Reciprocal recurrent selection (RRS) has been successfully applied to maize breeding for more than 60 years. Our objective was to assess the relative efficiency of the genotypic value and the effects of general and specific combining abilities (GCA and SCA) on selection. The GCA effect reflects the number of favorable genes in the parent. The SCA effect primarily reflects the differences in the gene frequencies between the parents. We simulated three traits, three classes of populations, and 10 cycles of half- and full-sib RRS. The RRS is a highly efficient process for intra- and interpopulation breeding, regardless of the trait or the level of divergence among the populations. The RRS increases the heterosis of the interpopulation cross when there is dominance, and it decreases the inbreeding depression in the populations and the genetic variability in the populations and in the hybrid. When there is not dominance and the populations are not divergent, the RRS also determines population differentiation. The half-sib RRS, which is equivalent to selection based on the GCA effect, is more efficient than the full-sib RRS based on the genotypic value, regardless of the trait or the level of improvement of the populations. The full-sib RRS based on the SCA effect is not efficient for intra- and interpopulation breeding.