<p> Soybean yields have remained below 2.0 t ha⁻¹ for over 30 years in Asian countries; however, few studies have analyzed national statistical data in detail to explore the underlying factors. This is the first study to analyze precisely and comprehensively factors involved in the historical soybean yield suppression in Japan by using national production-cost and other statistics, coupled with a crop model. Panel data of 5–6 land-size sections for 1987–2018 were prepared to analyze the regression of actual yield against 13 production input-related variables, model-predicted potential yield, and agricultural social capital score by each data class for the whole country (i.e., Japan), two climatic regions (temperate and subarctic), and two field types (converted paddy and upland). Although potential yield rose by 8% for the studied 32 years at the country/national level, actual yield declined by 18%, and this was primarily attributed to trends in converted paddy fields in the temperate region. Panel regression analyses showed that for this field type and region, a total of 14–21% actual yield reduction occurred with decreased water management input, as well as increases in weed control failure and converted paddy fields, resulting in a 16% reduction at the country level. In contrast, the increased potential yield, mainly due to climatic changes (warming), contributed to improving actual yields by 7–8% in the region and country, thereby partially offsetting the above yield reduction. Against our expectation, no actual yield improvement was observed with increased agricultural social capital score, but rather a 19% reduction for the country. Implications of the results for policymakers and scientists aiming to improve country-level soybean yields in Japan were discussed, as well as the importance of potential yield as an explanatory variable and that of spatial error model for panel data of land-size sections. </p>