and Anchovy Engraulis encrasicolus on the west coast of South Africa. Juvenile survival was poorly estimated, but was lower and more variable than adult survival. Fidelity to the locality of origin varied over time, but was high in adults at Robben and Dassen islands (≥ 0.88) and above 0.55 for juvenile and immature Penguins at all localities. Movement occurred predominantly during 1994–2003 and was indicative of immigration to Robben and Dassen islands. Our results confirm that a prolonged period of adult mortality contributed to the observed decline in the African Penguin population and suggest a need for approaches operating over large spatial scales to ensure food security for marine top predators.