Content uploaded by Richard Langlois
Author content
All content in this area was uploaded by Richard Langlois
Content may be subject to copyright.
Explaining Vertical Integration: Lessons from the American Automobile
Industry
Richard N. Langlois, Paul L. Robertson
Journal of Economic History, Volume 49, Issue 2, The Tasks of Economic History (Jun.,
1989),
361-375.
Your use of the JSTOR database indicates your acceptance of JSTOR’s Terms and Conditions of Use. A copy of
JSTOR’s Terms and Conditions of Use is available at http://www.jstor.org/about/terms.html, or by contacting
JSTOR at jstor-info@umich.edu or at (3 13) 764- 13 14. No part of a JSTOR transmission may be copied,
downloaded, stored, further transmitted, transferred, distributed, altered, or otherwise used, in any form or by any
means, except: (1) one stored electronic and one paper copy of any article solely for your personal, non-commercial
use, or (2) with prior written permission of JSTOR and the publisher of the article or other text.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or
printed page of such transmission.
Journal of Economic History is published by Economic History Association. Please contact the publisher for
further permissions regarding the use of this work. Publisher contact information may be obtained at
http://www.jstor.org/journals/eha.html.
Journal of Economic History
01989 Economic History Association
JSTOR and the JSTOR logo are trademarks of JSTOR. For more information on JSTOR contact
jstor-info@umich.edu.
0
1996 JSTOR
http://www.jstor.org/
Mon Feb 9 10:52:29 EST 1998
Explaining Vertical Integration: Lessons
from the American
A
utomobile Industry
RICHARD N. LANGLOIS AND PAUL L. ROBERTSON
The early history of the American automobile industry provides fertile hunting
grounds for theorists seeking corroboration of various, conflicting theories of
vertical integration. An examination of the whole history suggests that no single
theory always fits the facts perfectly. A complete explanation must combine
specific theories in a way that is attentive to such factors as industry life-cycle,
demand, economies of scale, and appropriability. If there is any “general”
theory, it lies in the set of “dynamic” transaction-cost approaches rather than in
the asset-specificity approach now dominant.
VERTICAL INTEGRATION: THEORY
0ne starting point for a theory of vertical integration goes back at
least to Adam Smith: the division of labor. Writers in this tradition
include Allyn Young and George Stigler, who assume that the firms in
an industry are initially vertically integrated and that increasing output
leads to differentiation as various stages of the production process are
spun off into specialized concerns.
1Whereas small firms in industries
with limited output might need to undertake the production of interme-
diate goods because outside suppliers would not find it profitable to
manufacture on such a limited scale, an expansion of the output of final
products could permit specialized firms to take over the production of
intermediate goods. This differentiation is especially important where
there are different levels of economies of scale for the various produc-
tion
stages-
where, for example, the most efficient level of output of an
intermediate product is greater than that needed by any single manu-
facturer of the final good. As an example of this kind of differentiation,
Young cites the early printing industry, which over time evolved into
not only the modern printing industry, but also into firms turning out
wood pulp and paper, inks, type, and other inputs into printing.
While there are certainly industries in history that have fit this
pattern, we can just as easily point to industries in which the pattern has
been quite different. Indeed, as we will suggest shortly, the automobile
The Journal of Economic
History,
Vol. XLIX, No. 2 (June 1989). 0 The Economic History
Association. All rights reserved. ISSN 0022-0507.
Richard
Langlois
is Associate Professor of Economics, University of Connecticut, Storrs, CT
06269-1063. Paul Robertson is Head of the Department of Economics and Management, University
College, University of New South Wales, A.D.F.A., Campbell, ACT 2600, Australia.
The authors would like to thank Lee Alston, Fred Carstensen, Daniel Raff, John D. Robertson,
and John R. Robertson for helpful comments.
’
Allyn
A. Young,
“Increasing Returns and Economic Progress,” Economic Journal, 38 (Dec.
1928),
pp. 523-42; and George J. Stigler,
“The Division of Labor Is Limited by the Extent of the
Market,” Journal of Politicd
Economy,
59 (June
1951),
pp. 185-93.
361
362 Langlois and Robertson
industry is a potential counterexample: firms were quite differentiated at
an early stage in the industry’s life but became much more integrated as
output expanded.
What makes the approach from the division of labor incomplete is that
it considers only one component of cost: the raw technological costs of
manufacturing. Since Ronald Coase’s famous article in the 1930s,
economists have been alert to the importance of transaction costs in
explaining the structure of economic organization.2 This often fuzzy
category is intended to comprise such less tangible costs as those
involved in search, negotiation, and monitoring. Oliver Williamson is
the leader of a group of writers who argue strongly that the pattern of
vertical integration in an industry reflects a minimizing of the sum of
production and transaction costs, not just production costs alone (as in
the Young-Stigler model)
.
3
In order to use the concept of transaction cost meaningfully in an
explanation of economic organization, one has to specify the nature and
sources of the costs one has in mind. One tradition in the literature has
pointed to the costs of measuring the output of a stage of production. In
the well-known formulation of Armen Alchian and Harold Demsetz, for
example, technological indivisibilities in team production make it costly
to monitor the marginal products of individual team members, suggest-
ing some form of internal organization as a less costly alternative to
market contracts between the members
.
4
Another approach has emphasized the costly possibility that parties
to an arm’s-length contract-a large buyer; a supplier, monopolistic or
otherwise; organized or hard-to-replace laborers-might threaten to
withhold the services of the assets they control in order to appropriate
a larger share of the quasirents of cooperative production. The potential
for “hold-ups” of this sort makes for higher expected costs ex ante than
would be the case if the assets involved were jointly owned. The threat
of hold-up is least when the assets involved are generalized ones, thickly
supplied and therefore easy to replace. Conversely, the threat of
hold-up, and the cost of arm’s-length contracting, is greatest when
assets are highly specific, thinly supplied and hard to replace.
This asset-specificity version of transaction-cost theory has been
arguably the dominant approach in the study of vertical integration. Its
proponents have claimed the support of economic history in general,
2
Ronald H. Coase, “The Nature of the Firm,” Economica (new series), 4 (Nov. 1937), pp. 386-
405.
3
Oliver E. Williamson, The Economic Institutions of Capitalism (New York, 1985), esp. p. 103.
4
Armen Alchian and Harold Demsetz, “Production, Information Costs, and Economic
Organization,”
American Economic Review, 62 (Dec. 1972). pp. 777-95. For a more general
formulation of this measurement-cost approach, see Yoram Barzel, “Measurement Costs and the
Organization of Markets,” Journal of Law and Economics, 25 (Apr.
1982), pp. 27-48; and Steven
N. S. Cheung, “The Contractual Nature of the Firm,”
Journal
of
Law
and
Economics, 26 (Apr.
1983),
pp. 1-21.
Vertical Integration in the Auto Industry 363
with the case of the auto industry figuring prominently in the evidence.
Nonetheless, there remains reason to think that this approach may not
offer a complete picture of the nature of and motives for vertical
integration. For one thing, there may be other sources or types of
transaction costs at work, some of which may not be easily subsumed
under the rubric of measurement costs. In particular, the transaction
costs of economic change may help shape the pattern of vertical
integration in an industry. Explanations that center on the costs of
economic change, including both the quantitative change of a rapidly
growing market and the more qualitative change of technological
and organizational innovation, are what we might call “dynamic”
transaction-cost theories.
Morris Adelman was one of the first writers to associate vertical
integration with economic change.5 In a rapidly growing industry, he
argued, suppliers of intermediate goods may not be able to expand
quickly enough to meet the needs of the producer of final goods, thus
,
motivating that producer to integrate backwards.6 Morris Silver has
more recently recast this “bottleneck” explanation in informational
terms.7 Economic change, he argues, often involves not merely quan-
titative growth but also qualitative change-innovation in the Schum-
peterian sense. Existing firms with the expertise to help an innovator
could find the strangeness of a new idea so overwhelming that they
would dismiss it out of hand. Rather than devoting effort to catching the
attention of suppliers and converting them to the idea, the innovating
firm could find it cheaper to acquire the capabilities necessary to
produce its own inputs or handle its own marketing. After the innova-
tion has become more familiar and outsiders have been able to assimi-
late the innovator’s ideas, later entrants would not face the same
bottlenecks and could draw more freely on outsiders. The value of
external economies, a concept much appreciated by Alfred Marshall
and his British followers, therefore varies according to whether a firm is
an innovator or a follower.
In some cases, vertical integration is effectively forced on innovative
firms, who would have preferred to economize on managerial resources
by buying complementary products and services on the open market. In
other cases, however, we might say that integration results from an
entrepreneurial decision to supersede an existing network of comple-
mentary activities. The innovations of refrigerated meat packing in the
nineteenth century or containerized shipping in the twentieth super-
seded inferior systems that might easily have been judged perfectly
5 Richard N. Langlois,
“Economic Change and the Boundaries of the Firm,” Journal of
Institutional and Theoretical Economics, 144 (Sept. 1988), pp.
635-57.
6
Morris Adelman, “Concept and Statistical Measurement of Vertical Integration,” in Business
Concentration and Price Policy (Princeton,
1955),
pp. 318-20.
7 Morris Silver, Enterprise and the Scope of the Firm (London, 1984).
364 Langlois and Robertson
efficient by the standards of neoclassical competition theory. But in the
supersession of those systems lay even greater (internal) economies?
Disintegration using Marshallian external economies and vertical inte-
gration are active strategic alternatives open to the entrepreneur.
A principal argument for the superiority of vertical integration over a
network of decentralized producers is the potentially greater ability of
the large firm to implement innovation, especially large-scale innovation
of a systemic sort. It can be equally well argued, however, that
backward integration might retard change by tying firms to a single
source of innovation when access to the market would have provided a
wider range of choices. As was the case in the British shipbuilding
industry before 1914, the presence of a large number of competing
suppliers can be an important external economy if they are all commit-
ted to improving their products and providing other design services to
benefit their customers, especially to the extent that the latter are small
and unable to afford research into improving a wide range of inputs.’
Indeed, William Abernathy has used the automobile industry as pre-
cisely an example in which vertical integration led to stagnation because
it isolated firms from innovation, especially product innovation, they
might have obtained in a more decentralized system.”
One would in general expect improvements in process technology
that depend on economies of scale for their viability to come only after
customers have accepted a dominant product technology. When there
are several variations of a product competing in a limited market in the
early stages of the product life-cycle, producers may not invest in a
capital-intensive technology that will pay only at high levels of output.
On the other hand, it is possible that the final choice of a product
variation will be made on the basis of process innovation that reduces
prices to a point at which other considerations become largely irrelevant
to customers. Because of the increased uncertainty faced by a producer
who adopts a capital-intensive process technology when the nature of
the product is still in flux, however, the tendency would be for firms to
wait for a resolution of consumer preferences before committing them-
selves to mass production.
8 William Lazonick has lately put great emphasis on this point in presenting a theory of the
vertically integrated firm that is otherwise much in the spirit of the Adelman-Silver view. See
William Lazonick,
“The Social Determinants of Technological Change: Innovation and Adaptation
in Three Industrial Revolutions” (paper presented at the Second International Conference on the
History of Enterprise,
Terni,
Italy, October 1-4, 1987); and William Lazonick, “The Causes and
Consequences of the Modern Corporation: Innovation and Adaptation in the Theory of the Firm”
(paper presented at the U.C. Intercampus Group in Economic History Conference, University of
California, Santa Cruz, April 29-May 1, 1988).
9Sidney Pollard and Paul Robertson, The British Shipbuilding Industry,
1870-1914
(Cambridge,
MA, 1979),
p. 92.
For a general theory of innovation in a network of suppliers and producers, see
Eric von Hippel, The Sources of Innovation (Oxford, 1988).
10
William Abernathy, The Productivity Dilemmu: Roadblock to Innovation in the Automobile
Industry (Baltimore, 1978).
Vertical Integration in the Auto Industry
365
Moreover, the value of external economies can differ significantly
between innovator and follower. Whereas an innovating concern might
make do by purchasing goods or services developed for other purposes
by outside suppliers or distributors when capital is limited, adaptive
followers would not be faced with the same disability and could plug
into existing networks with few if any penalties. As David Teece
suggests, the advantages (or disadvantages) of innovative leadership
will depend on, among other things, the ease of imitation in the industry
and the nature of the assets complementary t o the innovation.11 If
imitation is easy, such as when patents or trade secrets do not
effectively protect the innovator, an innovating firm may be at the mercy
of competitors (or even suppliers) who can enter quickly and take cheap
advantage of the capabilities created at high cost by the innovator. To
profit well in such circumstances, the innovator would have to own
many of the assets complementary to the innovation. This need not
imply vertical integration in any meaningful sense, however, since in
principle the innovator need only take financial positions in the com-
plementary assets-long positions in those likely to appreciate and short
positions in those likely to depreciate. Only when the complementary
assets are highly specific to one another do problems of “hold-up”
necessitate joint ownership and control.
THE AUTOMOBILE INDUSTRY: HISTORY
Robert Paul Thomas has divided the early history of the American
automobile industry into four periods: the pre-1900 era of invention; the
era of product development from 1900 to 1908; the era of rapid
expansion from 1908 to 1918; and the era of replacement demand from
1918 to 1929.12
In the era of invention, the focus of all manufacturers, if they could be
called that, was almost exclusively on product innovation. Here the
industry came closest to the pattern suggested by Young and Stigler.
Car makers like Winton, Olds, and Ford were craft shops, all highly
integrated in the sense that they made, or more accurately, improvised,
most of their own parts. One well-known example is Henry Ford making
his first cylinder out of a piece of pipe. What distinguishes the
automobile industry from such earlier examples as printing or machine
tools is that this early vertical integration and use of crafts techniques
disappeared very quickly once genuine commercial production began in
11 David J. Teece, “Profiting from Technological Innovation: Implications for Integration,
Collaboration, Licensing, and Public Policy,” Research Policy, 15 (Dec.
1986),
pp. 285-305.
12
Robert Paul Thomas, An Analysis of the Pattern of Growth of the Automobile Industry, 1895-
1929 (New York, 1977), pp. 6-8. For a more detailed account of the history of vertical integration
in the auto industry, see Paul L. Robertson and Richard N. Langlois, “Innovation and Vertical
Integration in the American Automobile Industry,
1900-1940”
(working paper, economics and
management dept., University College, University of New South Wales, Jan. 1989).
366 Langlois and Robertson
earnest. By the turn of the twentieth century, the capabilities to
manufacture parts adaptable to the automobile already existed in the
American economy, and these external economies obviated the early
stage of vertical integration that had been thrust on many nineteenth-
century industries. At the same time, the novel and experimental nature
of the product made financing difficult and left firms starved for capital,
further working against large-scale integration. Most importantly, per-
haps, the novelty of the product meant that a dominant design paradigm
had not yet coalesced in the minds of customers, making a commitment
to large-scale integrated production extremely risky.
In many respects, the expansion of the industry in the first two
decades of the century resembled a textbook example of the benefits of
external economies. Early in its development, the industry became
concentrated in Michigan, Ohio, and Indiana. Localization allowed
assemblers to take advantage of external economies that flowed from
concentrations of suppliers. Although the degree of vertical integration
varied from firm to firm, virtually all automobile companies began as
assemblers rather than manufacturers. Early cars could be easily put
together from components developed for other purposes, such as
bicycle wheels, or from variations on known themes, such as wooden
bodies. Particularly in the United States, the term “horseless carriage”
was an accurate description of the design concept that lay behind many
cars. They were seen initially as improvements on existing forms of
transport rather than as a new product that needed to be thought out
from scratch. As a result, finding willing components manufacturers and
dealers does not seem to have been a problem, although raising capital
was.
In the era of rapid expansion, the design of automobiles began to
coalesce into a more or less unified paradigm. One important move
toward standardization came with the replacement of basic American
designs with more advanced and complex French ideas. Although
Americans originally conceived of cars as buggies without horses (like
the curved dash Oldsmobile runabout), the French thought in terms of
a road locomotive. French autos had heavier, multicylinder engines
mounted at the front, multigear transmissions, and differentials, all
attached to a steel frame. The preference of wealthy customers for
French imports induced American firms to begin manufacturing cars on
the French pattern in 1901. After 1905, the preference for cars on the
French model was reflected in a shift of demand patterns toward more
expensive cars. This left an enormous unfilled demand for a model that
delivered the benefits of the French concept at a price comparable to the
Oldsmobile. Ford’s success with the Model T was based on precisely
that combination of modern features and low price.
As is typical in the product life-cycle, the radical changes in product
characteristics that occur in an early era gave way to incremental
Vertical Integration in the Auto Industry
367
product change. The focus of innovation began to shift to process
innovation, including the development of mass automobile production
using the moving assembly line, which was the ultimate key to the
success of the Model T.
Mass production involved Ford in the large-scale design of machine
tools, at least some of which also seem to have been manufactured by
the firm. Although Thomas contends that the automobile industry was
well served by outside machine-tool firms in its early days, the greater
extent to which Ford took the division of labor when he developed mass
production and the assembly line called for a far wider use of special-
purpose machinery than had been necessary for other firms.13 To meet
their new needs, Ford and his engineers were forced to develop many of
these devices themselves. As Silver hypothesizes, this was not because
the technical advances in the tools were beyond the understanding of
independent tool makers; rather, it was because only the men at Ford
understood the uses to which the new machines were to be put. It would
have cost more for Ford Motor Company to explain the process of
automobile -manufacture to the tool makers (a process that was in any
case still evolving in the minds of Ford and his assistants) than it did to
diversify into tool design on an ad hoc basis. The Ford techniques could
have diffused to outside suppliers and indeed quickly did so, but the
massive demand for the Model T generated by Ford’s low-price policy
made even small delays costly. In this sense, integration was a strategy
forced upon
decision. Ford as much as it was a conscious entrepreneurial
There is also more to the story than the radical newness of the Ford
machines. Central to the innovation of mass production is its systemic
character-its organization and timing of production activities-which
had the effect of making vertical integration self-reinforcing in certain
ways. Because much of the assembly activity was collected in one
place, Ford engineers were able to perceive opportunities for grasping
economies of scale that would not have been apparent to decentralized
parts suppliers. An example of this is the development of metal-
stamping techniques. Pressed-steel frames, which were much stronger
than their structural-steel predecessors, were generally adopted in the
industry by 1909. A couple of years before, however, the managers of
the John R. Keim Mills, a producer of pressed steel in Buffalo, had
approached Ford with a proposal to supply axle housings. Ford agreed,
and by 1908 Keim was supplying a variety of parts. Engineers from the
two firms cooperated in perfecting the pressing process; Ford invested
heavily in machinery for the Buffalo plant, and eventually purchased
Keim in 1911. When the workers at the Keim plant struck the following
13
Thomas, Pattern of Growth, p. 269. Compare David A. Hounshell, From the American
System to Mass Production, 1800-1932 (Baltimore, 1984), p. 233.
368
Langlois and Robertson
year, Ford ordered all of the machinery removed from Buffalo and
shipped to Highland Park. The technology sparked the interest of
Ford’s engineers, and metal-stamping was soon adapted to producing
crankcases, axles, housings, and even bodies. Thus integration initially
motivated by a desire to avoid hold-up problems and supply disruptions
led to further integration because of the possibilities for innovation it
opened up in the Ford environment.
It is important to emphasize that integration at Ford during that period
was a phenomenon of novelty and innovation. As soon as the outlines
of the assembly-line innovation were clear, it diffused rapidly to
independent parts suppliers, a process aided by Ford’s openness to
trade journalists. Moreover, Ford engineers soon learned that the
production process could be efficiently decentralized.14 Once the inno-
vation of mass production of parts became assimilated, and as the extent
of the market
grew,
centralization became more costly and less bene-
ficial. At Ford, the ensuing decentralization took place within a verti-
cally integrated structure for largely historical reasons.
Significantly, the one other firm to attempt a high level of integration
before 1920 was also run by a strong-minded visionary. In his two forays
into General Motors, William Crapo Durant brought together not only a
large number of automobile assemblers but also a collection of compo-
nents suppliers.
Like Ford, Durant believed strongly in the latent demand for auto-
mobiles, especially at the lower end of the price range. Unlike Ford,
however, Durant’s innovations lay not in production of the vehicles but
in the creation of a marketing and distribution network. From the point
of view of process technology, General Motors (GM) was very much a
follower. For both of these reasons, integration at General Motors was
of a very different sort than that at Ford. Under Durant, there was no
systematic policy behind the expansion of General Motors, and no
attempt to integrate the various divisions into a coordinated manufac-
turing group on the Ford model. It was only after a decade of wrestling
with their legacy that Durant’s successors were able to rationalize the
firm and move toward genuine integration.
Vertical integration at GM also arose from motives that were quite
different from those at Ford. Alfred Chandler and others have cited as
a principal reason for vertical integration Durant’s desire to ensure
adequate and timely supplies.
15 This is no doubt part of the story. But
an arguably more important motive was Durant’s desire to appropriate
the benefits of his vision and his marketing innovations by taking
positions in as many assets complementary to auto manufacture as
possible. Durant was by no means a passive stockholder in these
14
Henry Ford with Samuel Crowther, My Life and Work (Garden City,
1923),
pp. 83-84.
I5 Alfred D. Chandler, Strategy and Structure (Cambridge, MA, 1962), p. 116.
Vertical Integration in the Auto Industry
369
enterprises. Capital markets for automotive securities were not yet
developed enough to allow appropriation by passive investment: the
first appearance of automotive stock on the New York exchange was the
sale of General Motors voting trust certificates in 1911.
By the
192Os,
the market for cars had changed radically. As both
product and process innovation became less important, cost relation-
ships between automobile manufacturers and potential suppliers be-
came increasingly prominent. By the end of that decade, a modest
reduction in the degree of vertical integration had begun, based on a new
appreciation of the flexibility that could be derived from outside
suppliers. Between 1922 and 1926, the importance of components
purchased from outside suppliers had declined from 55 percent to 26
percent of the wholesale value of American motor vehicles? Even in
dollar terms, the value of components had declined despite a near
doubling in the total value of finished vehicles. By 1927, however,
purchases of some components were already beginning to increase, and
vertical integration became progressively less important in the early
years of the Depression.”
This is surprising in view of earlier trends. It is especially interesting
for the early years of the Depression, when one would expect a
capital-intensive firm to protect its own output of components at the
expense of independent firms. The explanation for this pattern involves
several key factors: the growth of the market for spare parts; organiza-
tional innovation; and the introduction of the annual model change.
Despite the increasing degree of vertical integration in the early
192Os,
the aging of the national fleet of automobiles led to a proliferation
of parts firms to supply the replacement market. While employment in
automobile plants proper decreased by 43,628 over the period from 1923
to 1925, employment in parts firms grew by
64,628.‘*
Because of the low
investment required to produce a limited range of items, entry was easy
and most of the new parts firms were small or medium sized. Compe-
tition among them reduced costs to levels that frequently could not be
reached by the automobile firms themselves, particularly as the need for
flexibility grew near the end of the decade.19 As novelty became an
important selling factor, firms were forced to change models more
frequently. Moreover, the number of models produced by each firm
l6
Lawrence E. Seltzer, A Financial History of the American Automobile Industry (Boston,
1928), p. 59.
”
Norman G. Shidle, “Trend Toward More Car Models Helping Outside Suppliers,” Automo-
tive Industries (July 30,
1927),
p.
146; and Harold Katz, The Decline of Competition in the
Automobile Industry,
1920-1940
(New York,
1977),
p. 260.
”
Seltzer, Financial History, p. 50.
l9
Eva Flugge,
“Possibilities and Problems of Integration in the Automobile Industry,” Journal
of Political Economy, 37 (Apr. 1929). p. 166.
370 Langlois and Robertson
multiplied. This reduced the economies of scale open to automobile
manufacturers, giving further advantage to the small parts firms.
Organizational innovation also helped shape the pattern of vertical
integration in this era. In the twenties, the principal innovation had the
effect of shifting the margin between firm and market (to use Coase’s
metaphor) in the direction of the market. Following the scares of the
early decade, firms sought to keep their inventories small by purchasing
only amounts that were immediately necessary. This led to the adoption
of “hand-to-mouth’
’
purchasing, a more colorfully named predecessor
of the now-fashionable “just-in-time” practices. And, if modern propo-
nents of this technique are correct, the adoption of hand-to-mouth
purchasing had efficiency advantages over and above its ability to
economize on inventories.
But the most important factor influencing the pattern of vertical
integration in the twenties and thirties was the annual model change and
the product innovation that accompanied it. The used-car market
opened up competition that new-car makers had not faced before. In
order to give their new cars an advantage over used models, car makers
turned to increased product innovation. The decade of the twenties
amounted to, in effect, what we might think of as a discontinuity in the
automobile’s product cycle. Instead of a continued evolutionary pace in
product and process innovation, relatively radical product innovation
had again become necessary. And here, as in the early days of the
industry, those firms able to draw on the product innovations of
decentralized suppliers gained at least a temporary advantage. It was
thus the small, less integrated firms who were best able initially to
succeed in the era of replacement demand. Abernathy believes, for
example, that Chrysler’s “strategy of design flexibility and shallow
vertical integration proved very successful in the prewar period, when
the rate of technological change in the product was rapid.“*’
By contrast, Ford had the greatest difficulty in adapting to the
changed conditions. Despite notable evolutionary changes in the Model
T in the
mid-1920s,
Ford proved incapable of making a smooth
transition to a new model. The firm was obliged to suspend production
for nine months and faced severe teething problems thereafter. The
Model A contained a number of important components that either were
not manufactured by Ford or were so different from those of the Model
T that the existing plant and machinery could not be converted.
The new regime thus had the effect not only of promoting the fortunes
of firms employing relatively less vertical integration but also of
prodding the highly integrated to become less so. “Major product
innovations,” as Abernathy argues, “destroy old paths of backward
vertical integration and create opportunities for new ones. Product
2o
Abernathy, Productivity Dilemma, p. 37.
Vertical Integration in the Auto Industry
371
innovations thus generally reduce the degree of backward integration.“*’
This certainly seems to have been the case at Ford. Moreover, despite
the legend that Ford did not keep records, the firm knew as early as the
mid-1920s that some components could be purchased more cheaply than
Ford manufactured them. This message was reinforced by the transi-
tional problems that the firm experienced. By the early
1930s
Ford
executives were convinced that a smooth introduction of a new model
was more likely if the burden of change were spread over a large number
of firms.
Of all the firms in the industry, however, General Motors was in the
best position to take advantage of the new economics of replacement
demand. Unlike Ford, GM had not optimized its production process for
the single-minded manufacture of one unchanging model. Under Ford
defector William Knudsen, GM installed a system of “flexible” mass
production that allowed for model changes without trauma.** GM was
also less integrated than Ford, and in a much looser way, which made
it easier to take advantage of outside suppliers’ ideas. Moreover, the
company had always had something of a strategy of product innovation
and model change. Dating back to Durant’s early days at Buick, GM
chose not to compete by reducing price on a standardized model, and
instead improved the performance and amenity characteristics of a car
with a more or less constant price. At the same time, General Motors
was also in a better position than its smaller rivals to appropriate the
benefits of product innovation and model change. GM possessed a
richer package of complementary assets, notably its large distribution,
dealer, and consumer-finance network. This gave the firm an ability to
benefit, in a classic way, from both its own innovations and those
developed by smaller competitors and independents.
And when GM lacked some of the relevant assets, it moved to acquire
them. The most important example of this was the acquisition of Fisher
Body, a pioneer in the development of the closed body. Although this
was still a relatively unpopular option at the time, GM decided to tap
into the technology by acquiring 60 percent of Fisher in 1919. At the
same time, GM signed a ten-year agreement to buy all of its closed
bodies from Fisher, and in return insisted on stringent contract terms in
order to avoid opportunistic behavior. As overall demand grew and
preferences shifted toward closed bodies, General Motors became
discontented with its arrangement with Fisher, which had become an
extremely important supplier. GM believed the price was too high in
view of the lower unit capital costs that Fisher Body incurred, and
resented Fisher’s refusal to locate its body plants adjacent to GM
assembly plants, a move that would have benefited General Motors but
2’
Ibid., p. 64.
22
Hounshell, American System, pp. 265-66.
372 Langlois and Robertson
would also have limited Fisher’s flexibility in dealing with other
customers. Ultimately, General Motors took over the remainder of
Fisher Body in 1926 in order to protect its interests.23
Benjamin Klein, Robert Crawford, and Alchian explain this episode
narrowly in terms of the highly specific assets Fisher and GM had to
commit to the venture and the attendant threats of hold-up and
expropriation of rents. But it was not the specificity of the assets that
made integration look better to GM than contractual alternatives. As we
saw, the later years of the twenties and the early thirties saw a trend
toward increased dependence on suppliers for parts, and many of these
suppliers certainly employed assets highly specific to the task and
produced components critical to the assemblers’ flow of production.
What led to high transaction costs in the case of Fisher was the very
rapid shift in demand toward the closed body. In contrast to Ford’s
experience with the Model T, General Motors was not in the forefront
of innovation in this instance. Closed bodies had been available for
years and had also been adopted by competitors. In the long run, GM
might well have been able to rely on outside body manufacturers as all
other major auto firms, including Ford and Chrysler, were able to do to
some extent in the 1930s. In the
mid-1920s,
however, GM’s larger
volume and emphasis on flexibility made it especially vulnerable to
hold-up problems. Because it was not an innovator, GM had to compete
with other auto makers for scarce closed bodies as demand accelerated.
If it could not keep pace with rising consumer demand for closed bodies,
GM stood to lose the benefits that its policy of up-to-date styling was
supposed to provide. This might in fact have led to a permanent loss in
market share if the public bought other brands by default. Rather than
pay a premium to guarantee a reliable supply in a seller’s market, GM
decided to invest a relatively small sum to gain full control of a supplier
in which it already held a majority interest.24
Note that although sunk cost may be sunk, GM’s prior investment
made vertical integration into body building cheaper than for other firms
that needed to begin from scratch. As the additional investments
required to overcome the problem of hold-up were greater for other auto
makers, those firms were more willing to incur whatever transaction
costs derived from association with independent body manufacturers,
23
Benjamin Klein, Robert Crawford, and
Armen
Alchian,
“Vertical Integration, Appropriable
Rents, and the Competitive Contracting Process,”
Journal of Law and Economics, 21 (Oct.
1978),
pp. 308-10.
24
In a recent re-examination of the Fisher Body case (which came to our attention after this
paragraph was written), Klein presents an interpretation substantially in accord with our own.
Benjamin Klein, “Vertical Integration as Organizational Ownership: The Fisher Body-General
Motors Relationship Revisited,”
Journal of Law, Economics, and Organization, 4 (Spring
1988),
pp. 199-213.
Verticul Integration in the Auto Industry 373
especially since the car makers could also offset these costs to a degree
with the external economies of a competitive market.
Detailed estimates of vertical integration are not available for the
years of the
193Os,
but Harold Katz believes that the use of outside
suppliers was reduced as firms learned to cope with complexity and
change.25
The effect of reduced innovation on the degree of vertical
integration is an area in which theory is not well developed. In the case
of automobiles during this period, however, there were exogenous
factors that tended to militate in favor of reintegration. One of these was
the market for spare parts, an area over which car makers had an
incentive to retain control in a period when replacement demand was
growing. As original equipment and replacement parts were jointly
produced-indeed, were identical-they could be turned out more
cheaply if producers could operate on the larger scale allowed by sales
to both markets. Moreover, replacement parts have traditionally carried
a higher profit margin than original equipment. Because automobile
manufacturers forced their dealers to use only authorized replacement
parts, they were in a position to pad their profits by producing parts
internally.
Another factor favoring integration was organizational innovation.
The corporate reorganization that Alfred Sloan put in place at GM in the
1920s is now hailed as a pioneering application of the multidivisional
structure (the M-form). It was a structure well adapted to the institu-
tionalized change of the annual-model strategy: it decentralized day-
to-day decision-making to an extent that approached that of the market,
while retaining a centralized strategic and coordination ability.26 The
M-form innovation thus arguably reduced the costs (increased the
benefits) of internal organization, and helped move the firm-market
margin closer to the firm.
TOWARD A THEORY OF APPLICABILITY
The early history of the American automobile industry suggests that
no single theory adequately explains vertical integration in all periods
and’in all its guises. If there is any candidate for a “general” theory, it
is among the “dynamic” transaction-cost theories. These theories
explain the early integration at Ford and GM, and, in an important
sense, they also explain the Fisher Body episode of 1926, which has
been a keystone example for the asset-specificity view: it was the
“disequilibrium” of a rapid rise in demand for closed bodies, not highly
specific assets as such, that, as Klein now agrees, “moved the contrac-
25
Katz, Decline of Competition, p. 258.
26
Williamson, Economic Institutions, pp. 279-85.
374 Langlois and Robertson
tual arrangement outside of the self-enforcing range and made it
profitable for Fisher to hold up General Motors.“*’
The search for a general theory of vertical integration is perhaps less
interesting than seeking a schema that determines when various partic-
ular explanations are applicable. What this history of the automobile
industry suggests is that certain levels of integration are more likely to
prevail at specific stages of the industry life-cycle and under specific
conditions of demand, economies of scale, and appropriability . We can
summarize these relationships in a tentative way as follows.
In the early stages of an industry’s development, when product
technology is still in flux and markets are small, innovating firms can be
expected to eschew vertical integration if possible. In part this is
because these firms will not need enough of many of their inputs to
produce them efficiently; moreover, small innovating firms rarely have
excess financial or managerial resources to devote to multiple stages of
production. If outside suppliers cannot be found, however, because the
needs of the innovating firm are either too specialized or too difficult to
communicate, then vertical integration must occur if the innovation is to
succeed. This is particularly likely when the innovation is in the nature
of a systemic rearrangement of production technology or organization.
As product innovation slows down, new entrants will not face the
same need to integrate vertically because there will now be sufficient
economies of scale and knowledge of the product to permit the
existence of independent suppliers. Integrated innovators may spin off
intermediate stages at this point and rely on outside suppliers shared
with later entrants, or they may decide they cannot recoup their
investments through selling off suppliers and therefore remain more
heavily integrated than new competitors.
A reduction in the rate of product innovation as an industry matures
increases the scope for process innovations leading to increased econ-
omies of scale. Firms that embark on strategies of mass production may
find it desirable to increase either forward or backward vertical integra-
tion at this stage in order to protect their sources of supply and
distribution networks so they can achieve economies of throughput and
rapidly amortize their heavy fixed investments. The extent of vertical
integration should nevertheless be tempered when the firm can take
advantage of markets in which transaction costs are moderate.
Further innovation may be undertaken either through internal devel-
opment or by relying on outside suppliers. Economies of scale, both in
*’
Klein, “Vertical Integration,” p. 202. In this formulation, it appears that economic change,
“uncertainty” in Klein’s language, plays as important a role as the degree of asset specificity in
explaining vertical integration. This has been noticed before. See Richard N. Langlois, “Internal
Organization in a Dynamic Context: Some Theoretical Considerations,” in M. Jussawalla and H.
Ebenfield, eds., Communication and
Infer.mation
Economics: New Perspectives (Amsterdam,
1984), pp. 31-33.
Vertical Integration in the Auto Zndustry 375
the production process and in the R
&
D function, may be important in
determining the choice. In particular, changes in the relative economies
of scale of intermediate and final products may make it desirable for a
vertically integrated firm to revert to independent suppliers. When a
change in process technology creates economies of scale for interme-
diate goods that exceed the requirements of any manufacturer of a final
good, then a shift to outside suppliers can be profitable. A move in the
opposite direction can also bring about disintegration because a reduc-
tion in the relative economies of scale of producing an intermediate good
may eliminate the advantage gained from backward integration and
open up attractive possibilities for playing competing suppliers off
against one another. This may increase the range of intermediate-
product innovations available to manufacturers of final goods, as well as
yield cost advantages, if the supplying firms are capable of undertaking
research and development. Otherwise, the final-goods producer may
continue to bear the responsibility for innovation in intermediate goods
even if it buys from independent suppliers. History may also play an
important role here. A firm that starts out highly integrated may develop
a bias toward certain kinds of innovations (for example, process
innovations of a systemic sort) that further reinforce its integrated
structure. And a firm that is less integrated may rely on innovation from
decentralized suppliers in a way that reinforces a less integrated
structure, at least in the short run.
