We applied temporal symmetry capture–recapture (TSCR) models to assess the strength of evidence for factors potentially responsible for population decline in bottlenose dolphins (Tursiops truncatus) in Doubtful Sound, New Zealand from 1995 to 2008. Model selection was conducted to estimate recruitment and population growth rates. There were similar levels of support for three different models, each reflecting distinct trends in recruitment. Modeling yielded low overall estimates of recruitment (0.0249, 95% CI: 0.0174–0.0324) and population growth rate (0.9642, 95% CI: 0.9546–0.9737). The TSCR rate of population decline was consistent with an estimate derived from trends in abundance (lambda = 0.9632, 95% CI: 0.9599–0.9665). The TSCR model selection confirmed the influence of a decline in the survival of calves (<1 yr old) since 2002 for population trends. However, TSCR population growth rates did not exceed 1 in any year between 1995 and 2008, indicating the population was declining prior to 2002. A separate reduction in juvenile survival (1–3 yr old) prior to 2002 was identified as a likely contributing factor in the population decline. Thus, TSCR modeling indicated the potential cause of the population decline in Doubtful Sound: cumulative impacts on individuals <3 yr old resulting in a reduced recruitment.