A driving stress of the M w 5.8 reverse-faulting Awaji Island earthquake (2013), southwest Japan, was investigated using focal mechanism solutions of earthquakes before and after the mainshock. The seismic records from regional high-sensitivity seismic stations were used. Further, the stress tensor inversion method was applied to infer the stress fields in the source region. The results of the stress tensor inversion and the slip tendency analysis revealed that the stress field within the source region deviates from the surrounding area, in which the stress field locally contains a reverse-faulting component with ENE-WSW compression. This local fluctuation in the stress field is key to producing reverse-faulting earthquakes. The existing knowledge on regional-scale stress (tens to hundreds of km) cannot predict the occurrence of the Awaji Island earthquake, emphasizing the importance of estimating local-scale (< tens of km) stress information. It is possible that the local-scale stress heterogeneity has been formed by local tectonic movement, i.e., the formation of flexures in combination with recurring deep aseismic slips. The coseismic Coulomb stress change, induced by the disastrous 1995 M w 6.9 Kobe earthquake, increased along the fault plane of the Awaji Island earthquake; however, the postseismic stress change was negative. We concluded that the gradual stress build-up, due to the interseismic plate locking along the Nankai trough, overcame the postseismic stress reduction in a few years, pushing the Awaji Island earthquake fault over its failure threshold in 2013. The observation that the earthquake occurred in response to the interseismic plate locking has an important implication in terms of seismotectonics in southwest Japan, facilitating further research on the causal relationship between the inland earthquake activity and the Nankai trough earthquake. Furthermore, this study highlighted that the dataset before the mainshock may not have sufficient information to reflect the stress field in the source region due to the lack of earthquakes in that region. This is because the earthquake fault is generally locked prior to the mainshock. Further research is needed for estimating the stress field in the vicinity of an earthquake fault via seismicity before the mainshock alone.