Does foreign direct investment (FDI) benefit the domestic firms of the host country? And if so, through which channels are these benefits received? These long standing and important questions have attracted much interest among economists. Strictly from a theoretical perspective, the answer remains unclear: FDI can either help or harm domestic firms, depending on the intensity levels of various effects. Ultimately the net effects of FDI have to be evaluated empirically, but recent empirical studies have found both positive to negative effects, depending on the focus, the data and the method used.1 This article contributes another piece of evidence to this issue by focusing on a potentially important (but largely neglected) determinant of spillovers from FDI to domestic firms,2 namely the productive advantage foreign firms have over domestic firms within a sector and region.
According to Blomström et al. (2001), “foreign investors make available (directly or indirectly) appropriable technology to host country businesses. Appropriable technology should be viewed broadly as any tangible or intangible resource that can generate economic rent for host country firms, … by improving total factor productivity.” By this we should believe that technological spillovers will depend both on the technological edge foreign firms have over domestic firms, and on the geographical proximity of foreign firms to domestic firms. Hence, our variable of interest will be the productivity of foreign-owned firms, and in particular the difference in productivity between foreign and domestic firms in a sector-region. Since the work of Caves (1974), economists have honed their attention on the concentration of foreign-owned firms (measured as the share of FDI capital or share of FDI workers) within a sector and/or a region as the source (and proxy) of potential spillovers.3 In an influential paper, however, Aitken and Harrison (1999) have shown that properly controlling for unobserved region-specific effects eliminates most of the effect of FDI density on domestic firms' productivity. This may occur because certain local characteristics that attract FDI4 (such as the presence of a high-quality labor force, large local markets, good infrastructures and good administration) also enhance the productivity of domestic firms, generating a spurious correlation between these two variables. Once these factors are properly controlled for, no correlation survives.
This result, however, does not necessarily imply that there are no spillovers from FDI. The density of FDI may simply not be the right source of (and proxy for) spillovers. Rather the productivity advantage enjoyed by foreign firms may potentially be a more important source of spillovers. While the largest concentration of foreign firms normally occurs in regions and sectors where domestic firms are already highly productive, the largest effect of foreign firms on domestic ones could be in backward regions, where FDI concentration is small but their productivity effects are large due to the scope for technological catch-up. The effects of this mechanism are labeled the “Veblen–Gerschenkron” effect (or VG effect for brevity) in acknowledgement of the early intuitions of Veblen (1915) and Gerschenkron (1952).
The original formulation of the VG effect, originally proposed by Findlay (1978), states that technologically disadvantaged regions are more likely to benefit from spillovers from FDI, and may as a consequence experience stronger productivity growth relative to more advanced regions. Using a new data set of domestic and foreign-owned manufacturing firms in Italy and Germany for the period 1993–1999, we test the hypothesis that the productivity edge of foreign-owned firms in a sector and region is an important determinant of productivity growth for domestic firms in the same sector and region. The positive effect of the productivity of foreign-owned firms on the subsequent growth of domestic firms in a sector and region will be referred to as the Veblen–Gerschenkron (VG) effect. On one hand, some recent studies (Haddad and Harrison, 1993, Kokko, 1994, Sjöholm, 1999 and Castellani and Zanfei, 2003) have emphasized the beneficial effects of FDI on domestic firms due to intra-industry technology spillovers and subsequent technological catch-up. On the other hand, studies on the diffusion of technological knowledge based on patent and innovation data (Jaffe et al., 1993, Audretsch and Feldman, 1996 and Peri, 2005) have emphasized the high degree of localization that characterizes these flows. Our analysis combines this regional dimension of FDI spillovers with the importance of the productivity-advantage of foreign-owned firms in generating them. Such a method allows us to identify the VG effect while controlling for sector-specific and firm-specific determinants of productivity and growth.
A second novel contribution of our study is the computation and use of as many as five measures of total factor productivity at the firm level, each based on different methods and different assumptions. Because measures of total factor productivity are always indirect (they are based on the calculation of a “residual” from a production function) they can become polluted by systematically unobserved errors. Our measures of firm-level TFP address three important problems, namely the endogeneity of foreign ownership, unobserved heterogeneity of firms and selection of the sample. Each measurement method addresses one problem at the exclusion of the other problems. However, using each of these TFP measures, we find significant and similar estimates of the VG effect. This substantially reduces our concerns for spurious results.
The final novel element here is that our paper focuses on the interesting cases of Italy and reunified Germany. Our new data set contains geographic identifiers that allow us to locate firms within 1 of 103 Italian Provinces or 16 German Regions (Bundesländer). Italy and Germany have well-known issues of uneven regional development, and the cross-sectional variations in productivity across firms are quite large. Our study exploits these differences and provides some evidence on the role of foreign investments in promoting productivity growth in less-developed regions. In particular, Germany provides a very interesting case study, since the unification and subsequent inflow of firms from Western Europe put eastern firms in touch with new technologies after 50 years of almost complete isolation. East German firms had exhibited substantial technological backwardness upon the sudden inflow of western firms, generating the ideal conditions for strong VG effects.
Our estimates show the presence of positive and significant VG effects using each of our productivity measures. These effects may generate non-negligible productivity growth for local firms over the long run. Their magnitudes are estimated to be larger for German than for Italian firms (although the estimates for the German sample are less precise and such difference is not statistically significant). To provide an idea of the magnitude of such an effect, our baseline estimates imply that the presence of foreign-owned firms that were on average 50% more productive than local ones, would induce a long-run productivity gain for local firms of around 8% for a typical German region and around 3% for a typical Italian region. Because the productivity of the average foreign-owned firm was, in some cases, two or even three times as large as the productivity of local firms (at the beginning of the considered period) the VG channel could account for a sizable part of the productivity growth of domestic firms during the 1990s. We interpret this as evidence that the presence of highly productive FDI had a stimulating effect on the productivity of local firms.
The rest of the article is organized as follows. 2 and 3 describe our empirical model and the estimation strategy. Section 4 describes the data for Germany and Italy. Section 5 presents the basic estimates, with particular emphasis on the coefficient that captures the VG effect, and performs several checks of the robustness of these estimates. Section 6 concludes the article.
