Abstract: The proportion of traded goods in world output has been rising steadily over the past several decades. When we look at specific products exported by the advanced industrial nations, increasing export specialization is evident. Such specialization cannot be explained by conventional notions of comparative advantage, nor entirely by the new trade theory based on economies of scale. Rather, a significant proportion must be due to technological or “absolute” advantages on the part of the specialized exporter, and a significant dimension of technological advantage is product-based and renewed through learning, giving rise to dynamic economies of variety as a source of export specialization. Industries characterized by such product-based learning and absolute advantage tend to have important developmental effects on their host economies because they earn quasi-rents. Such industries also tend to be organized into production networks combining the advantages of specialization and flexibility, which are key to technological learning. These export-oriented absolute advantage industries tend to be found in one or a few subnational regions of their host countries. In this way, the global economy may be thought of as consisting, in important part, of a series of “technology districts.” Unlocking the organizational secrets of technological learning in these places is now a key task for understanding the dynamics both of these localities and of the global economy as a whole. I give examples from studies in France, Italy, and the U.S.

Index

Six Propositions on Trade, Flexibility, Technology, and Regional Development

Two images dominate our thinking about the changes sweeping across the world of industry over the last two decades. First, the forms of production organization that characterized the most dynamic industries of the postwar period in the advanced economies, i.e., mass production in the consumer durables sectors and their associated capital goods, are no longer as significant for economic growth, change, and capital accumulation as once they were. The social science literatures are replete with accounts of the restructuring of the Chandlerian-Galbraithian firm, the spread of programmable technologies, the shortening of product cycles, the deepening of contracting

* This paper is based on research carried out while the author was a Fellow of the German Marshall Fund, and the support of the Fund is gratefully acknowledged. Additional support came from the International Studies and Overseas Programs of UCLA and the Academic Senate of the University of California, Los Angeles Division. The paper draws on work from a book in progress. I would like to thank Andrew Sayer for his comments on an earlier version; Eric Sheppard for his detailed comments and criticisms; the participants at the Lemnos Seminar and at the University of Lille; and Erica Schoenberger for her helpful suggestions.

and subcontracting relations, the revival of small and medium sized units of production, and an emphasis on quality as much as price in competition. Second, firms and production systems have patterns of input sourcing (capital, labor, intermediate goods, and technology) and marketing that now are planetary in scale.

In certain ways, these images are not mutually consistent. An extensive case study literature documents increased organizational decentralization and heightened competition, united under the rubric of production flexibility. This same literature repeatedly calls attention to a certain geographical reconcentration of production - what some have called the “resurgence of regional economies.” ¹ At the same time, our minds are flooded by impressions of global firms and production systems no longer identifiable with any particular region or nation and of a world economy governed by a few globe-girdling firms. Couched in these terms, moreover, such images encourage different, sometimes incompatible, priorities for economic and regional policy.

The images may, nonetheless, be refined to capture two interdependent aspects of contemporary industrial reality. I argue in this paper that the leading edges of economic activity are highly identified with production systems that are flexible in the specific, narrow sense of being organized to carry out continuous product innovation. Production systems engaged in such “product based technological learning” (henceforth PBTL) account for important and increasing percentages of world exports; they are an essential element in the globalization of economic relations. Yet the key parts of such PBTL industries tend to be highly concentrated in distinctive subnational regions, in what I call “technology districts.”

What follows should be regarded as a series of hypotheses advanced on the basis of existing case study literatures and theoretical work in several closely related fields, but not yet supported by sufficient quantitative or qualitative information to accept or reject with certainty. I employ an explanatory architecture to make my case that systematically draws together four important themes in contemporary economic-industrial research: the organization of production systems; technological change; the emerging system of global trade flows; and the relationship between specialized production regions and the global economy. The argument can be summarized in six propositions, detailed in the main body of the paper:

(1) Trade specialization is evident when we look at the specific products and clusters of cognate products exported by the advanced industrial nations. Specialization in a product exists when a nation’s share of world trade in it is greater than the nation’s overall share in world trade. Each country has relatively few such products, and dynamic approaches show that, in terms of export specializations, the industrial structures of these countries are not converging.

(2) There are a number of reasons for trade specialization. One is that some product markets or parts of production systems are not amenable to standardization or routinization. In such cases, even the existence of global markets does not call forth a single, uniform best practice. Rather, technological dynamism in products, based on continuous “learning,” becomes a necessary part of certain forms of “best practice,” and firms that generate this practice (i.e., products) effectively create their own markets. PBTL is a sufficient condition to generate trade specialization in cases of absolute (technological) scarcity of the product. PBTL is necessary but not sufficient in other instances, where price considerations also play a significant role in determining the pattern of sales.

By contrast, countries not specializing in international trade in a product frequently have laggard versions of the sector, i.e., products that are not state-of-the-art in quality terms.

1. See Sabel (1989) for a comprehensive review of this literature.

It is likely that an increasing proportion of world exports rests on PBTL. For single countries, the proportion is highly variable, but it should be higher for those countries with large export surpluses or rapid per capita income growth. The latter occurs because technological scarcity generates quasi-rents and thus has more favorable developmental outcomes for the exporter than do other forms of export specialization.

(3) In such PBTL systems (whether in sectors where product technology is highly uncertain, such as the high technology industries, or in dynamic versions of a sector characterized by relatively “mature” technologies), vertical integration of the production system tends to be inhibited by firms’ need to avoid “lock-in” to a given technology; as a result, technologically dynamic industries typically are organized in the form of production networks, based on elaborate shifting divisions of labor, whether between firms or between units of a single organization. Elaborate and shifting networks are a principal organizational means of achieving the flexibility that forms the corollary of PBTL.

(4) Networks also may exist in industries not characterized by technological dynamism, particularly where simple market uncertainty encourages vertical disintegration; thus, not all networks are examples of PBTL-based international specialization industrial complexes.

(5) The existence of networks and the geographical agglomeration of important parts of the production system are positively related in technologically dynamic industries. In the case of traded, PBTL industries in France, Italy, and the United States, I observe that key collections of physical capital, labor, and information resources are almost always highly geographically concentrated in one, or a few, subnational regions of their respective leader countries. Key parts of today’s global economy may thus be thought of as a mosaic of these dynamic regions, or what I call “technology districts.”

(6) The technological learning that characterizes these agglomerations rests, to an important degree, on the conventions of the regional production system. Conventions are rules mobilizing resources and maintaining them in situations of mutual engagement. The most important dimensions of resource mobilization and engagement in PBTL systems are the functions and qualities of different sorts of transactions among firms, between firms and their labor markets, and between users and producers. PBTL is also dependent on the ways in which conventions guide and differentiate the mobilization of specific factor supplies in the form of public goods. In all, conventions organize the specific transactional structure in terms of: (a) patterns of resource mobilization; (b) forms of collective order in labor markets and interfirm relations; (c) definitions of product quality; (d) relations between key innovating groups and other groups in the production system; (e) roles of regional and local third parties in harmonizing preference structures; and (f) ideologies and cultures of local economic actors. The PBTL system’s evolutionary path is therefore deeply rooted in its surrounding political economy. In other words, although the PBTL is effective because it conforms to certain conditions prevailing in the global economy, in many ways the specific forms of PBTL and the specific roots of PBTL systems can be accounted for only by recourse to territorially based and noneconomic forces, especially the conventions that structure the participation of agents in PBTL production complexes. Stated differently, territorially bounded conventions define the qualitative basis of the external economies of PBTL systems. The conventions of these networks are markedly different from those of their laggard counterparts.

Let us now look at these propositions in greater depth, drawing on examples from research on flexible production systems and trade in the United States, France, and Italy.

Export Specialization and Technological Dynamism

The two most statistically striking changes in the advanced economies over the last two decades are (1) the decline of sectors rooted in the “second industrial revolution” (the metalworking and mechanical sectors) and their replacement as motors of growth by commodity chains centered on the “third revolution,” in electronics and certain services, and (2) the internationalization of markets. Between 1967 and 1986, for example, gross output of the world’s economies increased by a factor of 6.5 (to 15,730 billion 1986 dollars), while trade multiplied by a factor of 10 (attaining $3,127 billion). The share of trade thus increased from 12 percent to 20 percent.

The rise in trade is manifest in an extraordinary variety of sectors. Trade in goods (except energy) increased from 8 percent to 12 percent of world output and trade in services from 1.5 percent to 2.5 percent (Lafay and Herzog 1989). Among the goods-producing sectors, the increase is surprisingly widespread. Those filières (commodity chains) having stable or increasing shares of world trade include metalworking, machinery, vehicles, electronics, chemicals, textiles, woodworking, and paper. Those declining significantly in trade shares include non-ferrous metals, electrical machinery, steel, and food processing.

The Increase in Trade Specialization

The dramatic rise in international trade is also associated with widespread increases in (inter- and intraindustry) export specialization by the world’s advanced industrial economies. In a world where trade grows more rapidly than production, mastery of tradeable production offers the greatest growth multiplier to producer firms, regions, and nations.

The analyst needs a high level of commodity disaggregation to appreciate this specialization. A trend toward convergence in broad patterns of production and exporting of different countries has been well identified at the broad (equivalent of 2- or 3-digit) industry level. Against this, however, intraindustry product specialization and trade have increased as a result of specialization of exchanges within the same broad sectors (Gerstenberger 1990; Greenaway and Milner 1986). This is a consequence of complex intermediate inputs in some industries and of product differentiation in others. As a result, when trade is examined at the level of 5-digit SITC products, the similarity of export vectors of the main industrial countries has decreased steadily since 1978 (Gerstenberger 1990).

Trade and Product-Based Technological Learning

A large and growing share of world trade consists of exchanges unexplainable by traditional theories of comparative advantage (Tyson 1987). Drawing on recent theoretical work on technological innovation, industrial organization, and trade, a coherent logic can be put together for why certain products are made in relatively few places.

Three general explanations exist for the persistence of trade: factor cost differentials, economies of scale, and technological mastery or scarcity. In the standard trade theory, comparative advantage emerges either from the range of natural resources present in a given place or from reductions in relative costs of production owing to the choice of activities best adapted to the local factor mix. Yet when a product is subject to the conditions of perfect competition on which the standard Ricardian theory rests, there are ample reason for expecting that, in today’s world, the proportion of its output that is traded will steadily diminish. This result, precisely the opposite of that predicted by the standard theory, emerges because the world of production has changed fundamentally since the time of Ricardo. We now live in a world where factors of production for technologically stable products are

not endowed, but produced as intermediate inputs. Almost any developed country making the effort can become as efficient as the next country in a technologically stable manufacturing sector. As Vernon (1979) intimated, with respect to the efficiency of production processes, the diffusion time for standardized technologies has progressively diminished, because of the domination of mass production by global corporations whose networks have spread and deepened over the past few decades and because of interfirm and intercountry technological imitation. For example, as Amendola, Guerrieri, and Padoan (1991) show, technological variation among major countries tends to diminish steadily after major technological revolutions. The result is convergence in the costs of production of different countries in markets for standardized goods. Put another way, the hardware applied to a given production task in the industries of the major industrial countries is increasingly similar. Yet, as Pavitt and Patel (1990) also demonstrate, major (multiproduct, international) firms are consistently more diversified in their possession of production technologies than in their product ranges. As a result, even in the recent period (characterized by the technological scarcities generated by the microelectronics revolution), the coefficient of trade specialization of the major industrial countries has increased much more than the coefficient of technology specialization. ²Moreover, Amendola et al. (1991) show that cumulative mechanisms appear to drive the change of technological profiles.

In contrast, the “new trade theory” argues that trade patterns are largely the consequence of economies of scale in production and therefore that many markets are imperfectly competitive (Krugman 1990). These are not the standard, static economies of scale of conventional theory. Intraindustry trade or subindustry trade is now the rule, owing to the increasing complexity, specialization, and differentiation of many products. Economies of scale are thus a key factor, while even in the biggest industrial economies and at a global level, many industrial markets have seen their levels of product differentiation rise in recent years, so that the global economy as a whole is now characterized by much greater product variety than previously. The new trade theory thus shows that many broad industrial markets have become more contestable since the end of the 1960s, even in the face of similar factor prices, rapid technology diffusion, and imperfect competition. Indeed, in spite of the fact that in many sectors and countries domestic production is concentrated in fewer hands than before, world market concentration has tended to fall.

While this explanation is powerful and widely applicable, it is nonetheless incomplete in certain respects. First, minimal optimal scale economies in the production of many goods - especially intermediate industrial inputs and certain design- or fashion-based goods - are believed to be stable or dropping, owing to the increasing flexibility (programmability in particular) of capital equipment (Coriat 1990; Gold 1981; Dertouzos et al. 1989). Trade is nonetheless rising in a number of these markets, as in automobile parts. Second, trade appears to be rising even in product markets where economies of scale have traditionally been quite low (most evidently in traditional industries such as clothing and textiles). Third, there is a critical ambiguity with respect to the significance of economies of scale in small, open economies. Take the case where a nation’s

2. There were, and are, some obvious exceptions to this reasoning. The role of natural resources remains critical in the energy, nonferrous metals, and possibly the wood-paper sectors, whereas for the labor-intensive textile filiere, the availability and price of labor are likely to be determinant for certain products. In the others, these classical Ricardian factors have become increasingly irrelevant.

tastes demand large quantities and specific qualities of a certain good and where that nation’s economy is small and open (e.g., Belgium’s degree of openness is 45 percent, but that of the U.S. about 10 percent). Belgium might produce a certain kind of beer for its own specialized tastes, but if minimal optimal scale economies are greater than domestic demand, it might then export some production. If exports in that sector make up a very small share of world trade, the suggestion is that trade specialization has to do with economies of scale. But if Belgium exports a relatively high share of the world’s production of certain kinds of beer, trade specialization cannot be attributed to mere economies of scale. Even though the need to export to attain economies of scale might have “pushed” the Belgians to export in the first place, world market demand now “pulls” production along because of the high quality of the Belgian product. We are thus in a realm where the mere existence of scale economies often explains little about why trade specialization comes into existence or persists. Fourth, where a strong export position is due to scale economies, its developmental consequences for the area of origin will be strongly influenced by the threat of imitation from technological convergence, i.e., steep ongoing price-cost reductions and heavy capital requirements (or subsidies in the case of agriculture).

In markets not characterized by high barriers to entry, or in those where barriers to entry are present but export specialization is stable or increasing, something must make advantage possible in the face of rapid imitation (catch-up) by potential competitors. Many analysts now suggest that this force is technological mastery, which gives rise to “absolute advantages.” The latter are different from comparative advantages in the sense that they reflect productivity gaps between countries or regions that are so great that they are insensitive to almost any set of changes in input prices (Dosi, Pavitt, and Soete 1990). Such absolute advantages might be due either to superior process technology or to product technology. The latter, however, appears particularly important for increases in product variety over time, i.e., increases in the range and speed of product innovation such that, at any given moment, certain centers of scarce skills exist that promote trade even in the presence of easily attainable minimal scale economies. “Economies of variety over time” means the ability of a production system to turn out a changing array of outputs in the general product field in which it is specialized so as to outrun the catch-up effect from ever more rapid imitation and convergent productivity. This capability can be accomplished with or without static economies of scope-variety (i.e., making many products at the same time).

Any viable explanation of the persistence of trade specializations through these dynamic economies of product variety over time must therefore identify the conditions that allow certain industries in certain places to build up advantages that, over and over again, permit them to turn out products not easily or rapidly imitated. In other words, in contrast to our observation above that standardized or generic technologies are increasingly rapidly diffused in today’s world, other cases experience the powerful counterforce of “destandardization” and respecialization, redefining the competitive process in terms of technological advantage rather than mere costs or scale. “Technology,” in this context, means the products, processes, capital, and knowledge applied to productive activity. Technological advantage exists when the actors in a given place possess products, processes, and attendant knowledge that permit them to produce better things than other places, or simply permit them to produce goods and services not elsewhere available.

What is behind such destandardization-respecialization? I argue for an endogenous form of dynamic technological advantage that I call “product-based technological learning” – PBTL - by which I

mean the ability continuously to reinvent, differentiate, improve, and reconfigure products through a dynamic redeployment of key, specialized production skills and equipment. “Learning” is a metaphorical term for activities that generate dynamic economies of variety. Examples include Silicon Valley’s mastery of specialized semiconductors, Italian domination of men’s clothing, lamps, and ceramic tiles, Hollywood’s continuous innovation in motion pictures, London’s ongoing stream of innovations in financial services, France’s position in women’s clothing, and Germany’s strength in machine tools. Some part of the contemporary increase in export specialization is thus likely associated with technological dynamism in products; successful specializers are masters of product innovations that cannot be immediately or easily imitated.

An Historical Divide?

The existing technological innovation and evolutionary economics literature is overwhelmingly oriented toward the engineering- and science-intensive industries, i.e., the high technology sectors and their principal oligopolistic firms. If learning-based dynamic economies of variety were limited to these high technology oligopolists, it would be easy to claim that the contemporary surge in export specialization is a temporary phenomenon related to the electronics revolution and its application to other sectors. In this “infant industry” perspective, by destandardizing knowledge and production processes for a short time, microelectronics permits a few countries that master state-of-the-art base technologies and applications to gain world market shares. But it also predicts that the transition from the postwar mass production economy to a new system will be achieved when standardization, imitation, and convergence of industrial structures return. In other words, learning is a one-time advantage for the first movers in the microelectronics age. This view converges nicely with standard comparative advantage theory, of course, since advantage in the form of imperfect competition is now just a temporary deviation from the course of perfect competition. The problem with this view is that world trade is rising even in traditional industries where the application of microelectronics does not, strictly speaking, revolutionize production (see Lafay and Herzog 1989, 62). Trade is rising across a remarkable variety of industries, and it is declining (in terms of share of world trade) only in the non-energy resource and agricultural sectors and the most traditional raw materials processing sectors, such as steel.

Leontief (1953) prefigured contemporary developments by identifying the paradox of “contrary factor intensities,” such that rich countries were specialized in labor-intensive products and poor countries in capital-intensive outputs. He concluded that skill- and knowledge-intensity was a principal determinant of output and trade patterns. Vernon (1966) followed with the product cycle concept, attempting to show why, in organizational terms, the technological and skill cores of an economy were reinvigorated through product pioneering. Vernon (1979) later updated the product cycle model in an international context, pointing to ever more rapid international catch-up effects - both between rich and poor countries, and among rich nations - through technological imitation and diffusion. He suggested that multinational enterprises could only continue to dominate world markets through international location and technology licensing, or what we now recognize as “triad” locational strategies in the world’s three major markets. Many supposedly “skill and knowledge intensive” products are now very widely diffused (e.g., chemicals).

Neither Leontief nor Vernon, however, could fully anticipate today’s central dynamic: the changing definition of skill and knowledge intensity and its effects on the organizational and locational dynamics of industry. Skill- and knowledge-intensity is no longer as securely tied, in a given sector, to the internal resources of one or

a few major firms as it was when Leontief and Vernon wrote, when technological advantage, once achieved, was effectively embedded in internal economies of scale and stabilized for a certain period of time. Skill- and knowledge-based competition is now more dependent than before on resources that lie outside major firms (and hence are not fully appropriable by them). Partly as a result of the latter, such competition now takes a different temporal form than it did 30 years ago; instead of being intermittent and creating medium-term advantages, it has become more continuous, with much briefer temporal advantages. The reduction of temporal advantages is not now due only to rapid imitation, in other words, but to real skill- and knowledge-based competition. The evidence is thus consistent with the view that we have entered a period in the development of capitalism in which trade and specialization are not enjoying temporary surges, but are outcomes of the technological dynamism of specializers, which has become a broad, endogenous characteristic of the logic of best practice. Indeed, the contemporary microelectronics revolution, rather than destandardizing a fixed array of outputs for only a brief time, appears to have made possible a continual and ongoing respecialization and redefinition of outputs. This ongoing differentiation of products has the effect of unsettling the division of labor on an ongoing basis, much in the way described by Young’s (1928) seminal discussion. Hence, the respecialization effect dwarfs the catch-up effect associated with standardization of products, leading to what Freeman (1991, 37) suspects might be “a permanent shift in industrial structure and behavior.”

Developmental Effects: PBTL versus the Rest

Of course, PBTL is not the only form technological change can assume; as evolutionary macromodels of technical change have shown, rapidly changing average production or product technologies may be due to imitation or selection, as well as innovation (Dosi et al. 1988; Webber, Sheppard, and Rigby 1991). But the trade and market share advantages enjoyed by more rapid imitators are likely to be shorter lived than those of producers engaging in PBTL because, by definition, imitators are subject to rapid competition from elsewhere, i.e., to a catch-up process already in motion.

Moreover, the developmental consequences of imitation, while important to maintaining shares of trade and employment by raising average productivity to world standards, are quite different from the developmental potentials associated with PBTL. In the case of imitation without significant scale effects, competition must necessarily be cost-price based, such that the pace of capital accumulation is limited. Where this is achieved through technology-intensiveness, capital requirements will be high and employment effects limited; where it is achieved through the use of cheap factors of production, especially labor, wages will be limited. In the case of imitation with significant scale effects, capital requirements are also necessarily high, and employment creation effects must be limited; moreover, technological catch-up makes it probable that advantages will be short-lived unless large quantities of capital are pumped into the production system to outrun imitators.

The PBTL system’s products, by contrast, tend to earn supernormal returns (quasi-rents). The PBTL-based specialized exporter, in other words, creates the de facto definition of world “best practice” by inventing or perfecting products. There is a powerful motivation to engage in PBTL, for not only does it offer the possibility of export-led growth, but the qualities of that growth are more attractive than for the other two cases. Quasi-rents (from a high price-cost margin) can be distributed in the form not only of high returns to capital and further physical capital accumulation through reinvestment, but also in the form of high-wage jobs in the labor markets where technologically dynamic

industries are present. Every major industrial country will necessarily have some export specializations for each of these three reasons. The aggregate developmental outcomes of export specialization thus depend on the particular mix for each nation. All other things being equal, however, the wage- and capital-accumulation effects of PBTL-based exports are more attractive than those of cost- or scale-based exports.

Index

List of Tables

Table 1: Trade Composition and Trade Ratios for Main Industrialized Countries by Typology of Industrial Sectors

Table 2: Top Fifty U.S. Industries Ranked in Terms of World Export Share, 1985

Table 3: Top Fifty Italian Industries Ranked in Terms of World Export Share, 1985

Table 4: Top Fifty French Industries Ranked in Terms of World Export Share, 1985

Table 5A: The Roots of Export Specialization: PBTL vs. the Rest, United States

Table 5B: The Roots of Export Specialization: Learning/Economics of Variety vs. the Rest, Italy

Table 5C: The Roots of Export Specialization: Learning/Economics of Variety vs. the Rest, France

Table 5D: Totals for Three Countries

Table 6: The Degree of Country Specialization in HTO, DIC, and PMM Industries, 1985

Table 7: Technological Districts in the United States

Index

next page

AAP Homepage