CLUSTER THEORIES (1): SPATIAL EXTERNALITIES
A recent New Yorker magazine piece on the bunching of new restaurants in a particular block in Brooklyn invokes Alfred Marshall’s century-old theory of industrial districts. (Marshall 1890, 1920, pp. 267-277.) Whether for restaurants today or for Manchester’s cotton mills in the 19th century, the same processes can be seen.
Marshall’s basic point about why companies in the same industry congregate still holds: industrial districts enjoy the same economies of scale that only giant companies normally get. Specialized suppliers arrive. Skilled workers know where to come to ply their trade. And everyone involved benefits from the spillovers of specialized knowledge. As Marshall put it, “The mysteries of the trade become no mysteries, but are as it were in the air.” (James Surowiecki 2000, p. 68, emphasis added.)
Four Advantages of Localization
Something “in the air,” the first of the four ingredients in Marshall’s district theory, can be construed to refer to knowledge spillovers (as in conversations between people working for different companies). A second ingredient is the common pool of the factors of production the local cluster uses, especially labor. The greater their supply, the lower the costs for each firm in the industry. Third, as inputs become more specialized, they tend to become more productive, a la Adam Smith’s legendary pin factory.
On these three supply-side counts each firm in the area will have lower costs than if it operated in isolation. But in addition, and on the demand side, comparison shopping becomes easier for the customer when the retail firms are bunched—as in the Brooklyn restaurants.
All this is above and beyond any savings in transportation costs between suppliers and buyers. When we combine the two sets of forces, Marshall’s spatial externalities and the savings on transportation costs when activities are concentrated, the result can be combined under the heading of agglomeration economies. (See Chapter 5 of Hoover and Giarratani in this text.)
Localization in Marshallian industrial districts happened on both side of the Atlantic in the late 19th century. In a masterful essay on Marshall, Paul Krugman points out that a special 1900 Census monograph, “The Localization of Industries,” identified 15 specific concentrations in the U.S., among them
Collars and cuffs, localized in Troy, New York; leather gloves, localized in the two neighboring New York towns of Gloversville (sic) and Johnstown; shoes, in several cities in the northeastern part of Massachusetts; silk goods, in Paterson, New Jersey; jewelry, in and around Providence, Rhode Island; and agricultural machinery, in Chicago. (Krugman 1991, p. 61.)
He observes that in seemingly every example, localization resulted from a combination of an initial accident in a particular location, after which “cumulative processes took over.” (Technically, by the way, this is the non-neoclassical language of positive feedback and path-dependency.)
A Neoclassical Update
A good update on spatial externalities comes from Mills (1992). A skeptic on labor pooling, he lists three other families of causation, which I paraphrase here:
(1) Input-output clustering. With or without economies of scale, if firm A’s input is used to produce firm B’s output and is costly to ship, cost savings can be attained by proximity, and it would be natural to find both A and B locating in the same metropolis. Whether labor can be viewed in this context is open to question. Natural resources (A) and resource-processing (B) qualify. Consider climate, as illustrated by aircraft production in places with mild climates. In contrast, oil and gas are cheap to transport, so processing can occur far from oil and gas fields.
(2) Comparison shopping clusters. In New York’s high-fashion garment district, buyers converge from all over the country to examine the wares and negotiate face-to-face over a whole range of price, design, and production details. While it might seem that such communication could be done via information technology, something about the cluster makes it advantageous for buyers to come in person.
(3) Information clusters. Firms that deal with similar kinds of information may cut costs or maximize their access to information by locating in the same metropolis. This arises when information has public-good attributes (with zero marginal costs) or is a joint product linked to supplier-customer communications. Another set of issues concern the distinctions between (a) quantitative vs. qualitative and, more cogently, (b) ambiguous vs. unambiguous information.
Ambiguous information in Mills’ view is information that requires face-to-face access and hence contributes to clustering. It is any information that must, in effect, be “negotiated” to establish its meaning.
This point is fundamental in that it establishes a continuing need for face-to-face access, despite the coming of the Internet.
The Difference a Century Makes (Not Much)
Now, just for the record, let’s go back and compare Marshall and Mills on two of these three advantages of localization. Look what Marshall had to say about comparison shopping, an issue virtually never (except in Mills) cited in the voluminous literature:
So far we have discussed localization from the point of view of the economy of production. But there is also the convenience of the customer to be considered. He will go to the nearest shop for a trifling purchase; but for an important purchase he will take the trouble of visiting any part of the town where he knows that there are specialty good shops for his purpose. Consequently shops which deal in expensive and choice objects tend to congregate together; and those which supply ordinary domestic needs do not. (Marshall 1920, p. 273.)
And what about conversations? For Mills, face-to-face access is a must when the information to be communicated is ambiguous, a matter of calibration or negotiation. What kinds of information did Marshall think benefited from proximity? “When an industry has thus chosen a locality for itself,” Marshall writes,
…The mysteries of the trade become no mysteries; but are as it were in the air, and children learn many of them unconsciously. Good work is rightly appreciated, inventions and improvements…have their merits promptly discussed: if one man starts a new idea, it is taken up by others and combined with suggestions of their own; and thus it becomes the source of further new ideas. (Marshall, p. 271.)
The next point to be noted concerns the uncertain role of trust between competing firms as a factor conducive to spatial clustering.
In The Second Industrial Divide, Piore and Sabel (1984) highlighted "flexible specialization" as the developmental stage succeeding "Fordism" or mass production. They emphasized the virtues of the "Third Italy" and its industrial clusters specializing in high-fashion, design-intensive goods. They saw virtuous networks emerging among rival firms, which manage to cooperate around activities of mutual benefit such as training, marketing, and research.
By the same token, Bennett Harrison concluded that the difference between Marshall’s model and this Italian version was trust—an outgrowth of experience. As he put it, this interpretation of the new wellsprings of regional growth in Italy proceeded “from proximity to experience to trust to collaboration to enhanced regional growth."
The Italian studies inspired scholars to look for parallels in clusters in other countries. A recent study of Danish textiles, for example, finds another industry cluster marked by trust and cooperation. (Sven Ileris 2000.) In Europe, it seems, trust based on experience may permit rival firms in the same industry to cooperate on certain common fronts.
In the U.S., however, the parallels have proved elusive. Did Silicon Valley measure up to Harrison's criterion? Alas, no. The spectacle of Valley firms such as Intel suing each other at every opportunity (usually over intellectual property rights) rules this particular cluster out as an example of post-neoclassical trust and harmony. In Harrison’s view, firms that rely on each other through informal agreements and cumulative collaboration do not wind up in court.
Similarly, a companion chapter in this textbook points out that clusters in The Third Italy may be culture- or country-specific. As Bergman and Feser observe, trust is relatively easy among firms and their suppliers. A much rarer phenomenon—except in Italy—is trust between competitors. On this and other counts Bergman and Feser designate the Italian clusters as a distinctive type, which they dub Italianate.
A controversy in the cluster literature concerns the roles of localization economies (a la Marshall) and the more diverse urbanization economies that come with large city size, often identified with Jane Jacobs (1969, 1984).
For example, Mills’ 1992 account outlined above had been prompted in part by the findings of Henderson (1988) and O hUallachain and Satterthwaite (1988) “that localization economies are more important than urbanization economies. That means that growth of employment within a sector tends to depend more on the size of the sector than on the size of the metropolitan area” (Mills 1992, p. 3).
As it happened, an influential journal article by Glaeser et al. that same year (1992) reached an opposite conclusion. The authors found that “industries grow slower in cities in which they are heavily over-represented” (p. 1129). These findings tended to cast doubt on localization economies as a source of growth. By implication, they also supported Jacobs’ arguments for diversified cities as seedbeds for innovation and rapid growth.
The resolution of this issue can be found in a product-cycle or industry-maturity perspective. As a recent World Development Report puts it,
Whether an industry benefits most from urbanization or localization economies depends on how innovative it is. New, dynamic industries are likely to locate in large urban centers where they can benefit from the cross-fertilization provided by diverse actors. Older, mature industries concentrate in smaller, more specialized cities, where congestion costs are low and localization economies can be high. (World Bank 2000, p. 117.)
When an industry matures, in other words, its initial centers become vulnerable to competition from newcomers and specialization becomes not an advantage but a burden.
Porter: Rivalry and Information Flows
By the same token, however, innovation can maintain a specialized cluster’s competitive advantage. This is a useful light in which to view Porter’s influential version of cluster theory, as introduced in The Competitive Advantage of Nations (1990) and refined in two more recent articles. (Porter 1998, 2000.) The approach derives from management theory, but it retains a neoclassical emphasis on competition and rivalry rather than trust.
As Porter puts it, "Successful firms are frequently concentrated in particular cities or states with a nation" (1990, p. 29). Offering detailed maps of the Italian and German cluster landscapes (pp. 155-156), he cites these additional examples as well:
British auctioneers are all within a few blocks in London. Basel is the home base for all three Swiss pharmaceutical giants. Danish windmill producers are centered in Herning. In America, many leading advertising agencies are concentrated on Madison Avenue in New York City. Large-scale computer manufacturers Control Data, Cray Research, Burroughs (now part of Unisys) and Honeywell all are headquartered in or near Minneapolis, Minnesota. Pharmaceutical and related companies…are based in the New Jersey/Philadelphia area. (Porter 1990, p.155.)
Porter explains such agglomerations in terms of his "diamond." Its four corners are demand, factor conditions, rivalry-strategy, and industry clusters. That is, an industry becomes internationally competitive because of
· favorable home conditions in the markets it sells to,
· the quality of its factor inputs,
· the competitive pressures encouraging excellence within its industry, and
· the supplier and customer linkages specific to the industry, which in practice are often traced out within specific urban agglomerations.
On this last point, a tight geographical locale intensifies (1) information flows and (2) rivalry between competing firms. Local media, banks, universities, bars, and eateries enhance communication. The spatial proximity of rivals—whether they communicate and cooperate or not—spurs competition and innovation. Trust between rivals does not enter into it.
In the spirit of Porter’s emphasis on information flows, financial services in London and New York City provide convincing evidence of the continuing value of face-to-face communication.
In a May 1998 article, for example Ben Edwards assessed London's staying-power as a financial center. ("Capitals of Capital: Financial Centres Survey," The Economist, 347:8067, p. 8.) Some of his analysis smacks of Marshall's writings a century ago on the industries of northern England, highlighting the role of specialized skills. "Developing financial markets requires a wide range of talents, and clusters make it easier to co-ordinate them. Lawyers must ensure... Accountants must check.... As long as these people prefer to meet in person to co-ordinate their work, there will be a need for financial centres."
Even more to the point: "But even if business must be done centrally, why do it next to your competitor?" (Edwards, p. 8.)
The answer to this question, a matter of localization economies, not urbanization economies, draws on Porter's theme of information flows. Being near your competitors and mutual suppliers enhances your knowledge of their operations, a prod to innovation on your part. It also permits raids on their employees, who may have just the skills you are looking for. ("So, in New York, Wall Street investment banks routinely poach credit analysts from their rating-agency neighbors, Standard & Poor's and Moody's….")
Two distinct logics of location are emerging in London and other financial centers. One image is hub-and-spoke. The image refers not to the relationships between large and small firms, but to the location of activities within a firm. "Hub businesses are centralised: strategic planning, project management, product development, and risk-taking activities such as trading and cooking up exotic financial derivatives. Spoke activities—such as sales, marketing and company analysis—keep the business in touch with the customer and with good information. With globalisation and improved communication, spoke operations are becoming leaner."
The second location principle is to scatter whatever operations can be scattered to lower-cost sites. In a classic pattern reminiscent of Raymond Vernon's 1960 analysis of New York City, back-office (i.e., administrative and number-crunching) activities can be housed in remote locations where wages and land rents are low and commuting is easy. Only the functions requiring frequent face-to-face contact (mergers and acquisitions, raising capital, fund management) must be kept in the center. A mixed case is trading, which in the age of the Internet is up for grabs.
Two questions can be posed here about Porter’s cluster analysis. One concerns the potential perils of clustering for member firms, the other the role of government.
On the first matter, Porter himself now recognizes that participation in an established cluster may not always help a firm adapt to new circumstances. In this recent passage, for example, it is hard to miss the influence of Porter’s Harvard Business School colleague, Clayton Christensen (1997):
When a cluster shares a uniform approach to competing, a sort of groupthink often reinforces old behaviors, suppresses new ideas, and creates rigidities that prevent adoption of improvements. Clusters also might not support truly radical innovation, which tends to invalidate the existing pools of talent, information, suppliers, and infrastructure. In these circumstances, a cluster participant….might suffer from greater barriers to perceiving the need to change…. (Porter 2000, p. 24, emphasis added.)
This is a pivotal issue. Proponents of cluster-based development strategies tend to assume that fostering specialized clusters is good for an area’s growth prospects. But students of industrial history might well differ.
After all, what became of Marshall’s prototypical industrial district, 19th century Manchester? The answer, in a phrase, is that “Industrial evolution is a history of cruel fates.” (Rothschild 1973, p. 191.) Marshall’s “something in the air,” in other words, may have become an obstacle to change.
The list only begins with textiles in Manchester or steel in Pittsburgh, mature centers that failed to meet the challenge of new competitors located elsewhere. (Chinitz 1961.) More recent episodes include Boston’s Route 128 in the 1980s. Its minicomputer firms proved blind to the need to move on to PCs. Similarly, in the Porter passage quoted earlier we find Minneapolis cited as the home of Cray, Burroughs, and Unisys. But just as with minicomputers, “big-iron” supercomputing proved to be a shaky base.
For both Route 128 and Minneapolis-St. Paul what has proved decisive to future development is the capacity to shift from dying bases to new sources of growth, as given by universities and financial institutions at least as much as by existing firms.
A second ambiguity is Porter’s attitude toward government policy. On the one hand, he applauds state, local, and national initiatives informed by cluster theory. (See Table 1 in Porter 2000, p. 31.) In his view, “clusters should represent an important component of state and local economic policy.” (Porter 2000, p. 29.) And indeed they do. The influence of cluster theory for policy design can be seen in this useful set of links on policies in individual states, as compiled by the Humphrey School of Public Policy at the University of Minnesota.
But in the same article, Porter takes pains to differentiate cluster strategy from industrial policy—which is bad, he writes, because it entails picking winners, a zero-sum game. “Although industrial policy aims to distort competition in favor of a particular location, cluster theory focuses on removing obstacles…. The emphasis in cluster theory is not on market share but rather on dynamic improvement.” (Porter 2000, p. 28, emphasis added.)
This will strike some readers as a fine distinction. To skeptics, cluster theory sometimes looks like a vehicle for state and local government officials in search of a targeting rationale.
For example, Terry Buss offers “The Case Against Targeted Industry Strategies.” His view: “Targeted industry studies use poor or inappropriate data, deeply flawed social science methods, and simplistic mathematical models in producing targets. Targets themselves tend to be dubious” (Buss 1999, p. 343).
Why, Buss asks, are targeting strategies, including cluster-based targeting, so widely practiced? Not because of their scientific merit, but for political reasons. Impressive analytics can be drummed up on demand to justify inherently political proposals. And why, when so many targeted industry strategies have failed, do states and localities continue to rely on them? Partly because they have the appearance of scientific backing, but mainly because of a herd effect. Once some states and localities develop targeting strategies, others feel compelled to follow suit.
Implicit in any such critique of targeting is the view that when it comes to new-enterprise development, market forces provide the best mechanism for picking winners and weeding out losers. Elementary as that proposition sounds, it remains a useful touchstone for sorting out economic from political dimensions in state and local economic development policies, including cluster-based strategies.
Skepticism about targeting is the context for Porter’s distinction between generalized cluster strategies and local industrial policies. But what does it mean to say as he does that a cluster-based strategy aims not for market share but for “dynamic improvement”?
Static spatial externalities depend on proximity among knowledge workers in a given setting at a given time. “It is argued that knowledge for innovations resides in the communication between skilled (knowledge) workers, and that this is dependent on their geographical proximity.” In other words, as studies of static externalities tend to confirm, “the capacity to receive knowledge spillovers is influenced by distance from the knowledge source.” (Echeverri-Carrol and Brennan 1999, p. 29.)
But while the effects may register over time, the analysis can be viewed as inherently static.
Nothing that has been described in this module on neoclassical perspectives treats history as anything but a prologue. To that extent, the conversation remains open to other approaches, whether from evolutionary economics, path-dependency perspectives, increasing-returns models, or network theories.
In particular, what are dynamic agglomeration economies? How do they shape the positions of competing centers of innovation? And how do they condition the relative importance of proximity and long-distance networks?
Such questions are explored in the companion module on post-neoclassical cluster theories.