Genetic diversity is the source of adaptive evolution and is crucial for population persistence under various environmental changes. Because perennial plant populations consist of individuals in different life history stages (e.g., seedling, juvenile, flowering), it is important to know which stage classes particularly possess large genetic variation and thus play key roles in maintaining genetic diversity. In this study, we developed a theoretical model for a perennial plant to predict stage-wise genetic diversity, and then applied the model to seven wild populations of a perennial herb Trillium camschatcense in the Tokachi region, Hokkaido. We estimated the parameters of the model, which were stage-specific survival probabilities and fecundities, as well as population size, by the four-year of demographic census survey, and the estimated values were substituted to the model to yield the theoretical prediction. In parallel, we empirically obtained stage-wise genetic diversity using double-digest restriction site-associated DNA sequencing (ddRAD-seq). Although the theory predicted equivalent levels of genetic diversity among stages, the empirical data showed that juvenile stages possessed high genetic diversity. The results suggest that wild populations do not satisfy any assumptions of the model, such as random mating, and consequently their genetic diversity is strongly subjected to life history stage.