Declines of imperiled small mammals are often attributed to predation without investigating the relative influence of survival and reproductive parameters on population growth. Accordingly, declines in the endangered Key Largo woodrat Neotoma floridana smalli (KLWR) population have been attributed to predation by feral cats Felis catus, Burmese pythons Python molurus bivittatus and raccoons Procyon lotor. We estimated survival, recruitment and realized population growth rates from four capture–mark–recapture studies to determine if the pattern of demographic variation was consistent with predation as the primary cause of KLWR declines. Additionally, we evaluated the KLWR captive-breeding and release program by comparing survival of captive-born and released KLWRs to wild-born KLWRs. The realized population growth rate of wild-born KLWRs had a strong pattern of covariation with recruitment; covariation between the realized population growth rate and apparent survival was negligible. Consistent with demographic theory, our results suggest that KLWR population dynamics were driven primarily by variation in recruitment, and that periodic reductions in recruitment led to population declines. We found that the survival curve and the first month (S1) and first 3-month (S1–3) survival estimates for the wild-born KLWRs [S1 = 0.929 (0.890–0.968); S1–3 = 0.942 (0.919–0.965)] were considerably higher (χ2 = 33.9, 1 d.f., P < 0.001) than released KLWRs [S1 = 0.521 (0.442–0.600); S1–3 = 0.561 (0.493–0.629)]. Low survival rates from predation limited the success of the captive-breeding and release program. This study illustrates the importance of pre-release conditioning of captive-bred animals and the importance of considering reproductive parameters in conjunction with survival estimates to understand the drivers of population decline.