Do Keiretsu Really Hinder FDI into Japanese Manufacturing?

Abstract

This paper examines an issue that has received considerable comment but little analysis. It has often been argued that the presence of the keiretsu in Japan has been instrumental in deterring multinational firms from entering Japan, though evidence for this is patchy. We present some new analysis of this issue, thereby evaluating the effects of keiretsu on inward investment penetration in Japan. In contrast to previous work in this area, our results suggest that there is little relationship between inward FDI and keiretsu networks, once one controls for endogeneity and unobservable heterogeneity. The results illustrate some important interaction effects between keiretsu and other explanatory variables that explain differences in inward investment penetration.

Key Words:

• Keiretsu
• FDI
• Market Structure

JEL classifications:

• F23
• L60 L41

Notes

1. The financial system has generally been reformed to facilitate an environment of mergers and acquisitions in Japan. Changes in the accounting system have led to a requirement to disclose the value of equity holdings at market value. This has resulted in massive sales of shares by bank-centred keiretsu and a stringent requirement for consolidation, making it difficult for firms to manage their earnings by allocating gains and losses among keiretsu group members.

2. Many foreign firms that are categorised as manufacturing companies are, in fact, associated with service and distribution activities rather than production per se.

3. Between 1999 and 2002, the Big Six were merged to the Big Four (Mitsubishi, Mitsui-Sumitomo, Mizuho, and UFJ) following major restructuring of Japan’s financial sector.

4. The membership is not mutually exclusive. For example, Hitachi is affiliated with the President Clubs of Fuyo, Sanwa, and DKB. We would note that some large manufacturing firms such as Honda, Sony, and Panasonic do not hold membership.

5. Our approach of estimating the system of dynamic panel equations is in the spirit of Holtz-Eakin et al. and Rosen (1988), thereby using lagged values as instruments to generate orthogonality conditions on differenced data. Notice that in the linear context we are working with, the 3SLS estimator can be derived as a GMM estimator from the orthogonality conditions implied by the set of instrument (see Theorem 5 in Cornwell et al., 1992)

6. They find that HK firms who hold President Club membership have a greater probability of engaging in FDI through mutual learning.