Agglomeration and foreign direct investment

The spatial clustering of foreign direct investment (FDI) is clearly visible in the location of multinationals investing in the United States, the European Union, China, and other regions. This agglomeration is at least partly the result of policy, as in China’s special economic zones, but spatial concentration is also characteristic of domestic firms and of FDI in economies with few controls. These observations suggest that market forces, as well as policy, lead to clustering.
That new establishments tend to go to the same locations as earlier entrants suggests that productivity rises with the level of economic activity, especially as firms often must pay higher land prices to locate in clusters. If such productivity-enhancing effects, or agglomerative economies, exist and spill over to domestic activities, a case may be made for government incentives to multinationals to induce local affiliate production. Indeed, dozens of countries favor foreign direct investment through tax breaks and subsidies. Through these incentives, governments hope to begin a selfreinforcing processwhereby subsidized early entrants attract additional investment.
To better design such policies, researchers have sought evidence that agglomerative economies exist and, if they do, the extent of their benefits to local productive factors. Location-choice studies seek to measure the attractiveness of local characteristics for foreign investors and thus provide a way to estimate the self-reinforcing power of FDI. Virtually all location-choice studies find that the existing stock of foreign investment is a significant predictor of the location a multinational will choose for new local affiliates.However,most countries receive a relatively small number of new multinational affiliates in a given year and for these projects there is often limited information, constraining our ability to identify the specific sources of agglomerative economies.
Head and Ries (1996) observe a relatively large number of investment projects, 931 equity joint ventures in 54 Chinese cities from 1984 to 1991. Their study is noteworthy for its carefulmodeling of the agglomerative process, emphasizing local input sharing as the source of positive firmspillovers. Using conditional logit analysis to estimate the likelihood that a particular city is chosen as the investment site, Head and Ries find that agglomerative economies greatly magnify the direct impact of government incentives. Their simulation analysis suggests that two-thirds of the gains from incentives can be attributed to the self-reinforcing nature of earlier investments. Not all locations gained equally, however, as cities considered attractive for other reasons, such as infrastructure and industrial base, gained the most. Similarly, Devereux, Griffith, and Simpson (2007) find that firms are less responsive to government subsidies in areas where there are fewer established plants in their industry.
Evidence that past investment increases the likelihood of new investment does not necessarily imply the existence of agglomerative economies. Agglomeration arises because there are benefits to locating near similar firms and because certain locations have natural advantages features of a location that are independent of firm location decisions. A common example of how natural advantages influence location choice is the North American steel industry, which concentrated in the Great Lakes region largely because of the location of iron ore and coal deposits. Inmeasuring the extent of agglomerative economies, researchers confront an identification problem: Are firms choosing a common location because its inherent characteristicsmake themmore productive or are they more productive because they have all chosen the same location?
Head and Ries (1996) try to separate the roles played by natural advantages and agglomerative economies in two ways. First, they include in their logit analysis a set of variables that attempt to control for local characteristics that influence firm productivity, particularly infrastructure. Second, they allow for spatially correlated errors by including provincial fixed effects. These two approaches are standard in the literature, and data limitations often make it difficult to do more to avoid bias caused by omitted local characteristics or endogeneity. For example, it is often impossible to include fixed effects at the same geographic scale as the unit of location choice (e.g., city fixed effects in the Head and Ries study) because they cannot be estimated for regions that received no investment. However, to fully control for all features of a location that attract investment is impossible, and even in themost careful studies omitted variables likely remain a problem.
Some studies have tried to assess the relative attractiveness of various kinds of prior investment for new entrants. Examining Japanese investment in the United States electronics industry from 1980 to 1998, Chung and Song (2004) ask whether firms agglomerate with their competitors or with their own prior investments. They find that firms tend to colocate only with their own prior investments, with the exception of firms that have little of their own experience, who do tend to colocate with competitors.
More recent work emphasizes the role of trade costs and market access as an alternative explanation for foreign direct investment clustering. Head and Mayer (2004) develop a theoretical model in which firms prefer to locate where demand is highest and serve smaller markets by exporting. They confront the data with this hypothesis, measuring market potential by a term that weights demand in all locations by its distance from the proposed investment site. Head and Mayer use standard logit techniques to analyze the European regions chosen as the sites of 452 Japanese investments. They decompose existing investment in each region into three firm counts distinguished by their relatedness to the new entrant: domestic establishments in the same industry, Japanese affiliates in the same industry, and Japanese affiliates with the same parent or network. They find that all three measures of prior investment have a large and positive influence on the likelihood that a region will be chosen by a new entrant, with this effect larger the closer the relations between firms. Thus there are strong agglomeration effects even when controls for market potential are included in the analysis.
An important issue for policy is whether domestic productivity is enhanced by the presence of foreignowned firms. Most productivity-spillover studies are of specific industries or are case studies, both of which are limited as a guide to policy. Haskel, Pereira, and Slaughter (2007) offer evidence on domestic spillovers from FDI using a plant-level panel of all UK manufacturing firms from 1973 to 1999. Several previous studies using plant-level data find a negative or insignificant effect of industry-level foreign direct investment on local productivity. The UK data are unique in that they cover the whole of manufacturing in a developed country. Haskel, Pereira, and Slaughter estimate plant-level productivity and regress it on industry-level FDI, controlling for inputs and the level of competition. They estimate that a 10- percentage-point increase in foreign presence in a UK industry raises the total factor productivity of that industry’s domestic plants by about 0.05 percent. They compare the value of these estimated spillover effects to per-job incentives offered in specific cases and find that these expenditures outweigh the benefits.
Haskel, Periera, and Slaughter (2007) use a variety of methods to deal with identification problems. In addition to explaining variation in gross output, they time-difference the data, explaining the change in output as a function of changes in inputs and foreign industry presence. Thismethod accounts for plant-specific effects. The authors also include time, industry, and region fixed effects in their regression analysis. They also worry about the possibility that changes in industry foreign direct investment levels are correlated with changes in domestic productivity, and use instrumental variable techniques tominimize endogeneity bias. Their findings provide the strongest evidence to date that foreign investment does raise domestic productivity, but more work is needed before we have a clear guide to policy.
In sum, locations aremore attractive the larger the existing stock of foreign investment, especially when the existing investments are by firms that are closely related (same industry, nationality, or parent firm). Government incentives are a significant determinant of multinational affiliate location choice but incentives aremost effectivewhen a location is desirable for other reasons.Although recent evidence suggests that foreign-owned firms enhance the productivity of local establishments, the value of these domestic spillovers appear to be less than the incentives used to attract foreign investment.
See also location theory; New Economic Geography; technology spillovers