The distinction between lateral and backward integration is somewhat arbitrary. 1 shall include in the former the supply of components, body panels, and the like and reserve backward integration for more basic materials.
1. A Case Study
Klein, Crawford, and Alchian’s (1978, pp. 308-10) treatment of the bilateral exchange relationship between Fisher Body and General Motors in the 1920s is illustrative of lateral integration. The basic facts are these:
- In 1919 General Motors entered a ten-year contractual agreement with Fisher Body whereby General Motors agreed to purchase substantially all its closed bodies from Fisher.
- The price of delivery was set on a cost-plus basis and included provi- sions that General Motors would not be charged more than rival automobile Price disputes were to be settled by compulsory arbitration.
- The demand for General Motors’ production of closed body cars increased substantially above that which had been As a consequence General Motors became dissatisfied with the terms under which prices were to be adjusted. It furthermore urged Fisher to locate its body plants adjacent to GM assembly plants, thereby to realize transportation and inventory econo- mies. Fisher Body resisted.
- General Motors began acquiring Fisher stock in 1924 and completed a merger agreement in 1926.
The contracting relation between General Motors and Fisher Body thus moved through three stages. More or less autonomous contracting evidently worked to the satisfaction of the parties in the wooden body era. Specialized physical assets were needed, however, to support the distinctive body designs that attended the shift to the metal body era. A condition of greater bilateral dependency thereby resulted. Efficient contracting principles required that a new contracting structure be crafted in response. Price adjustment by formula and dispute settlement by arbitration were thus expressly provided. Unanticipated demand and cost realizations nevertheless placed this bilateral contracting relation under strain. Additional strains were in prospect, moreover, if Fisher Body were to accede to General Motors’ request that site-specific investments be undertaken. Faced with the prospect that both operating and investment decisions would be out of alignment during much of the rapid growth stage of development, bilateral governance eventually gave way to unified governance.
To be sure, transaction cost reasoning does not predict this sequence in detail. The observed succession of changes—from classical through bilateral to unified contracting—in response to the progressive deepening of transaction- specific investments, during a period when adaptive needs were great, is nonetheless consistent with the overarching argument. The transaction cost hypothesis would have been contradicted, moreover, had the contracting relation remained constant in the face of these changes. Rival theories of eco- nomic organization are mainly silent on these matters.
2. Statistical Model Estimations Using Field Data
Kirk Monteverde and David Teece (1982) have recently studied the degree to which General Motors and Ford are integrated backward into components and e economic factors that are responsible. They examined 133 component groupings, which, they observe, “include most of the major items that go into a comp ete vehicle (1982, p. 207). Probit techniques are used to estimate a og i e 1 ood function with eight independent variables, the most important 0 w ich are (1) engineering effort in designing the ith component; (2) a inary variable indicating whether the component is specific to the manufac-turer or not; (3) a dummy variable for the company (Ford or GM); and (4) a series of subsystem dummies (engine, chassis, ventilation, electrical, body). Their principal findings were: (1) “the variable chosen as a proxy for transaction- specific skills (‘engineering’) is highly significant”; (2) “components specific to a single supplier [are] candidates for vertical integration”; (3) “General Motors is more integrated into component production than is Ford”; and (4) the subsystem dummies were not significant taken alone, but taken together “significantly contributed to the explanatory power of the model” (1982, p. 212). They concluded that transaction cost considerations, especially industrial knowhow that is specialized to a particular firm, are important in defining the efficient boundaries in the two corporations.
An independent study by Gordon Walker and David Weber partially confirms and extends this assessment. Their sample consisted entirely of relatively simple parts that are inputs’to the initial assembly stage of automobile manufacture (1984, p. 381). They examined the effects of asset specificity (measured, albeit, rather indirectly), uncertainty, and scale economies. Their data consisted of sixty decisions made by a component division for which a committee met to evaluate the merits of make or buy. Microanalytic observations from component purchasing, manufacturing engineering, product engineering, and sales were elicited. The data were then analyzed using the unweighted least squares procedure of Joreskog and Sorbom (1982). Although they interpret their results as “mixed” in transaction cost terms and attribute most of the explanatory power to “comparative production costs,” they acknowledge that this may be due to limitations in both the data and the model for which subsequent work is indicated (Weber and Walker, 1984, p. 387). Further efforts to plumb the factors that are responsible for comparative production cost differences between buyer and supplier will, I submit, disclose added transaction cost features. (Diseconomies of.small scale in an integrated firm are often explained by the contracting difficulties that an integrated firm experiences in selling product to rivals.)
Erin Anderson and David Schmittlein (1984) examine the organization of marketing. They examine whether or not the sales force in the electronic components industry is integrated by estimating a logistic function. They likewise obtain mixed effects: “integration is associated with increasing levels of asset specificity, difficulty of performance evaluation, and the combination of these two factors,” but frequency and uncertainty measures turn out not to be significant (Anderson and Schmittlein, 1984, p. 385). One of the ramifications of both the Walker and Weber and the Anderson and Schmittlein articles is that added theory is necessary if the needs of empirical studies of transaction cost issues are to be met.
Thomas Palay has recently studied transportation transactions between manufacturers and railroads. Although most rail shipments are unexceptionable, in that shippers are contracting for standardized services, some pose special railroad car design and/or handling problems. An example is the “high cube” cars that are specialized to automobile parts shipment. The cars, larger and more expensive than standard box cars, are transferable among automobile manufacturers without sacrifice of value. However, the racks used to secure auto parts in transit, at a rack cost of between five thousand and fifteen thousand dollars per boxcar, are designed to the needs of each manufacturer, hence are nonredeployable. Although initially the carriers owned both the high cube cars and the racks, problems developed with respect to the latter. Consonant with the theory, racks are now mainly owned by the shippers (Palay, 1981, pp. 117-18).
Tank cars and covered hoppers to move chemicals were the most spe-cialized and required the greatest investment of the transportation transactions that Palay examined:
This equipment is generally built to the particular substances being moved… Glass or rubber lining, specialized pressure volumes, and damage control equipment are but a few examples of the unique equipment employed… Utilization patterns simply make it too costly to attempt to modify a car to handle a new product each trip, and cleaning involves expensive facilities and technologies. The cost of specialized tank cars ranged from 50 to 100 thousand dollars [1981 pp. 129-30]
The “highly idiosyncratic nature of the rail equipment has led to shipper ownership of its own tank and covered hopper cars” (Palay, 1981, p. 134).
Scott Masten’s recent examination of the organization of production in the aerospace industry and of the government’s procurement policies is similarly corroborative:
Overall, the data on the aerospace system support the contention that design specificity and complexity are necessary, if not sufficient, conditions for the breakdown of cooperation in market-mediated exchanges and the subsequent integration of production within the firm. In addition, the procurement policies professed by the government provide supportive detail not yet available in the formal analysis, such as the effects of uncertainty on the scope of contractual agreements and the relevance of the absolute value of investments on the need for specialized governance structures. [Masten, 1984, p. 417]
Finally, Paul Joskow’s (1985) ambitious study of vertical integration and long-term contracts for the supply of coal for coal-burning electric utilities demonstrates that governance structures line up with the needs of transactions in a discriminating way. The correspondence between transaction cost reasoning and microanalytic economic organization in practice is strongly borne out by Pablo Spiller’s (1985) empirical study of vertical mergers as well.
Source: Williamson Oliver E. (1998), The Economic Institutions of Capitalism, Free Press; Illustrated edition.