We tested experimentally the effects of genetic variation in host population on the host-parasitoid system persistence. The experimental systems consisted of one parasitoid wasp species (Heterospilus prosopidis), one bean weevil species (Callosobruchus chinensis), and one bean species, of which only the host species (bean weevil) was genetically manipulated. As control treatments with low genetic heterogeneity in the host population, we used two bean weevil strains (Kyoto and Niigata strains) which have several contrasting ecological traits. For the high genetic heterogeneity treatment, hybrid bean weevils which were generated by crossbreeding the strains Kyoto and Niigata were used. In the multiple-generation experiments, all three treatments had different patterns of extinction. The control treatment with the Niigata strain was very prone to system extinction, whereas the treatment with hybrids showed coexistence of constituent species in almost all replicates and also showed stabilized population dynamics. The other control treatment, using the Kyoto strain, showed intermediate proneness. To interpret the results of multiple-generation experiments, we conducted several short-term experiments. The different persistence patterns between the two control treatments were explained by the shapes of the host-finding abilities of wasps and the growth rate of the bean weevils. The mean values of many ecological traits of hybrid lines were not different from those of the Kyoto strain, but their variability increased. These outcomes corresponded well to the prediction of models by Doebeli [J Theor Biol (1997) 188:109-120]. We discuss the mechanisms in which the variability in host species population was effective for the prolonged system persistence.