We have assessed the determinants of the choice of integration, relational contracting (keiretsu sourcing) and market sourcing by seven Japanese automobile manufacturers (OEMs) with respect to 54 components in light of contract economics. Our major findings are the following. First, the specificity and interdependency of a component significantly promotes vertical integration over keiretsu and keiretsu over market, consistent with transaction cost economics. Second, interdependency is a more important consideration for the former choice than for the latter choice, and the reverse is the case for specificity. This suggests that the hold-up risk due to specific investment can be often effectively controlled by a relational contracting based on keiretsu sourcing, while accommodating non-contractible design changes may often require vertical integration. Third, while higher testability of a component makes the effects of specificity significantly smaller, it also promotes the choice of keiretsu sourcing over market sourcing. One interpretation of this last result is that while higher testability improves the contractibility of the component with high specificity, it simultaneously enhances the advantage of keiretsu sourcing since it provides more opportunities for the supplier to explore new information for a collaborative exploitation with an OEM.