The effective population size is influenced by many biological factors in natural populations. To evaluate their relative importance, we estimated the effective number of breeders per year (Nb) and effective population size per generation (Ne) in anadromous steelhead trout (Oncorhynchus mykiss) in the Hood River, Oregon (USA). Using demographic data and genetic parentage analysis on an almost complete sample of all adults that returned to the river over 15 years (> 15 000 individuals), we estimated Nb for 13 run years and Ne for three entire generations. The results are as follows: (i) the ratio of Ne to the estimated census population size (N) was 0.17–0.40, with large variance in reproductive success among individuals being the primary cause of the reduction in Ne/N; (ii) fish from a traditional hatchery program (Htrad: nonlocal, multiple generations in a hatchery) had negative effects on Nb, not only by reducing mean reproductive success but also by increasing variance in reproductive success among breeding parents, whereas no sign of such effects was found in fish from supplementation hatchery programs (Hsupp: local, single generation in a hatchery); and (iii) Nb was relatively stable among run years, despite the widely fluctuating annual run sizes of anadromous adults. We found high levels of reproductive contribution of nonanadromous parents to anadromous offspring when anadromous run size is small, suggesting a genetic compensation between life-history forms (anadromous and nonanadromous). This is the first study showing that reproductive interaction between different life-history forms can buffer the genetic impact of fluctuating census size on Ne.