Comments on Revisiting Complementarity between Japanese FDI and the Import of Intermediate Goods: Agglomeration Effects and Parent-firm Heterogeneity

Abstract

One of the contributions of this paper is the creation of unique panel data by combining Chinese Custom Trade data, China’s An-nual Survey of Industrial Firms, Toyo Keizai’s Overseas Japanese Companies, and Tokyo Stock Research data by identifying both parent and affiliated firms for seven years from 2000 to 2006. Adopting a discrete-time hazard model used by Hess and Persson (2012), the authors lead to a conclusion that firms in agglomerated regions with more foreign affiliates shorten its duration, while small and medium-sized (SME) firms import for a longer duration. My first comment is on the use of Toyo Keizai’s Overseas Japanese Companies. These data provide the year of establishment of the FDI company. However, the paper does not use this establishment year data for further analysis. The behavior of FDI companies can de-pend on the year of establishment year, especially with regard to survival analysis. The behavior of FDI companies founded in 1990 and in 2000 must be different. My second comment is the period coverage. There is similar survival analysis for FDI and trade nexus that addresses much longer periods. For example, Clausing (2000) used data for 18 years from 1977 to 1994, and Head and Ries (2001) used 25 years of data from 1966 to 1990. This paper uses a much shorter period—seven years from 2000 to 2006. Is this period setting long enough to assert that this conclusion is valid? In addition, it focuses on the negative sign of coefficients of important explanatory variables such as Diff, Upstream, Process , and Exports , and the positive signs of variables such as Alabor, KL-ratio, Agg-local-r , and Plarge . It may be more informative to include in the analysis things such as how many years it takes to reduce 50 percent or 80 percent reduction of intermediate imports. Nevertheless, this paper creates fine, rich parent-affiliate panel data derived from four dif-ferent databases. The framework of analysis is compact and