Pragmatic Collaborations: Advancing
Knowledge While Controlling Opportunism
SUSAN HELPERa, JOHN PAUL MACDUFFIEband
(aDepartment of Economics, 400 Wickenden Hall, Case Western Reserve
University, Cleveland, OH 44106. Email: email@example.com,
b2000 Steinberg-Dietrich Hall, Wharton School, University of Pennsylvania,
Philadelphia, PA 19104-6370. Email: firstname.lastname@example.org and
cColumbia Law School, 435 West 116th Street, New York, NY 10027, USA.
This paper starts from the observation that firms are increasingly engaging in
collaborations with their suppliers, even as they are reducing the extent to which they
are vertically integrated with those suppliers. This fact seems incompatible with
traditional theories of the firm, which argue that integration is necessary to avoid the
potential for hold-ups created when non-contractible investments are made. Our view is
that pragmatist mechanisms such as benchmarking, simultaneous engineering and ‘root
cause’ error detection and correction make possible ‘learning by monitoring’—a
relationship in which firms and their collaborators continuously improve their joint
products and processes without the need for a clear division of property rights. We argue
that pragmatic collaborations based on ‘learning by monitoring’ both advance
knowledge and control opportunism and thus align interests between the collaborators.
1. On the (Im)Possibility of Collaboration:
Two Views of the Firm
It is now widely observed that firms more and more collaborate with their
suppliers, even as they reduce vertical integration with them. This observation
seems incompatible with the standard theory of the firm, which argues that
integration is necessary to avoid the potential for hold-ups created when
non-contractible investments are made. In this paper, we start with this
apparent paradox (documented with examples both current and historical) in
order to describe a non-standard type of firm—a novel organizational form,
© Oxford University Press 2000
Industrial and Corporate Change
neither market nor hierarchy, that addresses the problem of opportunism
while at the same time advancing learning and innovation. From this
juxtaposition of two views of the firm, we draw theoretical explanations of
the behavior of the non-standard firm, explore the implications for the
governance of such firms and identify the limits of traditional organizational
building blocks such as contract and property.
The standard theory of the firm, albeit a stylized portrait, refines and
generalizes a picture of economic activity sketched by Adam Smith, realized
in the mass production US economy starting in the late 19th century and
rendered as history by Alfred Chandler Jr (Chandler, 1962, 1977) and others.
Its central theme is that the firm, and property in general, exist to reduce the
hazards of collaboration that could not efficiently be overcome in market
exchange. The hazards arise typically whenever (potential) partners can use
their control over resources indispensable to the joint venture to extort
benefits in excess of those provided in the partnership agreement or where
some can play on the ignorance of others to claim the benefits due for having
performed their obligations without actually having done so.
These problems are pervasive because humans are by nature ignorant and
guileful and they use their guile to extract advantage from one another’s
ignorance. At the limit they exchange information only when bargaining over
the distribution of the potential benefits, tainting what they say and the
bargains they strike. Ownership of assets (understood as the exclusive right
to determine their use unless otherwise contracted), combined with the legal
rights linked to asset ownership, provide powerful instruments for limiting
the extortion and deception that daunt cooperation (Holmstrom,
1997).1Examining collaboration across firm boundaries is an indispensable
method for discovering just what these instruments are and when and how
they work, and don’t work. The key is learning how to understand economic
behavior under circumstances when self-interest threatens to trip itself up and
deprive the parties of the gains of cooperation.
The non-standard theory arises from observations of the inherent sociability
of human behavior and the development of reciprocity norms between and
among individuals and groups. Its empirical inspiration is twofold: (i) the
(Japanese-inspired) ‘lean’ firm that successfully challenged the dominance of
mass producers in the automobile and other industries starting in the
mid-1970s; (ii) the (primarily US-based) high tech start-ups in computers,
1By ‘standard’ theories of the firm, we mean to include such works as Williamson, Hart and Chandler
(Chandler, 1977; Williamson, 1985; Hart, 1995). For an analysis of differences between transaction cost
and property rights theories see Whinston (Whinston, 1997). For a discussion of other incentive
mechanisms besides ownership see Baker and co-workers and Holmstrom and Roberts (Baker et al., 1998;
Holmstrom and Roberts, 1999).
semiconductors,softwareand biotechnologythatin thissameperioddeveloped
and provisional nature of all understanding, from the simplest verbal exchange
to the most complex co-development project. As a consequence, interlocutors
and partners must cooperate in pursuit of mutual intelligibility as a condition
for self-understanding. In this view, because of the mutual vulnerability
resulting from their ignorance of the world, humans are by nature at least as
disposed to be cooperative in order to learn as to be guileful. Once the
cooperative exploration of ambiguity begins, the returns to the partners from
further joint discoveries are so great that it pays to keep cooperating.
Certain corporations, of various stripes and drawing on multiple sources of
inspiration, have formalized and developed this potential for learning through
cooperation. They do so by introducing what we call ‘pragmatic mechanisms’:
disciplines that reveal the ambiguities of current product designs, production
processes and organizational boundaries. At the same time, they orchestrate
joint inquiry, among collaborating individuals, groups and organizations, of
these ambiguities. In the process of inquiry, each collaborator can con-
tinuously monitor the performance of the (relevant) others, while learning
from them and acquiring skills that can be redeployed in other joint ventures.
The overall result, which we will call learning by monitoring, becomes the
basis for pragmatic collaborations.2
From the perspective of the non-standard view, the significance of these
pragmatic mechanisms is that they relax constraints of endemic ignorance
(the systematic incapacity to learn enough to be invulnerable in exchange)
that the standard view regards as constitutive of human nature. Firms exist
in this non-standard world because it is necessary to fix the points of view,
expectations and responsibilities of the collaborators in order to evaluate and
revise them through collaboration.3The exploratory results from any one
collaboration become the basis for adjustments as necessary to facilitate the
next joint activity. Put differently, the de facto joint residual control over assets
deployed in the common project becomes the basis for deliberation over the
shape of the next project, rather than a source of unproductive haggling.
The non-standard firm uses a number of pragmatic mechanisms to create
and maintain the conditions under which two or more firms can sustain
collaboration. These mechanisms, including benchmarking, simultaneous
2For related views in the organizational theory literature see Schon and Weick (Schon, 1973; Weick,
1995). The roots of the pragmatist perspective, as we use it here, lie in philosophical writings by von
Humbolt, Hegel and Dewey and in the anthropology of Mauss and the early Bourdieu (Bourdieu, 1977).
Also, see Sabel for an earlier exposition of ‘learning by monitoring’ (Sabel, 1995).
3See Kogut and Zander for more on the firm as a source of identity that guides the coordinated action
of its members (Kogut and Zander, 1996).
engineering and systems of error detection and correction (as described
below), help overcome both the problems of bounded rationality and of
In the rest of this paper, we use the example of the automobile industry to
argue that collaborative development is now (and has been for crucial his-
torical periods) much more central to the activity and organization of firms
than the standard view allows. We then deepen and extend the non-standard
view to account for the institutionalization of ongoing incremental (and
cumulatively radical) innovation among collaborators. Our project is both
positive (we show evidence that firms have in many cases adopted these prac-
tices) and normative (we argue that more firms should adopt these practices).
The argument is organized in four additional sections. The second section
presents three examples of collaborative supplier relations, all drawn from the
automotive industry: (i) Japanese keiretsu from the post-war period to the early
1990s; (ii) current subcontracting trends in the US; (iii) a reinterpretation of
the purchase of Fisher Body by General Motors in 1926, often cited as the
paradigmatic example of the emergence of the modern, vertically integrated
The automotive industry is a logical choice for multiple reasons. First, we
can draw upon our extensive fieldwork in the industry for a detailed account
of collaborative processes and pragmatic mechanisms.4Second, the
make-or-buy decision in this industry has been the locus classicus of empirical
and theoretical work on the nature and boundaries of the firm since Coase
(1937). We believe it is especially powerful to find evidence of collaboration
in the setting where vertical integration is understood as the archetypal
response to the hazards of collaboration.
The third section presents a systematic explanation of such collaborative
innovation by sharpening the contrast between the two types of firms,
standard and non-standard. We emphasize how, for each type, the division of
labor and governance structure are linked to a particular notion of cognitive
possibilities, i.e. a distinct view of the relationship between the context of our
problems and the activities we should undertake to solve them. Put
differently, we contrast a familiar class of (standard) firm that resigns itself to
operate within routines as a condition for efficient action and an unfamiliar
(non-standard) one that achieves efficiency precisely by systematically
questioning routines without vitiating them as guides to action.
This section also explores in more detail the advantages of pragmatic
collaborations. We consider both the efficiency and effectiveness of these
4This paper draws on (but for space reasons does not repeat) extensive field work by the authors in the
auto industry (e.g. Sabel, 1995; MacDuffie, 1997; MacDuffie and Helper, 1999).
collaborative arrangements, particularly with respect to the production of
useful knowledge, and we also describe how the arrangements control
opportunism by aligning the interests of the collaborators.
In the fourth section, we look at issues of dynamics: how a pragmatic
collaboration gets started and what can undermine it. We argue that the claim
that the non-standard firm can thrive only under certain institutional con-
ditions (e.g. manufacturing in Japan) has been undermined by the successful
diffusion of the non-standard approach to a wide variety of contexts. We
conclude that the process of disciplined joint inquiry, using the pragmatist
mechanisms, can actually generate the conditions necessary to maintain and
nourish a collaboration, even where the institutional environment appears
highly prone to concerns about ‘hold-ups’.
In the same section we address open questions in the non-standard view
regarding the boundary of the firm and its determination of strategy.
Notwithstanding the experimental character of the new collaborations, there
are emergent answers to these questions. We find that non-standard firms
make strategic choices using the same disciplined scrutiny of assumptions that
is brought to bear on the choice of products, production processes and
organization design. The non-standard firm develops a series of institutional
devices, ranging from novel metrics connecting the performance of operating
units and corporate wholes to new ways of linking managerial career paths
and corporate reorganization, to support the definition and investigation of
strategic choices though collaborative problem solving. Thus strategy, and
boundaries, can be seen as joint products of the operation of the non-standard
The fifth section provides a summary and conclusion. Note that in this
paper we take the ‘standard’ view of the firm from economic theory as a
starting point. We identify challenges to that view from observations of
current business developments and reinterpretations of business history, and
advance a ‘non-standard’ view of the firm as a better way to explain those
observations. In addressing a central question about bilateral collaboration
between firms, e.g. why don’t concerns about ‘hold-ups’ prevent collab-
oration?, we choose not to address another set of questions, drawn from
economic sociology, about the role of networks of firms in the production of
knowledge and the generation of rents (Powell et al., 1996; Dyer and Singh,
1998; Gulati, 1998; Kogut, 1998). In the concluding paragraphs of the paper
we suggest how our future work could address these latter questions.
2. A Profusion of Surprising Collaborations: Repainting the
Picture of the Firm
In this section we draw on three pieces of evidence of collaboration from the
automotive industry worldwide. The first, Japanese subcontracting in the
post-war period, was the focus of considerable scholarly attention during the
1980s. Collaboration between large firms and their subcontractors was widely
seen as a key source of Japanese competitive strength, yet was often viewed as
culturally and historically idiosyncratic. Yet the same pragmatic mechanisms
are now found in very different cultural settings. Second, we draw on primary
data from the USA and Canada, collected by Helper in 1993, showing that
car manufacturers today collaborate more closely with suppliers than they did
20 years ago. Yet the suppliers not only remain independent but also are able
to reduce their own vulnerability to customers by diversification. These new
relationships represent, in our view, a hybrid form of organization, one where
the ability to participate in learning is at least as important a criterion for
partner selection as low bids on piece price.
Third, we examine the takeover of the Fisher Body company by General
Motors in 1926, arguing that the takeover was unlikely to have been
motivated by standard concerns about the alignment of the interests of
principals and agents. More broadly, we show that the transaction occurred
in a period in which the major automobile producers, especially Chrysler, were
building collaborative relations with suppliers strikingly similar to those
observed currently in the USA and elsewhere. Fisher Body was a pioneer and
master of such relations. This revisionist view confronts the long-held
affection among economists towards Fisher Body as the prototypical example
of vertical integration to overcome problems of hold-ups.5
2.1 Japanese Subcontracting
The literature on the close relationships between Japanese automakers and
their suppliers is extensive (see, for example, Dore, 1983; Smitka, 1991; Sako,
1992; Nishiguchi, 1994) and we will only briefly summarize it here.
Close relationships between automakers (particularly Toyota) and their
suppliers evolved in the post-war period. These suppliers were typically either
financially independent of their customers or had less than 10% ownership by
their customers. These relationships included a number of practices that seem
untenable from the point of view of the standard theory, since they left parties
5For a collection of challenges along similar lines see a special issue of the Journal of Law and Economics,
vulnerable to opportunistic behavior. These practices included: broad, open-
ended contracts; provision by suppliers of detailed cost data to their
customers; heavy investment by suppliers in customer-specific assets (such as
location and the stationing of engineers at customers’ facilities); provision by
customers of free technical assistance to suppliers, assistance which usually
had the effect of improving the parts supplied to other customers; joint
product design efforts in which the contribution of customer and a number
of different suppliers (often rivals) was neither separately identified nor
Yet, these relationships not only proved to be stable, they also led to
excellent performance. Suppliers contributed substantially to Japanese
automakers’ higher quality (Cusumano and Takeishi, 1992), lower inventory
(Nishiguchi, 1994) and more efficient design (faster lead times and fewer total
engineering hours) (Clark, 1989).
The Japanese success at collaboration was initially attributed to factors
unique to Japan. For example, Williamson concluded ‘The hazards of trading
are less severe in Japan than in the United States because of cultural and
institutional checks on opportunism’ (Williamson, 1985, p. 122). Dore
argued that the cultural factor of ‘goodwill’ made opportunism a ‘lesser
danger’ in Japan (Dore, 1987, p. 173). Candidates for institutional checks
included supplier associations, life-time employment for the employees of
large firms and bank monitoring of corporate performance that encouraged
long-term collaboration. These and other scholars believed that economies
that could not count on the loyal workforce and patient capital that this
culture and set of institutions produced could not build the collaborative
institutions that depended on these as foundations. Thus, the prospects
seemed slight for the diffusion of the Japanese system outside its territory of
2.2 US Supplier–Customer Relationships
In the individualistic USA, strict institutional safeguards were held to be
necessary to support investment in specific assets. For example, Monteverde
and Teece argued that assemblers would need to vertically integrate ‘when the
production process, broadly defined, generates specialized, nonpatentable,
know how’ (Monteverde and Teece, 1982, p. 206). The existence of such
knowledge creates the possibilities for hold-ups even when the assembler
holds title to the specialized equipment, for example expensive dies or jigs,
used by the supplier. This hold-up of know-how, they further argue, is most
often generated in connection with the development of new parts. Hence the
greater the design effort associated with part development, the greater the
likelihood that design and production will be accomplished in-house.
Looking at 133 automotive components supplied to Ford and General
Motors in 1976, Monteverde and Teece show that ‘the development effort
associated with the design of any given automotive component is . . .
positively related to the likelihood of vertically integrated production of that
component’ (Monteverde and Teece, 1982, p. 212).
More current investigations of customer–supplier relations in the auto-
mobile industry reveal a different world. We draw on a survey of suppliers,
both independent and vertically integrated, conducted by Helper in 1993. A
questionnaire was sent to the divisional director of sales and marketing of all
firms listed in an industry directory of major first tier suppliers (both domestic
and foreign owned) to manufacturers of cars and light trucks in the USA and
Canada (see Appendix for more on the survey). Although the Helper data
does not permit replication of the Monteverde and Teece study, it does show
that the relation identified in the earlier work no longer exists.
To begin with, vertically integrated6establishments are no more likely to
be engaged in design work in general or design work on technically
demanding parts than independent firms. Fifty percent of the vertically
integrated establishments in the Helper sample did ‘the majority of engin-
eering hours’ on their product design on the current model, whereas 40% of
the independent suppliers did. This difference is not statistically significant.
On the previous model, the shares were 35 and 22%, respectively. Thus
vertical integration is today neither necessary nor sufficient for participation
in design and design participation has been growing faster among
independents than vertically integrated units. A similar picture emerges if the
focus is narrowed to producers of the most technically demanding parts;767%
of all vertically integrated establishments are in this category, as against 48%
We also examine the relation between design capability and integration by
looking at the most successful of the co-designers: what we will call the ‘super
suppliers’. These ‘super suppliers’ have the following characteristics: they do
‘the majority’ of engineering for current model parts; they have won more
than one award for supplier excellence (the first one is today often the
indispensable ticket to participation in the supply chain at all); they have at
least two customers.8
6We defined those firms who reported 30% or greater ownership by an automaker as vertically
7These were parts that respondents scored as 4 or 5 on a 5 point scale of ‘technical complexity’.
More generally, the Helper study reveals an extraordinary increase in
communications between customers and suppliers, independent of formal
status as independent or vertically integrated. The volume of face-to-face, fax,
phone and Email exchanges was huge and increasing. All types of interaction
occurred, on average, between on a daily and weekly basis.
Contacts for the purpose of ‘joint efforts to improve the product or process’
were strikingly frequent. Ford suppliers had such contacts with their customer
once a day for these purposes, while suppliers to the other firms had such
contacts every 2 days or so.
Another sign of more intimate collaboration is the increasing use of resident
engineers. In 1989 only 5% of the independent firms and 23% of the
vertically integrated units reported deploying them; in 1993 the shares had
risen to 17 and 38%, respectively. These engineers spend almost all of their
time at the customer’s design and/or manufacturing facilities, working to
resolve quality problems or to improve designs. These engineers’ salaries are
paid entirely by the suppliers. However, both customer and supplier can
assign tasks. While this ambiguous control of engineering assets would be
thought in the standard theory to lead to unproductive haggling, in practice
the infusion of ideas from multiple sources can lead to faster resolution of
quality problems, particularly those with multiple causes (Nishiguchi, 1994;
author’s interviews at General Motors, 1992).
Collaboration appears to be especially marked among the group of
diversified, award winning designers, the super suppliers. We will see below
that firms in this group expect to get help from the customer in case a
competitor develops a better or cheaper product during the life of the
contract. They learn more from their customers than those who are less
successful (and who presumably would have more to learn). They do this
without reliance on vertical integration or elaborate contracts. Note also how
different these relationships are from the economists’ model of perfect
competition, since a customer is not free to switch instantaneously among
many sellers without losing a partner it has learned how to learn from.
The success of Japanese firms in the US, and the adoption of many
‘Japanese’ practices here, makes it difficult to argue that vertical integration
or detailed contracts are necessary to support collaboration between firms.
Now we want to show that collaboration is not even novel in the US auto
industry. Despite the emphasis in the standard view of the historical
importance of the threat of opportunism in the rise of the modern firm, recent
8This last constraint is surprisingly unbinding; it rules out none of the vertically integrated suppliers
and less than 1% of the independents. While there is no direct evidence in the Helper data that firms with
more customers are more profitable than others, a careful investigation of Japanese auto suppliers did show
that these firms earn higher returns (Nobeoka, 1995).
writing portrays broad segments of the US automobile industry as having
anticipated key elements of the current style of collaborative customer–
supplier relations in the first decades of this century. Thus, it is also hard to
argue that technological factors (such as the diffusion of computers) drove the
development of these organizational forms.
2.3 Retelling the History of Vertical Integration in the US
The takeover by General Motors of the Fisher Body Corporation in 1926 has
become the canonical example of the logic of the standard argument.9In the
version provided by Klein, Crawford and Alchian (Klein et al., 1978), Fisher
refused to build its stamping plants next to GM assembly plants, because it
feared that dedicated plants would have difficulty serving Fisher’s non-GM
customers. Fisher also refused to change a cost-plus pricing agreement in its
original 1919 contract, an arrangement that proved unexpectedly lucrative
due to unforeseen increases in demand for the closed steel bodies made by
Fisher. In this view, the purpose of GM’s purchase was to resolve an agency
problem arising from asset specificity and an insufficiently complete contract.
GM valued control of dedicated resources enough that it was willing to buy
control of Fisher from its owners. Indeed the whole point of the story is that
control is all that mattered: each side was presumed to know the capacities
and intentions of the other and the only question, decided by title of
ownership, was to whose purposes resources and wills would be directed.
But this account is incorrect, regarding both the Fisher–GM relationship
and the larger context of customer–supplier relations in the industry.10Three
corrections are particularly important : (i) Questions of residual control played
at most a subordinate role in the transaction; (ii) Fisher, like many other large
suppliers, met GM’s needs for dedicated resources without abandoning
collaboration with other customers; (iii) in the 1920s and 1930s, US auto
firms were more intent on extending and regularizing collaborative relations
with suppliers than resolving agency problems through vertical integration.
Beyond these corrections, we offer a different interpretation of the history. We
argue that the aim of the Fisher Body transaction was to enlist the aid of the
Fisher brothers in constructing such a collaborative system within GM. From
this perspective, the point of the deal would have been to transfer expertise
9See for example Hart (Hart, 1995), which draws extensively on Klein and co-authors’ account in
describing a ‘property-rights’ theory of the firm.
10We are very grateful to Thomas Marx of General Motors for generously providing us with source
materials from his own research into Fisher Body.
rather than control or, more specifically, to transfer expertise in the
collaborative organization of expertise rather than grant exclusive control as
a substitute for collaboration.
Our revisionist approach begins by noting discrepancies between the
standard account and some facts. First, as Klein, Crawford and Alchian
register in a footnote, GM had owned 60% of Fisher since 1919. From 1919
to 1924, the stock was governed by a trust with two of the trustees repres-
enting GM and two representing Fisher. However, this trust agreement
expired in 1924, 2 years before GM bought 100% of Fisher.11GM’s majority
ownership should have been sufficient to give GM a strong voice in Fisher’s
decision-making after 1924. Indeed, the pricing agreement was revised in
1924, to require that Fisher’s prices be set in the same manner as at GM’s
wholly owned divisions (Chandler and Salisbury, 1971, p. 576; Freeland,
Second, Fisher had already begun to co-locate plants with GM facilities
under the trust arrangement. A 1924 report for stockholders states:
Wherever, in the United States and Canada, there is an important
passenger car plant of Buick, Cadillac, Chevrolet, Oakland, or
Oldsmobile, there is, or will be close by, a plant of Fisher Body, adequate
to meet the demand for closed bodies of high quality.
Much of this plan had already been carried out.
In 1923, GM undertook a major expansion of Chevrolet chassis
production facilities and opened Fisher body plants adjacent to each new
Chevrolet factory. During that year pairs of Chevrolet chassis plants and
Fisher body assembly plants began joint operations in Norwood, Ohio;
Janesville, Wisconsin; and Buffalo, New York.(White, 1991, p. 56)
At this time Fisher also had body plants in Flint (near Chevrolet and Buick),
Oakland (near Pontiac) and Lansing (near Oldsmobile), all in Michigan.12
This building boom resulted from informal discussions, not from installation
of institutional safeguards: The authorized biography of William Knudsen,
the architect of Chevrolet, reports that ‘Early in Knudsen’s career with
General Motors, he persuaded Fisher to build body plants adjacent to
Chevrolet factories’ (Beasley, 1947, p. 141). Knudsen had come to GM in
1921, 5 years before GM bought complete formal control of Fisher.13
11GM board minutes, September 25, 1919 (Chandler and Salisbury, 1971).
12For the locations of Fisher’s plants and their dates of opening see the 1 June 1933 summary of the
Fisher Body Corporation, division of General Motors (Pound, 1934). For location of GM plants at the time,
see the GM annual report, 1920, p. 12.
13Why do Klein, Crawford and Alchian report otherwise? They seemed to have relied on Alfred P .
Third, the advantages of asset specificity in the form of co-location were
sometimes offset by other considerations, as the major automobile makers
routinely continued to have bodies made by outside companies for decades
after GM bought Fisher. Even GM procured some bodies from the Budd
Company (Schwartz, forthcoming). Ford obtained bodies from Budd, Murray
and Midland Steel and made some in-house. About two thirds of the
outsourced total was provided by Briggs. In 1929 it even signed a renewable
5 year lease to make these bodies in Ford facilities. Output from the rented
Ford plant was sold to other manufacturers as well (White, 1991).14
A. O. Smith supplied GM, Ford and other automobile makers with frames
starting in the 1920s, and still provides these customers today. Indeed, Ronald
Coase, the intellectual forebear of the standard theory, visited an A. O. Smith
factory in Milwaukee in the 1930s, drawn by the firm’s reputation as one of
the most efficient manufacturing companies of its day (Coase, 2000). Its
evident success as an independent producer caused him to reject in advance
the idea that vertical integration was the most efficient solution to problems
of interest alignment: the view that later came to dominate the literature on
transaction cost economics. (We will return to this theme below.) There are
many other contemporaneous examples of such dependence of customers
upon independent suppliers for key parts.15
These examples draw attention to larger changes in industrial organization
of the period that transform the context and thence the significance of the
standard account. In the revised picture, we present the pre-1940 auto
industry as an independent discoverer, avant la lettre, of what are now called
Japanese methods by US automobile producers.16Supplier–customer
relations were Japanese-like in the sense that they involved long-term
associations, joint product development and attempts to minimize inventory
while producing a variety of products.17
Sloan’s testimony in US versus du Pont, GM, et al. in 1953. In this testimony, given almost three
decades after the events took place, Sloan does say that ‘the Fisher Body Corporation was unwilling to put
an investment in these [body] assembly plants’, but this statement is (i) contradicted by the facts above
and (ii) only a minor part of Sloan’s explanation of why GM bought Fisher. He gives more weight to
problems with Fisher’s cost-plus contract and to the fact that the Fishers ‘. . . were very capable people
and . . . we needed that kind of talent at General Motors at the time, and needed it badly, and I was very
anxious to have them come in and help us in the broader problems of General Motors’ (US District court
trial transcript no. 49-C-1071, pp. 2908–2909).
14‘Body by Briggs,’ Special-Interest Autos, Nov./Dec. 1973, 24–29. In 1953 Briggs was bought by
15For example, GM bought more than half of its brakes from Bendix (a company in which GM had a
25% ownership) and Bendix sold GM more than half its output. Ford was substantially dependent on out-
siders for rubber, aluminum, wheels, brake shoes, assembled bodies and radiators (Schwartz, 2000, p. 13).
16This account draws on Helper (1991), Hochfelder and Helper (1996) and Schwartz (forthcoming).
17Relationships were different from the Japanese model as it first became known in the West in the
1970s and 1980s in that the subcontractor was not necessarily the junior partner and that the system was
At the turn of the century suppliers were larger and more established than
assemblers; in many cases they had long produced for other industries, such
as carriages and bicycles. These suppliers were important sources of both
innovation and working capital (for details see Helper, 1991). Later, ideas and
financing flowed both ways. For example, at Ford in the 1909–1914 period
‘the Company was not then averse to purchasing virtually all of its materials
and parts from independent producers’. The automobile maker shared its
growing management expertise with its suppliers.18
Joint design was also a feature of this period (although with power relations
reversed to a surprising degree). For example, Timken advertised that
‘Timken axles in your car, no matter what its size or price, were selected and
installed only after many conferences between Timken engineers and the
engineers of the car builder’. Timken admonished these car builders that they
could not include Timken axles ‘merely to furnish a selling point; they must
be built in—not tagged on’; Timken often insisted on modifications in the
car’s design before it would allow the axles to be installed.19
Between 1915 and 1925 there was a shift towards vertical integration, as
Ford built his own parts plants and General Motors acquired a number of part
of new car value. After 1926 there was a return to outsourcing (and no reversal
of thisgeneral strategyuntil after World WarII).Bytheearly1930stwo-thirds
of Ford’s cost of production came from outside suppliers. At GM the share was
lower, but largely because of the acquisition of Fisher Body, which accounted
of total value) from outsiders as of 1938 (Schwartz, 2000).
The return to outsourcing was a response by the established producers,
of independent parts suppliers. The first was new competitors, independent
firms that assembled standard parts in leased facilities to enter the market
cheaply, without the high fixed costs associated with product development and
capital investments. The second threat was used cars, repaired and sometimes
improved with parts provided directly by suppliers and installed by small firms
with low fixed costs. Once the depression started, these threats were joined by
less self-consciously organized, more ‘feral’ as Schwartz describes it (Schwartz, forthcoming, p. 1).
18‘The Ford Motor Company purchased materials for its components-makers, reorganized their
manufacturing processes, supervised their larger policies, and, in some cases, aided them in financing
production. The Company became so dependent upon the production of its specialized suppliers that its
own operations were frequently within thirty minutes of suspension because of tardy deliveries of parts or
materials’ (Seltzer, 1928, pp. 89–90 and 100).
19Advertisements in The Auto Era, February 1916 and August 1917, inside back cover (Hochfelder and
Helper, 1996, pp. 4–6).
an opportunity: automobile makers found attractive prices from outside
suppliers who were desperate to fill their factories.
To counter the threats, established makers increased the pace of innovation.
of used cars compared with new ones. GM led the way with the introduction of
the annual model change in the early 1920s, with Ford, the holdout among the
major producers, finally following suit with the switch to the model A in 1927.
The drive for innovation, however, led to deepening collaboration with the
parts suppliers, not, as the standard account might suggest, their margin-
alization. The costs of development were simply too high for even the richest
of the assemblers to bear alone; hence the efforts to reduce these costs by
co-developing components of all types with suppliers. The suppliers, in turn,
reduced costs by co-developing variants of each product with other major
customers and by drawing on experience gained in the aftermarket.
The upshot was the proliferation of collaborative arrangements that
strikingly anticipate the non-standard customer–suppler relations reflected in
the Helper data reported above (Langlois and Robertson, 1989; Schwartz,
2000). ‘By the mid-1920s,’ Schwartz writes, ‘GM had developed and imple-
mented a policy much like that of Toyota: seeking out more than one source
for all its components, limiting the number of sources to two or three
companies, and establishing long-term relationships with these firms’
(Schwartz, 2000, pp. 72–73). The corporation’s relations with Libbey-Owens-
Ford for glass products and with automobile body makers such as Budd
(despite the amalgamation of Fisher) were instances of this policy.20
Similarly, Hounshell notes that ‘Ford went to outside suppliers more and
more during the early years of the Depression’ (Hounshell, 1984, p. 300). The
resulting collaboration was crucial to the success of such large design changes
as the changeover to the V-8, as well as the solution of the thorny problems of
machine design and production layout typically accompanying introduction of
new components.21Accounts of Ford operations from the mid-1930s portray
such co-development as crucial to the firm’s operating routines.22
The same drive to innovation that led to reliance on outside suppliers also
led to creation of a ‘hand-to-mouth’ inventory system, a precursor of just-in-
time in which almost no stocks of parts were held in reserve against supply
20After selling its internal supplier of glass, National Plate Glass, to Libbey-Owens-Ford, GM sourced
two thirds of its windshields from the new owner.
21Hounshell recounts the co-development of a new gas tank for the model A by Ford and the Gibb
Company. Henry Ford developed a radical new design, but existing equipment could not seal the tank at
production volumes. Gibb, which manufactured welding equipment, sent technicians to the plant.
Together with Ford engineers and production workers they redesigned the tank, welder and the assembly
process concurrently (Hounshell, 1984, pp. 284–285).
22Barcaly, cited in Schwartz (Schwartz, 2000, p. 74).
disruptions.23The switch to hand-to-mouth inventory in turn reinforced the
reliance on inventive collaboration with outsiders.24
The embodiment of these developments was the Chrysler Corporation.
Its founder was Walter Chrysler, who made his name while rising to be
president of Buick-GM, only to quit after falling out with William Durant,
the initial architect of General Motors. Prototypes of the first Chrysler
model were co-developed by a team of three in-house engineers working
with component suppliers, mostly in Detroit, but as far away as Ohio and
Indiana. Every component of that first model was outsourced and assembly
was in a rented facility. The finished product included many features previ-
ously available only on luxury models that cost three times as much as the
Chrysler, suggesting that the co-development efforts achieved remarkable
Chrysler’s second major model innovation, the Plymouth, carried the logic
of collaboration a step closer to the modern system traced above. Again the
design was co-developed with suppliers. However, for the Plymouth the prod-
uct plan called for ongoing and substantial improvements in the efficiency of
the production process after the model was introduced. The Plymouth
entered the market in 1929 in the low-to-mid price range. In the following
three years, joint efforts by suppliers and assembler reduced production costs
by 30% and the well-equipped Plymouth began to compete with entry-level
Ford and Chevrolet models. Here the innovative capacities of collaborative
supply relations were made manifest to a whole industry.
To complete the reinterpretation of the Fisher Body transaction, we note
that Fisher Body was an industry leader in collaborative supply arrangements.
During World War I, Fisher and Hudson had co-developed the first
cost-effective closed automobile body. This development attracted the atten-
tion of General Motors and led to its initial acquisition of 60% of Fisher stock.
But the Fisher–Hudson collaboration continued despite the change in formal
ownership, and with the additional help of machine tool makers the two firms
developed techniques for stamping large body panels from wide sheets of
steel. This so reduced welding and other manufacturing costs that a sedan
could be produced for the price of an open bodied car (Schwartz, 2000, p. 72).
In 1923 Fisher Body collaborated with the Chrysler design team to produce
the curved fenders and bumpers that would eventually come to be associated
with the idea of streamlining.25
23Inventory turns at the Ford River Rouge plant were in the range 35–40 per year in the early 1920s,
comparable with the levels achieved by the best Japanese producers 60 years later. (Schwartz, 2000).
24However, the ‘hand-to-mouth’ system did not include JIT’s link between ever tighter inventories and
25(Schwartz, 2000, p. 71). Fisher also collaborated with competitors: Charles Fisher and Walt Briggs
if GM needed them and GM was showing no sign of caring deeply about asset
specificity (as demonstrated by its increasing use of outsiders in crucial supply
relations). So is it not possiblethat GM bought theoutstanding sharesof Fisher
Body in order to acquire the expertise of the Fisher brothers in organizing its
(GM’s) operations rather than to impose its will on their company?
There is a surprising amount of evidence in favor of this view. First,
contemporary accounts support it. For example, Pierre Dupont, in a letter to
a board member describing negotiations with the Fishers, states that in return
for stock, in 1924 two of the brothers had:
. . . agreed to sever their connection with the Fisher Body corporation
and become members of the General Motors executive committee, for
the purpose of acquainting themselves thoroughly with General Motors
operations and difficulties, without any allotment of specific duties. This
I think is a step forward. As it has not been possible for Alfred Sloan to
give personal attention to all of the ramifications of the Corporation, I
think the Messrs. Fisher are especially well adapted to assist him in
ferreting out troubles promptly and applying remedies.26
Similarly, the GM annual report for 1926 asserts:
Many benefits will accrue through the consolidation of the two
properties. Of even greater importance [than operating economies], is
the bringing into the General Motors operating organization in closer
relationship, the Fisher brothers, through whose constructive ability,
foresight, and energy the institution bearing their name has been built
up to the dominating position it now holds.
The brothers were famous as team players,27with expertise in just the mix
of design, production28and management skills that were succeeding at
Chrysler. Thus their accession to GM might well have looked to contem-
poraries more like an effort to reorganize the latter than an assertion of the
power of a principal over its agent.
This interpretation accords with that of Chandler and Salisbury, who write
were ‘fast friends’ who would spend ‘many an evening together’ and ‘talk shop, compare notes, and
even help each other on sticky business decisions.’ Charles Fisher served for a long time on Briggs’s board
of directors (Body by Briggs, p. 24).
26Letter to Sir Harry McGowan, 10/21/24, Defendants’ trial exhibit, Fisher Body Corporation, 366 US
316, 5/22/61. See also the quote from Sloan above.
27The Fishers ‘were team players both at GM and in their family business ventures and their practice
of discussing business matters each day at luncheons and making decisions as a family was among their
most famous characteristics. The Fishers collaborated so closely, a GM vice president once joked, that when
one of the brothers cut himself while shaving, they all bled’ (White, 1991, p. 61).
28In particular they had engineering talent which allowed them to see how to develop machinery which
would produce at high volume pieces durable and precise enough to withstand the speed of a car (Fisher
Body: Its Contribution, p. 4).
that the desire to gain access to the Fishers’ management skills was the motive
for GM’s increase of its ownership of Fisher Body from 60 to 100%. They do
not mention asset specificity as a consideration (Chandler and Salisbury,
1971).29Finally, the Fisher brothers were paid in stock and this form of
payment is compatible with the intention of drawing them into the active
management of the corporation.30
What was it that the Fishers did that made them so valuable? They had
considerable design and engineering expertise, as mentioned above. But most
important for our story, they helped GM adopt inside the corporation the
collaborative style that Chrysler was using with outside suppliers.31Put
differently, the Fishers contributed greatly to the multidivisional corporate
structure and system of coordinating committees that Alfred Sloan so
masterfully developed to manage GM. To repeat, this analysis stands the
standard account on its head. Instead of seeing the ‘takeover’ as a response to
the threat of opportunism, it presents the amalgamation as an effort to
construct a variant, suited to the conditions of GM, of a collaborative supplier
system most clearly articulated in the 1920s by Chrysler.
Buying Fisher Body was only one way for GM to achieve this goal; legal
counsel could probably have devised instruments that accomplished the same
ends without a change in ownership. Our point is that in designing their
firm’s organizational structure, managers consider effects of that structure on
learning as least as much as they do effects on opportunism. In the Fisher case
there is no evidence in the extensive records available that the brothers
behaved opportunistically in the sense of ‘self-interest-seeking with guile’
(Williamson’s definition; Williamson, 1985). They did not withhold infor-
mation or seek to mislead GM. They did, however, want to maximize their
incomes. Aware of this, GM changed the financial incentives offered to the
Fishers in order to get their full attention. This is a very different account from
the opportunism story: instead of buying Fisher Body because they did not
trust the Fisher Brothers, GM bought Fisher Body because they trusted the
Fisher Brothers so much that they wanted them intimately involved in
managing all of GM’s assets.32To review the argument so far, we find that
29See also Beasley, who says that ‘[Fred J.] Fisher’s was a compelling voice’ within GM in the early 1920s
(Beasley, 1947, p. 141).
30From this perspective the fusion with Fisher anticipates the fusion of GM with EDS in the 1980s. In
both cases, on this interpretation, the corporation amalgamates a major supplier in order to learn how to
organize its own affairs. On the relation between GM and EDS see Levin (Levin, 1989).
31For example, at Cadillac, before (and especially after) the merger, Lawrence Fisher worked closely with
supervisors and paint supply company personnel to improve the paint process, relying heavily on practices
formalized later as Total Quality Management, such as organized experimentation and employee
involvement. One outcome was to reduce drying time for bodies from 21 to 7 days. (Lawrence Fisher
testimony, US versus Dupont 353 US 586 (1952), fol. 998–1025.
32Coase points out that the Fisher Brothers essentially controlled the GM executive committee in the
current customer–supplier relations in both the Japanese and the US auto
industries are not like those depicted in the standard account. We further find
that the paradigmatic case of the standard logic has surprising affinities with
non-standard developments today. We turn now to a more systematic
comparison of old and new forms of organization and assumptions about the
possibilities and mechanisms of cooperation that underpin them.
To review the argument so far, we find that current customer–supplier
relations in both the Japanese and the US auto industries are not like those
depicted in the standard account. We further find that the paradigmatic case
of the standard logic has surprising affinities with non-standard developments
today. We turn now to a more systematic comparison of old and new forms
of organization and assumptions about the possibilities and mechanisms of
cooperation that underpin them.
3. Two Ideas of the Corporation
In this section we return to the contrast between two ideas of the corporation,
primarily to elaborate our portrait of the non-standard firm. We approach this
task from multiple perspectives. First, we present an ideal-typical sketch of
the standard firm as treated in the economics and industrial organization
literature, in order to highlight differences from the non-standard firm that is
our primary focus here. This sketch is not entirely conceptual; as we note
below, the standard view gained widespread acceptance because it captured
with notable fidelity key empirical features of the post-war era of large
corporations operating under relatively stable economic conditions. Still, we
do not pretend to give the standard view equal time here, for its parameters
and tendencies are already well documented. Second, we offer a description of
the pragmatic mechanisms that characterize both the internal operations and
the collaborative activities of the non-standard firm, drawing in part on the
empirical examples highlighted in the previous section and in part on our
field-based investigations of such firms. Third, we address our central
performance claim, that the non-standard firm is, in collaboration, able both
to advance knowledge and control opportunism.
3.1 The Standard Firm
At the origin and core of the standard view are industrial organizations with
1920s, hardly what one would expect GM to grant to a group viewed as nefarious (Coase, 2000; see
also Brooks and Helper, 2000).
distinct characteristics: centralized, hierarchical and vertically integrated.
Goals set by headquarters were achieved by hierarchically ranked, specialized
subunits, all part of a single organization. For much of this century these
features were seen as expressing basic and incontrovertible principles of
effective human action.
efficient. Adam Smith gave powerful reasons for thinking that coordinated
effort by one top-of-the-widget maker and one bottom-of-the-widget maker
can produce more per unit of time and other resources than two whole widget
makers working separately. The greater the subdivision of labor, the greater
the efficiency; furthermore, the greater the extent of the market, the less the
risks of dedicating resources (in the form, say, of single purpose automatic
machines) to the highly specialized tasks of finely divided labor.
Given extensive markets, Smith’s ideas culminate in the familiar notion of
economies of scale: the greater the production volume, the lower the unit cost.
Separation of conception and execution and the centralization of the former
at the top of a corporate hierarchy follow immediately. The precondition of
efficiency in this case is a superintendent with comprehensive knowledge of
market possibilities and production techniques to design the product and
initiate subdivision of production into specialized tasks, each of which can be
further decomposed by subordinates.
The second principle, in many ways a corollary to the first, is that this
efficient specialization creates vulnerabilities. The more highly subdivided the
production of any product, the tighter the connections or the more com-
plementary the relations among the single components of the production
process. Conversely, the less likely it is that any of those components can be
put to use in other production processes. (Consider, again, the example of an
automatic machine performing one operation on a single part of a particular
model of a given product.)
Owners of highly specialized, complementary resources cooperate,
therefore, at great risk. Whoever invests first in the joint project can be ‘held
up’ by a partner who simply refuses to commit the complementary resources
(without which the initial investment is worthless) except under terms more
favorable than originally agreed. But the threat of expropriation deters the
initial investor, so the joint project is paralyzed by the prospect of the
vulnerabilities it creates. Vertical integration is the organizational answer to
this danger of opportunism. If a single owner has exclusive control over all
the phases of production, and has so specialized the division of labor for each
phase that there are very few buyers of the relevant goods and services, the
The first principle is that the division of labor is
possibility of the owner being held up disappears.33The third principle is an
ingredient of the first two. It ties the most basic features of organizations to
the limits of human cognition. It acknowledges our manifest inability to
perform, in anything like the time available, the calculations necessary to
assess the costs and benefits of the choices plausibly open to us at any
moment. To act, given this bounded rationality, we must economize on our
limited attentiveness by making use of the expedients of habit and the
subdivision of complex tasks into simpler ones. By habit, we take crucial
elements of our situation so for granted that we don’t need to be attentive to
what we are assuming and how it shapes our further thoughts. We break
problems down into chunks whose separate solutions are within our cognitive
grasp and which can then be fitted together into a comprehensive solution to
the original question. When problems are sufficiently complex as to require
collaborative solutions, centralization and hierarchy are called upon to
partition problems into manageable chunks. But they also ensure that
subordinates, who, by definition, know things their superiors cannot, cannot
make self-interested use of their expertise.
The third principle is an ingredient of the first two. It ties the most basic
features of organizations to the limits of human cognition. It acknowledges
our manifest inability to perform, in anything like the time available, the
calculations necessary to assess the costs and benefits of the choices plausibly
open to us at any moment. To act, given this bounded rationality, we must
economize on our limited attentiveness by making use of the expedients of
habit and the subdivision of complex tasks into simpler ones. By habit, we
take crucial elements of our situation so for granted that we don’t need to be
attentive to what we are assuming and how it shapes our further thoughts.
We break problems down into chunks whose separate solutions are within our
cognitive grasp and which can then be fitted together into a comprehensive
solution to the original question. When problems are sufficiently complex as
to require collaborative solutions, centralization and hierarchy are called upon
to partition problems into manageable chunks. But they also ensure that
subordinates, who, by definition, know things their superiors cannot, cannot
make self-interested use of their expertise.
Routines (the organizational equivalent of habits) likewise do double duty.
They establish connections among the parts and simultaneously place limits
on the operation of each part as necessary to maintain the integrity of the
whole. Thereby they limit the possibilities for self-dealing that specialization
33On the increases in bargaining power resulting from limiting other agents’ alternatives see Baker,
Gibbons and Murphy and Marglin (Baker et al., 1998; Marglin, 1974).
The cognitive gains from hierarchical specialization and routinization,
moreover, are mutually reinforcing. The more routinized a task, the easier it
is to learn. This, as Smith observed in his ‘pin factory’ example, explains the
almost superhuman dexterity of operators performing simple, repetitive jobs.
However, the more routinized the operations (the more, say, it consists of a
few repetitive movements of the hand), the easier it is to decompose it yet
further. Smith counted this possibility of further simplification as another
source of the efficiency gains of the division of labor.34The principles
associated with the standard firm derive their power, in large part, from the
undeniable efficiency gains associated with specialization. Yet these very
principles carry with them a caveat about inevitable constraints; they suggest
that organizations, no less than persons, are condemned to knowledge traps
of their own making. To know we must specialize, yet in specializing we come
to be defined by what we unknowingly take for granted. Hence, the true price
for organizations of gains through specialization (beyond the risk that a shift
in demand will devalue dedicated equipment) is a kind of institutional
self-oblivion. To pursue its ends effectively, the org- anization must stop
inquiring why those ends are its ends or why it pursues them as it does. When
routines become entrenched as the inevitabilities of common sense, the
organization is the prisoner of its history, choosing within the forgotten limits
imposed by its initial choices.
The principles associated with the standard firm derive their power, in large
part, from the undeniable efficiency gains associated with specialization. Yet
these very principles carry with them a caveat about inevitable constraints;
they suggest that organizations, no less than persons, are condemned to
knowledge traps of their own making. To know we must specialize, yet in
specializing we come to be defined by what we unknowingly take for granted.
Hence, the true price for organizations of gains through specialization
(beyond the risk that a shift in demand will devalue dedicated equipment) is
a kind of institutional self-oblivion. To pursue its ends effectively, the org-
anization must stop inquiring why those ends are its ends or why it pursues
them as it does. When routines become entrenched as the inevitabilities of
common sense, the organization is the prisoner of its history, choosing within
the forgotten limits imposed by its initial choices.
Partial antidotes exist. The struggle to survive given scarce resources selects
the organizations with routines most suited to the demands of their environ-
34Smith suggested that this ongoing simplification might be accomplished either by attentive workers
or ‘philosophers’ specializing in this very task. In practice, the standard approach leads most naturally to
engineers in the simplifyingrole, given how well the fine-grained breakdown of a task facilitates the process
of fully automating it, and how well automation provides both standardization of routines and hierarchical
ments. Even if particular organizations cannot re-orient their activities to
accommodate change, some organizations will do very well indeed if their
routines and specialized expertise are well suited to the demands of their
environment. Furthermore, in anticipation of their own congenital rigidity,
organizations can establish counter-institutions. For example, internal
research and development laboratories are established to renew crucial aspects
of current routines by routinizing the creation of knowledge. But the very fact
that the process of renewing routines must be placed outside the operational
core acknowledges that this defect of organizations cannot be corrected from
within. Such counter-institutions, by their very existence, thus ratify the view
that the astonishing accomplishments of hierarchically specialized institutions
are necessarily associated with the danger of stultification.35For the years of
the post-World War II expansion, these potential cognitive and economic
costs were so vastly outweighed by the benefits of specialization that these
latter alone came to be taken for granted as defining the logic of efficiency.
However, starting in the mid 1970s, for reasons we will not consider here, the
stable markets for standard goods on which this system of production had
rested became fragmented and volatile and some of the costs of specialization
were suddenly manifest and onerous (Piore and Sabel, 1984). In volatile and
fragmented markets, the prospect of amortizing a huge initial investment in
the design of highly complex products and production systems required to
achieve economies of scale was dauntingly risky. Firms that responded to
foreign competition with bold projects could easily miss their market’s
window of opportunity and be left with nothing but write-offs to show for
their temerity. Firms that responded cautiously saw advances in products and
processes passing them by, making obsolete once vital capabilities.
For the years of the post-World War II expansion, these potential cognitive
and economic costs were so vastly outweighed by the benefits of specialization
that these latter alone came to be taken for granted as defining the logic of
efficiency. However, starting in the mid 1970s, for reasons we will not
consider here, the stable markets for standard goods on which this system of
production had rested became fragmented and volatile and some of the costs
of specialization were suddenly manifest and onerous (Piore and Sabel, 1984).
In volatile and fragmented markets, the prospect of amortizing a huge initial
investment in the design of highly complex products and production systems
35Indeed, the common opposition of firm and market typically assumes that the market overcomes
cognitive stultification in the firm by introducing a variety of relationships and perspectives (Kogut, 1998).
Yet arms length market relationships rarely provide fertile ground for the pooling of perspectives (or, put
differently, for the process of making tacit knowledge explicit and shareable) that we identify as critical to
pragmatic collaborations. Exposure to such market relationships may destroy the blinders that firms
inevitably acquire, but won’t necessarily lead to learning and insight.
required to achieve economies of scale was dauntingly risky. Firms that
responded to foreign competition with bold projects could easily miss their
market’s window of opportunity and be left with nothing but write-offs to
show for their temerity. Firms that responded cautiously saw advances in
products and processes passing them by, making obsolete once vital
For a time this Hobson’s choice seemed a cruel fact of organizational
nature. US companies forgot the collaborative possibilities contained in their
own past and tried merely to cut costs without changing their centralized
ways. But in the last decade, under continuing competitive pressure, US firms
have increasingly come to understand, adopt and further develop alternative
forms of production. At the heart of this approach is a reinvention of crucial
aspects of customer–supplier relations, aspects that formalize ways of
interacting that were emergent in the years before World War I and from the
mid 1920s to World War II. Efficiency is still attained, but now quality, speed
and the ability to manage product variety are achieved as well. And the
forgetful rigidity of the standard approach is no longer the price paid for
3.2 The Non-Standard Firm
The non-standard firm is federated, not centralized: decisions of higher level
entities are crucially shaped by the decisions of their constituent units. The
federation is open, not vertically integrated: components or services crucial to
the final product of one firm can be provided by independent companies and
the firm’s internal specialized producers can provide outsiders with crucial
inputs. These outward differences are the result of distinctive principles of
efficiency and governance; these in turn are rooted in a new understanding of
cognitive possibilities. Inverting the logic of the standard approach, routines
become accessible to deliberate evaluation without subverting them as guides
to normal activity.
utilized by the ‘non-standard’ firm. The fundamental unit of the new firm is
the team or work group. This unit has the responsibility to achieve goals
mutually agreed upon with its collaborators, by means that are mutually
determined through group deliberation. Thus, unlike the specialized
subordinates in the hierarchy of a mass producer, the work group is free to
change its internal organization and to choose inputs (tools, engineering
services, components and so on) from either inside or outside the company
Here we elaborate on the pragmatist mechanisms
that furnishes its members. The choice to ‘buy’ externally rather than ‘make’
internally is not biased by the fact that the relevant product or service is
available inside the collaborating firm(s); ‘inside’ resources must outcompete
‘outside’ resources to be selected. To the extent that it has this autonomy, the
work group functions as if it were an independent firm, whatever its formal
Coordination within and across these groups is by novel methods of iterated
goal setting, inspired by organizational breakthroughs in Japan, but no longer
limited to Japanese firms. These methods establish a first idea of what to
produce (and how) through an exacting survey of current products and
processes. This benchmarking, supplemented by assessments of what new and
unproved techniques might become available for use, disrupts established
expectations of what can be done. It thus casts pragmatic doubt on the
advisability of current methods and setting the stage for exploration of the
Design follows a disciplined, decentralized process known as simultaneous
engineering. Each subunit responsible for a constituent component proposes
modifications of the initial plan, while also considering the implication of like
proposals from the other subunits for its own activities. Provisional designs
are thus evaluated and refined and the cost of each attribute is compared with
its contribution to functionality using the techniques of value analysis/value
Once production begins, systems of error detection and correction use
breakdowns in the new routines to trigger searches for weaknesses of the
design or production process that escaped earlier examination. As in prag-
matism, the continuous adjustment of means to ends (and vice versa) is both
the means and end of collaboration among the producers.
Moreover, the exchanges of information required to engage in bench-
marking, simultaneous engineering and error detection and correction also
allow the collaborators to monitor one another’s activities, closely enough to
detect performance failures and deception before they lead to disastrous
consequences. Ultimately, these information exchanges lead the actors to
convergent understandings of the world they are exploring. Because it ties
mutual assessments of reliability to joint explorations of capability, we speak
of the system of collaboration as a whole as learning by monitoring.
Thus, for example, the new van design team in an automobile firm sets the
36But note that there is great variation across and even within industries with regard to the de facto
powers of these groups and it is rare for teams to have explicit powers to hire and fire, although they often
exercise substantial informal control in such matters. The reasons for the usually quite limited powers of
teams are beyond the scope of this paper. On the temptations of owners to sacrifice learning for monitoring
see Babson (Babson, 1995).
general performance characteristics of the vehicle by benchmarking the best
features of current vans and exploring which innovations under development
can be incorporated into its design. It next decomposes these general goals,
again by reference to leading examples and comparison of possibilities, into
subtasks such as the design of an engine, or heating, ventilation and air
conditioning system, and chooses a specialist team from inside or outside the
parent company to realize the initial specifications.
The separate project teams elaborate all the subsystems concurrently,
applying to that task the same kind of evaluation of competitors’ successful
efforts and developmental possibilities used in the van team’s first round of
benchmarking. In addition, they benchmark the production processes central
to their eventual products to ensure that the methods employed will meet or
surpass the efficiency of their most capable competitors. Engine plants, for
instance, will compare their prospective performance, measured in units of
output per unit of inputs, with the actual performance of plants making
similar engines at similar production volumes and certified, by warranties
offered to the final consumer, to be of similar reliability.
This requires companies and units to pool data on the actual performance
of key processes, and this they do, either through bilateral information
swapping or often by creating industry institutes that rank each company by
process, on condition that each inquiry be accompanied by a full description
of the inquirer’s own current results. Sharing such proprietary performance
data with competitors would once have been unthinkable. But in volatile
markets, companies realize it is simply too risky to assume that one’s current
processes, no matter how much they improve on past practice, are com-
petitive, let alone superior.37Then the initial overall goals are modified by the
methods of simultaneous engineering, e.g. the engine design group may find
a way to better its target specifications or to cut its manufacturing costs if it
can persuade other component groups that design characteristics should be
Refinement of the eventual design continues by means of just-in-time
production methods and the error detection and correction methods
associated with it. In just-in-time production, parts are supplied to each work
station only as needed: ideally, one at a time. This renders disruptions and
defects immediately visible. Breakdowns at one station halt production by
disrupting the flow of parts to downstream operations; similarly, defects
37Note that benchmarking need not yield just an effort to copy a competitor’s product. Creative
benchmarkers find ways to use good performance in other industries to challenge themselves. For example,
several manufacturers have studied L.L. Bean’s order fulfillment system and Xerox chose to study Lego for
precision plastic molding.
introduced in one manufacturing step make it difficult or impossible to
accomplish the subsequent ones correctly.
To ensure the flow of production, therefore, the source of disruption or
defect must be identified as a failure of workmanship or an imperfection of
design or operating organization. Such inquiries typically require tracing long
causal chains back to improbable origins by an insistent series of questions,
sometimes called the ‘five why’s’. For example:
• Why is machine A broken? No preventive maintenance was performed.
• Why was the maintenance crew derelict? It is always repairing machine B.
• Why is machine B always broken? The part it machines always jams.
• Why does the jam recur? The part warps from heat stress.
• Why does the part overheat? A design flaw.
Thus error detection and correction, like benchmarking and simultaneous
engineering, reveals possibilities for improvement in unexpected (mis)con-
nections among the parts of complex endeavors; the cumulative effect of these
results is captured in improvements in the benchmark standards for various
The master cognitive innovation of this new type of firm is embodied in
precisely these apparently modest, even commonsensical, institutions. For
benchmarking, simultaneous engineering and error detection methods like
the ‘five why’s’ are procedures for doing just what the standard view of
effective action given bounded rationality says cannot be done: routinely
questioning the suitability of current routines. Whether in the initial
specification of new designs (benchmarking), the concretization of these
approximations (simultaneous engineering) or in the course of their practical
application (error detection), this disciplined inquiry of routines occurs at just
those times when self-interrogation seems most valuable but most difficult.38
These mechanisms oblige the actors to search for solutions in a circum-
scribed space of possibilities (the set of best current or potential designs, the
activity chains that might have caused a particular breakdown) whose exact
contours and contents they could not have anticipated. The outcome of the
search is thus likely to be sufficiently unfamiliar and disconcerting to force
re-evaluation of habitual responses. The new firm is thus a member of a new
class of institutions defined not by the fixed routines to which they are
oblivious, but rather by the routines they use for interrogating and altering
38It is natural for firms to fall into stable patterns of interaction with their closest collaborators (Gulati,
1998). This may, over time, lead towards the same cognitive stultification that we associated with the
internal workings of the firm. Our argument is that the pragmatist mechanisms, through the introduction
of doubt about past practices and decisions, serve to prevent such narrowing of perspective, whether within
the firm or between close collaborators.
their routines. Think of the new institutions as pragmatist: they system-
atically provoke doubt, in the pragmatist sense of an urgent suspicion that
habitual beliefs are poor guides to current problems.39
It is group discussion of problems that renders the resulting flood of
alternatives tractable. Group discussion meets an immediate objection to
problem solving through extensive collaboration rather than hierarchical
decomposition of tasks: the geometric explosion of pairwise contacts that such
collaborations on standard assumptions entail. If A must consult first with B,
then with C, and the latter must then meet by themselves, the sheer number
of consultations is unmanageable unless the group is minuscule. If, however,
the collaborators meet together (a possibility not contemplated in the
standard view of bounded rationality), one meeting substitutes for many,
economizing on participants’ time.
More importantly, group discussion pools the diverse capacities and
experiences of its members to judge the alternatives produced by bench-
marking, simultaneous engineering and problem solving searches. Thus the
new van team convenes specialists in engine and transmission design as well
as in styling, marketing and manufacturing to discuss proposals about the
target market in relation to desired engine performance. Each proposal
illuminates the others and all are seen in light of the diversified knowledge of
the group; both the group and its members are enlightened by the interplay
of diverse disciplines and projects. The upshot is to reveal possibilities that
would remain obscured if those same proposals were scrutinized one by one
or jointly by a lone evaluator.40
3.3 Advantages of Pragmatic Collaborations
We suggested above that the guiding principles associated with the
non-standard approach offered collaborating firms the opportunity both to
advance knowledge and to control opportunism. Here we address the
advantages of pragmatic collaboration explicitly.
Sources of Efficiency and Effectiveness
Despite their manifestly
39For examples of how NUMMI (the GM–Toyota joint venture) handles the tension between standard
operating procedures and continuous improvement, see Adler et al. (1999).
40Placing group discussion at the center of our notion of pragmatic collaboration raises questions about
the nature (and possible limits) of firm capabilities for this activity. We believe that the quality of these
discussions will be affected by the absorptive capacity of each party (Cohen and Levinthal, 1990) and that
repeated interactions may make this capacity partner-specific (Dyer and Singh, 1998). We also
acknowledge that there may be a limit to the number (and nature) of different collaborations that a firm
can successfully manage at one time, i.e. that a firm’s collaborative capacity can be exceeded (Bensaou,
1999; Prahalad, conference discussion).
demanding goals, the pragmatist methods described above can be effective
even if their initial results are modest. Recall that the aim in benchmarking
and simultaneous engineering is simply to reveal sufficiently large differences
between current and potential performance to provoke debate about the
possibilities of improvement and, subsequently, about the improvement of
methods of improvement themselves. Reaching agreement on the initial
characterizations of designs, production methods and error detection systems
should be feasible because these are understood as starting points: provisional
and perfectible, not definitive.
Recall, too, that the ensemble of ‘learning by monitoring’ mechanisms does
not aim to produce an exhaustive, fully replicable characterization of the
products or processes to which they are directed. Benchmarking does not
produce laboratory protocols by which successful experiments can be
reproduced elsewhere. Rather, it reveals feasible goals and indicates the
feasible set of means for obtaining them. Error detection systems (which can
themselves of course be benchmarked) are then used to determine how to
adapt the indicated means to the local setting to achieve the goal. Thus the
early characterization of means and ends provide not just starting points but
also a basis for organizing exchanges of experiences among collaborators. The
exchanges, in turn, result in learning that allows adjustments after the initial
rounds to either the design or production process.
This two-fold information pooling (of plans and perspectives) yields
efficiency gains of a distinctive kind. Where the hierarchical decomposition of
tasks leads, as we saw, to economies of scale, information pooling yields
economies of scope: the greater the variety of projects undertaken, the less
costly it is to undertake yet another variety of those projects.
One source of these gains is suggested directly by the cognitive properties
of the new institutions. Comparisons among unfamiliar (in part) alternatives
(competing designs, various possibilities for realizing these, alternative
explanations of the origins of defects) reduces the likelihood of insular,
self-absorbed decisions while also reducing the risks of discovering much later
the costly shortcomings of particular decisions.
A second source of efficiency gains is the self-reinforcing character of
disciplined information pooling itself. Just as the decomposition of tasks
facilitates further decomposition, so the methods of collaborative inves-
tigation of ambiguity lead, within and among work groups, to increasing
dexterity in the use of those methods. There are corresponding increases in
the scope of alternatives that can be canvassed and the depth to which their
implications can be examined.
The cumulative effect of such efficiency gains allows firms that have
mastered these pragmatist disciplines to overturn the verities of the earlier
mass production system, transforming the traded off desiderata of that world
into mutually reinforcing attributes of the new one. For example, it was a
truism of mass production that exploration of many design alternatives
hindered timely and rigorous pursuit of any one. But the experience of firms
in technologically sophisticated industries with extremely short product life
cycles shows, on the contrary, that pursuit of many alternatives is the best way
of understanding the advantages and disadvantages of each, and so
contributes to selection of the best current possibilities.
Similarly, in mass production a decrease in efficiency was taken to be the
price for an increase in quality. Isolated efforts to increase accuracy, by use, for
example, of inspectors at critical operations, seemed inevitably to interfere
with the automaticity of production, reducing the throughput of the system
per unit time and decreasing productivity. Coordinated efforts to increase
accuracy (except as a by-product of the increasing decomposition of tasks)
seemed unmanageably complex. However, the error detection and correction
methods of ‘learning by monitoring’ reveal defects in the organization of
production that remained hidden under less exigent conditions. Elimination
of these defects affords possibilities for raising overall efficiency, by
minimizing downtime due to repairs, introducing delicate automation
equipment whose operation depends on maintenance of tight tolerances,
reducing the re-work of botched products, that, judging by the relative
performance of competing firms using the old and new methods, were simply
unavailable in environments more tolerant of fault.41
an alternative solution to the problem of opportunism. That problem arises
in mass production, as noted above, as a direct consequence of hierarchical
specialization. Resources specific to one project in such a system have only
scrap value if put to other use and expertise is so fragmented and specialized
that the doings of one actor or group are inscrutable to others. Hence the
temptations of hold-up and deception, to which vertical integration and the
corresponding direction by authority and incentive are a response. The new
institutions, in contrast, so transform the conditions of cooperation that
incitements to trickery can be countered by the very exchanges of information
required for the exploration of ambiguity.42
Pragmatist information pooling also provides
41The same automotive assembly plants rank highest on both productivity and quality according to a
global survey by MacDuffie (1995).
42Repeated dealings have been cited as a sufficient condition for the build-up of reputations for avoiding
opportunistic behavior (Axelrod, 1984). However, in an environment which has been characterized by
opportunism in the past, long trading relationships may give the parties more experience with each other’s
For one thing, the pooling of proposals and perspectives breaks down
distinctions between mutually ignorant specialists, each tempted to exploit
the ignorance of the other. Where hierarchy produces the information
asymmetries of mutual ignorance, learning by monitoring in effect creates an
information symmetricizing machine in which actors must keep one another
abreast of their intentions and capacities. In simultaneous engineering and
error correction by the ‘five whys’, for example, actors must teach each other
important elements of their respective specialties and reveal the logic of their
intentions in order to make themselves comprehensible.
For another thing, the master resource in the new system is the ability to
re-deploy resources fluidly. As noted above, the novel search routines and
problem solving disciplines help develop this flexibility by breaking apart
static procedures. Equally important is the capacity to re-use a high (and
increasing) percentage of the capital equipment committed to one project in
subsequent ones, by reprogramming the computers that guide its operation
or changing one type of tool-bearing module for another. The greater a work
team’s command of the search routines, the problem solving disciplines and
the re-configuring of flexible equipment, the more accomplished it becomes
at the re-deployment of any resource. The effect is that product-specific
resources are ‘de-specified’, coming increasingly to resemble general purpose
assets, and thus no longer the instruments or object of hold-ups.
Interest Alignment in the New Collaborations
look at how ‘learning by monitoring’ and pragmatist mechanisms of infor-
mation pooling help to align interests between collaborators. If the assets
needed for the new forms of collaboration are general purpose and must be
acquired in part by collaborating, then firms should be eager, not hesitant, to
begin working by the novel methods. Their chief reservation would be fear
of engaging an incompetent or unreliable partner; but the information
exchanges intrinsic to ‘learning by monitoring’ would alert them to this
danger before the consequences were ruinous.
For example, the same process that allows firms and their internal or
external suppliers to agree on the definition of a subsystem or its components
allows joint evaluation of target prices, target rates or return for collaborating
partners; even a target rate of productivity improvement to be expressed in
periodic price decreases. Simple sharing rules may result, e.g. the supplier
Next we take a closer
untrustworthiness and make it more difficult to establish trust. For example, in the auto supplier data
cited above, US firms who have more years of selling products to their principal customer have significantly
less trust in that customer (Sako and Helper, 1998). Learning by monitoring offers suppliers a mechanism
to collect information on whether or not they should believe their customer’s claims that they have changed
keeps at least half the gains from innovations leading to productivity increases
in excess of the target rate, with the share falling at a conventional rate as the
innovation matures. Persistent performance failures, such as a supplier’s
inability to meet price reductions or improvements in product performance in
the medium term, can be penalized by a reduction in the supplier’s share of
the customer’s total purchases of the affected product.
At any moment, therefore, collaborators can compare actual performance
with expected performance and analyze any divergences against the backdrop
of the extensive common knowledge of possibilities that allowed agreement
on the goals in the first place. Suppliers that do exceptionally well in one or
more rounds of the product cycle can then be delegated more extensive
responsibility in the co-design of subsequent models; those that do excep-
tionally poorly will eventually be dropped from the pool of collaborators.
A complementary mechanism also works to align interests in this setting.
Consider the possibility that pragmatic, problem solving deliberation loosens
the hold of interest by fitfully darting, as it were, beyond its reach. Solutions
are uncovered bit by bit as the inadequacies of customary answers are traced
back towards their source and unfamiliar territory is steadily charted. But
self-interest depends as much in its calculations of advantage on settled
expectations as bounded rationality depends on routines in its searches. It
can no more evaluate the surprising outcomes of pragmatic searches than
bounded rationality can anticipate them.
Such uncertainty, even if temporary, about the potential advantages and
disadvantages of one collaborative advance can favor pursuit of the next. The
value to all parties of the current, partial innovation (measured as improve-
ments in the performance of current problem solving mechanisms) will likely
be increased substantially by the next partial step forward. In addition, the
continuous exchange of operating information among the collaborators
reduces the risk than any party can use the novel arrangements for self-
dealing. The de-specification of assets, continuous monitorability and simple
gainsharing rules all offer assurances against the fear of hold-up. Thus once
provisionally workable solutions to common problems are at hand, there are
incentives to pursue them to increase the mutual advantage, rather than to
halt the search to re-negotiate the division of gains from the initial advance.
Elaboration of each solution creates in turn new surprises and renews the
pursuit of further improvements. In time, as emerging solutions change what
the actors do and how they rely on one another, their very ideas of what is
possible will come to reflect these entanglements. ‘Self’-interest will come to
take as the starting point for subsequent calculations the collaborative
surprises of practical deliberation that it was formerly assumed to undermine.
The upshot of all these mechanisms acting together is that the construction
of Japanese-type production systems does not presuppose the existence of
long-term relations, because the system in the course of its operation produces
We can see some of these mechanisms at work by looking at the group of
prize winning ‘super suppliers’ defined above. Firms in this group comprise
about 10% of the Helper sample. They are clearly engaged in discussion with
their customers, with whom they speak by phone on average daily, three times
as often as the other suppliers in the sample. More than half of this time is
spent providing technical assistance to the customer (again a significant
difference between the super suppliers and the sample as a whole). However,
the distribution of benefits from this exchange is not one-sided; these firms
are also more likely to report that in dealings with their main customer they
learned much that will help in dealings with the others they supply.
This activity is not supported by clearly drawn property rights. It is usually
impossible to say who owns partly finished designs or the rights to determine
use of a resident engineer’s time. There is no evidence that these dealings are
governed by elaborate contracts, often considered a substitute for vertical
integration as a governance mechanism, that distribute the burdens between
the parties in case of a long list of contingencies and provide for arbitration
in the case of disagreements. Rather, the response to problems is, as the
foregoing suggests, trying to solve them together. The purpose of such formal
agreements as there are between the parties is to provide a broad framework
of rights and responsibilities, rather than a detailed attempt to foresee all
contingencies.43We will see below that in an important sense these agree-
ments should not be considered contracts at all.
Only if the problem solving mechanisms themselves break down does the
relation fail. Fewer than 9% of the super suppliers expect a customer to
abandon them if a competitor offers to provide the customer with a product
identical to theirs at a lower price. Sixty-two percent of the super suppliers
expect, instead, that the customer would help them match the rival’s efforts.
But such help is not viewed as unconditional: Twenty percent of the super
suppliers expect the customer to gradually reduce the supplier’s market share
if there is no improvement. The message is clear: failure in the short term will
be tolerated if it becomes an occasion for learning that eventually results in
Thus the reduction in asset specificity and the improvement in information
43An analogy to individual behavior may be useful. ‘Some people acquire resources by theft, but mainly
they buy them. People do take precautions against being robbed, but it would be a mistake to analyze the
system by focusing exclusively on the precautions’ (Ronald Coase, personal communication, July 1998).
flow do not mean that adoption of learning by monitoring leads to a
replacement of hierarchy by more market-like forms. In the model of perfect
competition, actors minimize the prices they pay for commodities with given
attributes by switching frequently among many potential sellers. In learning
by monitoring, in contrast, agents seek to continuously improve products and
processes by choosing partners who are good at learning and by staying with
them even if their prices are temporarily higher.
4. Dynamic Considerations
Above we argued against the claim that the non-standard firm can thrive only
under certain institutional conditions (e.g. manufacturing in Japan), saying
that it has been undermined by the successful diffusion of the non-standard
approach to a wide variety of contexts. In this section we look a little more
closely at these dynamic issues. What causes learning by monitoring to arise
and, at times, to fade away? How does the use of pragmatic mechanisms
change parties’ attributes?
4.1 The Rise and Decline of Learning by Monitoring
Even in a short-term oriented, individual interest-maximizing society like the
USA, learning by monitoring has been spreading, despite the absence of
Japanese-style collective institutions. We believe that the process of
disciplined joint inquiry can actually generate the conditions necessary to
maintain and nourish collaboration, even where the institutional environment
appears highly prone to concerns about ‘hold-up.’ Firms can develop a high
trust equilibrium over time (Aoki, 1984), by starting with projects that have
a high pay-off and only a small vulnerability. If all goes well, the parties can
try larger projects.
For example, Donnelly Corporation, an American manufacturer of mirrors,
was an early supplier to Honda’s US manufacturing operations. It first started
off supplying a relatively standard rear view mirror, using excess capacity in
an existing plant. The pragmatic mechanisms gave both sides a chance to
observe each other’s trustworthiness. Donnelly saw that the large amounts of
technical assistance provided for free by Honda helped lower its costs, and
Honda observed Donnelly’s willingness to make changes and steady
improvements in quality. Gradually they moved on to bigger projects: first a
dedicated plant for Honda, then an entirely new product (side mirrors) and,
finally, a new, complicated process (an automated paint line) and joint design
with Honda engineers in Japan. Honda did not make a specific guarantee to
Donnelly, but did say that if the supplier performed well at these steadily
more difficult tasks, Honda would see to it that Donnelly’s business grew
(MacDuffie and Helper, 1999).
At the beginning of this process, opportunism may be almost as high as in
a ‘standard’ firm, but is better controlled through the pragmatic mechanisms.
Eventually, managers realize that these mechanisms are sufficiently effective
as means of monitoring that if they cheat, they will get caught, so they desist
in attempting to be opportunistic. Because these mechanisms advance
knowledge, they increase the pay-off of cooperation, as well as reducing that
of opportunism, so people get used to cooperating. This strengthens their rule
of thumb that cooperation is good. Again they are less drawn to opportunistic
strategies. Ultimately, learning by monitoring may create actors who value
cooperation as an end in itself. These actors would be betraying their identity
if they cheated, regardless of the prospect of detection. Were this to occur,
learning by monitoring would have produced a culture of trust that would
greatly resemble the one once assumed to be its precondition.
Similarly, the degree of asset specificity may evolve over time. Initially one
firm’s skills at discussion may be specific to one partner, but may become
general over time. For example, Donnelly was so intensively involved in
learning the ‘Honda Way’ that its other customers suffered in the first years.
Eventually, however, Donnelly found that the techniques were useful for other
customers and cost and defect levels fell sharply. Furthermore, Donnelly now
has many other customers for the painted side mirrors that they had never
made before getting the Honda contract.
As Donnelly has extended its side mirror business to other customers, there
is some risk to Honda of spillover of the value created in the collaboration.
However, the people at Honda most passionately involved with this
collaboration seem convinced that, on balance, its benefits outweigh its costs.
The most frequent view we heard was, ‘We all get better this way.’ The belief
that the entire industry benefits from this kind of collaboration can be a
powerful motivator for further innovation. This vision of ‘enlarging the pie’
can help to sustain collaborations even in the face of pervasive worries about
opportunism or spillover.
The history of the US auto industry, however, reminds us that if conditions
change dramatically, firms may decide it is in their interest to defect from
collaborations. As Ford and GM grew, they increasingly suffered from
spillovers to other firms, without any corresponding benefit. As they became
more powerful in consumer markets, the alternative of pitting suppliers
against each other became more attractive, since this strategy protected the
automobile maker’s rents from being shared with suppliers and consumers did
not have many alternatives if quality was low. So the automobile makers
abandoned the pragmatic mechanisms for competitive markets in the 1950s,
1960s and 1970s. When the Japanese entered the US market, conditions
changed again. US automobile makers no longer had rents to protect from
suppliers and found that higher quality was a way to attract consumers.
Under these conditions, they rediscovered (primarily through careful study of
Japanese automobile makers) the pragmatic mechanisms (Helper and Levine,
4.2 Strategic Use of Pragmatic Mechanisms
A key concern of the literatures on both technical change and managerial
strategy is whether being good at one kind of innovation makes a firm bad at
other kinds. In particular, it might be that learning by monitoring is a type of
incremental change that hinders radical innovation or prevents dramatic
changes in strategy (Abernathy, 1978; Porter, 1980).
Search routines that break through the constraints of habitual responses to
design and operational problems may not be well suited to higher level
monitoring. What will happen when these pragmatist mechanisms confront
decisions in the province of corporate governance? For example, evaluating
the viability of whole lines of business, choosing among very different but
plausible long-term development goals or responding to threats or oppor-
tunities facing the corporation as a whole? There are both empirical and
theoretical reasons to think that traditional mechanisms of corporate
governance (banks with controlling stakes or dispersed shareholders) will not
shine at this more encompassing task.
The leading example of the inadequacy of standard governance forms to
this task is surely the dismal performance of the large Japanese banks in the
last decade (Aoki and Patrick, 1994; Aoki and Saxonhouse, 2000). As
contingent corporate monitors, they are supposed to take control of corpora-
tions in those contingencies when sitting managers demonstrate incapacity.
In practice, during the recessions of the 1990s these banks have demonstrated
scant capacity to act. Firms under their supervision have wasted free cash flow
or made major strategic miscalculations about capacity, expansion of product
lines and distribution channels.
The general problem is contained in the following paradox. The infor-
mation generated by the ‘learning by monitoring’ system for its day-to-day
and medium-term needs is also necessary to understand how to correct large
errors when they occur. Yet this information is not readily available to the
‘contingent’ monitors (main banks or shareholders) who intervene in such
emergency situations. Only agents monitoring the (non-standard) corpora-
tion day-to-day, i.e. participating in its routine project selection and evalu-
ation procedures, could know enough of its highly decentralized operations to
correct large errors in an effective way. But just such agents are discredited
when the errors come to light and outsiders, whatever bundle of interests they
are trying to maximize, simply cannot learn fast enough to be useful. From
this perspective, therefore, the differences between ‘patient capital’ banks
(with their ‘voice’ view of the corporation as a community) and impatient
shareholders (ever ready for ‘exit’) are less important than the similarities in
In principal, at least, the necessary link between daily operations and
project selection could be established by a straightforward extension of
‘learning by monitoring’. Recall that simultaneous engineering, benchmark-
ing and error detection all depend on drawing the participants beyond the
circle of familiar habits and routines by exposing them to unfamiliar projects
and prospects. The diversity of their viewpoints allows evaluation of the
novelty, which in turn allows re-examination of their differences. The
immediate products of this deliberation are choices among competing
solutions to problems and organizational reforms that could prevent a similar
Why assume that reflection on possibilities stops here? As competing
patterns of problem solving and solutions to problems emerge they can be
formulated as coherent, if previously unimagined, alternatives. In choosing to
favor one pattern of problem solving or a particular solution over another,
moreover, a firm is simultaneously making a choice and refining or
re-interpreting the criteria by which the choice is made. This kind of choosing
and choosing how to choose is familiar from law and other interpretive
disciplines, where the decision to apply a precedent changes the subsequent
understanding of how the precedent is to be applied. Equivalent choices by a
firm define its strategy.
Thus conceived, a firm’s strategy would be a joint product of its cumulative
problem solving. Determining the nature or identity of the firm is just the
highest level of choice among competing alternatives that is the structuring
principle of the new firm. Fundamentally, the firm, like the products it makes
and the way it makes them, is a design problem.
But is a solution to the governance problem on these lines feasible? Recall
that the discussion of ‘learning by monitoring’ proceeded from institutional
mechanisms (benchmarking, error detection, just-in-time) to general
principles of pragmatist information pooling and self-scrutiny. The principles
are credible because they can be embodied in mechanisms that perform what
might (without such demonstrations) be thought impossible tasks, for
instance probing the value of routine assumptions without destroying all
routines. Unless we know, for instance, that benchmarking new products is
manageable, we might suspect that the exploration of little known
alternatives paralyzes rather than accelerates decision making. In reaching to
the strategic, we raise new questions of feasibility. What are the mechanisms
carrying the pragmatist logic to the strategic level? Are there institutional
analogues to benchmarking at the level of strategic discussion? If there are,
how are they connected to the lower level mechanisms of institutional self-
There do indeed seem to be such mechanisms, albeit experimental ones.
They are much less well characterized than the disciplines discussed earlier
and there is less reason to be confident of their ultimate effectiveness.
Consider the following three examples of strategic benchmarking. Together
they suggest that firms are intent on addressing the problem of extending
learning-by-monitoring institutions to the strategic level, despite being
uncertain of how (or even whether) it can be done.
The first is Capstone, a ‘corporate incubator’, created by the chairman of
Ford, Alex Trotman, in 1996 to foster Ford’s future leaders by intensifying
their focus on strategic possibilities, especially those outside the automobile
industry (General Electric has a similar program and Toyota has just instituted
one). In the Capstone program four teams of about six managers spend
roughly half a year studying critical questions of corporate strategy,
proceeding through a series of benchmarking evaluations of alternatives and
concluding with specific recommendations for change. The first such
recommendation (at least as reported publicly) was a pilot program, modeled
on developments in mass retailing, to consolidate Ford and Lincoln Mercury
dealerships into new entities to be jointly owned by Ford and local dealers.
Generalizing the Capstone model, executives at Ford and elsewhere would
compete for promotion by proposing (potentially) competing corporate
strategies after extensive benchmarking of like choices in their own and other
industries, bringing strategic assumptions to the surface while keeping
continuing operations close to hand.
A second mechanism, the proliferation of performance metrics, focuses
internally rather than externally. The central aim of these metrics is to find
measures that reveal the performance of subunits in relation to the perform-
ance of the corporate whole. The utility of individual metrics is, however,
limited. If performance on a single metric told most of the story of all the
subunits’ contribution to overall performance, the center could just set good
performance on that metric as the corporate goal. In that case there would be
little need for substantial decentralization of authority. But if decentralization
is warranted (and the assumption in non-standard firms is that it is, because
subunit goals diverge and change), then the likelihood of finding truly
comprehensive and common indicators of success is limited (Meyer and
What individual measures cannot do, baskets of measures can (Kaplan and
Norton, 1992). Discussion of which measures to add and which to remove
from the basket brings to light differences in strategic orientation among the
operating units of the decentralized non-standard firm that are then accessible
to debate and choice. Thus in a brochure marketing a complex proprietary
performance metric called ‘total shareholder return (TSR),’ the Boston
Consulting Group stresses that:
The process of mapping the strategy and resulting value drivers often has
as much benefit as the quantitative analyses of alternative actions or
strategies. It serves as a catalyst for surfacing opinions or assumptions
and provides a forum among the operating management team for
gaining consensus on action. . . . [Indeed], a properly designed and
implemented value management program . . . creates a common
language between line and staff, and between corporate and business
units. It provides a clear link between strategy and TSR performance.
(Boston Consulting Group Brochure, pp. 22 and 24–25)
A third example, combining features of the first two, are US venture
capitalists. They operate not only in high-tech industries but, increasingly, in
the re-structuring of mature sectors through ‘leveraged build ups’ and other
novel devices that allow for profound restructuring of the firm’s relation to its
customers, use of information and organization of production. In choosing
portfolio firms, and from time to time dropping or adding portfolios, venture
capitalists engage extensively in the kind of strategic benchmarking
associated with Capstone. In deciding which portfolio firms to abandon and
which to continue supporting, they make strategic use of performance
metrics. More generally, venture capitalists combine managerial advice and
guidance, rooted in something close to day-to-day knowledge of the
decentralized, pragmatist firm, with expertise in finance. They are, moreover,
too much like managers (or employees or suppliers) to be contingent
monitors, whether shareholders or bank monitors.
These examples do not prove that it is institutionally feasible to extend
pragmatist principles to questions of strategy and corporate governance. But
clearly there do exist actors in the corporate realm who recognize problems in
the current connection between the operational and the strategic, i.e.
limitations in the standard view of the firm, and are consequently motivated
to develop more effective, deliberative alternatives. To the extent that they
do, they become an important force behind the diffusion of the non-standard
5. Conclusion: Some Implications for Economic and Organizatinal
Theories of the Firm
The structure of the firm in economic and organizational theory is typically
seen as the result of an efficiency calculus. Economic goals and environments
are taken as given and rational agents choose the organizational form that
minimizes the costs of attaining the result in the environment. Thus, for
Coase, firms choose between markets and organizations, preferring the former
until the cost of a marginal transaction there exceeds the cost of the same
transaction in the latter (Coase, 1988). More recent, but closely related,
theories portray maximizing agents constructing firms of different configura-
tions from basic building blocks such as contract and ownership (Hart, 1995;
Baker et al., 1998). Classic theories of organization, especially those focused
on the firm, premise similar choices, though often with a richer menu of
possible institutional solutions to the problems posed by particular environ-
ments (Selznick, 1949; Gouldner, 1954). In different ways, all of these
theories recognize the existence of a turbulent realm beyond the reach of
organization. This is the realm of the ‘informal’ [see Barnard, for whom
management of the ‘informal’ to achieve the tasks of the formal organization
was in many ways the pre-eminent function of the executive (Barnard, 1947)
or the ‘adhocracy’ (Mintzberg, 1993) presumed to be of secondary importance
to the world of the dominant organizational forms].
Our findings dissolve the distinction between the formal and the informal
(or the organized and the unorganizable). We have seen that learning-by-
monitoring firms can establish elaborate and robust organizations in
environments too volatile for standard forms, but on condition that they
make it routine to question their own routines. These firms are far more
structured than informal groups, but their structures are by design much
more fluid than those of traditional formal organizations.
At a minimum the existence and rapid diffusion of these firms cast doubt
on the sufficiency of the menus of organizational forms and building blocks
available in current academic theories. Thus the new forms of collaboration
are neither markets nor hierarchies (Powell, 1990). In contrast to a hierarchy,
there is no principal among the collaborators who can definitively partition
tasks for the others. Moreover, the collaborators’ positions within the new
arrangements are contestable in way that the places of hierarchical
subordinates would not be. In contrast to a market, the collaborators do not
merely signal each other through prices. They jointly explore what they want
to do even as they are doing it.
This joint exploration, and its implications for subsequent projects, marks
this new collaboration off from familiar contractual and ownership relations.
In normal contractual relations parties are presumed to undertake distinct
tasks and to be sufficiently knowledgeable about them to deliver certain
goods or services under agreed conditions without continuing consultation
from the others. In co-development arrangements of the kind discussed
above, however, these conditions are not met. On the contrary, what each
party wants depends in part on what the others do. Think of an engineer,
co-located from a supplier to a customer, who suggests how the latter can
re-configure a production line to make better use of a part and suggests to
the former how the part might be altered to make optimal use of the new
line. We are a long way from arms length contracting, indeed contracting of
any standard kind.
We are far from familiar forms of ownership as well. Ownership, recall, is
residual control: the right to dispose of an asset insofar as its disposition is not
subject to contract. Put another way, residual control is the owner’s right to
fill gaps in an existing contract to his/her liking or to determine how to
re-deploy an asset once contracts controlling its current use have run their
course. Now, however, extending the example just given in the direction of
the discussion of project selection and governance above, suppose that, thanks
to the engineer’s suggestions, the customer and supplier agree to develop
their next generation of products on the basis of the re-configured line and
the re-designed part. Joint control of the assets in the new collaboration
shades into joint residual control, and thus a novel form of ownership.
Read this in light of our partial reconstruction of the history of the US
automobile industry. These novel forms and instruments of organization
suggest a broader corrective to the ecological views of organizational choice.
Even though economic agents are aware of organizational models (albeit less
fully than the theories noted above suggest), they are often much less sure
about the organizational choices they face than the theories make them out
to be. The reason is that the environments they face are more complex and
ambiguous than the theories suppose, and more subject to non-marginal
change by agents’ actions. Because of this deep uncertainty, actors must do as
much strategizing as maximizing (Sabel 1996; Sabel and Zeitlin, 1997;
Zeitlin, 2000). Under some broad range of conditions (current ones included),
they must be at least as much concerned with shaping the context they are in
as with pursuing what seems to be their advantage within any context.
Self-interested adjustment to conditions taken can thus proceed together
with efforts to find or create a more advantageous set of constraints, in
disregard of Schumpeter’s classic distinction between adaptive reactions
(which accept given constraints as binding) and creative ones (which do not)
[see Schumpeter and the discussion of Schumpeter’s ideas in Lazonick
(Schumpeter, 1961, especially p. 60; Lazonick, 1983)]. If our re-interpretation
of the Fisher Body story is correct, GM’s purchase of the second chunk of
Fisher stock was strategizing in this sense, as have been the surprisingly
frequent shifts in supplier strategy in the automobile industry generally. Just
as actors will sometimes create new organizational forms when available ones
are unsuited to their contexts, so they will sometimes strategize by combining
existing forms in novel hybrids.
A close look at current customer–supplier relations in the US automobile
industry reveals many such accommodations. Ford, for example, adopts some
but not all of the collaborative practices described above. It enters long-term
contracts and provides modest support for supplier development. In return it
demands, and gets, information on their production processes, but not on
their cost structure. Thus the contrast between standard and non-standard
forms will be less stark in reality than in theory, but for reasons that the
non-standard theory can digest more easily than the standard one.
To summarize, we have put forth a theory of the firm which differs from
the standard theory in many ways, from underlying assumptions about
human cognitive processes and the nature of the environment to a
characterization of the organizational processes necessary for pragmatic
collaborations and, finally, to conclusions about managerial actions, the
significance of property rights and the boundaries of the firm.
We begin with the assumption that firms (and the individuals within them)
value collaborative learning as a valuable way to circumvent the constraints of
bounded rationality, particularly when faced with high uncertainty and
volatility, more than they fear the risks of opportunism associated with such
collaboration. Given that the information relevant to solving problems often
resides in a wide variety of people, located both within the firm and in the
disciplined joint inquiry; the ongoing search for insights and new perspectives
throughout the network of firms engaged in collaboration. The central goal of
as guides to action; they are called ‘pragmatic’ because they systematically
provoke doubt. Next, we identify the features of these pragmatist mechanisms
that allow them both to advance knowledge effectively while also providing
safeguards against opportunism that help keep the collaboration going, a
phenomenon we call ‘learning by monitoring.’
Dialogue within and across groups of individuals with proximate
knowledge of a particular problem is, we argue, crucial to achieving the
benefits of pragmatic collaborations and therefore requires that managers act
more as discussion leaders than contract administrators. Under these
conditions, the joint ownership of assets can lead to improved performance,
counter to the predictions of the standard theory. In addition, we claim that
firms faced with a large number of ongoing and potential pragmatic
collaborations make decisions, under the guise of selecting which projects and
which collaborators, that are really about the boundaries of the firm and
reflect the firm’s strategy (albeit often an emergent strategy). Boundaries are
often placed (or repositioned) so that those for whom intense, frequent
discussion is most productive are on the same project, which may or may not
mean within the same firm.
As mentioned above, this paper primarily contrasts the standard and
non-standard views of the firm in order to address shortcomings of economic
theory as applied to collaborative activity between firms. We anticipate that
future papers will address networks of firms, drawing on perspectives from
economic sociology and organizational theory. This future work will argue
that network theories, while strong at characterizing the characteristics of a
network at one point in time, have been less successful at explicating the
dynamics of networks. Our view is that the careful study of ‘learning-by-
monitoring’ through pragmatic collaboration can demonstrate how a network
of firms develops and applies mutually beneficial knowledge while also being
able to resolve how the gains accruing to that knowledge are distributed.
From this perspective, the results of each collaborative project (i.e. an iteration
in the joint inquiry that binds firms in the network together) allow predictions
as to the evolution of network structure and capabilities.
We thank George Baker, Ronald Coase, Bob Gibbons, Ha Hoang, Oliver
Hart, David Levine, Gerry McDermott, Marshall Meyer, Jim Rebitzer, Mark
Roe, Sidney Winter and Jonathan Zeitlin for their extremely useful
comments. Thanks are also due to participants in seminars at: MIT (Econ-
omics of Organizations and Operations Management Seminars); Stanford
Business School; Harvard Business School; UCLA (History and Industrial
Organization Workshop); Wharton (Jones Center Workshop and ‘New Forms
of Organization’ Conference); University of Wisconsin (Sociology and
Industrial Relations Workshop). Earlier versions of this paper were titled ‘The
Boundary of the Firm as a Design Problem.’
Abernathy, W . J. (1978), The Productivity Dilemma: A Roadblock to Innovation in the Automobile Industry. Johns
Hopkins University Press: Baltimore.
Adler, P . S., B. Goldoftas and D. I. Levine (1999), ‘Flexibility vs. Efficiency? A Case Study of Model
Changeovers in the Toyota Production System,’ Organization Science, 10, 43–68.
Aoki, M. (1984), The Cooperative Game Theory of the Firm. Oxford University Press: New York.
Aoki, M. and H. Patrick (eds) (1994), The Japanese Main Bank System: Its Relevance for Developing and
Transforming Economies. Oxford University Press: New York.
Aoki, M. and G. R. Saxonhouse (eds) (2000), Finance, Governance, and Competitiveness in Japan. Oxford
University Press: New York.
Axelrod, R. (1984), The Evolution of Cooperation. Basic Books: New York.
Babson, S. (ed.) (1995), Lean Work: Employment and Exploitation in the Global Auto Industry. Wayne State
University Press: Detroit, MI.
Baker, G., R. Gibbons and K. Murphy (1998), ‘Implicit Contracts and the Theory of the Firm,’ Harvard
Business School Working Paper.
Barnard, C. I. (1947), The Functions of the Executive. Harvard University Press: Cambridge, MA.
Beasley, N. (1947), Knudsen: A Biography. Whittlesey House: New York.
Bensaou, B. M. (1999), ‘Portfolios of Buyer–Supplier Relationships,’ Sloan Management Review, 40(4), 35–44.
Brooks, R. and S. Helper (2000), ‘Comment on Fisher Body,’ Journal of Economic Perspectives, 14, 122–123.
Bourdieu, P . (1977), Outline of a Theory of Practice (trans. R. Nice). Cambridge University Press: Cambridge.
Chandler, A. D. (1962), Strategy and Structure. MIT Press: Cambridge, MA.
Chandler, A. D. (1977), The Visible Hand. Harvard University Press: Cambridge, MA.
Chandler, A.D. andS. Salisbury(1971), PierreS.DuPontandtheMakingoftheModernCorporation. Harper and
Row: New York.
Clark, K. (1989), ‘Project Scope and Project Performance: The Effect of Parts Strategy and Supplier
Involvement on Project Development,’ Management Science, 35, 1247–1262.
Coase, R. H. (1937), ‘The Nature of the Firm,’ Economica, 4, 397–405.
Coase, R. H. (1988), ‘The Nature of the Firm: Origin,’ Journal of Law, Economics, and Organization, 4, 3–17.
Coase, R. H. (2000), ‘The Acquisition of Fisher Body by General Motors,’ Journal of Law and Economics,
Cohen, W . and D. Levinthal (1990), ‘Absorptive Capacity: A New Perspective on Learning and Innovation,’
Administrative Sciences Quarterly, 35, 128–152.
Cusumano, M. and A. Takeishi (1991), ‘Supplier Relations and Supplier Management: A Survey of Japanese,
Japanese-Transplant, and U.S. Auto Plants,’ Strategic Management Journal, 12, 563–588.
Dore, R. (1983), ‘Goodwill and the Spirit of Market Capitalism,’ British Journal of Sociology, 34, 459–482.
Dore, R. (1987), Taking Japan Seriously: A Confucian Perspective on Leading Economic Issues. Stanford University
Press: Stanford, CA.
Dyer, J. and H. Singh (1998), ‘The Relational View: Cooperative Strategy and Sources of Interorganizational
Competitive Advantage,’ Academy of Management Review, 23, 660–679.
Freeland, R. (2000), ‘Creating Holdup Through Vertical Integration: Fisher Body Revisited,’ Journal of Law
and Economics, XLIII, pp. 33–66.
General Motors (1924), Fisher Body: Its Contribution to the Automotive Industry, p. 8. General Motors: Detroit.
Gouldner, A. (1954), Patterns of Industrial Bureaucracy. Free Press: Glencoe, IL.
Gulati, R. (1998), ‘Alliances and Networks,’ Strategic Management Journal, 19, 293–318.
Hart, O. (1995), Firms, Contracts, and Financial Structure. Clarendon Press: Oxford.
Helper, S. (1991), ‘Strategy and Irreversibility in Supplier Relations: The Case of the US Automobile
Industry,’ Business History Review, 65, 781–824.
Helper, S. and D. Levine (1992), ‘Long-term Supplier Relations and Product Market Structure,’ Journal of
Law, Economics, and Organization, 8, 561–581.
Hochfelder, D. and S. Helper (1996), ‘Joint Product Development in the Early American Auto Industry,’
Business and Economic History.
Holmstrom, B. (1999), ‘The Firm as Subeconomy,’ Paper Presented at the Make vs. Buy Conference, Columbia
Law School, February. Journal of Law, Economics and Organization, 15, 74–102.
Holmstrom, B. and J. Roberts (1998), ‘The Boundaries of the Firm Revisited,’ Journal of Economic Perspectives,
Hounshell, D. (1984), From the American System to Mass Production, 1800–1932: The Development of
Manufacturing Technology in the United States. Johns Hopkins University Press: Baltimore.
Kaplan, R. S. and D. P . Norton (1992), ‘TheBalanced Scorecard: Measuresthat Drive Performance,’ Harvard
Business Review, January/February, 71–79.
Klein, B., R. A. Crawford and A. Alchian (1978), ‘Vertical Integration, Appropriable Rents, and the
Competitive Contracting Process,’ Journal of Law and Economics, 21, 403.
Kogut, B. (1998), The Network as Knowledge, Wharton School Working Paper.
Kogut, B. and U. Zander (1996), ‘What Firms Do: Coordination, Identity, and Learning,’ Organization
Science, 7, 502–518.
Langlois, R. N. and P . L. Robertson (1989), ‘Explaining Vertical Integration: Lessons from the American
Automobile Industry,’ Journal of Economic History, 49, 361–375.
Lazonick, W . (1983), ‘Industrial Organization and Technological Change: The Decline of the British Cotton
Industry,’ Business History Review, 57, 491–516.
Levin, D. P . (1989), Irreconcilable Differences: Ross Perot vs. General Motors. Little, Brown: Boston, MA.
MacDuffie, J. P . (1995), ‘Human Resource Bundles and Manufacturing Performance: Organizational Logic
and Flexible Production Systems in the World Auto Industry,’ Industrial and Labor Relations Review, 48,
MacDuffie, J. P . (1997), ‘The Road to Root Cause: Shop-Floor Problem-Solving at Three Auto Assembly
Plants,’ Management Science, 43, 479–502.
MacDuffie, J. P . and S. Helper (1999), ‘Creating Lean Suppliers: Diffusing Lean Production Through the
Supply Chain,’ in J. Liker, P . Adler and M. Fruin (eds), Remade in America: Transplanting and Transforming
Japanese Production Systems. Oxford University Press: New York.
Marglin, S. A. (1974), ‘What Do Bosses Do?’, Review of Radical Political Economics, 6, 33–55.
Meyer, M. and L. G. Zucker (1989), Permanently Failing Organizations. Sage Publications: Newbury Park.
Mintzberg,H. (1993), Structure in Fives:DesigningEffectiveOrganizations. Prentice-Hall: Englewood Cliffs, NJ.
Monteverde, K. and D. J. Teece (1982), ‘Supplier SwitchingCostsandVertical Integration in theAutomobile
Industry,’ Bell Journal of Economics, 13, 206–213.
Nishiguchi, T. (1994), Strategic Industrial Sourcing: The Japanese Advantage. Oxford University Press: New
Nobeoka, K. (1995), The Influence of Customer Scope on Supplier’s Performance in the Japanese
Automobile Industry, Working Paper, Research Institute for Economics and Business Administration, Kobe
University, Japan, December.
Piore, M. and C. Sabel (1984), The Second Industrial Divide. Basic Books: New York.
Porter, M. (1980), Competitive Strategy. Free Press: New York.
Pound, A. (1934), The Turning Wheel: The Story of General Motors Through Twenty-five Years 1908–1933.
Doubleday: Garden City, NY .
Powell, W . (1990), ‘Neither Market Nor Hierarchy: Network Forms of Organizations,’ in L. L. Cummings
and B. M. Staw (eds), Research in Organizational Behavior, Vol. 12, pp. 295–336. JAI Press: Greenwich, CT.
Powell, W ., K. Koput and L. Smith-Doerr (1996), ‘Interorganizational Collaboration and the Locus of
Innovation: Networks of Learning in Biotechnology,’ Administrative Sciences Quarterly, 41, 116–145.
Sabel, C. F. (1995), ‘Learningby Monitoring: The Institutions of EconomicDevelopment,’ in N. Smelser and
R. Swedberg (eds), Handbook of Economic Sociology, pp. 137–65. Princeton University Press: Princeton, NJ.
Sabel, C. F. (1996), ‘Intelligible Differences: on Deliberate Strategy and the Exploration of Possibility in
Economic Life.’ Rivista Italiana degli Economisti, Anno 1, 55–80.
Sabel, C. F. and J. Zeitlin (1997), ‘Stories, Strategies, Structures: Rethinking Historical Alternatives to Mass
Production,’ in C. F. Sabel and J. Zeitlin (eds), Worlds of Possibility: Flexibility and Mass Production in Western
Industrialization, pp. 1–46. Cambridge University Press: Cambridge.
Sako, M. (1992), Prices, Quality, and Trust: Inter-Firm Relations in Britain and Japan. Cambridge University
Sako, M. and S. Helper (1998), ‘Determinants of Trust in Supplier Relations: Evidence from the Automotive
Industry in Japan and the United States,’ Journal of Economic Behavior and Organization, 34, 387–417.
Schon, D. A. (1973), Beyond the Stable State. Norton: New York.
Schumpeter, J. (1961), The Theory of Economic Development. Oxford University Press: Oxford.
Schwartz, M. (2000), ‘Markets, Networks, and the Rise of Chrysler in Old Detroit, 1920–1940,’ Enterprise
and Society, 1, 63–99.
Schwartz, M. (forthcoming), The Rise and Fall of Detroit. Sage: New York.
Seltzer, A. (1928), A Financial History of the American Automobile Industry. Houghton Mifflin: Boston, MA.
Selznick, P . (1949), TVA and the Grass Roots: A Study in the Sociology of Formal Organization. University of
California Press: Berkeley, CA.
Smitka, M. (1991), Competitive Ties: Subcontracting in the Japanese Automotive Industry. Columbia University
Press: New York.
Weick, K. E. (1995), Sensemaking in Organizations. Sage: Thousand Oaks.
Whinston, M. (1997), On the Transaction-Cost Determinants of Vertical Integration, Working Paper,
White, R. B. (1991), ‘Body by Fisher: The Closed Car Revolution,’ Automobile Quarterly, 29(4), 46–63.
Williamson, O. (1985), The Economic Institutions of Capitalism. Free Press: New York.
Zeitlin, J. (2000), ‘Americanizing British Engineering? Strategic Debate, Selective Adaptation, and Hybrid
Innovation in thePost-War Reconstruction, 1945–60,’ in J. Zeitlin and G. Herrigel (eds), Americanizationand
its Limits: Reworking US Technology and Management in Post-War Europe and Japan, pp. 123–152. Oxford
University Press: Oxford.
Appendix: Data from Helper Survey of Suppliers Download full-text
The data come from a survey of automotive suppliers Helper conducted in
spring and summer 1993. The survey was sent to the divisional director of
sales and marketing at automotive suppliers in the USA and Canada. The
focus of this survey was information about relationships with customers and
product characteristics. Because many companies supply their customers with
several different types of products and their relationships with their customers
differ by product, respondents were asked to answer the questionnaire for
their most important customer regarding one product which was typical of
their company’s output and with which they were familiar.
The sample chosen for the North American questionnaire included every
automotive supplier and automaker component division named in the Elm
Guide to Automotive Sourcing (available from Elm Inc., East Lansing, MI).
This guide lists the major first tier suppliers (both domestic and foreign
owned) to manufacturers of cars and light trucks in the USA and Canada.
Each respondent who had not yet responded to the survey received three
mailings over the course of 2½ months.
The responses were far above the norm for business surveys. The response
rate was 55% for the sales manager survey, after taking into account those
firms which were unreachable (the surveys sent to them were returned
undelivered) and those which were not eligible to answer the survey (they
were not first tier automotive suppliers or they specialized in supplying for
heavy trucks and buses).
Firms answering the survey are quite representative of the population in
terms of size of firm and location, as compared with data from the Elm Guide
and from County Business Patterns for SIC 3714 (automotive parts) and 3496
(automotive stampings). The respondents were experienced and know-
sledgeable; they averaged more than 18 years in the automobile industry and
more than 11 years with their company.