Vertical Integration and Production

We use two very large datasets to obtain some of the first systematic micro evidence on how production in vertically integrated firms differs from that of unlinked producers in the same industries. The Economic Census lets us observe virtually every establishment in the private economy, and crucially, the organizational structures of the firms that own them. The Commodity Flow Survey microdata offer a random sample of plants’ shipments, offering an unusually comprehensive glimpse into patterns of goods flows within firms. We find that overall levels of vertical integration were relatively stable during our 1977-97 sample period, though they declined somewhat after 1987. Within industries, vertical integration is related to differences in productivity, size, and capital intensity. These differences mostly embody persistent differences in the plants that are launched by or brought into vertically integrated firms, but to some extent also reflect changes that formerly unintegrated plants experience upon integration. Matching of plant “types” appears to be important in vertical integration: the correlation in these plant-level measures is stronger in firms with vertical structures than those that are purely horizontally structured. In terms of goods flows, we find that surprisingly few shipments of upstream integrated establishments go to any other plant in their firm. Across integrated plants, the modal fraction of their value of shipments that is intra-firm is zero; the median is 2.6 percent. We interpret this evidence as suggesting that vertical integration is often about transfers of intangible inputs like managerial oversight rather than physical ones. To the extent that such intangible inputs are complementary to physical inputs, equilibrium assignment entails the allocation of the best intangible inputs across multiple establishments in vertically related product categories.