Location and Trade Patterns
Figure 9: Stylized Scenarios of Location and Intra-Industry Trade
Ultimately, to explain concrete patterns of location and trade, we would need to be able to link industry
characteristics (such as scale, market structure, factor mix etc) to industrial organization (size and structure of
units, division of labor, etc), and then to locational structure (linkages, spatial transactions costs, agglomeration
economies), all the while taking into account the patterns of comparative advantage and history. For the
moment, however, there is widespread agreement that it is extremely difficult, if not impossible, to explain
locational concentration of industries according to industry characteristics: most attempts to do so are
inconclusive (Brulhart and Torstensson, 1996; Amiti, 1999). Moreover, though theory has a few things to say
about how industry characteristics might affect location patterns, different characteristics push in different
directions, and there is as yet no unified theory on the relationships that would allow specification of an
empirical model. For the moment, we must be more modest in our ambitions. One way to begin is to carry out
a thought experiment about the organizational and linkage structures which correspond to the observed levels of
locational concentration and intra-industry trade, as is depicted in Figure 9.
Cases B and D are relatively straightforward. In B, the combination of low locational concentration
and low intra-industry trade implies that most trade will be in final outputs and that it will be highly
asymmetrical (with exporters and importers not overlapping much), and that there will be a high level of
self-supply without international trade. The development of local input structures could come about for different
reasons, however. In the EU printing and publishing industry (342) for example, it is because of the local
demand structure; whereas in the glassware, furniture, wood products, and paper industries (362, 332, 331, 341),
it is probably because location follows natural resources. The wearing apparel industry (322) for the OECD is,
somewhat surprisingly, found in this group, and we surmise that this reflects the progressive development of
local input structures in countries receiving foreign direct investment, and thus the general expansion of possible
locations for that sector, leading to the decline in both IIT and locational concentration. Agglomeration
In D, a higher degree of locational concentration could in principle come about for many reasons: big
plants, a high level of agglomeration of the input-output chain (proximity relations between firms in a more
fragmented division of labor), or a selective comparative-advantage based locational process. They would
lead to final output trade and to major asymmetries between exports and imports, and hence to the statistic of
low intra-industry trade. The industries found in this group by the direction-of-change analysis19 are mostly
labor-intensive, nondurables industries. The decline in intra-industry trade is most likely explained by the
increasing adequacy of local supply chains. But why is overall locational concentration rising, especially since
these are industries with considerable potential for product variety and hence for the existence of many locations?
The case study literature often emphasizes the "high end" versions of these industries, such as the famous cases
of Italian industrial districts, which are based on highly specialized local resources and relationships (Storper and
Salais, 1997). But locational concentration at a world scale could also be due or to comparative advantage in
location of mass production of standardized versions of products. This would be concretely manifested in
relocation of these industries to low-wage labor markets. The footwear industry might manifest both of them.
In both the EU and the OECD, it exhibits declining IIT but is becoming more locationally concentrated,
especially in the EU (see Figure 2, sector 324). Footwear produced in Europe is generally of high quality,
concentrated in industrial districts. In the OECD, the concentration tendency is weaker because lower-quality,
mass produced shoes are subject to economies of scale, on one hand, but there are many places which have the
environments appropriate to this kind of activity.
19 For the OECD: 314,323, 324, 356, 321, 331, 383; for the EU: 322, 323, 324, 332, 381.
For destination A, version A-1 of Figure 9 shows how lower locational concentration and higher
intra-industry trade could be the result of international cross-shipment between the intermediate goods complexes of
different countries. Each country, in other words, could specialize in some kind of intermediate output, and could
also have local sourcing relationships (thus, local proximity relations as well). This specialization could be due
to scale effects in intermediates which are freed up by market integration, or they could be due to comparative
advantages or geographically-differentiated technological-knowledge advantages. A second version (A-2)
tells a rather different story which could lie behind the same statistics. It pictures an industry where
essentially separate national industrial complexes are competing through international cross-market
penetration.20 Another version of this pattern is market-oriented foreign direct investment, where a large
foreign firm invests in another OECD country, while continuing its backward sourcing of intermediate inputs
from the country of origin. This could correspond to two rather different on-the-ground situations. One is the
now-common situation where globalization brings into a given market many more versions of very similar
products, i.e. greater horizontal product differentiation. Another is the greater quality-based differentiation of
products in a narrowly-defined final use category which comes about as more foreign firms enter a given
consumer market through exporting. In the case of A-3 production systems, smaller, competing firms or
complexes engage in the variety- or quality-based competition strategies alluded to above and generate
cross-market invasion within the industry (for both final and intermediate outputs), pushing up the intra-industry trade
ratio. In all three types of production system organization and geography, complex competitive strategies on the
part of pre-existing national complexes may prolong locational spread or even increase it in the face of
globalization. The results of the direction-of-change analysis for the OECD (Table 5) and the EU (Table 6)
conform well to these different scenarios. For the OECD, the industries moving toward category A (more IIT,
less locational concentration)21 are generally based on assembly or process production, capital-intensive in nature.
They produce non-specialized outputs at high scale. This result fits only partially with the predictions of the
New International Economics, because although internationalization might permit scale economies in
intermediate goods to flourish and hence to increase international IIT, the overall scale of markets seems to allow
scale thresholds to be reached many times over and hence to lead to diminution of overall locational
concentration (scenario A1). This result might also come about because of increasing international
cross-hauling (A2 or A3).
For category C, locational concentration could be encouraged by increases in scale (a big firm or some
big production units) or because of spatial proximity relations between many modestly-sized firms and
production units, or a combination of both – big firm with many small and medium-sized suppliers clustered
around it. These differences could be detected using the Ellison-Glaeser index, but it requires detailed plant-level
data which are not available here (Ellison and Glaeser, 1997). Moreover, in the case of significant
agglomerations of small-and medium-sized firms, a wide variety of potential causes could be at stake, ranging
from standard input-output linkage costs to soft factors such as technological spillovers or complex relational
content to inter-firm linkages; all are now subjects of major theoretical and empirical literatures.
the sense of trade between places in products within the same SIC category, and true intermediate (non-final) outputs.
The cases that we found in category C22 would seem to fit the first scenario better, as we depict in C-1.
This case is consistent with standard theory, where a big firm or a big production complex in one country is
sourced internationally, and in turn where its outputs, whether intermediate or final, are cross-shipped to these
sourcing countries, possibly to its own branches or to competitor or upstream or downstream firms. The sectors
found here are generally capital-intensive and scale-oriented in final output production, and probably have scale
increases in intermediate production. The C-2 diagram fits the pottery and china industry (361, EU and OECD)
and the transportation equipment industry (384, EU): the central firm or cluster of firms is geared to the
production of quality- and variety-differentiated outputs, but to do this, it relies on international sourcing. The
intra-industry trade that results is asymmetrical in the sense that it involves the locational centers receiving
variegated outputs, which are then recombined into a range of quality- and variety-differentiated final outputs.
Europe versus the OECD and the USA23
There are certain differences between the EU and the OECD in the direction-of-change analysis. For
example, the apparel (322) and furniture (332) industries are in category B for the OECD, but in D for the EU.
They are becoming less locationally concentrated for the OECD, but more so for the EU. This is probably due
to the one-time effect of European integration, but it cannot be said whether the integration is favoring scale
economies or some other competitive advantage of the winner locations. The wood products (331) and plastics
(356) industries are in group D for the OECD, but B for the EU. In other words, IIT is declining for both
(maturation of local supply chains in low-cost countries), but these sectors are dispersing in the EU and
concentrating in the OECD. This is likely due to the increased importance of scale-oriented production in the
OECD as a whole, and of variety in the EU, but detailed case studies would be necessary to confirm this
hypothesis. The transportation equipment (384) and non-ferrous metals (372) are in category A for the OECD,
but C for the EU. This may possibly be due to the effets of European integration, making themselves felt
through the retirement of old capacity in Europe, and the continuing effects of foreign direct investment and
development of international divisions of labor in the OECD as a whole, i.e. due to different starting points.
There is only one "diagonal" difference between the OECD and EU (i.e. all the others are consistent on IIT, and
only differ on locational concentration). The electrical machinery industry (383) shows declining IIT and
increasing locational concentration for the OECD (group D), but increasing IIT and decreasing locational
concentration for the EU (group A). We suspect that this is again due to ongoing foreign direct investment and
maturing local supply chains in the OECD as a whole, and the double effects of European integration (increasing
IIT) and some process that is permitting a wide variety of locations to flourish in Europe faced with integration,
but our data do not permit us to identify what this might be.
Levels of specialization and concentration in European countries are generally much lower than those of
regions within the United States (Kim, 1995; Commissariat Général du Plan, 1999). Although there are certain
23 In separate exercises, using similar but not identical approaches, there is strong overlap between the descriptions of locational behavior of industries in Europe which are reported here and in Midelfahrt-Narvik, Overman, Redding and Venables, 2000.
measurement problems in making this assertion, because US states are smaller units of observation than
European nations on average, research which corrects for these differences still agrees with the overall
conclusion. Krugman (1991) suggested that increasing European integration would lead to an "Americanization"
of Europe's economic geography. Our evidence does not lend support to that view, nor do the measures of
specialization carried out by Kim (1995) or Midelfahrt-Narvik et. al (2000). All show that the specialization
levels of American states are on average much higher than that of European countries, even though the
dispersion rates for individual sectors are also much more higher than in Europe. From the early 1970s to the
late 1990s, Europe had only slight, and statistically non-robust, increases in specialization and locational
concentration. Their results are broadly consistent with ours, though some of the magnitudes are different.
The European economies have different industrial histories from American regions. Whereas in the
United States, industrialization took place in an economy under constant, but integrated territorial expansion, in
Europe it occurred behind barriers of trade protection, language and state support. As a result, there are many
highly developed production complexes, consisting of firms with strong technological and organizational
competencies and a high level of institutional coordination, in different regions and countries. Their capacities to
adjust to globalization appear to permit the A-2 and A-3 locational patterns to persist, so that European regions
are not following the American pattern of more specialized regional economies and more locationally
concentrated industrial sectors. This result could be accounted for two ways suggested in the literature. On one
hand, Krugman and Venables (1996) simulate a process where transportation/transactions costs are neither
distribution of activities is rather homogeneous, as in Europe, then agglomeration economies are never
sufficiently strong to generate a new equilibrium characterized by a high level of polarization and specialization.
By contrast, if activity is initially strongly unevenly distributed, as in the USA, then further integration reinforces
polarization. Thus, not only do agglomeration economies count; so do starting points.
The second explanation shares their concern with starting points in reinforcing initial patterns, but the
mechanism of adjustment has to do more with the effects of institutions and strategies. Specifically, detailed
research on intra-industry trade in Europe, which analyses products according to their quality levels, has shown
that European firms are developing quality- and variety-based competition to a greater extent than in the OECD
as a whole (Fontagné et al, 1997; Commissariat Général du Plan, 1999). These strategies may permit firms and
agglomerations to follow location pattern A-2 (Figure 9), maintaining a high degree of international dispersion,
as reflected in the HE index. Crucially, it is highly probable that increased intra-industry trade in many
industries is based not simply on locational restructuring around scale-intensive intermediate goods, but on
quality or variety-based competition in these intermediate goods markets as well as in final outputs. This is the
distinction between "horizontal" and "vertical" intra-industry trade referred to earlier (Greenaway and Milner,
1984; see also Eaton and Kierzkowski, 1984); Jaskold-Gabszewicz et.al., 1981). In addition to this distinction,
it should be remembered that intra-industry trade refers both to international cross-hauling of final outputs in
contestable markets and to international production sharing via a division of labor in a filière (and both can be
divided into horizontal and vertical forms). A key dimension of such evolutionary dynamics is how mature
industries in Europe restructure themselves by using internationally-generated knowledge and how companies
develop and maintain capabilities that permit them to engage in variety-based competition (Eaton and Kortum,
1997, 1999; Langlois and Foss, 1997).
By contrast, in sectors not amenable to quality differentiation of outputs, international and even
sub-national concentration are more likely, as sub-national firms attempt to compete in the intersub-national environment by
maximizing economies of scale, while foreign direct (inward) investors concentrate at a small number of
locations to serve their new markets, also by maximizing scale economies. Though we cannot measure
sub-national locational concentration with this data set, other recent research in the EU confirms that many sectors
are becoming more locationally concentrated at the sub-national level, probably by adopting the C-1 location
pattern at the national level (Thisse, 1998; Veltz, 1996; Commissariat Général du Plan, 1999). Europe's
economic geography may exhibit very little aggregate locational concentration and specialization at the
inter-national level, but it appears to be going through a major wave of metropolitan concentration at the sub-inter-national
level.
Thus, the interpretation advanced here is consistent with the notion found in the literature that the
evolutionary dynamics of existing production complexes may strongly influence what happens after market
opening, and not simply convergence toward a universal global optimum (Dosi, Pavitt and Soete, 1990; Fujita,
Krugman and Venables, 1999).
Caution is necessary here. Even though these data represent a relatively long time period, we do not
know where they stand relative to the current process of trade growth and global locational change. Do the
rather that EU integration is just beginning, with the real changes convergence with the OECD and the USA
--yet to come? More sophisticated work, which can link the observed national and sectoral trends to causal
models of location, is needed to make more sense of these descriptions.
Many Questions Remain Unanswered
There are many empirical and theoretical issues left open by a statistical exercise at this geographical
scale. Our data do not permit an examination of plant-level versus place-level concentration of industry. It
would be essential, in examining further the emerging patterns of geographical concentration, to know the extent
to which they are generated by bigger and fewer plants as opposed to clustering of firms. This would require
construction of international data similar to those used by Ellison and Glaeser (1997) for the USA, and this
would have to be done at a greater level of sectoral disaggregation than they were able to do.
Moreover, as noted, because our data use national territories as their geographical units, we cannot know
to what extent the national patterns – which indicate moderate concentration at the OECD level and a mixed
story at the European level – mask regional (subnational) spread or concentration within countries. As noted,
some authors claim that European economies are going through a wave of metropolitan concentration of industry,
to the detriment of a more even intra-national spread of production (Rodriguez-Pose, 1998; Veltz, 1996; Thisse,
1998). If this is the case, our finding here, that Europe is not going through a major wave of territorial
consolidation of manufacturing would be an indirect result of each country's capacity for building metropolitan
industrial systems, which in turn lie behind their stable shares of production in each sector. This leads to the
notion of a network of metropolitan agglomerations as the key spatial architecture of the global economy, and
the questions of geographical evenness and unevenness and specialization are, in effect, redirected to the
subnational scale (Veltz, 1996). More detailed data are needed to resolve these issues.
Since the data used here concern only the OECD countries, i.e. advanced economies, it is also possible
that they mask wider patterns of spread or concentration and intra- versus inter-industry trade. For example, if
routine production is being located outside the OECD in ever-increasing proportions (or in the lower-cost OECD
countries) then it might be generating spread tendencies on a world scale; but then again, market integration
might also be leading a global consolidation of production, as old national development policies are dismantled
and production systems in different countries are integrated into world wide commodity chains. Furthermore,
this analysis concerns only manufacturing (and a subset of it, at that) and not the major locus of employment and
output in advanced economies, which is services. Though many services are untradeable or less tradeable than
goods production, they are undergoing major changes as a consequence of market integration, and future work
should certainly be extended to them. Finally, there are many also many unanswered questions about the
observed tendencies toward convergence and divergence of incomes at different geographical scales and the
patterns of location, specialization and trade which might underlie them. All of this suggests the need for
patterns of location, specialization and trade which might underlie them. All of this suggests the need for