Harvesting for food or sport is often non-random with respect to demographic state, such as size or life stage. The population- level consequences of such selective harvesting depend upon which states are harvested and how those states contribute to population dynamics. We focused on a form of selective harvesting that has not previously been investigated in an experimental context: sex-selective harvesting, a common feature of exploited, dioecious populations. Using simple metapopulations (two patches connect by dispersal) of sexually dimorphic Bruchid beetles in the laboratory, we contrasted the effects of female-selective, male-selective, and non-selective harvesting over six generation of population dynamics. We also tested the ability of a harvest refuge (one patch of the metapopulation free from harvesting) to mitigate the effects of harvesting, and whether refuge effects interacted with sex selectivity. Sex-selective harvesting significantly perturbed operational sex ratios and harvest refuges dampened these perturbations. Metapopulations assigned to male-selective and non-selective treatments were able to fully compensate for harvesting, such that their dynamics did not differ from non-harvested controls. Only female-selective harvesting led to significant reductions in population size and this effect was completely offset by dispersal from a harvest refuge. A two-sex model confirmed that population dynamics are more sensitive to female vs. male harvesting, but suggested that higher levels of male harvest than included in our experiment would cause population decline. We discuss the roles of density-dependent competition and frequency-dependent sexual processes in the population response to sex-selective harvesting.