by Masaru Inaba and Keisuke Otsu, discussion paper KDPE 1705, March 2017.
Non-technical summary
The objective of this paper is to construct a dataset of Japanese prefecture level production, income and expenditure data and analyze the Japanese regional growth and business cycle features. The 47 prefectures are analyzed individually and also as 10 regional groups; Hokkaido, Tohoku, Kanto, Chubu, Kinki, Chugoku, Shikoku, Kyushu and Okinawa.
Our dataset is based on the System of National Accounts (SNA) from the Cabinet Office Economic and Social Research Institute (ESRI). From expenditure data, we construct series of the prefectural per capita GDP, private consumption, private investment, government consumption and government investment in 2000 yen over the 1955-2008 period. From income data we construct series of the prefectural labor income share and depreciation rate over the 1975-2008 period. We construct the prefectural net capital stock series over the 1975-2008 period in 2000 yen from the ESRI Prefecture Private Capital Stock data and Private and Public Sector Balance Sheet data along with the SNA data on investment and depreciation. We construct the prefectural total hours worked series over the 1975-2008 period from the Research Institute of Economy, Trade and Industry (RIETI) R-JIP database and the SNA employment data. Finally, we construct the prefectural total factor productivity (TFP) series from the prefectural output, net capital stock and total hours worked series along with the average prefectural labor share.
In terms of regional growth, we find that over the 1955-2008 period, the Tohoku region and Okinawa region, which had the lowest average income per capita, experienced the highest growth. This is evidence of regional convergence in which poor regions grow faster than rich regions so that the income levels of all regions converge to similar levels over time. We formally test this using the framework introduced by Barro (1991) and find that convergence exists in the prefecture level in Japan over the 1955-2008 period. The convergence during the post oil-shock period 1975-2008 is less obvious but we still find regional convergence after controlling for prefectural characteristics such as TFP level gaps, population growth rates, private investment rates and TFP growth rates.
In terms of business cycles, we focus on the post oil-shock period 1975-2008 and find that the bilateral correlation of per capita output is negatively affected by the distance and the similarity of industrial structure and positively affected by the size of the total output of the pair. We further document that the bilateral correlation of output is higher than that of consumption in 847 out of the total 1081 pairs. This phenomenon frequently documented in open economy macroeconomic literature is puzzling since prefectures should want to smooth their consumption path against income shocks through borrowing and lending among each other. We decompose the consumption risk sharing into 3 steps: i) the net factor payments across prefectures capturing the income risk sharing through the capital market, ii) the government transfer across prefectures capturing the income risk sharing at the personal disposable income level, and iii) the consumption risk sharing of households through the financial market. The results show that the highest contributor to consumption risk-sharing is the government transfer which implies that financial market imperfection might be contributing to the low cross-prefecture correlation of consumption.
You can download the complete paper here.