Sectoral Productivity Convergence, Input-Output Structure and Network Communities in Japan

Abstract

We study sectoral productivity convergence through the input-output structure of the economy and its network representation. In particular, we study 106 production sectors in Japan over the 2003-2012 period and identify highly interconnected groups of sectors using community detection algorithms. We next characterize the dynamics of these communities by evaluating the evolution of productivity dispersion using parametric and non-parametric frameworks of sectoral productivity convergence. We find two dominant communities: The central members of community 1 are mostly service-related industries, while the central members of community 2 are mostly high-tech manufacturing industries. The convergence analyses indicate that the two communities have contrasting convergence-divergence patterns. Robust convergence patterns are reported for community 1. In contrast, when those two communities are pooled into the entire group of all 106 industries, it appears that strong diverging signs in community 2 are responsible for the weak divergence in labor productivity for all industries in Japan.