انتقال تولید در شرکت های چند ملیتی و حقوق و دستمزد آمریکا

This paper tests whether the transfer of production stages within US-headquartered multinational enterprises (MNEs) from US parents to foreign affiliates has contributed to within-industry shifts in US relative labor demand toward the more-skilled. There are two main empirical results. First, there is evidence of MNE transfer during the past 20 years. Second, regression analysis does not support the MNE hypothesis. MNE transfer tends to have small, imprecisely estimated effects on US relative labor demand. This finding is inconsistent with models of MNEs in which affiliate activities substitute for parent unskilled-labor-intensive activities.

The wages of more-skilled Americans relative to those of their less-skilled counterparts have been rising sharply since the late 1970s.1 Many economists studying this wage divergence have concluded that its primary cause was a within-industry shift in relative labor demand toward the more-skilled.2 Despite this consensus, there is still disagreement about what caused the within-industry demand shifts. These shifts are consistent with skill-biased technological change, as many researchers have pointed out. However, they are also consistent with explanations related to international trade. For example, Feenstra and Hanson, 1996a and Feenstra and Hanson, 1996bfind outsourcing to be correlated with skill upgrading. In this paper I test a different hypothesis for what caused the within-industry relative-demand shifts: the transfer of production stages within US-headquartered multinational enterprises (MNEs) from their US parents to their foreign affiliates. Like skill-biased technological change (SBTC) and outsourcing, MNE transfer can generate within-industry skill upgrading. There are at least two important reasons to test the empirical importance of MNE transfer. One is that little is known about it. In his survey of research on trade, FDI, and labor markets Baldwin (1995)(p. 55) finds that `there do not seem to be any studies of how the shifts in the pattern of US direct investment and direct foreign investment in the United States have affected relative wages'. This lack of knowledge matters because US parents account for the majority of total US manufacturing value added – 60% in 1989 and 59% in 1994. It also matters because MNE transfer differs from outsourcing in two important ways. First, MNE transfer considers only within-firm activity and excludes any arm's-length transactions between US and foreign firms. The Feenstra and Hanson, 1996a and Feenstra and Hanson, 1996boutsourcing encompasses both within-firm and arm's-length transactions. Second, MNE transfer does not require output to be sent back to the United States. Outsourcing defined as imported intermediate inputs captures only transferred production stages which return to the United States for further processing. MNE transfer encompasses foreign production either sold in that host market or exported to some third market. A second reason to study MNE transfer is policy. If outsourcing causes skill upgrading then trade barriers are a possible (albeit inefficient) policy option. If MNE transfer matters, however, then limits on outward foreign direct investment (FDI) are a policy option as well. To analyze MNE transfer I combine two data sources covering 1977 through 1994: publicly available data from the Bureau of Economic Analysis (BEA, 1977–1994) on US manufacturing MNEs, and data for all US manufacturing from the National Bureau of Economic Research's (NBER) manufacturing productivity database. The BEA data start in 1977, about when wage inequality within US manufacturing started rising. Importantly, in at least some years both data sources classify employees as either nonproduction and production workers. This allows me to distinguish between more-skilled (i.e. nonproduction) and less-skilled (i.e. production) workers.3 There are two main empirical results. First, there is evidence of MNE transfer during the past 20 years. In particular, between 1977 and 1994 foreign affiliates of US MNEs increased their share of MNE-wide employment, capital stock, and value added. However, this MNE transfer has not simply entailed rising absolute levels of affiliate activity. Importantly, affiliate employment declined during the sample period. Second, regression analysis of an industry-year panel does not support the hypothesis that MNE transfer has contributed to US skill upgrading within industries. Across all specifications, most measures of MNE transfer have small and imprecisely estimated effects on US relative labor demand. This finding is inconsistent with models of MNEs in which affiliate activities substitute for parent unskilled-labor-intensive activities. The paper has four additional sections. Section 2discusses related work. Section 3presents a set of facts about MNE transfer, while Section 4presents econometric evidence on MNE transfer and US labor-demand shifts. Section 5concludes.

This analysis of MNE transfer and skill upgrading has generated two main empirical results. First, there is evidence of MNE transfer during the past 20 years. In particular, between 1977 and 1994 foreign affiliates of US MNEs increased their share of MNE-wide employment, capital stock, and value added. However, this MNE transfer has not simply entailed rising absolute levels of affiliate activity. Importantly, affiliate employment declined during the sample period. Second, regression analysis a panel of industry-year observations does not support the hypothesis that MNE transfer has contributed to US skill upgrading within industries. Across all specifications, most measures of MNE transfer have small and imprecisely estimated effects on US relative labor demand. This finding is inconsistent with models of MNEs in which affiliate activities substitute for parent unskilled-labor-intensive activities. The finding that MNE transfer did not contribute to US skill upgrading is important in at least two ways. First, it belies the common perception that MNEs have recently downsized US parents while aggressively expanding affiliates. Second, it has an important policy implication. Restrictions on outward FDI (likely to be bad for aggregate welfare anyway) can do very little, if anything, to slow skill upgrading within US industries. Finally, this finding does not mean that MNEs did not contribute to US skill upgrading at all. Given the size of US parents, it would be surprising if MNEs played no role at all. What this paper does clarify is that this role was not primarily one of MNE transfer. For example, there is evidence that US parents have increasingly relied on outsourced intermediate inputs. For broad industry aggregates, the BEA census years have sufficient data to construct the same measure of outsourcing in Feenstra and Hanson (1996b). Of all intermediate inputs purchased by US manufacturing parents, the share that was imported from abroad rose from 6.79% in 1977 to 6.96% in 1982 and 10.94% in 1989. This matches the manufacturing-wide trend documented by Feenstra and Hanson (1996b). Also, foreign affiliates ship only about 40% of these imported inputs. The fact that even MNEs rely on arm's-length transactions for most of their outsourced inputs suggests the importance of the Feenstra and Hanson (1996b)measure. It should also be noted that I have not considered the possibility that MNEs contribute to skill upgrading through threatened (rather than actual) relocation abroad. In principle, MNEs might affect US wages by merely threatening to transfer operations abroad. Anecdotes of these `threat' effects are commonplace, but there remains a need for systematic analysis of their importance.