To assess the genetic diversity of Japanese native horse populations, we examined seven such populations using mitochondrial DNA (mtDNA) and microsatellite analyses. Four reference populations of Mongolian horses and European breeds were employed as other equids. In the mtDNA analysis, the control region (D-loop) of 411 bp was sequenced, and 12 haplotypes with 33 variable sites were identified in the Japanese native horses. The phylogenetic tree constructed by haplogrouping and using worldwide geographic references indicated that the haplotypes of the Japanese native horses were derived from six equid clusters. Compared with the foreign populations, the Japanese native populations showed lower within-population diversity and higher between-population differentiation. Microsatellite analysis, using 27 markers, found an average number of alleles per locus of 9.6 in 318 native and foreign horses. In most native populations, the within-population diversity was lower than that observed in foreign populations. The genetic distance matrix based on allelic frequency indicated that several native populations had notably high between-population differentiation. The molecular co-ancestry-based genetic distance matrix revealed that the European populations were differentiated from the Japanese and Mongolian populations, and no clear groups could be identified among the Japanese native horse populations. The genetic distance matrices had few correlations with the geographic distribution of the Japanese native populations. Based on the results of both mtDNA and microsatellite analyses, it could be speculated that each native population was formed by the founder populations derived from Mongolian horses. The genetic construction of each population appears to have been derived from independent breeding in each local area since the time of population fission, and this was accompanied by drastic genetic drift in recent times. This information will help to elucidate the ancestry of Japanese native horses.