The fall of global economic barriers, along with technological advances, has accelerated the international fragmentation of production processes. Firms strategically relocate their production systems to foreign countries where they can benefit from country-specific advantages. This incentive is referred to as the comparative advantage (or vertical) motive for foreign direct investment (FDI). An emerging question is which country-specific characteristics generate a comparative advantage that shapes the pattern of FDI. The literature has typically focused on factor endowments, such as skilled labor and physical capital, and found evidence that countries with an abundant factor endowment attract more foreign investors in industries that use the given factor intensively (e.g., (Antràs, 2003 and Yeaple, 2003)). (See Fig. 1, Fig. 2 and Fig. 3.)
Interestingly, in a separate strand of literature, the laxity of environmental regulations has also been assessed as a potential source of comparative advantage. The theoretical rationale is straightforward: polluting firms have an incentive to shift their production system to countries with lax environmental regulations to lower production costs. Classical Heckscher-Ohlin trade theory is then applied to predict that environmentally lax countries specialize in industries producing polluting goods, whereas environmentally stringent countries specialize in industries producing clean goods. 2 Unlike the factor endowments case, however, tests of this so-called pollution haven hypothesis (or pollution haven effect) have yielded rather conflicting and weak empirical evidence. (Brunnermeier & Levinson, 2004) summarize in their review that early literature up to the 1990s typically find no significant pollution haven effect, while later studies tend to find a statistically significant, but economically mild, effect of environmental regulations on industry composition.
One concern in the pollution haven literature is that it has heavily relied on empirical results from the most advanced countries, such as the U.S. (Eskeland and Harrison, 2003, Grossman and Krueger, 1994 and Hanna, 2010), Germany (Wagner & Timmins, 2009), and the U.K. (Manderson & Kneller, 2012). Although these countries impose the toughest environmental standards, they also retain potential confounding factors that may dampen the pollution haven effect. A typical approach to solve the problem is to use exclusion restrictions (Kellenberg, 2009, Levinson and Taylor, 2008 and Xing and Kolstad, 2002). However, the validity of exclusion restrictions is in fact often criticized (Broner et al., 2012 and Millimet and Roy, 2011). Another group of studies tries directly to disentangle such confounders. (Antweiler et al., 2001) and (Cole & Elliott, 2005), for example, posit that polluting industries, which tend to be capital intensive, may locate in advanced countries to exploit rich capital stocks despite the countries’ stringent environmental regulations. Both studies find statistically significant pollution haven effects when this capital-seeking incentive is sorted out. In other studies, (Ederington et al., 2005) and (Wagner & Timmins, 2009) argue that positive spillovers from industry agglomeration can be an important reason for polluting firms not to leave advanced countries.
Another crucial, yet overlooked, confounder is clean technology adoption. Environmental regulations may not only induce shift of production location, but also promote innovation and adoption of clean technologies (Popp et al., 2010). Firms employing clean technologies in response to the domestic environmental regulations, usually observed in the rich world, would have less incentive for outward migration. 3 In this spirit, (Dean et al., 2009) examine the pattern of FDI inflows across Chinese provinces to test whether foreign firms embodying less efficient abatement technologies are more responsive to inter-provincial differences of environmental regulations. Their finding confirms that only ethnically Chinese investors were significantly sensitive to the provincial differences, while non-ethnically Chinese investors who transferred relatively advanced technologies showed no significant response. Until now, no study examining the most advanced countries has yet addressed the clean technology issue.
This paper provides new evidence for the pollution haven hypothesis by investigating the pattern of South Korean FDI outflows to 50 host countries in 121 industries over the period 2000–2007. The Korean case is an advantageous setting for the clean technology issue. After experiencing severe environmental degradation accompanied by dramatic economic growth, South Korea newly adopted and amended almost its entire body of environmental legislation throughout the 1990s (OECD, 1997). However, the adoption of clean technologies lags behind the change in regulations, as Korean firms, mostly small- and medium-sized ones, still rely on old, dirty production technologies. In the subsequent review, ( (OECD, 2006), p. 1) stresses that “indicators of carbon, energy and some material intensities still remain among the highest in the OECD.” 4 Facing increasingly stringent environmental standards with limited access to clean technologies, the incentive for Korean firms to seek a pollution haven appears clearer than those in the most advanced countries.
To deal with other potential confounders, we employ a difference-in-differences (DID) type identification strategy, i.e., determinants of comparative advantage are identified by interaction terms between country and industry characteristics. 5 If environmental laxity is a determinant of comparative advantage, the pollution haven effect is identified through the interaction term between a host country’s environmental laxity (relative to home country) and industry’s pollution intensity. This approach enables us to disentangle opposing forces between environmental laxity and other determinants of comparative advantage. Furthermore, since our variables of interest are in interaction terms, we can control for all country- and industry-specific unobserved heterogeneity, in which case our model for FDI flow resembles the empirical model for trade flow used in (Romalis, 2004). Hence, we can apply the same model to trade data to see if a consistent behavior is observed in the pattern of trade.
After relevant issues are carefully treated, we find strong evidence for the pollution haven hypothesis. Countries with relatively lenient environmental regulations tend to attract more South Korean FDI in polluting industries than in non-polluting industries in terms of the total amount of investment (i.e., intensive margin of FDI). Economic significance is comparable to the disproportional effects of physical capital and skill endowment. This finding is robust to the inclusion of additional sources of comparative advantage. Note that, however, the same finding disappears in such cases when (i) industries are observed at a more aggregated level, (ii) physical capital is not included as a source of comparative advantage, or (iii) unobserved heterogeneity is not adequately controlled for. This highlights the importance of an elaborate identification strategy and a fairly disaggregated data. With the same identification strategy and data, we also find that the pollution haven effect is consistently observed in the patterns of new birth of foreign affiliates (i.e., extensive margins of FDI). Polluting industries show a disproportionally higher tendency to establish their new foreign affiliates in environmentally laxer countries than non-polluting industries do. When our model is applied to South Korean import data, we find that countries with lax environmental regulations tend to specialize in the production of polluting goods and export them to Korea.
Our paper contributes to the literature in a number of ways. First, the strong evidence from South Korea extends the validity of the pollution haven behavior to newly developed countries. In particular, the pollution haven incentive can be magnified in a transitional phase where clean technologies have not been adopted despite strengthened environmental regulations than in a phase where environmental regulations have already induced self-perpetuating clean technology innovation and adoption. This prediction is also in line with the recent literature on the directed technical change toward clean technology (Acemoglu et al., 2012 and Aghion et al., 2012). Secondly, even though the pollution haven hypothesis is testable by examining the pattern of FDI flow (Cole and Elliott, 2005, Eskeland and Harrison, 2003, Hanna, 2010, Javorcik and Wei, 2004 and Kellenberg, 2009), trade flow (Broner et al., 2012, Ederington et al., 2005, Grossman and Krueger, 1994 and Levinson and Taylor, 2008), or birth of plants (List et al., 2003), no study to our knowledge has looked into different types of industry activity at the same time. We investigate those three different industry activities to confirm a consistent behavior among one another and thereby provide a clearer picture of industry relocation occurring in a country. The third contribution is methodological. Our strategy is particularly useful when investigating FDI at the cross-country level, because the decision process of foreign investment is very sensitive not only to the industry structure of home countries but also to the characteristics of host countries, which generates a variety of country- and industry-level unobserved heterogeneity. 6
The remainder of this paper is organized as follows. Section 2 describes our data and discusses some important features of the data. In Section 3, we set up empirical models and address related econometric issues. Section 4 presents estimation results with their robustness checks. Section 5 concludes.
