Theory and Society

, Volume 38, Issue 3, pp 217–243

Modeling firms in the global economy



DOI: 10.1007/s11186-009-9083-7

Cite this article as:
Perrow, C. Theor Soc (2009) 38: 217. doi:10.1007/s11186-009-9083-7


I examine the apparent deverticalization of firms in the world economy and their adoption of relational contracting and modularization, necessitated by rapid product change, cheap and rapid transport, and new technologies. I argue that relational contracting is superseded by modularization when possible in the interest of more control over suppliers, and modularization in turn leads to consolidation, when possible, through buying up suppliers or making them captives. The result is increased concentration of economic power in the world economy, and examples of this are presented.

Modeling the new global economy is difficult. There is probably universal agreement that the economy as it existed up to the 1970s has disappeared. Gradually, then precipitously, the large monopolies and oligopolies that dominated the world economy downsized and lost market share, and a host of new firms emerged. Once dominant in the world economy, US firms lost much of their hegemony to Japanese, European, and then East Asian competitors. Since then we have seen extensive “deverticalization” of many of the biggest firms through outsourcing and divestures, as well as more rapid product change and new products and services. The process has inevitably created a landscape of more—albeit smaller—firms, more competition, and a greatly enlarged network of firm interactions. New technologies, rapid transport, and highly efficient communications and data processing have all played their role in bringing about this development. For most social scientist commenting upon these changes—I refer to them here as the New Economy theorists—they are benign; even the giant factories of the East are bringing more prosperity to their rural transplants. In particular, the sub-text of these theorists is that deverticalization of giant firms has not only led to lower consumer prices and more innovations, but to a deconcentration of economic power.

However, more firms do not necessarily lead to a deconcentration of power among firms. Nor is the networking among firms necessarily less exploitative. And the returns to other stakeholders, such as local communities, workers, and the environment have not increased. In particular, growing economic wealth has increased the number and importance of stockholders, and they are increasingly managed by pension funds, mutual funds, and other giant investor funds. This has led firms to an emphasis on shareholder values and short-run returns at the expense of re-investment of profits, collaborative customer–supplier relations, labor’s share, and community and environmental concerns.

The new configurations of firms and their interrelations have made it difficult to estimate the concentration of power in the economy, as has the emergence of Asian giants that are state-controlled or state-sponsored. Large firms may have more concentrated power than a simple market share statistic will reveal because of their de facto (as opposed to de jure) control over their suppliers or even customers. I argue that significant concentrations of power still hold, though its forms are different from the simpler economies of 40 to 50 years ago.

The argument in brief

Table 1 outlines the first part of the argument in terms of the variables of interaction—linear vs. complex interactions—and tight vs. loose coupling. I am drawing upon but, by introducing the interaction and coupling terms, modifying a most valuable classification scheme outlined by Timothy Sturgeon, which in turn is based upon the work of Sturgeon and other commodity chain theorists (who now prefer the term “value chain” to avoid the implication that only commodities are involved) (Sturgeon 2008).
Table 1

Four forms


Linear interactions

Complex interactions

Loose coupling

Generic market CELL 1

Relational contracting CELL 2

Tight coupling

Concentrated CELL 4

Modular contracting CELL 3

Prior to the 1970–1979 decade the world economy consisted largely of firms in Cells 1 and 4, a large number of small and moderate sized firms that had either substantial competition or dominated only small markets, and a small number of concentrated firms that dominated large markets. With the changes in technology, transportation, communications, and rapid economic growth, some firms in Cell 1 began to produce more complex and rapidly changing products that required more complex interactions within the producing firm and with a larger number of specialized suppliers.1 This moved some firms to Cell 2, relational contracting. More complex interactions required more intimate contacts among designers, engineers, and sales both within the firm and between the firm and its suppliers and customers. The complexity of the interactions became a burden in Cell 2, and some firms with appropriate products and technologies could reduce the complexity by modularization, represented by Cell 3. Here interdependencies are confined to the modules, but the modules are themselves independent of each other, linked only by standardized interfaces that govern the interactions of the modules. Modularity decreases the coupling of the system, giving suppliers that provide modules considerable independency.2 But squeezing out this remaining uncertainty—complexity within the modules, and the increased independency of suppliers—required more control over the modules and their makers. For some products this was achieved by further standardization of production and, most importantly, of suppliers, who now became captives of the lead firm. This allowed the lead firm to move to Cell 4, where it achieved market control.

Further consolidation in Cell 4 can be achieved by making not only primary suppliers and intermediary suppliers and warehouses captives, but retailers as well. The battle for dominance in Cell 4 firms is waged among giant producers, warehouses, and distributors (retailers), all of whom seek consolidation of each other. The dominance is unstable, however, in two respects. Hostile takeovers of firms in Cell 4 can result in selling off parts of the firm that then migrate to one of the other cells. Second, new products, processes, and firms emerge or move from Cell 1 into Cell 2 or 3, threatening the dominance of firms in Cell 4.

The scheme is highly stylized of course. Most firms remain in their cells, and movements do not have to be in the sequence I have described. Microsoft moved from Cell 1 directly to Cell 4 in terms of its market control of operating systems and browsers, but has modular relations with suppliers of other functions such as search and music in a competitive market. Furthermore, examination of any one firm in any cell will reveal some elements of all the other three cells (e.g., some modularity and some relational contracting, and perhaps arms-length bargaining; or market control in shrink-wrap software but not in “enterprise” software). There are no examples of pure types, but there are examples where one of the four configurations dominates.

The technological and the other developments mentioned might, by themselves, have been enough to move the world economy from being populated by just Cell 1 and Cell 4 firms to being populated by all four forms. New wealth, new products, and cheap and rapid transport and communications by themselves would have encouraged production systems based upon relational contracting and modularity. But the world economy up to the 1980s or so was stable because of two conditions that suddenly changed and made the new transformation more likely. First, the United States, with Europe and Japan playing secondary roles, dominated it. While the United States is still dominant, in many respects it is challenged by Asian players. Second, this pre-1980s dominance was accompanied by an accommodation in the United States between labor and capital, monitored by the banks, that reflected the interests of more stakeholders than just the shareholders, namely labor and community constraints. In Europe, this accommodation of multiple stakeholders was a matter of legislation. The governance system in the United States was eroded during the 1970s and 1980s with deregulation and the withdrawal of government support for labor. These changes appeared more slowly and incompletely in Europe, but they are proceeding. An article-length account of firm relations in a global economy must be highly schematic and overly path-dependent. (For a valuable critique of such overly path-dependent accounts, see Schneiberg 2007).

I first examine these societal changes, then discuss the issue of modularity, then expand upon the four Cells, and finally asses the degree of concentration or deconcentration in the world economy based upon the limited evidence that is available.

The governance structure and shareholder value

Under capitalism, competition will lead to monopoly if at least a few powerful firms seek to maximize profits, rather than see an adequate level of profitability as one goal along with others such as growth, stability, and satisfying a variety of stakeholders, such as investors, communities, customers, and labor. If a firm has the political and/or economic power to control a substantial part of the market, this will lead to higher profits, which will attract more investors and allow more growth and market control until a monopoly position is achieved. However, even if a firm tries to maximize returns to shareholders, there are constraints upon profit maximization. Labor must receive enough returns to reproduce itself and to consume the products; the environment must be protected enough to supply basic inputs to firms and allow the consumers to survive; and enough other firms must exist and prosper sufficiently to provide innovations that will either cut costs or create new products. Most of these constraints are long-run ones and imposed by the governance system of state regulations and a concentrated capital market that guides investment. Short run profit maximization can always be a goal, but because of these constraints, most firms had to settle for stability, survival, and adequate returns.

This began to change during the economic downturn, stagflation, and lowered profits in the 1970s, when US capitalism was dominant in the world. US business firm organized to reduce regulations, and the Carter administration (1977–1981) acquiesced. A merger movement flourished in the 1970, pushed by large banks that still were the source of most capital for business and industry, but large financial institutions are risk-adverse, and they promoted mergers that served their interest in stability. (I draw heavily upon the short summary of Mark Mizruchi and Howard Kimeldorf here (Mizruchi and Kimeldorf 2005; see also Mizruchi 2003)). Antitrust constraints were reduced during the Reagan presidency of 1981–1989. Mergers in the 1980s and 1990s increased, but were riskier; they were encouraged by the new phenomenon of corporate raiders seeking short-run gains. Mergers of firms within an industry had been favored by the banks, but the non-bank financial firms now favored horizontal mergers, linking unrelated industries, producing conglomerates. When the conglomerate movement failed to raise profits, the increasingly important financial investment experts punished diversified firms and rewarded those who focused upon core competencies (Davis et al. 1994).

With increasing national wealth, firms could raise capital through stock issues, drawing upon the increasingly large accumulations of funds in pension funds and mutual funds that the maturing baby-boomers were generating. The large banks were no longer the principal source of investment funds and no longer a stabilizing factor in finance. Firms had to pay attention to the investment experts. They not only sold off functions unrelated to their core competency, but outsourced more and more of what had been their core. This raised return on equity, increased dividends, and emphasized the short-run returns that financial analysts, easily buying and selling for short-term gains, were interested in. The incentives for top management to focus upon quarterly profits were increased by the practice of increasing the stock option part of top management remuneration, aligning their interests, it was argued, with those of stockholders, rather than other stakeholders or the long-term growth and stability of the firm. (For examinations of the shareholder value versus stakeholder values phenomenon, see Dobbin and Zorn 2005; Lazonick and O’Sullivan 2000; Williams 2000.)

This financial setting was well aligned with a predominant characteristic of the new economy, rapid product changes, new products, niche markets, and the geographically dispersed production and retail sites that the transport and communications revolution produced. Together they changed the economic landscape.

Maximizing shareholder value led to the great merger movement and the booming market for junk bonds. The large institutional investors, such as the pension funds (including TIAA-CREF) and the newly emerging mutual funds, forced corporations to maximize their returns upon equity and increase their stock value. They did this in part by selling off parts of the firm to generate capital for dividends, to improve return upon investment, and outsource many activities to suppliers who were forced to accept lower profits and bear the risks that the corporation was able to avoid (from downturns, excess inventory, competing products and the like), and lower labor costs by avoiding unionization and giving reduced or no benefits to employees.

The unequal distribution of wealth in the United States responded immediately to the doctrine of shareholder value since most households did not own any shares and the top 1% of shareholders held the vast majority of stock. Beginning in 1986, very high incomes began to increase rapidly, outstripping productivity growth. The richest 1% of tax filers claimed 80% of all income gains reported in federal tax returns between 1980 and 2005 (Levi and Temin 2007). The unequal distribution also increased as the salaries, bonuses, and stock options for top management increased, given that their interest in short run returns were now more fully aligned with the interest of stockholders. This helped make top management more concerned with cutting costs and improving profits in the short run than with long-term investments that would benefit the company, the shareholders, and other stakeholders. The structure of the financial institutions, now heavily deregulated, thus shaped the motives of top management in the largest firms. The academic inquiry into the New Economy, based in business schools, assumes that short run cost cutting is the primary motive of management, and an entirely legitimate one. The revisionist’s position (e.g., Williams 2000) argues that cost cutting had not always been the primary one; it became the primary one only with political changes that fostered financial deregulation.

I assume here that for many leaders of firms, the satisfactions of a business career lie in many areas; maximizing personal income is only one of many, and often not the most important one. For these leaders, stock values are mostly important as a constraint. Like the rest of us, these managers are embedded in a social life that constrains self-interest. However, there are enough individuals who, finding themselves in a context provided by deregulation, favor self-interested behavior and would pursue it vigorously. If these people hold powerful positions in big firms, in time this forces others to do likewise in order to compete and thus to survive. The combination of the context of under-regulated finance capitalism and the motivation to maximize their wealth (and power) will give these aggressive managers and their firms sufficient advantages to shape the context further to their liking. For example, an initial advantage the firm has can be aggressively exploited by buying up or forcing out competitors, and can eventually result in market domination if either the structure of the industry permits it or the firm’s actions change the structure enough to permit it.

This analysis of motivation for a, perhaps even small, segment of corporate elites, is quite different from the implicit one of New Economy theorists. For them, cutting costs is a (benign) universal value, and so are innovation and rapid product change, and presumably this is all to benefit the consumer. But the revisionists’ interpretation will better account for the goals of firms that will be outlined after a diversion in the next section to discuss the poster child of the New Economy theorists, modularity.

Introduction to modularity and complexity

The most important new term to be introduced into the discussion of the New Economy is modularity. I start with an extended account of modularity and all it is supposed to do in the New Economy.

Engineers are found of saying “complexity is the enemy of reliability.” But complexity can be managed with modularity. The idea is that those parts that must interact with each other in complex ways should be confined to a module whose only interaction with the rest of the system is through an interface. The benefits of such a design are that unanticipated interactions become fewer, easier to test for, and their effects are confined to the module. The design of the module is simpler and more amenable to innovation as long as the specifications for the interface are met.

A modularized system will be more loosely coupled than an integrated system (though loose coupling does not necessarily mean modularity). The failure of a module will not bring down other parts of the system, including other modules, and defective modules can be easily replaced. The advantages are illustrated in Herbert Simon’s celebrated article on the architecture of complexity in which he contrasts two imaginary watchmakers (Simon 1962). The first assembles a complete watch; the second assembles parts of the watch and then combines them. If there is an interruption the first watchmaker must start all over again whereas the second must only start over again on the particular module he was working on. If interruptions are frequent the second one is more productive.

However, Simon did not consider systems that are rarely "interrupted." In these cases integrated production would be more efficient since the modules in the second system all require interface designs, which can be difficult, and cannot utilize common modes—functions that can be utilized by more than one part. For this reason, mass production systems are highly integrated, whereas job shops and customized production systems tend to be more modularized. Producer-driven commodity chains tend to integrate production while customer-driven chains are more modularized.

An elegant definition of modularity is by John Paul MacDuffie and Susan Helper (MacDuffie and Helper 2006:424): “‘Modules’ as a basis for product architecture are defined as elements that are interdependent within, and independent across, whereas ‘integral’ product architecture is based on interdependence both within and across elements.” The module has an interface that allows it to connect to other parts, but is otherwise independent of all other parts and other modules in the system.

However, we shall see that this elegance can be misleading.

Modularity can be applied at any system level. The world economy can be described as made up of interacting modules, most of which will be nations then firms. At the same time a supplier firm will be a module to its customer firm, and any firm can be broken up into modules and each module into further modules down to the electron level or deeper. The scale must be defined. A large module may be made up of many smaller modules and those of still smaller modules until there is no difference between a part and a module. The appropriate scale to be used should be based on the points where system designers can make conscious decisions as to whether to emphasize integration or modularity. Thus, for our purposes, the definition of a module must include the opportunity to either modularize or integrate.

Building on a suggestion in a fundamental work on modularity in production systems by Carliss Baldwin and Kim Clark (Baldwin and Clark 2000), Akira Takeishi and Takahiro Fujimoto distinguish three types of modularity that are often confused: the product design level, where the design of the product is broken up into separate parts and then integrated; the production level, where there may be subassemblies that function as modules, or in-house production of complex components that can be “plugged in; and supplier or inter-firm modules, where the lead firm or customer buys modules built by others with a specified interface (Takeishi and Fujimoto 2001). Most of the literature concerns the last, inter-firm modularity. But Takeishi and Fujimoto use these distinctions effectively to contrast US and European automakers, which emphasize interfirm modularity and some production modularity, with their Japanese counterparts, who concentrate on product design modularity while paying almost no attention to production or interfirm modularity. Both use modularity, but in very different ways, and thus Japanese firms can be highly integrated and modularized, but integrated only in the area of product design, not in production or supplier relationships.

Is modularity a new phenomenon? Under some usages it is not. All economies, even very ancient ones, can be described in terms of modules. First, at the firm (product design) level, modularity could be crudely seen as just the old make or buy decision; modularization is choosing to buy, while integration is choosing to make. The new modularity theorists mean much more than that because they are dealing with highly technical products that require detailed specification of interfaces. Second, at the production and inter-firm level it is, again crudely, another word for specialization, the Adam Smith mantra, as Langlois and Robertson observe (Langlois and Robertson 1992). But specialization is ubiquitous and exists in both integrated and modular organizational forms. Modularity proponents emphasize the specialization that is required for novel technologies and applications, not the production of off-the-shelf components for mass markets. There is widespread agreement that there has been a huge growth of specialized firms selling complex products to a variety of customers, and the specifications of the customer suggest that we are speaking of modularity, not a buy rather than make decision. But the increasing number of firms does not mean they all are using the modularity strategy.

In rough terms the growth of the number of firms interacting is explained by the growth of economies (more organizations to interact) and new technologies (allowing the standardization and codification necessary to modularization). In addition, the growth of transportation means lowering the costs of outsourcing and assembling from diverse locations. To some extent (overestimated, I believe) the deverticalization of firms brings more firms into the economy. Regarding deverticalization, firms contract out, spin off parts, establish wholly owned subsidiaries, or split in twos or threes. Presumably the number of firms increases when this happens, but this does not necessarily mean there are fewer giant firms, or that the giants have shrunk in size very much. What happens is that there is more economic activity and an increasing amount of it is not within any one firm but between firms. There are also new industries, more new products, and shorter product cycles.

These developments apply not only to products but also to services. Services that were once provided within the firm, or within the nation, are now outsourced, leading to a burgeoning of “intermediaries,” such as huge Group Purchasing Organizations (GPOs) that sit astride large chunks of the commodity chain (Heckscher and Adler 2007:391–393). While I simplify things greatly here by speaking of customers and suppliers of products, it should be remembered that the services as well as products could be made or bought, and could be specialized or not.

In summary, modularity has many definitions, and can be applied to the design, production and supply functions. However, as we shall see, there is no clear evidence for its superiority over integration, and its supposed ubiquity may have more to do with the loose definitions employed and the simple growth in economic activity worldwide than with any movement to deverticalized organizations and deconcentration of economic power.

Two simple examples

Our exploration of modularity is not complete. First I want to frame the argument with two examples, one of the disadvantages of integrated versus modular systems, the case of Microsoft’s Windows operating system, and the other of the strengths of modular systems, networks of small firms. Then I more directly contrast the benefits and risks of modularity and integration.

Critics of Microsoft’s Windows operating system in all its iterations including the new one of Vista, argue that applications provided by Microsoft are integrated into the kernel of the operating system, whereas this is less true of Apple’s Macintosh operating system and very much less the case with open source systems such as UNIX and Linux, which emphasize modules rather than integration. Critics of Microsoft favor modularity over integration because it reduces interactive complexity and tight coupling, both of which increase its vulnerability to errors and especially to deliberate intrusions.

I will just briefly mention the obvious security problems with the integrated Windows system. Integration allows hackers, criminals, potential terrorists, and nations such as China to take control of Microsoft’s operating system, Windows, which powers over 90% of computers on the Internet. These intruders can gather sensitive data for industrial espionage and financial fraud, take control of industrial sites, and establish “botnets” that can launch denial of service attacks upon government facilities or hold commercial sites hostages to ransom payments. There is evidence for all of these vulnerabilities (Perrow 2008).

While we think of applications such as Excel or PowerPoint as a module that simply connects to the operating system—after all the software comes shrink-wrapped—these programs are in fact deliberately designed by Microsoft to integrate with the Windows kernel rather than ride on top of it as a module. The integrated system structure is probably easier to engineer, but more importantly, it prevents competitive products from running on the system. (The rush to market for competitive reasons also encourages sloppy code that invites intruders.) Rewriting Windows to be more modular would reduce the security problems. Or, business and government could use more modular operating systems that are available but have only a tiny market share. Either of these alternatives would be more effective than building more and more firewalls or the futile attempt to shut down ISPs that hackers use, though these are strategies our government is pursuing to provide security (Perrow 2007).

The second example illustrates the advantages of modularity in a particular form. It concerns the difference between a large integrated firm that produces and sells products, and a network of small firms from which the same products emerge (Perrow 1992). In the integrated firm, production and sales are independent divisions, but I would not call them modules because a high degree of interdependency between production and sales is designed in, or knowingly allowed, or inadvertent. If the firm breaks out into divisions because the products are sufficiently distinct that extensive firm-wide integration is not required beyond the setting policies and standards, we could call this a case of modularity at a high level of generality. (Often, instead of divisionalization, dependent subsidiaries are established primarily for tax purposes and little modularity exists (Prechel 2000).) In contrast, the network of small firms consists of modules that independently produce parts of the final product, or assemble parts into a subassembly, or do final assembly. In contrast to the divisionalized firm, all three forms of modularity are present. Marketing and distribution tasks may be undertaken by some of the producers or by a more specialized large firm or set of small firms. We tend to think of the integrated firm as more efficient than a network of small firms, but under a more ecumenical definition of efficiency than just short-run production efficiency, this is often not the case. There are economies of scale at the network level (sharing of skills among firms, flexibilities and redundancies in case of disruptions, common supply and distribution sources) and positive network externalities (education and training, government services, etc.).

Some problems and advantages of modularity and integration

When the product is quite small there is usually no alternative to integration. But where there is a choice between integration and modularity there are some advantages to integration. For the designer or engineer, an architecture that emphasizes modules requires extra work in terms of designing the interface of the module, which determines its interaction with other parts or modules. In an integrated system it is probably easier to make common mode connections for power, safety devices, process indicators, and so on. (However, these are often the source of unexpected and undesirable interactions (Perrow 1999).) In addition, I expect that a circuit design that is integrated would have less wiring than a modular design and thus be easier to design and construct and more efficient in the use of energy.

However, there are disadvantages to integration. The difficulty of testing is the paramount one. A proper test must be of the whole system after it is completed; testing the parts separately does not test the interaction of parts. If there is a failure it is difficult to determine the faulty part because each part is potentially interacting with many other parts. Once it is found, it is difficult to replace a faulty part without substantial disassembly, making the system inoperative for a long time. The same is true if a part wears out. It may be necessary to shut down the whole system, and have substantial disassembly and reassembly even if it is possible to determine which part is worn out. In an assembly line the whole line must be stopped, whereas in modular production, subassemblies can be carried out while the faulty module is replaced or repaired.

In contrast, in a system with many modules, the individual modules can be easily tested, and with good design the interface can be tested with the appropriate modules without testing the whole system. If the whole system fails at some point it is easier to find the source of failure, perhaps by unplugging a module and replacing it with a similar module, thus determining that the module had a unique defect that other similar modules did not have. Or, if the design of the module itself was faulty, this can be determined by testing it in its interface with all modules it interacts with. The larger the system the more important is the testing advantage of modularity.

The literature says that there are design advantages to modularity. While, as I have indicated, integrated designs are more elegant, efficient, and simple because of less need for interfaces with modules and more exploitation of common modes, designing a module should be simpler than designing its function into the integrated system. The tasks can be partialed out to specialists who need to know only about the immediate part or module their module interfaces with, rather than having to understand the whole system. (This allows more arms-length, nonreciprocal relations with suppliers. It is also probably truer of complex and frequently changing products.) With integrated systems there is always the problem of the unexpected interaction of parts, possibly because of a failure that defeats the safety system, or just because of an unusual, unforeseen, but quite legal interaction. (For this reason, integrated firms producing complex or non-standardized products have to make, rather than buy, components, or have to develop highly integrated, reciprocal relations with suppliers.) This is less likely to happen with module architecture because there are fewer occasions for unexpected interactions, and if they occur they are most likely to be within the module. Because the interface is crucial, of course, a poorly designed one can produce such interactions.

Flexibility would seem to be an obvious advantage of modularity and is one of the major reasons the literature on the New Economy celebrates modularity. With so many new products, new varieties of products, and customization, production and services must be more flexible than ever before.

Carliss Baldwin and Kim Clark, building on a number of discussions of modularity in the 1990s, include a useful example of modularity first published in a Harvard Business Review article in 1997, “Managing in an Age of Modularity” (Baldwin and Clark 1997). The computer industry, they say, is as revolutionary as the railroads were in the nineteenth century, leading the way to a new industry structure, modularity, which is more important than technology!

At the heart of [computers] remarkable advance is modularity—building a complex product or process from smaller subsystems that can be designed independently yet function together as a whole. Through the widespread adoption of modular designs, the computer industry has dramatically increased its rate of innovation. Indeed, it is modularity, more than speedy processing and communication or any other technology, which is responsible for the heightened pace of change that managers in the computer industry now face. And strategies based on modularity are the best way to deal with that change (Baldwin and Clark 2003:149) [Italics supplied].

The first significant modularity design in the computer industry was IBM’s System/360 in 1964. Both the processor (a key part of the operating system) and the peripherals and applications that IBM marketed had visible as well as hidden information in them. The visible information in the processor specified the interface that allowed any engineer in IBM to manufacture peripherals or software applications that could run on the processor. As long as the interface of the module conformed to the protocols, engineers were free to experiment with any design they wanted in their module. That competitors could do this so easily and extensively was not intended. (The hidden information was designed to prevent competitors from marketing their own applications, but it was limited.) Soon the 360 system had many competitors, which was good for the industry and, in the short and intermediate run, for IBM as well as it became the standard setter. As Baldwin notes succinctly, “The modularization was supposed to simplify the firm’s internal operations and improve customer relations, not invite competition. But, in the end, it did all three” (Baldwin 2007).

While Baldwin and Clark see modularity as the best way to deal with change, “it turns out that modular systems are much more difficult to design than comparable interconnected systems” (Baldwin and Clark 2003:152). IBM did not anticipate how difficult it was to build a modular system and might not have done so had they realized this. But they also did not anticipate the huge success that the 360 would have despite the emergence of competitors with compatible peripherals and applications.

An early example of modularity concerns the Pennsylvania Railroad in the mid-nineteenth century, discussed in Nasaw’s biography of Andrew Carnegie, and it suggests a motive for having supposedly independent suppliers that the modularity literature has not considered, but should (Nasaw 2006 Cpt 7). The Pennsylvania Railroad needed bridges as it expanded into new territory. It could have done what Alfred Chandler said industry was doing, and what Oliver Williamson recommends to reduce transaction costs, and formed a bridge division or bought an existing bridge company, ensuring reliability of delivery and quality. This would have been integration. Instead, Andrew Carnegie and the two top officials of the railroad formed an independent bridge company. Their bridge company needed iron, but rather than buy it on the open market and spur competition, they then formed an independent iron company to supply the bridge company that supplied the railroad. Rather than integration into one firm, we now have one firm and two independent modules the firm deals with, except that the three officials of the Pennsylvania Railroad also owned the two supposedly independent firms.

The advantage of modularity over integration had nothing to do with innovation or risk sharing or flexibility. The three owners had the railroad pay greatly inflated prices for the bridges, and had the bridge company pay inflated prices for the iron. Thus the owners made very large profits on their two independent companies, and while the Pennsylvania Railroad lost money by paying excessive prices, these losses were spread over all of the shareholders. The shareholders, who included Carnegie and the two top officials of course, in effect subsidized the profits of the iron and bridge companies that Carnegie and his “cronies,” as Nasaw labels them, controlled. This was not illegal at the time, but it was not publicly available information either since the ownership of the bridge and iron companies was deliberately disguised. This practice became even more extensive when the Pennsylvania Railroad gained control of the Union Pacific and led to near-bankruptcy of the Union Pacific.

Since the passage of the securities acts of 1934, which deemed such activity illegal, it has become much harder to disguise this activity. The governance environment makes a difference. However, there is evidence that a fair amount of deverticalization is undertaken to reap rewards from deceptive practices such as those that Carnegie engaged in, though this is nowhere even alluded to in the modularity literature. The research of Howard Prechel discloses true innovativeness in deceptive financial restructuring in recent decades by means of what could well be called modularity (Prechel 2000).

I will proceed through the four models I summarized at the outset, listing their details in the tables, but focusing most heavily upon the two novel ones, relational contracting and modularity. Then I conclude by examining the limited empirical evidence—regional surveys and case studies—that exists regarding concentration. Table 2 summarizes the four models.
Table 2

Four forms expanded


Linear interactions

Complex interactions

Loose coupling

Generic market CELL 1 Variety of volume, moderate to low sophistication, integrated production, variety of firm organization, low to moderate interdependency, high competition

Relational contracting CELL 2 Low to moderate production, sophisticated products, integrated, variety of firm organizations, high interdependency with suppliers, high trust, low opportunism, moderate competition

Tight coupling

Concentrated CELL 4 High volume, large firms, varied product sophistication, integrated production and firm organization, high opportunism, low competition

Modular contracting CELL 3 Varied production volume, moderate size, sophisticated products, modular production, bureaucratic or decentralized organization, moderate inter-dependency with suppliers, low opportunism, moderate to high market share and competition

Generic markets

Initially, say from 1950 to 1970 or so, two configurations or models predominated, differentiated by a firm’s degree of market control; they are labeled the generic market and the concentrated firm. The generic markets model consisted of many firms sharing the market, competing with one another on the standard basis of price, quality, service, and on new products or improvements of old products. Some used many suppliers while others were more self-sufficient, deciding to make rather than buy. Some had market dominance but only in niche areas. This represents the standard “market” of economists (Table 3).
Table 3

Generic market attributes

Cell 1: Generic markets

System (within firm; firm–supplier): Linear Interactions, Loose Coupling

Volume of production: High, Moderate, Low

Product sophistication: Moderate to Low

Production mode: Integrated, some modular

Firm Organization: Varied by/within firms (bureaucratic, decentralized, matrix)

Supplier relations: Low to moderate interdependency

Moderate to low trust, little community

Opportunism: Moderate for firm/supplier

Market: High competition, low market share for firms and suppliers

Returns from innovations: Few returns, not shared

Goal of Firm: Move to any other cell, most likely to 2 then 3, then 4

“Interactive complexity,” where many parts or components or organizations are not arranged in a linear sequence, but can have unanticipated interactions because of common mode functions or spatial proximity, are rare in these systems; they are largely linear, with predictable interactions. However, since they are not highly rationalized internally, nor are their relations with their suppliers or customers highly rationalized, the coupling of firm and supplier is loose—a firm may have more than one supplier and the suppliers more than one customer, and little interdependency between customer and supplier of the sort found in Cells 2 and 3. The production mode is more likely to be integrated—linear assembly—then modular, and the organization of the firm can be bureaucratic, decentralized, or even a matrix organization. Trust is not as marked as in Cell 2, relational contracting, and there are only moderate occasions for opportunism wherein either customer or supplier “holds up” the other, acquires intellectual property from the other, and drives hard bargains with arms-length relationships. Returns from innovations by either party are not shared, but appropriated by the innovator. Firms in the generic cell may move to Cells 2 or 3 as new products or design and production innovations appear, or a firm in Cell 1 may acquire sufficient market control to move to Cell 4.

Relational contracting

As new technologies and innovations increased in number and as transportation and communications became cheaper, new firms, or innovating generic firms, exploited new technologies, opened new markets or challenged the old ones. Large concentrated firms, under stockholder pressure, spun off parts of their integrated production facilities, and these new firms sought other customers and suppliers. Where the new technologies were “complexly interactive” (that is, not necessarily involving more parts or production processes, but involving more non-linear, less predictable, and less understood interactions among parts or processes), could not be disaggregated (remained tightly coupled), and required capacities that the firm did not possess, the firm came to rely upon specialized suppliers. Given the complexly interactive and tightly coupled nature of its production, it required close relationships with its suppliers, who participated in product design and joint innovations with the “lead” firm. This model is called “relational contracting” in the literature. (It is discussed extensively in Heckscher and Adler 2007.) (Table 4)
Table 4

Relational contracting firm attributes

Cell 1: Relational contracting

System (within firm; firm–supplier): Interactively Complex, Tightly Coupled

Volume of production: Moderate to Low

Product sophistication: High, Complex

Production mode: Integrated

Firm organization: Varied by/within firms (bureaucratic, decentralized, matrix)

Supplier relations: High interdependency, few suppliers

High trust, collaboration, community sense, tacit knowledge

Opportunism: Low for firm/supplier, but possible

Market: Moderate competition/market share for firm/supplier. Rapid product change/fast delivery

Returns from innovations: shared

Goal of Firm: Move to Cell 3 to reduce occasional opportunism by suppliers and reduce the costs of interdependency with suppliers

The major system characteristics of the firms and their relationship in this cell are non-linear interactions within and between firms, and tight coupling within and between firms. Both the lead firm (customer, in this relationship) and supplier are dealing with new or fast changing products and new technologies that require experimentation, perhaps customization, and present many uncertainties, including simultaneous engineering during design phases. The defect and rework rates will be high, design changes can be frequent and can even occur during production runs. The supplier can spend much time at the customer’s facilities, may have intimate knowledge of the customer’s own clients (which might be another firm in the commodity chain, or a retailer, a government client, etc.), and of proprietary secrets of the customer. The customer may have considerable knowledge of the supplier’s business and of the supplier’s other customers, if there are any. The customer may also hold an equity interest in the supplier firm. This runs the risk of creating opportunism by either party because of asset specificity for the supplier and lock-in for the customer. (A supplier that makes extensive investments in assets that are specific to one customer leads to the customer becoming dependent upon the supplier; the customer is “locked in.” But it also makes the supplier dependent upon the customer, since, being specific to that customer; the investments are worthless without the customer. Both sides have opportunities for opportunism.) Long term ties and trust will reduce the opportunities and motivations to exploit one’s position. However they will also disguise the possibility of the supplier using knowledge of the customers’ technology to improve its contracts with its other customers, or even, as happens, producing their own finished product or service in competition with their customer. Or, the customer, realizing that the supplier has large sunk costs in producing the product, can cut the supplier’s profits by lowering the price the customer will pay and make the threat credible by moves to change suppliers or make the product themselves.

None of these opportunistic moves is easy, however. Relational contracting is in itself an important sunk cost, providing benefits to both parties that are not easily transferred to other firms or parties. The trust, tacit knowledge, learning by monitoring and other virtues proclaimed by proponents of the relational model—and its variants such as the “pragmatic collaboration” model (MacDuffie and Helper 2006:433–434; Helper et al. 2000)—are not easily replaced with another partner, whether customer or supplier.

Regarding interactive complexity, it is not just that relational contracting arises where products are complex because there are many parts or components; a linear system can have many parts and is likely to. It arises because, since this cell produces sophisticated goods or services, the interactions of components cannot be fully programmed or even understood. Tacit knowledge will supplement formal knowledge and specifications; learning by doing is possible because of the close relationships of partners; and for new designs and technologies—characteristic of this cell—the relationships afford iterative trials. Interactive complexity may produce unexpected interactions that are fortuitous as well as damaging as design teams collaborate. Relational contracting would also be advantageous if there are rapid technological or product changes since “benchmarking”—evaluating your position and scanning the horizon for new opportunities and technologies—is easier if customer and supplier are in close, reciprocal contact.

Regarding tight coupling, products or services in this cell will be more integrated than modularized, and thus tightly coupled. Since tight coupling brings all the perils of a single failure cascading through the whole system, close contact and reciprocal adjustments are required between customer and supplier. (In this cell the relationships between a first tier supplier and its second and third tier suppliers are likely to have progressively fewer relational contracts.) Proponents of relational contracting, including the advocates of pragmatic collaboration, find examples in virtually all industries.

Examples of relational contracting practices exist in the literature, but it is hard to find examples of firms with these characteristics. Instead, relational contracting appears to exist along side of modularity and the traditional forms of transactions, such as arms-length bargaining, short-term contracting, and even vertical integration. One proponent of this model, Charles Sabel, cites as an example the huge conglomerate of 750 subsidiaries, Illinois Tool Manufacturing, which has bought up small companies and grown them, presumably, but no evidence is given, with relational contracting practices fostered by its large technology center. He also cites these practices in Cisco, one of the largest software firms, which others might nominate for Cell 4, a concentrated firm (Sabel 2007).

Relational contracting is appropriate for complex products that are frequently changing or still being redesigned, perhaps for custom applications. But it also is integrated production. Not the integrated production of a linear system where fully tested components can be assembled with full confidence that the whole will work, but the integration required of a complexly interactive system that can only be tested after full assembly. That is, components are linked to other components in multiple ways such that there are many interactions, some of which may be unforeseen.

Modular production, in contrast, reduces the number of interactions of different components by limiting the interactions to an interface that is carefully designed and controlled.


As noted in the previous section, complexity can be managed with modularity (Table 5). A modularized system will be more loosely coupled than an integrated system. The failure of a module will not bring down other parts of the system, including other modules, and defective modules can be easily replaced. This changes the firm’s relationship to its suppliers. The relations will be more loosely coupled, just as the internal structure of the customer firm will be more loosely coupled. There will be moderate interdependency between customer and supplier, and this allows the customer to use multiple suppliers, just as the suppliers may have multiple customers. Trust will be moderate rather than high, and there will still be some collaboration and sense of community. Whereas returns from innovation are likely to be rather equally shared in the relational contracting model, the sharing in the modular form will be more modest with the customer appropriating more of the profits, an important incentive to modularize.
Table 5

Modular firm attributes

Cell 3: Modular

System (within firm; firm–supplier): Interactively Complex, Loose Coupling

Volume of production: High, Moderate, Low

Size: Moderate

Product sophistication: High, Complex

Production mode: Modular, codified interfaces, supplier free to design module

Firm organization: Bureaucratic or decentralized

Supplier relations: Moderate interdependency, multiple suppliers/customers.

Moderate trust, some collaboration and community

Opportunism: Low for firm/supplier

Market: Moderate to high market share and competition. Rapid product change/fast delivery

Returns from innovations: Moderate sharing

Goal of Firm: Move to Cell 4 to make suppliers captives, more fully integrate production and increase market share

If the customer is able to reduce the sophistication and complexity of the product through design changes and stabilization of the modules supplied by the suppliers, it may more fully integrate production and make its suppliers captives, and thus move to Cell 4. It would gain market control and reduce competition, though more innovative products may threaten even firms in this cell, or they may lose the ability to adjust rapidly to changing markets. Indeed, a firm in the modular cell may not seek to lock in its relationship to suppliers to the extent that they become captives just because it anticipates rapid product or market changes that will require innovative and flexible suppliers.

Interactive complexity is not quite as high as in the relational contracting cell, but it is still higher than in the others. The lead firm may do some work “relationally” (sustained face-to-face interactions) with the suppliers of modules, but it is more likely to try simply to specify the output and interface conditions and allow the supplier to design and build the module (subcontracting parts of it to sub-suppliers in many cases). There is no need for face-to-face interactions or geographical proximity, as is the case in the relational model. But since the customer and some first-tier suppliers are responsible for the interaction of many modules, we can still expect considerable interactive complexity at the firm level for all parties, even though modularization reduces it.

Regarding loose coupling, individual modules themselves may be tightly coupled, but since the failure of one module does not bring about a cascade of failures such as can happen in an integrated system (like relational contracting and the captive firm models) this model is more loosely coupled. Furthermore, the customer may substitute one module for another without affecting the rest of the system, and substitute one subcontractor for another since the relationships are not as interdependent as in the relational cell. The modular cell favors industries that have complex products with rapid changes and short product lives. It is nonetheless hard to specify industries for any of the cells. As noted, the Japanese auto industry favors modularity at the design level only, and has captive firms, whereas European and US firms favor modularity at the production level and the interfirm level.

A good example of an industry that does favor modularity is the electronics industry. The development of industry-wide standards and the codification of knowledge in the electronics industry enable lead firms and highly competent suppliers to exchange rich information (such as detailed specifications) about transactions without need of deeply relational ties. “Turn-key” suppliers provide their clients with “a full-range of services without a great deal of assistance from, or dependence on lead firms” (Sturgeon 2002:455). However, the design and manufacturing of hard disk drives remains integrated and the customer-supplier relations take the relational contracting form. A prime example of modularity in the electronics industry is the emergence of “fabless” manufacturing, where the lead firm contracts out to semiconductor “foundries” who make the actual chips to specified designs which the lead firm them assembles. Xilinx, Apple, and Toshiba are examples of fabless manufacturing.

In modular networks asset specificity remains relatively low because there is “a highly formalized link at the inter-firm boundary, even as the flow of information across the link has remained extremely high” (468), and this is particularly true of the electronics industry. Sturgeon notes that the linkage between lead firms and these key component suppliers, which often work for multiple clients, enables external economies of scale that cannot be realized in relational networks. As Bair notes, compared with relational networks, modular networks are characterized by lower degrees of mutual dependence and a greater reliance on codified instead of tacit knowledge. “In a sense,” she cogently observes in a discussion of Sturgeon’s work, “Sturgeon is arguing that standards and codification mimic “trust”—they produce an outcome that is similar to what may be observed in long-term, relational networks, but via a different mechanism” (Bair 2008 Cpt 1).

Concentrated firms

Firms in cell 4 have substantial monopoly or oligopoly power (Table 6). They have rationalized production and distribution, and to the extent possible, innovation and design. This reduces interactive complexity and promotes linear systems in supplies, production, and distribution. They are vulnerable to the dangers of tight coupling, but market control reduces the dangers of external shocks from the environment, and fully dependent suppliers absorb the costs of internal mishaps. Internally, these technological aspects make it practical for them to have vertically integrated hierarchies, emphasizing authority relationships rather than the more expensive collaborative ones.
Table 6

Concentrated firm attributes

Cell 4: Concentrated firms

System relations (firm, and firm–supplier): Linear Interactions; Tight Coupling

Volume of production: High

Firm size: Large. Supplier size: Small to Moderate

Product sophistication: High, Moderate, Low

Production mode: Integrated, with some modularity

Firm organization: Bureaucratic, authoritarian, vertically integrated

Supplier relations: Suppliers dependent, arms-length bargaining, captives, many

suppliers. Low trust, low community

Opportunism: High for firm, low for supplier

Market: Low firm competition, with high market share. Supplier: high competition. Rapid product change/fast delivery in some products

Returns from innovation not shared with supplier

Goal of Firm: Remain in cell, maintain monopoly power

Where they choose to buy, rather than make, they control their suppliers, do not share intellectual property with them, or share any benefits from innovations made by their suppliers. Bargaining is arms-length and based upon short-run cost considerations. They seek to have alternative, dependent suppliers available. Trust is not as necessary as in any of the other models. What might be called “captive outsourcing” is possible, as, for example, the FedEx practice of forcing its truck drivers to purchase their vehicles from the firm so that, as independent contractors, no social security, health or pension costs are incurred, and the driver-owners absorb the risks of maintenance, breakdowns, and increasing fuel charges. The drivers continue to receive their workloads from the company and are as closely monitored as before. As independent contractors, they cannot form unions (Greenhouse 2008b).

Regarding linearity, the supplier provides the customer with standardized, or at least noncomplex goods or services, which can be obtained through arm’s-length bargaining. The interaction of customer and supplier is one of one-way dependency rather than interdependency. The coupling is tight both within the customer firm (e.g., continuous processing or assembly lines) and between customer and supplier (JIT delivery, highly specified components or supplies). The supplier is likely to have a highly dedicated production system and thus be highly dependent upon the customer (making it a captive relationship). The customer can shift to other suppliers at little cost, driving down the price and appropriating a good share of the profits that the supplier would otherwise enjoy. (The New Economy literature does not use terms such as profit; it only refers to the more neutral goal of “cost cutting.” However, organizational theorists occasionally speak of profits and consumer welfare when investigating the costs of vertical integration. Catherine de Fonenay and Joshua Gans write: “Vertical integration mitigates the hold-up problem faced by the monopsonist. It allows it to generate and appropriate a greater level of industry profits, at the expense of consumers” (de Fontenay and Gans 2004).)

That the customer can shift to other suppliers presents an anomaly in my scheme, since it affords the customer a degree of freedom. It is thus an instance of loose coupling at the supplier level, rather than tight coupling, since there are likely to be alternative suppliers it can shift to, thus lessening the dangers of holdup by the supplier. However, it remains tightly coupled at the producer and design level.

The dangers in this cell are small. The main risk is that of the failure of the supplier when no alternative suppliers readily available. But since this form generally involves noncomplex supplies there are likely to be other suppliers available. The customer may guard against the dangers of relying upon a single source by investing in the supplier, and if the customer finds it profitable or necessary it will simply integrate the supplier into its organization, in other words, to consolidate. Consolidation makes it easy to cut wages, reduce benefits, and de-unionize the new employees.

This is the most favorable position of the four in terms of interfirm relations. For the customer it has the great advantage of capturing the profits (or “rents” in the deceptive language of economics) of the supplier. It provides flexibility if the market or technology changes since existing suppliers can be cut loose and new ones found, though this does not present as much loose coupling as in the relational and modular cells. Finally, the customer is disassociated from any exploitation the supplier must engage in to meet the customer’s price. Trusting relations with suppliers or competitors are not necessary, so opportunism and self-interest on the customer’s side can prevail. It best fits the economist’s ideal of profitability, though of course there is little of the free market to be found here. In the conventional dichotomy of markets and hierarchy it is at the hierarchy pole.

Some caveats

Substantial field research has made it clear that firms rarely exhibit the ideal characteristics of a single model. Most firms have some short-term, hands off supplier relations and also long-term, fully interdependent relations. One part of a firm may depend heavily upon relational contracting while another part may not. There are always opportunities for opportunism; it is not limited to the concentrated firm. A customer in the relational contracting cell may learn enough about their suppliers’ modules to make them rather than buy them. A supplier may learn enough about the customer’s technology to incorporate it and sell it to a different customer. Even a large firm will have a mixture of trust and opportunism in its operations.

My scheme is over-schematized; no one firm or one customer–supplier relationship fits neatly into any one cell to the exclusion of relationships represented by the three other cells. Sturgeon puts it very well (despite a run-on sentence):

For example, a lead firm will typically forge market relationships for standardized goods, modular linkages in complex transactions when standards for exchanging codified information exist and are widely known, relational linkages with select partners when complex inputs are impossible to specify in advance and knowledge is not easily internalized, and captive relationships when smaller suppliers can be provided with sufficient knowledge to provide needed inputs and, at the same time, dominated in order to keep that knowledge from spreading to competitors. And, of course, firms must manage the value chain activities, and the linkages that exist within their own organizations (Sturgeon 2008:18).

Nevertheless, the four models (based upon Sturgeon’s work) are useful for assessing the extent to which deverticalization and new interfirm relations may lead to deconcentration of economic power in the global economy. In the next section, I examine a number of instances where deconcentration does not seem to have happened and instead we have the movement of firms to the concentration cell.

Examples of concentration

In the current literature on the “collaborative community” in the world economy, the concentrated firms are described as behemoths that are outmoded and doomed to failure because of their inflexibility (Heckscher and Adler 2007). Yet market control permits a large degree of inflexibility as well as super normal returns to capital. Their internal organization (integrated production) permits rapid product change if it is forced upon them, and their economic power allows them to buy up innovations and innovative suppliers if necessary. There is no evidence that they have disappeared from the economic scene, though in some cases they have outsourced to captive firms sufficiently to reduce their official employment size while keeping or increasing their revenue. Some giants, such as CISCO, are nearly “fabless” firms—selling products without engaging in any fabrication on their own part. Concentrated firms exist or are emerging in virtually all the economic sectors. It is the place to be. I will examine the limited evidence we have regarding the supposed deconcentration of economic power because of de-verticalization and outsourcing.

The case for modularity reducing concentration seems to be the strongest, but there are critics and skeptics. Baldwin and Clark give a very favorable review to the effort of Volkswagen to establish a new modularized truck factory in Resende, Brazil (Baldwin and Clark 2003). However, an article describing the operation shows it to actually resemble, the authors note, an early twentieth-century factory. In particular, the supposedly independent suppliers lost their independence and were taken over by Volkswagen (Ramalho and Santana 2002). Interestingly enough, Charles Sabel, a critic of modularity, describes the same factory as an illustration not of modularity, but of the “new organization” model of networks, with trust, co-design, learning by monitoring and so on, a variant of “relational contracting.” He cites as evidence the same article that Baldwin and Clark cite. But the article gives no more evidence of Sabel’s “new organization” model of networks than of modularity.

Sabel criticizes the modular form as “risky, even impossible” (Sabel 2007:120). He asserts its codified interfaces have led to significant failures of modularized firms such as Federal Mogul. Two other modularized firms, Dana and Tennaco, are described as “on the brink” of failure (129). Instead, he advocates “chunking,” a less specified, more exploratory collaborative relationship. John Paul MacDuffie and Susan Helper provide evidence from the automobile industry in the United States that supports Sable’s skepticism about modularity. The attempt to modularize, instead of reducing task interdependence and coordination between the customer—the automakers—and first and second tier suppliers has, if anything, increased the need for interdependence rather than reducing it. What was possible in electronics was not possible in the automobile industry (MacDuffie and Helper 2006:427).

Baldwin and Clark see increased competition and innovation as the result of the modularity that is occurring not only in the electronics industry but also outside of it. “As modularity becomes an established way of doing business, competition among module suppliers will intensify. Assemblers will look for the best-performing or lowest cost modules, spurring these increasingly sophisticated and independent suppliers into a race for innovation similar to the one already happening with computer modules” (Baldwin and Clark 2003:153). It is an appealing vision, but the race may spur not only innovation but also consolidation, which, when achieved, may dampen innovation.

Regarding consolidation, accounts of modularity in some industries or locations stress that lead firms preferred to have just a few first-tier suppliers, making these suppliers larger and more powerful. Thus it may be only in quite specific cases that there are a large number of suppliers to choose from to maximize performance or minimize cost. Furthermore, the sunk costs involved in choosing suppliers will be higher in proportion to the size of the supplier and the amount of services it provides. The literature generally considers that it is efficient for a supplier to take on more and more tasks all the way up to designing the final product. But the more powerful the supplier, the more captive the customer; the incentive of the supplier to take risks with innovations is likely to decline if the customer has fewer alternative suppliers to turn to, and has substantial sunk costs in its present supplier. Thus the modular movement appears to just make the old “make or buy” decision more complicated.

Nor does modularization necessarily promote an increase in the number of independent suppliers, thus reducing concentration. Baldwin and Clark have this revealing observation about consolidation and concentration:

Some automotive suppliers are already moving in that direction by consolidating their industry around particular modules. Lear Seating Corporation, Magna International, and Johnson Controls have been buying related suppliers, each attempting to become the worldwide leader in the production of entire car interiors. The big car manufacturers are indirectly encouraging this process by asking their suppliers to participate in the design of modules. Indeed, GM recently gave Magna total responsibility for overseeing development for the interior of the next-generation Cadillac Catera (Baldwin and Clark 2003:154; Baldwin and Clark 1997:154).

Since their writing, the industry has moved even more in the direction of consolidated suppliers and a stripped-down lead firm, as we shall see.

They may be correct in noting that the financial services industry is one that is ripe for modularization—since modularizing the servicing of stock and other securities should be easy—but so is consolidation easy for the same reasons. There is massive consolidation in this industry. Indeed, to take one of their cases, when Fidelity, a big mass-market provider of money management services, gives its 11 billion stock index funds to Bankers Trust (already a giant) to manage, one firm has “deverticalized” somewhat while another has increased its verticalization. In this example, the New Economy does not look so new in terms of consolidation. As the economy grows, so do big firms.

Sebastian Fixson, surveying 168 publications, finds dozens of definitions and conceptual schemes for modules, not all of them compatible. More alarmingly, there is little empirical confirmation of the superiority of modularization. While anecdotes of its ubiquity and presumed usefulness abound, Fixson’s survey of actual studies reveals only mixed results for benefits of modularity in the engineering literature, despite the great attention to the topic (Fixson 2007).

For example, according to Fixson, authors Zhang and Gershenson, building on their own modularity measures,

… study a number of small consumer products and do not find a general relationship between modularity and cost. In addition, it has been suggested that the number of modules affects the parts fabrication and assembly costs in opposite directions, hence the optimal number of modules needs to balance these two effects (92).

In summary, the case for modularity is not easy to make.3

Nor is the case of relational contracting any more compelling. I am sure it exists and, like modularization, needs to be included in any characterization of the world economy. While relational contracting and modularization are ancient organizational forms, they have never been so widespread. The question is, can they long withstand the ravages of short-run profit maximization? I am sure relational contracting is essential for some systems, and those who promote it give compelling examples. But it appears that it is expensive, hard to maintain, decays into arms-length modularization or worse yet, short-run cost pressures replace it with predatory practices.

Gary Herrigel surveyed several industries where relational contracting would be most likely to appear—high wage component manufacturing in Europe and the United States (Herrigel 2004). He does not find evidence that a collaborative community prevails or is even emerging in the New Economy. Instead, he finds that the Original Equipment Manufacturers (OEMs) pursue a wide variety of strategies with their suppliers. They can be cooperative and open on one component, and drive hard bargains on another. A single OEM will preserve substantial in-house capability and know-how in some areas but will outsource in others. Most important, the tendency appears to him to be to strengthen in-house capability rather than increasing modularity, thus a move to Cell 4, captive suppliers and vertical integration.

“OEMs are factionalized and different factions within the firm have different strategies for what decentralization should be and indeed, of what should be decentralized” (47). He finds that frustrated engineers scrapped the process of creating standards that modules need and instead design customer-specific or product-specific solutions. Furthermore, he cites evidence that suppliers are increasingly unable to provide anything but standardized and low-end goods, making them vulnerable to short-term abuse from OEMs.

The contradictions extend to firm structures in the relationship of customer and suppliers, and we find the emergence of something quite destructive of the collaborative community: consolidation among suppliers.

Many first-tier-suppliers have responded by embarking on a wave of vertical integration (through mergers, acquisitions, and joint-ventures) and geographic expansion to gain the ability to provide their customers with modules on a global basis. Thus we are seeing simultaneous trends toward vertical disintegration (by automakers) and vertical integration (among first-tier suppliers) that—in combination with globalization—is helping to create a new global scale supply-base capable of supporting the activities of final assemblers on a worldwide basis (49).

This vertical integration is a move from cells 1–3 to the concentration cell, 4, rather than a transition to a collaborative community; it suggests an economy based on self-interest rather than collaborative relations.

The contrast with the enthusiasm of earlier proponents of modularity is striking. For example, in 1996 Sanchez and Mahoney could confidently state that “We explain how advanced technological knowledge about component interactions can be used to fully specify and standardize the interfaces that make up a modular product architecture, creating a nearly independent system … of ’loosely coupled’ components” (63). This enthusiasm is repeatedly asserted by most contributors to the 2006 volume, The Firm as a Collaborative Community (Heckscher and Adler 2007). In contrast, Herrigel cites the research of Whitford and Zeitlan (which we will come to) that indicates a high degree of self-interested behavior on the part of OEMs. But surprisingly, despite his solid evidence of increasing concentration, he does not see a transition to vertically integrated firms, but only instability. He concludes:

It is a continuously self-reproducing situation of instability and the recomposition of relations within and between firms in high-wage regions of the global economy. There is no ‘‘transition’’ to a more stable, unitary set of relations. The noise is the thing (55).

Richard Appelbaum sees less noise than consolidation. He examines the production of apparel and textiles, toys, footwear, home electronics, and other consumer goods such as washing machines and sewing machines in East Asia, particularly China. “There is growing evidence that consolidation in consumer goods industries, with increasingly integrated production and distribution systems between large retailers and contractors, may be increasing the degree of vertical integration in global supply chains”(Appelbaum 2008). Initially, consolidation occurred in the large retailers, but Appelbaum sees it as also emerging in the giant contractors based in East Asia, where they are producing much of the world’s consumer goods. China, in particular, is anticipated to soon be producing one-half of the worlds’ branded consumer goods. “The power of the giant retailers—itself unforeseen by advocates of deverticalization—is being challenged not only by the power of giant producers, but even by giant wholesalers.” The area has seen increasing integration among its 39 largest retailers (which are among the world’s 500 largest corporations), allowing them to bring their suppliers under direct control. Retail consolidation in turn has favored large contractors and suppliers, and giant factories are making their appearance. Thus our concentrations cell should include not only producers that sell to the final consumer, but also producers that sell to retailers, and the retailers themselves. To give just one example, the world’s largest maker of branded athletic and casual footwear, Yue Yuen/Pou Chen, has 30 large clients in addition to being Nike’s biggest supplier; has 280,000 workers worldwide and 110,000 of them in just one complex; has grown by 50% in the past 5 years; and has a network of over 2,100 wholesale distributors and 640 retail outlets in China alone (Appelbaum 2008).

The electronics industry, ripe for modularity, is actually a good example of consolidation. The five largest contractors are all based in North America, four of them in Silicon Valley. They once had scores of contract manufacturers but now have consolidated their operations and are moving towards only three or four each. These consolidated suppliers were expected to capture 65% of the supply business by 2003. Singapore has a thriving electronics industry, with its giant OEMs controlled by or closely collaborating with the five North American contractors. Singapore’s flagship firm is the world’s third-largest electronic supplier and one of its contractors is the fourth largest contract chip manufacturer in the world. By 2005 mergers and acquisitions have reduce the large number of firms in this huge industry to three. Thus the concentration in electronics is in the four lead firms and in their suppliers as well (Sturgeon 2003).

An important survey of large OEMs in various manufacturing sectors in the United States gave little evidence of the collaborative relations that should be supplanting vertically organized huge firms that are decentralizing and outsourcing (Whitford and Zeitlin 2004). Unless the suppliers of the lead firms are themselves huge, they are largely dependent suppliers forced to cut costs by de-unionizing and lowering wages and absorbing all the risks that the large vertically integrated firms once had to bear. This is supported by the research of economist William Tabb, surveying the global economy. He finds that such things as relational contracting represent “a form of dominance not so different in impact from direct ownership (although with cost savings to the dominant partner who leaves the consequences of sudden market shifts and other uncertainties to the supplier)” (Tabb 2004:275). He continues: “Those who are successful will be the few, the proud, the gigantic. The process of consolidation should leave a half dozen or so dominant player in the world in each major industry segment. The very few auto producers, airline alliances, and pharmaceutical companies will then have vastly more market power” (Tabb 2004:282).


To conclude, the prospects for an emerging "collaborative community" appear to be quite limited. There are interesting and even exciting examples of collaborative communities utilizing relational contracting, but they are few. The somewhat less benign form of modularization with reasonably non-exploitative customer–supplier relations also exists, probably in larger numbers, but these are not the dominant firms. The dominant firms appear to be vertically integrated hierarchies with captive suppliers, at least in those areas where one would expect to see the appearance of new forms such as relational contracting and modularity—textiles, apparel, most consumer goods (from toys to durables), tractors and agricultural equipment, and of course, electronics. Vertical integration is appearing in all of these industries, and in others such as finance, energy, the media, transportation, and health care.

But even this is too simple. The largest firms include producers, retailers, and wholesalers. To Gereffi’s important distinction between producer-driven and buyer-driven commodity chains (Gereffi 1996) we should add wholesalers, with their warehousing and delivery services. Each of the three, producers, buyers, and wholesalers, are fighting for dominance, attempting either to integrate the functions of the other two, or to make them captives. At least the scale of this consolidation is without precedent. Although I have not given it sufficient thought, it may be that it is the first time in history that, on a widespread basis, the producer is not necessarily dominant, since wholesalers might dominate producers and retailers, or retailers dominate producers and wholesalers. The phenomenon is not limited to mass-produced consumer goods but extends to most goods and, importantly, most services. I think we are seeing something quite new here, and it is certainly not the collaborative community.

Another new development of the emerging global economy is the tight coupling of a product and service chain that is increasingly integrated around very large players. This allows disturbances to spread rapidly and makes recovery more difficult. Concentrated firms increase their profits by relying upon the cheapest supplier, but if that supply is interrupted by natural disasters or social disorders (or terrorists), they are in the position of Simon’s watchmaker who cannot tolerate interruptions. The generic market model, despite its inefficiencies, is more resilient. The redundancies of many producers and many suppliers will tolerate shocks. These issues are well explored by Barry Lynn (Lynn 2005).

The one thing that is not new, but as old as industrialization in the United States and now the developing world, is the exploitation of labor and the destruction of communities and the environment. For just a few references see, for the United States (Greenhouse 2008a; Rubenstein 2006); for Asia (Appelbaum 2008; Herrigel and Wittke 2008; Luthje 2004; MacDuffie and Helper 2006; Vind and Fold 2007; Whitford and Zeitlin 2004). The New Economy theorists do not go so far as to say that the inequalities of income and the marginalization of low-skilled labor, while regrettable, are necessary for economic growth (Shapiro 2008), but if they mention these problems at all it is only in passing.


Linear interactions” are those in expected and familiar production or maintenance sequence, and those that are quite visible even if unplanned. Think of an assembly line. “Complex interactions” are those of unfamiliar sequences, or unplanned and unexpected sequences, and are either not visible or not immediately comprehensible. Think of a chemical refinery or nuclear power plant (Perrow 1999:78).


Loosely coupled systems tolerate processing delays, changes in the order of sequences, and alternative methods; there is slack in resources, and buffers, redundancies, and substitutions are available. Tightly coupled systems are not tolerant in these ways (Perrow 1999:96).


Whenever we come across injunctions to “balance these two effects,” we should be cautious. It is similar to Aristotle’s frequent withdrawal to the weak notion of the “golden mean.” To choose the two endpoints of a dimension is in effect to rig the argument by determining the golden mean in advance. To avoid the cost of building modules we should assemble more parts; to avoid assembly costs we should build more modules. Therefore we should choose the mean of some, but not too many, modules. But what if we changed the endpoints to maximizing arms-length bargaining versus intimate contact with suppliers? The mean would now be maximum modularity. It provides some bargaining and some intimacy. But this ignores the costs of building modules and the costs of assembly.



Colleagues at Stanford’s Center for International Security and Cooperation provided help and encouragement, as did Jennifer Bair of Yale, and Jennifer Bernal, an undergraduate at Stanford, was a most able research assistant.

Copyright information

© Springer Science+Business Media B.V. 2009