Inbreeding has the potential to cause evolutionary changes in populations, although these changes are likely to drive populations to extinction through inbreeding depression and reductions in genetic diversity. We investigated the mating system and late-stage inbreeding depression (δ) in 10 populations of Magnolia stellata using nine microsatellite markers and evaluated the effects of population size and the degree of population isolation through inbreeding and inbreeding depression on the persistence of populations. The outcrossing rates were very similar (~0.7) among populations, but the correlations of paternity, fractions of biparental inbreeding and inbreeding coefficients at the seed stage (FS) varied among populations, suggesting that the level of outcrossing was similar among populations, while the quality of it was not. A significant negative correlation was detected between FS and population size. The average value of δ was 0.709, and the values in six of the 10 populations were significant. The values of δ differed among populations, although clear relationships with population size and the degree of population isolation were not detected. However, in one population, which was very small and located in the edge of the species’ range, we obtained a very low value of δ (–0.096), which may be indicative of purging or the fixation of deleterious alleles. Existing M. stellata populations that are small (and thus might be expected to have higher frequencies of inbreeding) and have large values of δ may be in danger of declining, even if the populations are located within the central region of the species’ range.