The spatial genetic structure is a topical issue in the studies of various aspects of ecology and evolution. Using the multilocus autocorrelation method with hypervariable microsatellite genetic markers, we investigated a fine‐scale pattern of genetic structure in 5 local populations of the common shrew Sorex araneus separated by distances of 300–1000 m (the Moscow chromosomal race). Spatial genetic autocorrelation analysis based on 5 microsatellite loci (expected heterozygosity >0.79) with 90 alleles revealed a consistent pattern of significant positive genetic structure. By testing the autocorrelation at multiple scales from 25 to 500 m, we found that positive spatial genetic structure is detectable in distance classes of <500 m. The weaker spatial genetic structure positively correlated with a higher ratio of nonresident individuals to residents’ activity (number of captures). In contrast to nonresident animals, the residents demonstrated prominent genetic structure. Genetic difference (FST) between the populations was significant (0.016–0.051) and comparable with that between populations of different races analyzed previously (0.016–0.038). FST was not associated with geographic distance. These demographic patterns allowed us to propose a scheme of genetic‐structure dynamics involving periodic appearance of more related local groups and renewal of allelic profiles from а common pool where the alleles are mixed. The scheme predicts fluctuating genetic structure and random similar differences among local populations.