1. The Observed Transformation
The principal development that induced forward vertical integration into dis- tribution to occur in the second half of the nineteenth century was the appearance of the railroads (Porter and Livesay, 1971, p. 55). To be sure, there were other significant technological developments, including the telegraph (Chandler, 1977, p.189), the development of continuous processing machinery (Chandler, 1977, pp. 249-53), the refinement of interchangeable parts manufacture (Chandler, 1977, pp. 75-77), and other technological developments that supported mass manufacture (Chandler, 1977, chap. 8). Without the low cost, reliable, all-weather transportation afforded by the railroad, however, the incentive to integrate forward would have been much less.
The extensive forward integration from manufacturing into distribution that occurred during the last thirty years of the nineteenth century was, however, highly varied. The differences are the conditions to be explained. From least to most, integration into distribution varied as follows: (1) none, in which case the prevailing wholesale and retail structure continued; (2) integration into wholesale but not retail; and (3) retail (which usually included wholesale). A temporal dimension is useful to bear in mind in all of this, since the factors that favor integration at one point in time may not continue indefinitely. Also, mistaken forward integration warrants attention. Presumably such integration errors are less likely to be imitated or renewed and for this reason will be less widely reported. But a predictive theory of forward integration should explain both failures and successes.
1.1. UNCHANGED DISTRIBUTION
The sectors of the American economy where independent wholesalers continued to serve as distributors of goods to independent retailers included “the complex of goods sold through retail outlets such as grocery, drug, hardware, jewelry, liquor, and dry goods stores” (Porter and Livesay, 1971, p. 214).
Whether independent wholesalers and retailers provided the full set of support services, however, or whether manufacturers performed some of these depended on product differentiation. As discussed below, the role of the middleman was reduced for branded products.
Manufacturer involvement in wholesaling functions of three kinds can be distinguished: preselling, inventory management, and facility ownership. The cigarette was the “orphan” of the American tobacco industry until 1880. The appearance of the Bonsack machine for the continuous processing of cigarettes in 1881 and its adoption by James Duke quickly changed that (Chandler, 1977, pp. 249-50). Duke cut the prices of cigarettes drastically to reflect manufacturing economies and coupled this with massive advertising (Chandler, 1977, pp. 290- 92). Although Duke continued to sell through jobbers and retailers, he also organized a network of sales offices in the larger American cities in which salaried managers worked to coordinate marketing and distribution (Chandler, 1977, p. 291). The appearance of continuous processing machinery and attendant economies of scale also gave rise to branding and subsequent efforts to presell product and manage distribution in “matches, flour, breakfast cereals, soup and other canned products, and photographic film” (Chandler, 1977, p. 289).
Whitman’s decision to use two different methods to merchandise candy is noteworthy. Wholesalers were bypassed in the sale of high-grade, packaged candies. Small, inexpensive bar and packaged candies, by contrast, were sold through the usual jobber and wholesale grocer network. Control of the whole- saling function for the former was arguably more important for quality control purposes. The high-grade items were “sold directly to retailers so that the company could regulate the flow of the perishable items and avoid alienating customers” (Porter and Livesay, 1977, p. 220), who were presumably prepared to pay a premium to avoid stale candy.
Ownership of wholesaling was reserved for products that required special handling, mainly refrigeration. Meatpacking and beer are examples (Chandler, 1977, pp. 299-302). Gustavus Swift was the‘leading innovator in meatpacking. He recognized that the practice of shipping live beef east involved considerable waste and proposed to eliminate this by slaughtering and dressing cattle in the Midwest and shipping the beef east in refrigerated cars. Implementing this transformation, however,’was not easy. It met with resistance both from Eastern butchers, packers, and jobbers (Porter and Live- say, 1971, p. 169) and from the railroad interests (Chandler, 1977, p. 300). In order to execute his strategy, Swift had to build his own refrigerator cars and ice houses and construct a network of branch houses that provided ‘ ‘refrig-erated storage space, a sales office, and a sales staff to sell and deliver the meat to the retail butchers, grocers, and other food shops” (Chandler, 1977, p. 300).
Integration into final sales and service represented a more ambitious variety of forward integration. Three classes of products can be distinguished: specialized consumer nondurables; consumer durables requiring information aids, credit, and follow-on service; and producer durables requiring the same.
Kodak photographic film is an example of the first kind. George Eastman developed a paper-based film to replace the glass plates then in use in the early 1880s. The film required a special camera, however, and developing the film was complex. Meeting little success among professional photographers, Eastman and his associates set themselves to developing the amateur market. “To sell and distribute his new camera and film and to service their purchasers, Eastman . . . created a worldwide marketing network” (Chandler, 1977, p. 297). He explained his decision to eliminate independent wholesalers as follows:
The wholesaler or jobber is a detriment to our business because a large proportion of it is in sensitized goods which are perishable… We have organized our distribution facilities so as to get the goods into the hands of the consumer as quickly as possible. Our sensitized goods carry an expiration date. Our own retail houses… have been educated to control their stocks very accurately so that the goods are kept moving. [Porter and Livesay, 1971, p. 178]
Consumer durables for which forward integration into retailing was at- tempted were sewing machines and, later, automobiles. Upon resolution of thé legal contests over patents in 1854, sewing machine patents were released to twenty-four manufacturers. Only three attempted to integrate forward, however, and only they remained major factors in the industry. The pattern adopted by Singer in the fourteen branches that it had opened by 1859 was to staff each with “a female demonstrator, a mechanic to repair and service, and a salesman or canvasser to sell the machine, as’well as a manager who supervised the others and handled collections and credits” (Chandler, 1977, p. 303).
Although automobiles were mainly sold through franchised dealers rather than company-owned outlets, the Ford Motor Company and others required their dealers to “supply full demonstrations and instructions for customers unschooled in the operation of the new vehicles. Furthermore, the dealers agreed to instruct consumers in the proper methods of caring for the cars and to keep on hand a supply of parts and a force of mechanics capable of repairing the autos” (Porter and Livesay, ‘971, p. 195). And, of course, independent wholesalers were eliminated entirely from the distribution process.
Alfred P. Sloan s explanation of why the automobile manufacturers de- cided to use franchises rather than own dealerships is interesting. He observes that
. . . automobile manufacturers could not without great difficulty have undertaken to merchandise their own product. When the used car came into the picture in a big way in the 1920s as a trade-in on a new car, the merchandising of automobiles became more a trading proposition than an ordinary selling proposition. Organizing and supervising the necessary thousands of complex trading institutions would have been difficult for the manufacturer; trading is a knack not easily fit into the conventional type of a managerially controlled scheme of organization. So the retail automobile business grew up with the franchised- dealer type of organization. [1964, p. 282]
Not only, therefore, did the retail sale and service of automobiles require that transaction-specific investments be incurred, but it also required, especially as trade-ins became more common, that judgments based on idiosyncratic local information be made. Centralized ownership reduced the incentive to exercise that judgment in a discriminating way and posed severe monitoring problems. Rather, therefore, than integrate fully into the retail sale of automobiles, an intermediate form, the franchised dealership, evolved instead.
Producer durables were distributed through two networks. Small, stan- dardized machinery was sold through commission merchants and jobbers. For products that were of special design, technologically complex, and quite expensive and for which installation and repair required special expertise, however, integrated marketing systems were developed (Porter and Livesay, 1971, pp. 183—84). Examples include Cyrus McCormiclc, who pioneered the development of integrated distribution for farm equipment and set the stage for others thereafter to imitate (Livesay, 1979, chap. 3). Office machines were another case where demonstration, sales, and service required specialized expertise for which franchised dealers were instrumental to success (Porter and Livesay, 1971, pp. 193-94).
The manufacture of textile machinery, sugar mill machinery, industrial boilers, and large, stationary steam engines also favored direct contact between buyer and seller (Porter and Livesay, 1971, pp. 181-82). The sale of electrical machinery posed special problems for customers that had “special needs and requirements that made standardization extremely difficult in the industry’s early years” (Porter and Livesay, 1971, p. 187). The sale, installation, and service of electric generators and related central station equipment required even closer attention. Forward integration in all of those areas was correspondingly extensive (Porter„and Livesay, 1971, pp. 180-92).
1.4. MISTAKEN INTEGRATION
Forward integration mistakes are not widely reported and, if reported, are not always recorded as mistakes. An exception is American Tobacco’s effort to use integration into the wholesaling and retailing of cigars as a device by which to expand its position in this market. Porter and Livesay record the effort as follows:
[American Tobacco] had much success in the nineties in extending its dominance from the cigarette business into other lines in the tobacco industry, including smoking tobacco, plug tobacco, and snuff. The cigar trade, however, proved much more difficult to conquer, primarily because it . . . was not subject to economies of scale in production. American Tobacco [turned instead, therefore, to forward integration) These efforts to move into the wholesale and even the retail end of the [cigar] industry proved very expensive, and American Tobacco endured substantial losses in its war on the cigar trade. [1971, p. 210]
Porter and Livesay also report that the “American Sugar Refining Co. engaged in a similar effort to drive its competitor John Arbuckle out of business by buying into wholesale and retail houses to discourage the sale of Arbuckle’s sugar. The attempt failed miserably and proved very costly” (1971, p. 211, n. 52).
. Pabst Brewing, Schlitz, and other large brewers purchased saloons in the late 1800s and rented them to operators as outlets for their brands of beer (Cochran, 1948, pp. 143-46). Whatever the merits it might have had at the time—which, except as a short-run expedient, appear doubtful (Cochran, 1948, p.199)—the shift from kegs to bottled beer rendered it nonviable.
2. Transaction Cost Interpretation
Chandler (1977, pp. 287, 302) and Porter and Livesay (1971, pp. 166, 171, 179) refer repeatedly to the “inadequacies of existing marketers” in explaining forward integration by manufacturers into distribution in the late 1800s and early 1900s. Presumably the same would be said of marketers in the 1950s when IBM integrated forward in the sale and service of computers. But to what do those inadequacies refer? Judging from the differential response, the nature or degree of severity evidently varied considerably.
The explanation advanced here is at best suggestive. It appears, howev-er, that scale economy, scope economy, and transaction cost factors are operative. In addition, a hitherto unremarked factor also appears: externalities.
Recall that the cognitive map of contract breaks transaction cost econo- mizing into two parts: the governance branch and the measurement branch. The governance branch is preoccupied with problems of harmonizing and adapting exchange where ex post bilateral trading is supported by investments in specific assets. Although the measurement branch plays a less substantial role in the analysis of problems dealt with in this book (partly because the problems selected for study have a bilateral trading quality), it is nevertheless important. The contracting complications posed by externalities have measurement origins.
Externality concerns arise in conjunction with a branded good or service that is subject to quality debasement. Whereas a manufacturer can inspect, thereby better to control, the quality of components and materials it purchases from earlier stage and lateral suppliers, it is less easy to exercise continuing quality controls over items sold to distributors.47 That is not a special problem if the demands of individual distributors are independent of one another. If, however, the quality enhancement (debasement) efforts of distributors give rise to positive (negative) interaction effects, the benefits (costs) of which can be incompletely appropriated by (assigned to) the originators, failure to extend quality controls over distribution will result in suboptimization.
Whereas scale economies accrue when cost savings are realized by adding apples to apples—formally, —economies of scope accrue if cost savings result when apples and oranges are joined— formally, C(X,Y) < C(X) + C(Y). Economies of scale are probably less relevant to a decision to integrate forward into distribution than are economies of scope. To be sure, firms that are very large in relation to the size of the market may be able to justify own distribution where smaller firms cannot—Kodak’s processing of film being an example. Often, however, retail goods are sold by independent retail outlets, even where producer market shares are very great— breakfast cereals and diamonds being examples.
Breakfast cereals, soup, bread, meat, and milk are thus sold from a noninte- grated grocery store. Diamonds are sold with other jewelry. Economies of scope (and/or the incentive deficiencies of integration) are evidently very large in relation to any prospective gains that would accrue to forward integration in such circumstances.
In consideration of the potential economies of scope to which market distribution modes have access, decisions to integrate forward presumably turn on benefits—or, inasmuch as our emphasis throughout has been on costs, on transaction cost savings—that market modes are unable to realize. The governance cost problems that arise in conjunction with specialized assets, including those associated with the wasteful dissipation of valued firm-specific reputation effects (which is the externality effect referred to above), thus warrant examination. Do they bear any systematic relation to the types of forward integration decisions reached by manufacturers in the late nineteenth century?
I submit that they do. Specifically, the forward integration decisions reported by Chandler take on the pattern shown in Table 5-1, where + + denotes considerable, + denotes some, ~ indicates uncertain, and 0 is negligible. The observed degree of forward integration is shown on the left column and the relative importance of economies of scope, externalities, and asset specificity are shown in each of the next three columns.
Thus forward integration is never observed if externalities and asset specificity are negligible (Class I), or if it does occur it is mistaken (Class VIII) and will be eventually undone. There simply are no governance or measurement purposes to be served in those circumstances. Limited integra-tion into wholesaling occurs in response to coordination and inventory man- agement needs (Classes II and III), but comprehensive integration is not observed until specialized investments, especially refrigeration, are required (Class IV).
Integration into final sales and service is mainly observed for consumer and producer durables where considerable knowledge is imparted at point of sale and specialized follow-on service is required (Classes VI and VII). Some products involve complicated, incentive issues, of which the decision to sell automobiles through franchised dealers rather than an employee network is an example. The issues here are discussed more fully in Chapter 6.
To be sure, the foregoing is a highly provisional assessment. The qualitative assignments (from + + through 0) are judgmental and based on description; other factors that have a bearing on forward integration have been ignored (see, however, section 7, below); and temporal elements, including customer learning, which may reduce the need for sales and service assistance, have been ignored. Often, however, refined assessments are not needed to evaluate discrete structural alternatives (Simon, 1978, p. 6)— which is to say that crude assignments are often adequate for pattern-seeking purposes.
Source: Williamson Oliver E. (1998), The Economic Institutions of Capitalism, Free Press; Illustrated edition.