The relationship between foreign direct investment (FDI) and international trade has long generated great interest. This interest is sharpened by the phenomenal growth in the magnitude of global FDI flows in the last two decades. FDI arises from decisions of multinational enterprises (MNEs) either to capture local markets abroad through horizontal investments in similar products (Horstmann and Markusen 1992; Markusen 1984) or to take advantage of lower cost factors offshore through vertical investments in labor-intensive assembly stages (Helpman 1984).
The knowledge-capital (KC) model of MNEs, as described in Markusen (2002) and initially estimated by Carr, Markusen, and Maskus (2001), is now a widely used empirical approach for explaining the location and production decisions of global firms in an integrated framework. This model provides predictions about the nature of firms that arise endogenously in response to changes in market sizes, country similarity, factor endowments, impediments to trade and investment, and various interactions among them. The firms can be national exporters or either horizontal or vertical MNEs. The model also can be used to analyze joint decisions between production for local markets and exports, making it a theory of both investment and trade (Markusen and Maskus 2002). Recently, it has been extended to the context of three factor endowments (Bergstrand and Egger 2007).
The contributions of this paper are threefold. First, we investigate whether the KC model predictions hold for sectoral investment. The analysis examines the relationship between relative endowments of knowledge-based factors and other FDI determinants on local affiliate sales and exports. Specifically, we estimate the model for all manufacturing industries (aggregated) and several manufacturing subsectors (chemicals and allied products, electronics, food and kindred products, industrial machinery and equipment, and transportation equipment). Surprisingly few previous investigations have applied the KC model to sectoral data.
Interest in industry-level analysis arises to the extent that firms in an industry display the KC model's basic characteristics of increasing returns associated with headquarters services and the potential to fragment production into different stages. Those headquarters services include marketing, research and development (R&D), strategic planning, and related centralized activities that generate knowledge. Because the resulting information advantages may be shared at low cost among multiple locations, knowledge becomes one essential determinant of strategic decisions by MNEs. A second is differences in factor intensities at different vertical levels of production.
All manufacturing industries share these characteristics to some extent, but they clearly vary in their reliance on R&D and marketing, on the one hand, and vertical fragmentation, on the other hand. For example, Brainard (1997) noted that basic determinants underlying the KC model could be related to industries of varying characteristics. Vertical specialization generates trade in intermediate goods, while if industries are characterized by different degrees of differentiated varieties, market size should make a difference across sectors in FDI activity. Moreover, the underlying heterogeneity of firms suggests that different industries would react in varying ways to changes in the basic determinants of the KC model (Helpman, Melitz, and Yeaple 2004). Accordingly, it becomes interesting to assess this variation, a question that has not been widely studied to date.
Second, we adapt the three-endowment KC approach of Bergstrand and Egger (2007), which they applied solely to aggregate FDI data, to explicitly account in the empirical work for the role of differences in relative physical-capital endowments on investment decisions at the sectoral level. Introduction of this third factor should offer a more robust explanation of industry-level MNE activities than relying only on skilled labor and total labor. …