Abstract Biased sex ratios among reproductive individuals are common in plants, but the underlying mechanisms, as well as the evolutionary consequences, are not well understood. The classical theory of Düsing and Fisher predicts an equal primary sex ratio at seed production, based on the selective advantage of the rare sex. Biased sex ratios among reproductive plants can arise from sexual dimorphism in survival and flowering. Sex ratio biases can also be present from the seed stage; in these cases, assumptions of Düsing's and Fisher's theory, for example, random mating or demographic equilibrium, are thought to be violated. We investigated mechanisms leading to female‐biased sex ratios in the arctic‐alpine dwarf willow Salix herbacea L. We studied sex ratios in three natural populations over 3 years as well as in 29 crosses (full‐sib families) under controlled conditions over four growth periods. We tested whether sex ratio was associated with habitat parameters (elevation and snowmelt time), or with germination, survival or flowering, and whether females and males differed in size or flowering that may cause observation bias. We detected a strong and consistent female bias, both in natural populations (sex ratio [proportion of females]: 0.71–0.82) and in our controlled experiment (overall sex ratio: 0.70–0‐72). Female bias became more pronounced with increasing elevation. Our data did not support sexual dimorphism in size or flowering. Family sex ratios varied largely (from 0.25 to 1), including many female‐biased families, unbiased families and two male‐biased families. Families with lower germination, seedling establishment, survival or flowering did not have stronger female bias, indicating that intrinsically higher survival or flowering in females does not explain overall female bias. Synthesis. Our results suggest that sex ratio bias in S. herbacea is already present in seeds and does not arise through intrinsic differences between sexes. Candidate mechanisms that can lead to both overall female bias and variation in sex ratio among families are meiotic drive or cyto‐nuclear interactions. The pioneer habit of Salix may lead to non‐equilibrium population dynamics that allow for the long‐term persistence of variable genetic sex ratio distortion systems that arise from genetic conflict.