The Fable of the Keiretsu
Yoshiro Miwa and J. Mark Ramseyer
Central to so many accounts of post-war Japan, the keiretsu corporate groups lacked economic substance from the start. Conceived by Marxists committed to locating “domination” by “monopoly capital,” they found an early audience among western scholars searching for evidence of culture-specific group behavior in Japan. By the 1990s, they had moved into mainstream economic studies, and keiretsu dummies appeared in virtually all econometric regressions of Japanese industrial or financial structure. Yet the keiretsu began as a figment of the academic imagination, and they remain that today. Regardless of the keiretsu definition used, cross-shareholdings within the “groups” were trivial, even during the years when keiretsu ties were supposedly strongest. Neither does membership proxy for “main bank” ties. Econometric studies basing “keiretsu dummies” on the available rosters produce predictably haphazard and unstable results. In the end, the only reliably robust results are the artifacts of the sample biases created by the definitions themselves.
Central to so many accounts of post-war Japan, the keiretsu corporate groups lacked economic substance from the start. Conceived by Marxists committed to locating “domination” by “monopoly capital,” they found an early audience among western scholars searching for evidence of culture-specific group behavior in Japan. By the 1990s, they had moved into mainstream economic studies, and keiretsu dummies appeared in virtually all econometric regressions of Japanese industrial or financial structure. Yet the keiretsu began as a figment of the academic imagination, and they remain that today. Regardless of the keiretsu definition used, cross-shareholdings within the “groups” were trivial, even during the years when keiretsu ties were supposedly strongest. Neither does membership proxy for “main bank” ties. Econometric studies basing “keiretsu dummies” on the available rosters produce predictably haphazard and unstable results. In the end, the only reliably robust results are the artifacts of the sample biases created by the definitions themselves.