The question of why and under what circumstances economic activities are vertically integrated (that is, brought under the governance of a single corporation, as opposed to being governed by contracts among several firms or persons) has remained a vexing problem in economics for over 70 years. Professor Bainbridge raises a puzzle [updated links: the original puzzle; and a follow-up] that seems particularly difficult to explain: the use of franchise contracts by Subway and of vertical integration by Starbucks. As he writes,
What bugs me about the Subway v. Starbucks problem is that I can't see any reason to believe that Subway's transaction cost schedule is going to differ from that of Starbucks. The businesses are essentially identical; yet, the business model is quite different.
Professor Bainbridge offers several possible reasons for the difference in governance structure, none appealing:
Professor Bainbridge comments that "behavioral economic concepts like path dependence and/or herd behavior [rather than] rational choice" may be relevant. Perhaps it's worth pointing out that path dependence does not require behavioral economics and is consistent with rational choice under transaction costs. If costs change over time, and if costs when the value network was being built favored a structure that later became disfavored, that structure might persist if the cost of switching structures is greater than the benefit from switching to a lower-cost structure.
Several bloggers, including Tyler Cowen, Alex Tabarrok, and Arnold Kling, suggested that Starbucks' strategy of building stores close together may have discouraged bidders for franchises, as they may have feared dilution of franchise value by the subsequent opening of Starbucks stores nearby. Eliminating that fear by granting franchise-buyers exclusive geographic rights would require Starbucks to change to an inferior strategy; and other contractual methods for reducing the fear may be more costly than vertical integration. This idea is attractive but leaves much unexplained, such as why a strategy of "flooding the zone" (in Kling's term) was valuable to Starbucks but not to other chains, and why the value of that strategy to Starbucks exceeds the (presumed) disadvantages of forgoing the franchise model.
Two major contributions to transaction cost economics, I think, shed valuable perspective on this puzzle. The first contribution I would associate with Harold Demsetz, the second with Richard Langlois.
1. The difference in costs between the vertically-integrated and franchise-contracted governance structures may not be large; it may not even be significant.
To understand why this is so, consider a corporation as a "nexus of contracts." That is, the corporation is an artificial person which can hold property and contract, and each employee (as well as every other factor) contracts with the corporation. An employee has a default "contract" consisting of the statutory provisions governing the relationship between employee and employer; a personal employment contract may modify or supplement those default provisions. So what is the difference between a franchise-owner and an employed branch manager? Each has a contract with the corporation, which may contain very similar terms. To each, the corporation issues certain requirements to guarantee a replicable brand experience across branches. In each structure, the corporation may have a right to change its brand requirements at will. In each structure, the manager's or franchise-owner's performance in upholding the brand experience is evaluated by the corporation, and in each case failure to uphold the brand experience can lead to termination of the relationship. The actual economic activities involved � requirements issued, monitoring, punishment for failure to perform � may be very similar; and, depending on how the franchise contract is written, the costs to a terminated manager or franchise-owner, and the rewards to a successful manager or franchise-owner, may be very similar. (For instance, employee-managers may receive as part of their compensation a share of profits in their store. Starbucks famously gave stock options to part-time clerks, so they may well have been generous in rewarding store managers.)
Thus, very similar activities may be governable by the two structures at very similar costs. In fact, it is not uncommon to see alternative governance structures at use in the same industry at the same time, with the companies using alternative structures experiencing very similar product costs and very similar levels of profitability. For instance, for many decades Ford was highly vertically integrated and Chrysler highly disintegrated, yet the two companies had similar success.
Suppose there were no employment law, so that every relationship had to negotiate its contract; call the contract the chain and a store manager would negotiate the "ideal contract." If that ideal contract would closely resemble the default contract established by employment law, then employer and employee can avoid bargaining costs by accepting the standard employment relation in lieu of a contract. As the ideal contract varies from that assumed by employment law, the standard employment relation becomes an imperfect governance structure; eventually the employer and employee are justified in bearing the costs of negotiating an employment contract that alters the standard terms. And as the nature of the relationship diverges still further from the standard employment relation, they may forgo an employment relationship altogether and rely entirely on contractual terms. In this case, they have ceased to be vertically integrated, and have become independent contractors.
From this perspective, the alternatives - vertical integration, or contracts - are not dichotomous, but ranges on a continuum. For some points on the continuum, the advantages of an employment relation over contracting may be significant; for others, the advantages of contracting over employment may be great; but for some, the value of the two structures may be equal.
It's possible that Starbucks and Subway are nearly indifferent as to which structure governs relations with store managers. If Starbucks and Subway were indeed indifferent between structures, it might be very difficult to pin down the specific economic factors which tilted Starbucks toward integration and Subway toward contracting - which straws, as it were, tipped the balance. Evidence suggests that Starbucks is nearly indifferent between structures: Starbucks currently has 1500 licensed franchises and 5700 corporate-owned stores. However, Subway, which has 17,500 franchises and only one company-owned unit which it uses for research and development, appears devoted to the disintegrated model.
2. Dynamic transaction costs may be more significant than continuing transaction costs.
Dynamic transaction costs are transaction costs which are of only transient importance. They are incurred during the period of entrepreneurship when the value network is created, not during the operating period when relationships are stable. In transaction cost economics as developed by Oliver Williamson, one would look to the operating period and ask which governance structure offers the lowest operating costs. However, the costs of entrepreneurship are certainly not negligible. If one structure is favored during the entrepreneurial period, but a different structure favored during the operating period, then the operating period may commence with a less-than-optimal structure. If the advantages of the optimal structure are great enough to justify the transaction costs involved in switching structures, then the value network will change structures early in the operating period; if not, it may retain a less-than-optimal structure.
According to Langlois and Robertson (1995), it has often been the case that vertical integration is sought during entrepreneurial periods, yet companies subsequently disintegrate after the operating environment becomes stable. Suppose Starbucks is an example of a company that, due to dynamic transaction costs, adopted a vertically integrated structure during its entrepreneurial period, and then finds itself in a stable operating environment in which it would be better off with franchise structure like the one Subway has. Still, it may not be worthwhile for Starbucks to make the transition, if the gain from an optimal structure would be small.
What might have been the source of these dynamic transaction costs? Suppose Starbucks was offering an innovative product whose value potential franchisees would have difficulty evaluating. Then during the period when the value of Starbucks' product was unproven, potential franchise-owners might underbid, in the opinion of Starbucks, for franchises. It might be more profitable for Starbucks to adopt a vertically integrated structure until the value of its innovation was proven.
There are historical analogues for this story. Morris Silver in Enterprise and the scope of the firm (1984) argued that Bessemer had this problem in selling his innovative steel process. Bessemer was initially dissatisfied with the license terms offered by existing steel producers - each wanted an exclusive license to Bessemer's patent, while he wanted to license his process widely so that all producers could adopt it immediately. Rather than grant an exclusive license, Bessemer opened his own steel mill to prove that his process reduced costs; once he had proven the process's value, steelmakers eagerly licensed his patent without exclusivity. In his autobiography Bessemer described his decision:
And yet, with all this newly developed power, I was paralyzed for the moment in the face of the stolid incredulity of all practical iron and steel manufacturers.... Each [of the large steel manufacturers at Sheffield] required an absolute monopoly of my invention if he touched [my process] at all. This I fully made up my mind to resist, by adopting the only means open to me - namely, the establishment of a steel works of my own in the midst of the great steel industry of Sheffield. My purpose was not to work my process as a monopoly, but simply to force the trade to adopt it by underselling them in their own market, which the extremely low cost would enable me to do.
Considering that the value of the Starbucks innovation is hidden from me even in retrospect (though it was certainly great, or Starbucks would not have been so successful in the marketplace), it would not surprise me if potential franchise-owners had difficulty appreciating that value in prospect. Certainly, the success of Starbucks surprised most observers. Prospective franchise-owners may have been better able to evaluate the Subway offering; there seems nothing particularly surprising about Subway's business performance.
The dynamic-transaction-cost story is consistent with the Subway corporate history. Subway was founded in 1965, and by 1974 had grown to a chain of 16 restaurants all owned and operated by the corporation. Only then did it turn to a franchising model. Now that Starbucks has proven the value of its product, it may turn increasingly to licensing. This shift may be underway: see here and here.
To me, the dynamic explanation seems more plausible than explanations built around monitoring during the operating period. There are notorious difficulties transacting in information about innovations during entrepreneurial periods; the problem of obtaining information about performance of contractors in a stable operating environment appears much more tractable.
Relationship economics is a version of transaction cost economics emphasizing the high costs and the considerable value of personal relationships. Does it have anything to contribute to the puzzle? It helps in two ways:
(1) Relationship economics emphasizes the costs of relationship-building during the entrepreneurial phase and thus directs attention to dynamic transaction costs, rather than operating transaction costs such as those due to opportunism, shirking and the like. In other words, it supports Langlois's approach. Within a dynamic transaction cost approach, it directs attention toward specifically those costs associated with creation of personal relationships.
(2) It provides some clues not only about which transactions will be executed, but also about the distribution of profits. Businesses seeking profits care both about finding and executing transactions that create profits, and about the distribution of profits among the transacting parties. Coase directed his concern to whether or not transactions were completed; by and large, he left the problem of how profits are divvied up to other methods of analysis such as game theory. But game theory needs a description of payoffs in order to explain the results of bargains; and relationship economics offers some clues as to what these may be. Relationship economics gives an explanation for profits and for the competitive advantage of various players in capturing those profits. These explanations show that temporary vertical integration during construction of a value network may be a useful strategy for capturing a larger share of the long-term profits generated by the value network.
I'll have more on these points later.
Copyright © 2003 Paul Jaminet. All rights reserved.