There is no doubt that hold-up problems are of central concern to business people. In negotiating joint venture agreements, venture capital contracts or any of a number of other business deals, much time is spent on building in protections against hold-ups. At the same time, such contracts are prima facie evidence that hold-up problems do not get resolved solely by integration of buyer and seller into a single party—the firm. Indeed, there seems to be something of a trend today toward disintegration, outsourcing, contracting out, and dealing through the market rather than bringing everything under the umbrella of the organization. This trend has seen the emergence of alternative, often ingenious solutions to hold-up problems.
The pattern of relations between Japanese manufacturing firms and their suppliers offers a prominent instance where the make-buy dichotomy and related theorizing have been less than satisfactory. Although the basic patterns apply in a number of industries (including, for example, electronics), the practices in the automobile industry are best documented (Asanuma 1989, 1992). These patterns have spread from Japan to the auto industry in the United States and elsewhere, and from autos to many other areas of manufacturing. These practices feature long-term, close relations with a limited number of independent suppliers that seem to mix elements of market and hierarchy. Apparently, these long-term relations substitute for ownership in protecting specific assets.
Two points of contrast in the treatment of specific investments between traditional U.S. practice and the Japanese model present particular problems for the received theory. The first concerns investments in designing specialized parts and components. Traditional U.S. practice featured either internal procurement or arm's-length, short-term contracting. Design-intensive products were very often procured internally (Monteverde and Teece 1982). When products were outsourced, the design was typically done by the auto-maker, with the drawings being provided to the suppliers. This pattern is what hold-up stories would predict, for the investment in design is highly specific and probably cannot be protected fully by contracts; thus, external suppliers will not make such relationship-specific investments, for fear that they will be held up by buyers after their investments are in place. In stark contrast, it is normal practice for Japanese auto firms to rely on their suppliers to do the actual design of the products supplied. The design costs are then to be recovered through the sale price of the part, with the understanding that this price will be adjusted in light of realized volumes.
A second contrast: Traditional U.S. practice has been that physical assets specific to an auto-maker's needs are owned by the auto-maker. This clearly applies in the case of internally procured items, but it also holds in cases where the assets are used by the external supplier in its own factory. For example, the dies used in making a particular car part will belong to the auto-maker, even though they are used in the supplier's plant on the supplier's presses. Again, this accords well with the transaction cost story of potential hold-up by the auto-maker. In Japan, in contrast, these specific investments are made by the supplier, who retains ownership of the dies. This would seem to present the auto-maker with temptations to appropriate the returns on these assets, once the supplier has made the relationship-specific investment. Moreover, because the Japanese auto manufacturers typically have a very small number of suppliers of any part, component or system, the supplier would also seem to be in a position to attempt opportunistic renegotiation by threatening to withhold supply for which there are few good, timely substitutes.
The Japanese pattern is directly at odds with transaction cost theory. Meanwhile, the divergence in ownership of the dies between the two countries presents problems for attempts to explain ownership allocation solely in terms of providing incentives for investment.
In Japanese practice, explicit contracting is not used to overcome the incentive problems involved in outsourced design and ownership of specific assets. In fact, the contracts between the Japanese auto-makers and their suppliers are short and remarkably imprecise, essentially committing the parties only to work together to resolve difficulties as they emerge. Indeed, they do not even specify prices, which instead are renegotiated on a regular basis. From the hold-up perspective, the prospect of frequent renegotiations over the prices of parts that are not yet even designed would certainly seem problematic.
The key to making this system work is obviously the long-term, repeated nature of the interaction. Although supply contracts are nominally year-by-year, the shared understanding is that the chosen supplier will have the business until the model is redesigned, which lasts typically four or five years. Moreover, the expectation is that the firms will continue to do business together indefinitely. There has been very little turnover of Japanese auto parts suppliers: over a recent eleven-year period, only three firms out of roughly 150 ceased to be members of kyohokai, the association of first-level Toyota suppliers (Asanuma 1989).
The familiar logic of repeated games, that future rewards and punishments motivate current behavior, supports the on-going dealings. An attempted hold-up would presumably bring severe future penalties. As importantly, the amount of future business awarded to a supplier is linked to ratings of supplier performance. The auto companies carefully monitor supplier behavior—including cost reductions, quality levels and improvements, general cooperativeness, and so on—and frequent redesigns allow them to punish and reward performance on an on-going basis. In this sense, supplier relationships in Japan are potentially less, not more, locked in than in the traditional U.S. model, where at the corresponding point in the value chain, the supplier is typically an in-house division or department.
Having a small number of suppliers is crucial to the Japanese system. It reduces the costs of monitoring and increases the frequency of transacting, both of which strengthen the force of reputation. Also, the rents that are generated in the production process do not have to be shared too widely, providing the source for significant future rewards. This logic underlies the normal "two-supplier system" used at Toyota. There is more than one supplier to permit comparative performance evaluation, to allow shifting of business as a reward or punishment, to provide insurance against mishaps, and perhaps to limit the hold-up power of each supplier, but the number is not chosen to minimize hold-ups.
The relationship is marked by rich information sharing, including both schedules of production plans necessary for just-in-time inventory management and also details of technology, operations, and costs. The auto-makers also assist the suppliers in improving productivity and lowering costs: Technical support engineers are a major part of the auto-makers' purchasing staff, and they spend significant amounts of time at the suppliers' facilities. All this in turn means that potential information asymmetries are reduced, which presumably facilitates both performance evaluation and the pricing negotiations.
Perhaps the major problem in the system may be that the auto-makers are inherently too powerful and thus face too great a temptation to misbehave opportunistically. Indeed, many Japanese observers of the system have interpreted it in terms of the auto-makers' exploitation of their power. One counterbalance to this power asymmetry is the supplier association, which facilitates communication among the suppliers and ensures that if the auto company exploits its power over one, all will know and its reputation will be damaged generally. This raises the cost of misbehavior. In this regard, the fact that Toyota itself organized an association of the leading suppliers for its Kentucky assembly plant is noteworthy (Milgrom and Roberts 1993).
An alternative solution to this imbalance would be for the auto-maker to own the dies, as in the United States. Here a property rights explanation may be useful: Under this arrangement, the supplier would not have the same incentives to maintain the dies, since it must be very hard to contract over the amount of wear and tear and its prevention.
Another significant shift in the organization of production is illustrated by Nucor, the most successful steel maker in the United States over the past 20 years. Nucor operates mini-mills, which use scrap (mainly car bodies) as raw material for steel production. After an initial technological breakthrough, Nucor started to expand aggressively (Ghemawat 1995). The strategy required much capital, and to save on capital outlays, Nucor decided to outsource its entire procurement of steel scrap. Traditionally, mini-mills had integrated backwards, partly to secure an adequate supply of raw material and partly because sourcing entails substantial know-how and so was considered "strategically critical". Chaparral Steel, another big mini-mill operator, continues to be integrated backwards, for instance.
In a break with the tradition, Nucor decided to make a single firm, the David J. Joseph Company (DJJ), its sole supplier of scrap. Total dependence on a single supplier would seem to carry significant hold-up risks, but for more than a decade, this relationship has been working, smoothly and successfully. Unlike in the Japanese subcontracting system, there are certain contractual supports. Prices are determined by a cost-plus formula to reflect market conditions, and an "evergreen" contract specifies that the parties have to give warning (about half a year in advance) if they intend to terminate the relationship. Even so, there is plenty of room for opportunism. Despite transparent cost accounting (essentially, open books), DJJ can misbehave, since realized costs need not be the same as potential costs. Asset specificity remains significant even with the six-month warning period, since a return to traditional sourcing and selling methods would be quite disruptive and expensive for both sides. Indeed, one reason why the partnership has been working so well may be the high degree of mutual dependence: Nucor's share of DJJ's scrap business is estimated to be over 50 percent.
The success of Nucor's organizational model has led other mini-mills to emulate and refine it. In England, Co Steel has gone as far as relying on its sole supplier to make ready-to-use "charges", the final assemblage of materials to go into the steelmaking ovens. The production technology for charges is quite complicated: About twenty or thirty potential ingredients go into each mixture, with the mix depending on the desired properties of the final product, and big cost savings can be had by optimizing the use of the different inputs. This activity entails much know-how and requires extensive information exchange with the steel plant to match inputs with final product demand. The charges must be prepared by the supplier on Co Steel's premises, both for logistical reasons and to facilitate information sharing. In transaction cost economics, such a cheek-by-jowl situation would be an obvious candidate for integration. Yet, the industry is moving in the direction of disintegration in the belief that specialization will save on costs by eliminating duplicate assets, streamlining the supply chain, and providing better incentives for the supplier through improved accountability.
Related experiments of "inside contracting" include Volkswagen's new car manufacturing plant in Brazil, where the majority of the production workers in the factory are employees, not of Volkswagen, but of subcontractors that provide and install components and systems on the cars as they move along the line. It is too early to tell whether other firms will return to inside contracting, which used to be quite common in the United States up to World War I (Buttrick 1952), and whether such a move will be successful. But evidently, even potentially large hold-up problems have not deterred recent experimentation.
Another illustration of close coordination without ownership is provided by airline alliances, which have proliferated in recent years. Coordinating flight schedules to take advantage of economies of scope requires the parties to resolve an intricate set of issues, particularly ones related to complex "yield management" decisions on how to allocate seats across different price categories and how to shift prices as the flight date approaches. Information and contracting problems abound, and it is hardly surprising that tensions occasionally surface. For instance, KLM and Northwest Airlines recently ran into a dispute that had to be resolved by dismantling their crossownership structure. But interestingly, this did not prevent KLM and Northwest from deepening their commitment to their North Atlantic alliance by agreeing to eliminate, over a period of years, all duplicate support operations in the United States and Europe. With the completion of this deal, KLM and Northwest have made themselves extraordinarily interdependent in one of the most profitable segments of their business. A 13-year exclusive contract, with an "evergreen" provision requiring a three-year warning before pull-out, is the main formal protection against various forms of opportunism, but undoubtedly the real safeguard comes from the sizable future rents that can be reaped by continued good behavior.
Why don't the two airlines instead integrate? Regulations limiting foreign ownership and potential government antitrust objections are a factor, as may be tax considerations. However, an explanation we have been given is that airline cultures (and labor unions) are very strong and merging them is extremely difficult. Pilot seniority is a particularly touchy issue.
In property rights theory, the boundaries of the firm are identified with the ownership of assets, but in the real world, control over assets is a more subtle matter."Contractual assets" can often be created rather inexpensively to serve some of the same purposes that the theory normally assigns to ownership: to provide levers that give bargaining power and thereby enhance investment incentives. What we have in mind here are contracts that allocate decision rights much like ownership; for instance, exclusive dealing contracts such as Nucor's, or licensing agreements of various kinds. Such "governance contracts" are powerful vehicles for regulating market relationships. With increased disintegration, governance contracts seem to have become more nuanced and sophisticated. They place firms at the center of a network of relationships, rather than as owners of a clearly defined set of capital assets.
BSkyB, a satellite broadcasting system in Rupert Murdoch's media empire, is an example of a highly successful organization that has created its wealth, not by owning physical assets, but by crafting ingenious contracts that have given it influence over an effective network of media players. Satellite broadcasting requires a variety of highly complementary activities, including acquisition and development of programming, provision of the distribution system (satellites, transmitters, and home receivers) and development of encryption devices (to limit reception to those who pay), all of which must be carried out before the service can be offered. Other similarly complex and innovative systems of complements, like electric lighting systems or early computer systems, were largely developed within a single firm. BSkyB instead relies on alliances with other firms. Topsy Tail is even more of a "virtual company". It employs three people, but has sales of personal appearance accessories (combs, hair clips and such) approaching $100 million. Topsy Tail conceives of new products, but essentially everything involved in developing, manufacturing, and distributing them is handled through an extensive contractual network. Benetton and Nike, to take some bigger and more conventional firms, also extensively rely on outsourcing and a small asset base. The critical asset in these cases is of course control of the brand name, which gives enormous power to dictate how relationships among the various players are to be organized.
Microsoft and the web of inter-firm relations centered around it provide another illustration. The stock market values Microsoft at around $250 billion, which is more than $10 million per employee. Surely very little of this is attributable to its ownership of physical assets. Instead, by leveraging its control over software standards, using an extensive network of contracts and agreements that are informal as well as formal and that include firms from small start-ups to Intel, Sony, and General Electric, Microsoft has gained enormous influence in the computer industry and beyond. We are not experts on Microsoft's huge network of relationships, but it seems clear that the traditional hold-up logic does poorly in explaining how the network has developed and what role it serves. If one were to measure asset specificity simply in terms of separation costs, the estimates for breaking up some of the relationships—say, separating Intel from Microsoft—would likely be large. Yet these potential losses do not seem to cause any moves in the direction of ownership integration.
A similar pattern can be observed in the biotechnology industry (Powell 1996). As in the computer industry, the activities of the different parties are highly interrelated, with different firms playing specialized roles in the development and marketing of different products. Most firms are engaged in a large number of partnerships; for instance, in 1996 Genentech was reported to have 10 marketing partnerships, 20 licensing arrangements, and more than 15 formal research collaborations (Powell 1996, p. 205). Significant relationship-specific investments are made by many parties, and potential conflicts must surely arise after these investments are in place. Yet the system works, thanks to creative contractual assets—patents and licensing arrangements being the oldest and most ingenious—but also to the force of reputation in a market that is rather transparent, because of the close professional relationships among the researchers.
This pattern, however, did not become standard until decades after the founding of the industry. Earlier, something akin to the practices associated now with the Japanese was the norm. See Helper (1991).
An alternative story is more in the line of Williamson's earlier discussions emphasizing inefficiencies in ex post bargaining. The useful life of a die far exceeds the one-year contracting period. If the supplier owned the die, changing suppliers would require negotiating the sale of the die to the new supplier, and this could be costly and inefficient.
Interestingly, Toyota followed U.S. practice in supplying the dies used by at least some of the suppliers to its Kentucky assembly plant (Milgrom and Roberts 1993).
Taylor and Wiggins (1997) argue that these long-term relations are also the means used in the Japanese system to solve moral hazard problems with respect to quality.
Baker et al. (1997) present a formal analysis of the choice between external and internal procurement, taking into account the important fact that long-term relational contracts can be maintained both within a firm as well as across firms.
Strikingly, as automobile electronics have become more sophisticated and a greater part of the cost of a car, Toyota has ceased to rely exclusively on its former sole supplier, Denso, and has developed its own in-house capabilities in this area. Arguably, this was to overcome information asymmetries and their associated costs (Ahmadjian and Lincoln 1997). In contrast, see the discussion of the effects of Ford's complete reliance on Lear for developing seats for the redesigned 1997 Taurus (Walton 1997).
See Segal and Whinston, 1997, for a model in the property rights spirit that is relevant to these issues.