Why Do We Observe a Regional Balassa-Samuelson Effect?

Author Name TOKUI Joji (Faculty Fellow, RIETI)
Creation Date/NO. October 2019 19-J-054
Research Project Analysis of the Regional-Level Industrial Productivity and Regional Production Networks
Download / Links


Drawing a scatter diagram between differences in service prices and productivity among Japanese regions reveals a positive correlation. Additionally, all students of international economics know of the famous Balassa-Samuelson Effect, which explains why poorer countries tend to have cheaper service prices than richer countries. The apparent resemblance of the two phenomena could justify naming the domestic phenomenon the regional Balassa-Samuelson effect. Are the economic mechanisms that achieve these phenomena the same in the domestic and international contexts? The international version of the Balassa-Samuelson effect explains the phenomenon by relying on the presumption that labor productivity of rich countries is much higher than that of poor countries in the tradable manufacturing sector, but not in the non-tradable service sector. Although this presumption seems to be realistic in the international context, it does not hold in the domestic context. When we compare domestic labor productivity between rich urban areas and the other areas, while the manufacturing sectors exhibit little difference in productivity, the service sectors exhibit a pronounced productivity gap. Therefore we consider two alternative hypotheses that may explain the regional Balassa-Samuelson effect: high land use costs or high labor costs in urban areas. To conduct this research we constructed consistent prefectural-level input-output tables, and estimated land use costs for each industry in each prefecture. We apply the Leontief price model and calculate regional price differences caused by differences in both land use cost and labor cost and we estimate a regression equation with regional service price index as dependent variable and calculated price differences caused by land use cost and by labor cost as explanatory variables. Using the estimated regression equation we decompose the sum of squares for the dependent variable into that related to land use cost and that related to labor cost, to find that the former account for only 20 percent and the latter account for 80 percent. Thus the impact on prices of high urban labor cost is more important in accounting for the regional Balassa-Samuelson effect.