sábado, 3 de março de 2012

Texto 21 - The economics of organization - the transiction cost approach portugues.docx

Your use of the JsToR archive indicates your acceptance of JsToR's Terms and conditions of use, available at http://www.jstor.org/page/info/about/policies/terms.jsp . JsToR's Terms and conditions of use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JsToR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publishercode=ucpress .
Each copy of any part of a JsToR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
JsToR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. we use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JsToR, please contact support@jstor.org.
The university of chicago Press is collaborating with JsToR to digitize, preserve and extend access to The American Journal of sociology.

The Economics of Organization: The Transaction Cost Approach1
Oliver E. Williamson
University of Pennsylvania
The transaction cost approach to the study of economic organization regards the transaction as the basic unit of analysis and holds that an understanding of transaction cost economizing is central to the study of organizations. Applications of this approach require that transactions be dimensionalized and that alternative governance struct ures be described. Economizing is accomplished by assigning transa ctions to governance structures in a discriminating way. The app roach applies both to the determination of efficient boundaries, as between firms and markets, and to the organization of internal transa ctions, including the design of employment relations. The approach is compared and contrasted with selected parts of the organization theory literature.
The proposition that the firm is a production function to which a profitm aximization objective has been assigned has been less illuminating for organization theory purposes than for economics. Even within economics, however, there is a growing realization that the neoclassical theory of the firm is self-limiting. A variety of economic approaches to the study of organization have recently been proposed in which the importance of internal organization is acknowledged.2 The one described here emphasizes
1 This paper has benefted from a number of discussions 1 have had with Wiiliam G, Ouchi, including those we had aL a Mini-Conference on Strategy, Marketing, and Organization (heid at the Graduate School of Management, UCLA, during April 1980 under the auspices o! Booz, Alien, & Hamilton) and at the rcccnt Conference on thc Economics o! Organization (held in Berlin in June 1980 under the auspices o! Lhe International Institute of Managemcnt). It has also bcnefited from a year-Iong dial ogue on these matters Lhat Ouchi and 1 have had with Paul Kaestle and William Alien. The paper also benefited greatly from remarks on an earlier version by Banri Asanuma and on a later revision by Herbert Simon. The assistance of AIS reviewers in reshaping the manuscript is also appreciated. Requests for reprints shouid be sent Lo Oliver E. Williamson, Dcpartment of Economics, University o! Pennsylvania, Philad elphia, Pennsylvanla 19104.
2 These inciude the neoclassicai theory of Lhe firm—which, however is relatively sparse in its organizational implications—managerial discretion theory (T3aumol 1959; Marris 1964; Williamson 1964), team theory (Marschak and Radner 1972), agency theory (Alchian and Demsetz 1972; Jensen and Meckling 1976), and the transaction cost approach (Coase [1937] 1952; Wllliamson 1975). Although 1 was aware, when 1 was
© 1981 by The Univcrsity o! Chicago. Ali rights reserved
0002-9602/82/8703-0002$01 .50
548 AIS Volume 87 Number 3

Transaction Cost Approach
transaction costs and efforts to economize thereon. More than most econ omic approaches, it makes allowance for what Frank Knight (1965, p. 270) has felicitously referred to as “human nature as we know it.”3
Economic approaches to the study of organization, transaction cost analy sis included, generally focus on efficiency. To be sure, not every interesti ng organizational issue can be usefully addressed, except perhaps in a minor way, in efficiency terms. A surprisingly large number can, however, especially if transaction cost aspects are emphasized. This is accompiished by making the transaction—rather than commodities—the basic unit of analysis and by assessing governance structures, of which firms and markets are the leading alternatives, in terms of their capacities to economize on transaction costs.
The transaction cost approach to the study of organizations has been applied at three leveis of anaiysis. The first is the overail structure of the enterprise. This takes the scope of the enterprise as given and asks how the operating parts should be related one to another. Unitary, holding company, and multidivisional forms come under scrutiny when these issues are addressed.4 The second or middle levei focuses on the operating parts and asks which activities should be performed within the firm, which outs ide it, and why. This can be thought of as deveioping the criteria for and defining the “efficient boundaries”5 of an operating unit. The third levei of anaiysis is concerned with the manner in which human assets are organized. The object here is to match internal governance structures with the attributes of work groups in a discriminating way.
OnIy issues of the two latter kinds are addressed in this paper.6 The study of both of these issues turns criticaily on the dimensionalizing of transactions. The antecedent literature from which the transaction cost approach derives is sketched in Section 1. The rudiments of the approach, inciuding the dimensionalizing of transactions, are then set out in Section Ir. Applications to the study of efficient boundaries are deveioped in Section III. Empioyment relation issues are addressed in Section IV. Comw orking on Markets and Hierarchies, that lt had a number of applications outsidc
economics, the book was directed at an economics audience. 1 was therefore gratified when organization theory speciaiists recognized merit ia the approach. 1 am espccially indebtcd to William Ouchi for bringing Lhe book to the attention of Lhe organization thcory audience (soe Ouchi 1977).
Knight’s remarks about the human attributes of economic agcnts have been widcly disrcgarded and attention lias been focused narrowly on Lhe risk-bearing aspects of Knight’s classic work.
1 have discussed thcse issues at Iength elsewhere (soe Williamson 1970, chaps. 2, 3, and 7; 1975, chaps. 8—9).
3Thc term “efficient boundaries” is borrowed from Ouchi (19800).
O For a discussion of the issues that arise at Lhe flrst levei, soe Lhe rcferenccs in n. 4.

American Journal of Socio!ogy
parisons with se!ected aspects of the organization theory literature and contrasts with “power” approaches to the study of organizations are made in Section V. Concluding remarks follow.
The transaction cost approach to the study of organizations relates to three relatively independent literatures. To be sure, there is considerable overlapping among them and they have not proceeded heedless of one another. The extent to which they deal with common issues, however, is rarely recognized.
Considering that economizing is central to the transaction cost approach, it is not surprising that an economics literature is among the antecedents. Also, inasmuch as interna! organizational issues are featured, the organizat ion theory literature makes an expected appearance. The third literature is less obvious: this is the contract Iaw literature in which contract is addressed as a governance issue.
Each of these literatures is large, and my summary of the intellectual progression in each is necessarily brief and omits important contributions. The 1930s witnessed significant advances in ali three areas. My sketch of the antecedents begins there.
The proposition that the transaction is the basic unit of economic analys is was advanced by John R. Commons in 1934. He recognized that there were a variety of governance structures with which to mediate the exc hange of goods or services between technological!y separable entities. Assessing the capacities of different structures to harmonize relations bet ween parties and recognizing that new structures arose in the sei-vice of these harmonizing purposes were centra! to the study of institutiona! econ omics as he conceived it.
Ronald Coase posed the problem more sharply in his c!assic 1937 paper, “The Nature of the Firm.” He, like others, observed that the production of fina! goods and services invo!ved a succession of early stage processing and assembly activities. But whereas others took the boundary of the firm as a parameter and examined the efficacy with which markets mediated exchange in intermediate and final goods markets, Coase held that the boundary of the firm was a decision variab!e for which an economic ass essment was needed. What is it that determines when a firm decides to integrate and when instead it relies on the market?
Friedrich Hayek’s 1945 article, “The Use of Knowledge in Society,” shed further insight. fie observed that the economic problem is relative!y uninteresting except when economic events are changing and sequential adaptations to these changes are needed. What distinguishes a high per5 50

Transaction Cost Approach
formance economy is its capacity to adapt efficientiy to uncertainty. Ait hough he did not state the issues in transaction-cost-economizing terms, such terms are impiicit in much of the argument.
The postwar market failure literature helped better to define some of the “failures” with markets that common ownership (the firm) served to overcome. Tt was not until 1969, however, that the underlying difficulties with markets were unambiguously traced to transaction cost origins. As Kenneth Arrow put it: “Market failure is not absolute; it is better to consider a broader category, that of transaction costs, which in general impede and in particular cases completely block the formation of markets”
(1969, p. 48).
The appearance of Chester Barnard’s book The Functions of lhe Execut ive in 1938 and of Herbert Simon’s explication of the Barnard thesis in Administrative Behavior in 1947 are widely recognized as significant events in the organization theory fieid. Purposive organization was emphasized, but the limits of human actors in bounded rationaiity respects and the importance of informal organization were prominently featured.
This stream of research was further deveioped by the “Carnegie Schooi” (March and Simon 1958; Cyert and March 1963). Hierarchical organizat ion and associated controls are traced to the limited capacities of human actors to cope with the complexity and uncertainty with which they are confronted. The organization is essentially viewed as a “problem-facing and problem-solving” entity (Thompson 1967, p. 9). But organizationai efforts are often myopic, and demands for control can and often do give rise to dysfunctional outcomes.
Aithough Alfred Chandier’s remarkabie book, Strategy and Structure (1962), had its origins in business history rather than organization theory, in many respects this historical account of the origins, diffusion, nature, and importance of the multidivisionai form of organization ran ahead of contemporary economic and organization theory. The mistaken notion that economic efficiency was substantiafly independent of internai organizat ional structure was no longer tenable af ter this book appeared.
James Thompson built on ali of the foregoing in fashioning his ciass!c statement of the organizationai problem in 1967. Both uncertainty and bounded rationality were featured. Moreover, implicitly, and sometimes explicitiy7 attention was fixed on efforts to economize on transaction costs. Core technologies, domains (or boundaries) of organized action, and the powers and iimits of market and hierarchical modes are ali recognized.
The legal literature to which 1 refer is concerned with contracting— especially the distinction between “hard contracting” (or biack-letter Iaw)
7 For example, Thompson’s proposition that “under norms of rationality, organizations group positions to minimize coordination costs” (1967, pp. 64—65) is in this spirit.

American Journal of Sociology
and “soft contracting” in which the contract serves mainly as framework. Karl Lieweilyn’s 1931 essay addressed these issues. He observed that transa ctions come in a variety of forms and that a highiy legaiistic approach can sometimes get in the way of the parties instead of contributing to their purposes. This is especially true where continuity of the exchange relation between the parties is highly valued.
Others who adopted and refined this theme inciude Steward Macaulay (1963), Lon Fulier (1964), Clyde Summers (1969), David Felier (1973), and Jan Macneil (1974). As Macneil puts it, the discrete transaction— “sharp in by clear agreement; sharp out by clear performance” (1974, p. 738)—is very rare in both Iaw and economics, and we deceive ourselves by treating it otherwise. What he refers to as “relationai” forms of cont racting—which may involve arbitration, coilective bargaining, and other types of obligationai market exchange—are becoming more important and need to be recognized.
A deepening awareness of transaction cost issues marks the progression of each of the literatures. Among other things, by the early 1970s it was becoming clear that the study of organizations was a comparative institut ional undertaking in which alternative governance structures—both within and between firms and markets—required explicit attention. Inasmuch, moreover, as the transactions of interest were not ali of a kind, differences among them would evidently have to be recognized. What were the dist inguishing attributes? Finally, although transaction cost economizing is an important and greatly neglected topic, such economizing cannot proceed regardiess of the production cost ramifications. Put differently, transaction cost economizing needs to be located within a larger economizing framew ork and the relevant trade-offs need to be recognized.
A transaction occurs when a good or service is transferred across a techn ologically separabie interface. One stage of activity terminates and ano ther begins. With a weii-working interface, as with a well-working mac hine, these transfers occur smoothiy. In mechanical systems we look for frictions: do the gears mesh, are the parts iubricated, is there neediess slfppage or other Ioss of energy? The economic counterpart of friction is transaction cost: do the parties to the exchange operate harmoniously, or are there frequent misunderstandings and conflicts that iead to delays, breakdowns, and other malfunctions? Transaction cost analysis supplants the usual preoccupation with technology and steady-state production (or distribution) expenses with an examination of the comparative costs of

Transactiori Cost Approach
pianning, adapting, and monitoring task compietion under alternative gove rnance structures.
Some transactions are simpie and easy to mediate. Others are difficuit and require a good deai more attention. Can we identify the factors that permit transactions to be classified as one kinci or another? Can we identify the alternative governance structures within which transactions can be organized? And can we match governance structures with transactions in a discriminating (transaction.cost-economizing) way? These are the neg lected issues with which organizational design needs to come to grips. These are the issues for which transaction cost analysis promises to offer new insights.
Behaviorai Assumptions
It is widely recognized—by economists, Iawyers, and others who have au interest in contracting—that complex contracts are costly to write and enforce. There is a tendency, however, to accept this fact as given rather than inquire into the reasons for it. As a resuit, some of the consequences of and remedies for costiy contracting are Iess weiI understood than would otherwise be the case.
What is needed, 1 submit, is more self-conscious attention to “human nature as we know it.” The two behavioral assumptions on which transa ction cost analysis relies that both add reaiism and distinguish this app roach from neociassicai economics are (1) the recognition that human agents are subject to bounded rationaiity and (2) the assumption that at ieast some agents are given to opportunism.
Bounded rationaiity needs to be distinguished from both hyperrationaiity and irrationality (Simon 1978). Unlike “economic man,” to whom hyp errationality is often attributed, “organization man” is endowed with less powerful anaiyticai and data-processing apparatus. Such iimited compet ence does not, however, impiy irrationaiity. lnstead, aithough boundedly rational agents experience limits in formulating and solving complex probi ems and in processing (receiving, storing, retrieving, transmitting) inform ation (Sirnon 1957), they otherwise remam “intendedly rational.”
But for bounded rationality, ali economic exchange could be efficiently organized by contract. (The economic theory of comprehensive contracting for unboundediy rational agents has been eiegantiy worked out.8) Given bounded rationaiity, however, it is impossibie to deal with compiexity in
8 The comprehensive contracting model is widely referred to as the Arrow-Dcbreu model. For a discussion and an interesting contribution to this Iitcrature, see Radncr

American Journal of Sociology
ali contractually relevant respects. As a consequence, incomplete contracti ng is the best that can be achieved.
Ubiquitous, albeit incomplete, contracting would nevertheless be feasibie if human agents were not given to opportunism. Thus, if agents, though boundedIy rational, were fully trustworthy, comprehensive contracting wouid still be feasible (and presumabiy would be observed). Principais wouid simply extract promises from agents that they would behave in the manner of steward when unanticipated events occurred, while agents would reciprocally ask principais to behave in good faith. Such devices wiil not work, however, if some economic actors (either principais or agents) are dishonest (or, more generaiiy, disguise attributes or preferences, distort data, obfuscate issues, and otherwise confuse transactions), and it is very costly to distinguish opportunistic from nonopportunistic types ex ante.
A different way of putting this is to say that while organizationai man is computationally less competent than economic man, he is motivationaliy more complex. Thus, whereas economic man engages in simpie self-interest seeking,9 opportunism makes provision for self-interest seeking with guile. Probiems of contracting are greatly complicated by economic agents who make “faise or empty, that is, self-disbelieved threats or promises” (Goffm an 1969, p. 105), cut corners for undisciosed personal advantage, cover up tracks, and the like.
That economic agents are simultaneously subject to bounded rationality and (at least some) are given to opportunism does not by itself, however, vitiate autonomous trading. On the contrary, when effective ex ante and ex post competition can both be presumed,1° autonomous contracting will be efficacious. Of these two, effective ex ante competition is a much easfer condition to satisfy: it merely requires that there be large numbers of quaiified bidders at the outset. The subsequent transformation of an exc hange relation involving large numbers to one involving small numbers during contract execution is what causes problems. Whether ex post comp etition is equally efficacious or breaks down as a result of contract execut ion depends on the characteristics of the transactions in question, which brings us to the matter of dimensionalizing.
9 As Peter Diamond has put it, standard “economic modeis . [treat] individuais as piaying a game with fixed rules which they obey. They do not buy more than they can pay for, they do not cmbezzie funds, and they do not rob banks” (1971, p. 31). Oniy recently has this standard presumption come under scrutlny, often by making aliowance for what insurance specialists refer to as “moral hazard,” which is a part icular form of opportunism.
10 Aithough large numbers of qualified bidders are frequently on a parity at the outs et, winning a bid and executing a contract often introduces a disparity between the qualifications of winners and those of nonwinners, with the result that bidding comp etition involving large numbers is not equaliy effective at the contract renewal int ervai. For a discussion, see Wiiliamson (1971; 1975, pp. 27-36; 1979b); and Klcin, Crawford, and Alchian (1978).

Transaction Cost Approach
As set out elsewhere (Williamson 1979b), the critical dimensions for des cribing transactions are (1) uncertainty, (2) the frequency ‘with which transactions recur, and (3) the degree to which durable, transactions pecific investments are required to realize least cost supply. Only recurrent transactions are of interest for the purposes of this paper;3’ hence attent ion will hereaf ter be focused on uncertainty and asset specificity, especiall y the latter.
Asset specificity is both the most important dimension for describing transactions and the most neglected attribute in prior studies of organizat ion. The issue is less whether there are large fixed investments, though this is important, than whether such investments are specialized to a part icular transaction. Items that are unspecialized among users pose few hazards, since buyers in these circumstances can easily turn to alternative sources and suppliers can seu output intended for one buyer to other buye rs without difficulty. Nonmarketability problems arise when the specific identity of the parties has important cost-bearing consequences. Transa ctions of this kind may be referred to as idiosyncratic.’2
Asset specificity can arise in any of three ways: site specificity, as when successive stations are Iocated in cheek-by-jowl relation to each other so as to economize on inventory and transportation expenses; physical asset specificity, as where specialized dies are required to produce a component; and human asset specificity that arises from learning by doing. The reason asset specificity is critical is that, once an investment has been made, buyer and seiler are effectively operating in a bilateral (or at Ieast quasi-bilateral) exchange relation for a considerable period thereafter. Inasmuch as the value of specific capital in other uses is, by definition, much smaller than the specialized use for which it has been intended, the supplier is effect ively “locked into” the transaction to a significant degree. This is symm etrical, moreover, in that the buyer cannot turn to alternative sources of supply and obtain the item on favorable terms, since the cost of supply from unspecialized capital is presumably great.13 The buyer is thus comm itted to the transaction as well. Accordingly, where asset specificity is great, buyer and seller will make special efforts to design an exchange that has good continuity properties.
The site-specific assets referred to here appear to correspond with those
11 For a discussion of the organizational consequences of occasionai, rather than rec urrent, contracting, see Wihiamson (197gb, pp. 246—54). Also sce n. 32 below.
‘2For earlier treatments of the economies of idiosyncrasy, see Willianison (1975, pp. 9—10, 27—33, 68—74; 1979b, pp. 238—45). Others who are persuadcd that idiosyncratic investments are crucjal to the undcrstanding of the economics of organization include Klcin et ai. (1978), Klcin (1980), and Tecce (1980).
13 For a somewhat related discussion of symmetry, see Thompson (1967, pp. 32—35).

American Journai of Sociology
Thompson describes as the “core technoiogy” (1967, pp. 19—23). Indeed, the common ownership of site.specific stations is thought to be so “natural” that alternative governance stz-uctures are rareiy considered. In fact, howe ver, the joining of separable stations—for example, blast furnace and rolling miii, thereby to realize thermal economies—under common owners hip is not technoiogically determined but instead reflects transaction-coste conomizing judgments.14 It wiii nevertheless be convenient, for the purp oses of this paper, to assume that ali site-specific stations constitute a technoiogicai core the common ownership of which wiii be taken as given. Attention is thus focused on eariier stage, later stage, and lateral transa ctions. The efficient governance structure for these turns on physicai asset and human asset specificity. Although these are often correiated, it wiii faciiitate the argument to treat them sequentiaiiy. Thus, physicai asset specificity is emphasized in Section III and human asset specificity is not introduced untii Section IV.
The treatment of efficient boundaries in this section deals with only a part, aIbeit an interesting part, of the fuil set of organizational issues. Oniy two organizationai alternatives are considered: either a firrn makes a component itseif or it buys it from an autonomous supplier. Thus rnixed modes, such as franchisfng, joint ventures, etc., are disregarded. 1 also take the core technoiogy as given and focus on a singie une of commerce— say the activities of a particular manufacturing division within a larger industrial enterprise. The object is to describe how the economizing decis ions which define the outer boundaries of this division are made.’5
Schernatic Description
Suppose that there are three distinct production stages which, for sites pecificity reasons, are ali part of the sarne firm. This is the technoiogicai core. Suppose that raw materiais are distinct and are naturally procured from the market. Suppose that two things occur at each production stage:
there is a physical transformation, and components are joined to the “main frame.” And suppose, finaiiy, that the firm has a choice between own dist ribution and market distribution.
Let the core production stages be represented by Si, S2, S3 and draw these as rectangies. Let raw materiais be represented by R and draw this
14 See Williamson (1971) and McKean (1971) for a discussion of alternative modes and an assessment of transaction cost consequences for site-specific transactions.
‘5The focus is on operating decisions of a firm or market kind. Both strategic dccis ions and interdivisionai asset sharing are ignored.

Transaction Cost Approach
as a circie. Let component suppiy be represented by C1.B, C2-B, C3-B if the firm buys its components and CI-O, C2-O, C3-O if it makes its own components. Draw these as triangles. Let distribution be given by D-B if the firm uses market distribution and D-O if the firm uses own distribut ion. Draw these as squares. Finally, iet a solid une between units repr esent an actual transaction and a dashed une a potential transaction, and draw the boundary of the firm as a closed curve that inciudes those activities that the firm does for itseif.
The closed curve that defines the efficient boundary of the firm iii figure 1 inciudes, in addition to the technical core, component C2 and the díst ribution stage, D. Components Cl and C3 and raw materiais are procured in the market. Obviously this is arbitrary and mereiy illustrative. It also oversimplifies greatiy. It is reiatively easy, however, to elaborate the schema to add to Lhe core, to consider additional components, to inciude several raw material stages and consider backward integration into these, to break down distribution. etc. But the central points would remam unc hanged, namely: (1) the common ownership of some stations—the core— is sufficientiy obvious that a carefui, comparative assessment is unneeded (site spedficity wili often characterize these transactions); (2) there is a second set of transactions in which own supply is manifestiy uneconomic, hence market supply is indicated (many raw materiais are of this kind); but (3) there is a third set of activities for which make-or-buy decisions can only be made af ter assessing the transformation and transaction cost consequences of aiternative modes. The efficient boundary is Lhe inclusive set of core plus additionai stages for which own supply can be shown to be Lhe efficient choice.
Fxc. 1.—Efficicnt boundary

American Journal of Sociology
A Simple Model
The crucial issue is how the choice between firm and market governance structures for decisions related to point 3 above are made. Transaction cost reasonhig is central to this analysis, but trade-offs between production cost economies (in which the market may be presumed to enjoy certain advantages) and governance cost economies (in which the advantages may shift to internal organization) need to be recognized.
The issues here are somewhat involved and are set out more fully and formally elsewhere.’6 The central points are these: (1) physical asset speciikity is never valued by itself but only because demand is thereby increased in design or performance respects;’7 (2) such valued demand consequences are often realized only at greater production expense (stand ardized items would be cheaper, often because scale economies could be more fufly exhausted); whence (3) the optimal choice of asset specificity requires that demand and production cost consequences be taken into account simultaneously; and (4) governance costs also vary with asset specificity, and these also have to be introduced into the calculus.
The choice between firm and market organization arises in this last connection. If assets are nonspecific, markets enjoy advantages in both production cost and governance cost respects: static scale economies can be more fully exhausted by buying instead of making; markets can also aggregate uncorrelated demands, thereby realizing risk-pooling benefits; and externa! procurement avoids many of the hazards to which internal procurement is subject.18 As assets become more specific, however, the aggregation benefits of markets in the first two respects are reduced and exchange takes on a progressively stronger bilateral character. The gove rnance costs of markets escalate as a result and interna! procurement supplants external supply for this reason.’9 Thus, the governance of rec urrent transactions for which uncertainty is held constant (in intermediate degree) will vary as follows: classical market contracting will be efticacious
16 The simple model sketched out here is developed more fully iii Williamson (1981).
Site specificity, in contrast, involves transportation and inventory cost savings, albeit by complicating the problem of mediating the exchange interface.
16For a discussion of bureaucratic hazards, see Thompson (1967, pp. 152—54) and Williamson (1975, pp. 117—31).
19 Actually, the naturc of the asset specfficity matters. 11 the assets in question are mobile and the specificity is due to physical hut not human asset features, market procurement may still be feasible. It can be accomplished by having the buyer own the specific assets (e.g., dies). He puts the business up for bid and awards it to the low bidder, to whom he ships the dies. Should contractual difficulties arise, however, he is not locked into a bilateral exchange. He reclaims the dies and reopens the bidding. This option is not available if the specific asscts ai-e of a human asset kind or if they are nonmobile. This “refincment” of transaction cost reasonfng illustrates how the approach can and should be dcvelopcd and its predictive powcr sharpcncd and testcd.

Transaction Cost Approach
whenever assets are nonspecific to the trading parties; bilateral or obligat ional market contracting will appear as assets become semispecific; and internal organization wilI dispiace markets as assets take on a highly specific character.
The axlvantages of firms over markets in harmonizing bilateral exchange are three. First, common ownership reduces the incentives to suboptimize. Second, and related, internal organization is able to invoke fiat to resolve differences, whereas costly adjudication is needed when an impasse dev elops between autonomous traders. Third, internal organization has easier and more complete access to the relevant information when dispute settling is needed. The incentive to shift bilateral transactions from markets to firms increases as uncertainty is greater, since the costs of harmonizing the interface vary directly with the need to adjust to changing circumstances.
At the risk of oversimplification,2° the essence of the foregoing argument can be shown graphically by expressing both production cost differences and governance cost differences as functions of asset specificity (A). Thus let tC = j(A) be the production cost difference between internal organizat ion and the market, tG = g(A) be the corresponding governance cost difference, and assume that these two functions have the shapes and relat ive locations shown in figure 2. So long as the vertical sum of tC + AG remains positive, market procurement enjoys the advantage. Indifference between governance structures obtains where AC + tG = O, namely, at
Á. Internal procurement enjoys the advantage for values of A that exc eed Á (since .C + G < O in this region).
Implicitly, this was the apparatus used in making governance structure assignments for the component and distribution stages shown in figure 1. Inasmuch as component C2 was taken out of the market and is supplied internally, while components Cl and C3 remam in the market, components Cl and C3 are presumably more standardized21 (Cl represents, say, arma-
20 The main simplification is that C (and possibly iG) is also a function of thc amount produced. Figure 2 can be thought of as a cross-section for a ftxed levei of output. Furthermore, Lhe optimal value of A will depend on both demand effects and absolute cost effects. Only cost differences are shown in the figure.
21 Transaction-specific investments are related Lo but need to be distinguished from the more familiar notion of standardization. Although many nonstandard goods and services are produced with the assistancc of nonstandard (specialized) assets, this need foi be the case. When iL Is not, the production of nonstandard goods or scrviccs with assets that involve little specificity poses few contracting problems. Thus, suppose that a glass mamufacturer is producing circular lenses for spotlights and supplies thcm to a large number of spotlight manufacturers. Suppose that one of the spotlight firms decides to add triangular and square spotlights to its lime. Such designs will be reco gnized as nonstandard, but they will pose special problems in contracting for lenses only if Lhe glass manufacturer has to dedicate spedal assets Lo thc production of thc nonstandard lcnses. If he can, with slight modification, produce them with existing plant and labor force, the fact of nonstandard dcsign poses no particular economic

American Journal of Sociology
ture wire and C3 a transistor) while C2 is more specialized (has distinctive chassis styling or performance features). Similarly, the decision to integ rate forward into distribution reflects the fact that the product cannot be marketed effectively through standard channels, presumably because specialized human assets are needed to seu and service the product and a bilateral employment relation develops as a consequence. In terms of figure 2, the values of A are Iow for Cl and C3 but exceed  for both C2 and D.
Two Examples
The transaction cost arguments set out above are of a normative kind:
what governance structure slzould be chosen. In contrast, the examples developed hei-e describe what has been observed. The critical question is not whether the appropriate governance structure was selected at the
obstacles. Tf, however, the glass manufacturer must be induced to incur specialized (transaction-spedfic) investments to produce the triangular or square lenses, a much more complicated contractual situation develops. The parties then have a stake in maintaining a continuing exchange relation (so tbat the specialized assets can be utilized effectively). Additional governance structure designed to sustain the relation and safeguard It against opportunism Is needcd.
FIO. 2.—Representative net production and governance cost differences

Transaction Cost Approach
outset but whether transaction cost factors, possibly manifested as diffic ulties that resulted from a maladapted structure, are responsible for the eventual configuration.
Automobile body manufacture.—Klein et ai. (1978, pp. 308—10) have examined the problems that arose when a bilateral exchange relationship between Fisher Body and General Motors was attempted in the 1920s. The basic facts are these:
1. In 1919 General Motors entered a i0-year contractual agreement with Fisher Body whereby General Motors agreed to purchase substant ially ali its closed bodies from Fisher.
2. The price for delivery was set on a cost-plus basis and included provisions that General Motors would not be charged more than rival aut omobile manufacturers. Price disputes were to be settled by compulsory arbitration.
3. The demand for General Motors’s production of closed body cars increased substantially above that which had been forecast. As a conseq uence, General Motors became dissatisfied with the terms under which prices were to be adjusted and urged Fisher to locate its body plants adjacent to GM assembly plants, thereby to realize transportation and inventory economies. Fisher Body resisted.
4. General Motors began acquiring Fisher stock in 1924 and completed a merger agreement in 1926.
Inasmuch as GM cars had distinctive body designs, the production of closed bodies required significant transaction-specific investments to be made. Site-specificity considerations reinforced this need. The transaction, moreover, was evidently beset by substantial demand and cost uncertaint ies. Since there was little to be gained from market procurement, while the governance costs of market procurement were predictably great, the transaction was one for which internal procurement was indicated. The strains that autonomous contracting experienced could thus have been anticipated, and the eventual reconfiguration from long-term contracting to common ownership is consistent with the basic transaction-cost-econom izing a.rgument.
Forward integration.—Chandler (1977) and Porter and Livesay (1971) report that extensive forward integration from manufacturing into dístribut ion occurred in the last 30 years of the l9th century. The reasons for this are several, including the appearance of infrastructure (in the form of the railroad, telephone, and telegraph) and a variety of manufacturing developments. But the response to these developments was anything but uniform. Forward integration included retailing for some commodities (e.g., farm equipment and sewing machines), extended only to wholesaling for others (e.g., tobacco and certain branded items), and was negligibie for

American Journal of Sociology
stiii others (e.g., packaged groceries and dry goods). What were the det ermining factors?
Tracing this differentíai response is beyond the scope of this paper but is reported elsewhere (Williamson 1980). Very briefly, the pattern appears to be this. Integration into retailing occurred only for commodities that required considerable point-of-saie information, possibly to inciude demons tration, and foilow-on service. Speciaiized human assets were evidentiy needed to provide such saies and service. Integration into whoiesaling occurred for commodities that were perishabie and branded. Forward int egration occurred because contracts to turn over inventory and destroy older stocks were neither seif-enforcing nor incentive-compatible, hence they piaced the manufacturers’ reputations at risk. Commodities that had none of these properties were soid through market distribution channeis because no speciai hazards were posed. This progression of forward int egration contingent on differential degrees of asset specificity and the differential hazards of opportunism is the principal implication of transa ction cost reasoning and appears also to be the main factor expiaining the selective degree of forward integration reported by Chandier.22
li wiil be convenient, for the purposes of this section, to assume that the transactions in question are site specific, whence internai organization is warranted. Mereiy to assign a transaction to an interna! governance struct ure does not, however, assure that the efficiency purposes of transaction cost anaiysis wiii be reaiized. It is necessary in addition to examine the human asset characteristics of the interna! transactions in question and to fashion the empioyment relation appropriateiy.
The sarne general principies appiy to the governance of human assets as appiy to the efficient organization of transactions in general. Thus to use a compiez striicture for governing simple transactions is to incur unn eeded costs, whiie to use a simple structure to govern a complex transa ction invites strain. The questions are, How are human asset differences best described, what are the empioyment reiation aiternatives, and what is the appropriate correspondence between them?
The discussion is in two parts. The first addresses the organization of human assets at the staff levei. The second deais with union organization, which appiies primariiy at the production levei.
22 AIfred Chandier advises me that he agrees broadiy with this interpretation of bis results.

‘1ransaction Cost Approach
Governance, General
Recali that transactions are described in terms of three attributes: freq uency, uncertainty, and asset specificity. The assets of interest here inv olve a continuing supply of services, whence frequency aspects wiII be suppressed and attention focused on the internal organizational aspects of uncertainty and asset specificity.
Tt will facilitate the argument to assume that transfers of goods and services across interfaces are not at issue. Internal governance is thus concerned entirely with intrastage activity. Inasmuch as physical assets are nonvolitional, transactions assigned to internal organization pose probI ems only in conjunction with human asset specificity.23
Note in this connection that skill acquisition is a necessary but not a sufficient condition for a human asset governance problem to arise. The nature of the skills also matters; the distinction between transactions pecific and nonspecific human assets is crucial. Thus, physicians, engin eers, lawyers, etc., possess valued skills for which they expect to be comp ensated, but such skills do not by themselves pose a governance issue. Unless these skills are deepened and specialized to a particular employer, neither employer nor employee has a special interest in maintaining a continuing employment relation.24 The employer can easily hire a substit ute and the employee can move to alternative employment without Ioss of productive value.
Mere deepening of skills through job experience does not by itself pose a problem either. Thus, typing skills may be enhanced by practice, but if they are equally valued by current and potential employers there is no need to devise special protection for an ongoing employment relation. Knowledge of a particular firm’s fihing system, in contrast, may be highly specific (nontransferable). Continuity of Lhe employment relation in the latter case is a source of added value.
Thus to the neoclassical proposition that Lhe acquisition of valued skills leads to greater compensation, transaction cost reasoning adds the followi ng proposition: skills acquired in a learning-by-doing fashion and imperf ectly transferable across employers need to be embedded in a protective governance structure, lest productive values be sacrificed if Lhe employm ent relation is unwittingly severed. The concern here is with what Knight has referred to as “the internal problems of the corporation, the protection
of members and adherents against each other’s predatory propensities”
23 Actuafly, this assumes away transfer pridng probiems, which can be tricky but take us away from our maio concerns.
24 This ignores transitionai problems that rnay be associated with job relocation. Ali employees experience these, 00 whicb account protection against arbitrary dismissal is sought. But the further guestion is what additional safeguards are warranted. This matter turns on human asset specificity.

American Journal of Sociology
(1965, p. 254). This poses a problem in the degree to which assets are firm-specific.
The internal organizational counterpart for uncertainty is Lhe ease with which Lhe productivity of human assets can be evaluated. This is essent ially the metering problem to which Armen Alchian and Harold Demsetz refer in their treatment of the firm (1972). Their argument is that firms arise when tasks are technologically nonseparable, the standard example being manual freight loading. As they put it (1972, p. 779): “Two men jointly lift cargo into trucks. Solely by observing the total weight loaded per day, it is impossible to determine each person’s marginal productivity.
• . . The output is yielded by a team, by definition, and it is not a sum of separable outputs of each of its members.”
When tasks are nonseparable in this sense, individual productivity cann ot be assessed by measuring output—an assessment of inputs is needed. Sometimes productivity may be inferred by observing the intensity with which au individual works; this is Lhe aspect emphasized by Alchian and Demsetz. Often, however, the assessment of inputs is much more subtie than effort accounting. Does the employee cooperate in helping to devise and implement complex responses to unanticipated circumstances, or does he attend to his own or local goals at the expense of others? Metering this, except over long observation intervals, can be inordinately difficult.
Human assets can thus be described in terms of (1) the degree to which they are firm-specific and (2) the ease with which productivity can be metered. The fact that Alchian and Demsetz consider only the latter exp lains the narrow construction of the employment relation in their assessm ent of economic organization.28 Both dimensions, however, are critical to au adequate assessment. Letting H1 and H2 represent Iow and high degrees of human asset specificity and M1 and M2 represent easy and difficult conditions of meterability, the following four-way classification of internal goveruance structures is tentatively proposed:
1. H1, M1: internal spot markei.—Human assets that are nonspecific and for which metering is easy are essentially meeting market tests cont inuously for their jobs. Neither workers nor firms have au efficiency int erest in maintaining Lhe association. Workers can move between employers without loss of productivity, and firms can secure replacements without incurring start-up costs. Hence no special governance structure is devised to sustain the relation. Instead, the employment relation is terminated
25 Alchian and Demsetz treat human assets as fungibie. Thus aithough incumbents niay continue to hold jobs for a considerabie period of time and may claim to be subject to an “authority relationship,” ali they are doing is continuously meeting bids for their jobs in the spot market under the Alchian and Demsetz scheme. See Alchian and Demsetz (1972, p. 777) and, for a discussion, Wiiiiamson (1975, pp. 66—69).
2 Alchian evidentiy agrees. See Klein et ai. (1979, p. 322, n. 49).

Transaction Cost Approach
when either party is sufficiently dissatisfied. An internal spot market labor relation may be said to exist. Examples include migrant farm workers and custodial employees. Professional employees whose skills are nonspecific (certain draftsmen and engineers) also fali into this category.
2. H,, M2: primitive team.—Although the human assets here are nons pecific, the work cannot be metered easily. This is the team organization to which Alchian and Demsetz refer (1972). Although the membership of such teams can be altered without ioss of productivity, compensation cannot easily be determined on an individual basis.27 The manual freight loading example wouid appear to qualify. This structure is referred to as a primitive team, to distinguish it from the relationai team, described below.
3. H2, M1: obligational market.—There is a considerable amount of firm-specific learning here, but tasks are easy to meter. Idiosyncratic techn ological experience (as described, for exampie, by Doeringer and Piore [1971, pp. 15—16]) and idiosyncratic organizationai experience (accounti ng and data-processing conventions, internalization of other complex rules and procedures, and the like) both qualify. Both firm and workers have an interest in maintaining the continuity of such employment reiations. Procedural safeguards will thus be devised to discourage arbitrary dia- missal. And nonvested retirement and other benefits will accrue to such workers so as to discourage unwanted quitting (for a discussion, see Mort ensen 1978).
4. H2, M2: relational team.—The human assets here are specific to the firm and very difficult to meter. This appears to correspond with the “clan” form of organization to which William Ouchi (1980b) has referred. The firm here will engage in considerabie social conditioning, to help assure that employees understand and are dedicated to the purposes of the firm, and employees will be provided with considerable job security, which gives them assurance against exploitation. Neither of these objectives can be realized independently of the other.
Relational teams are very difficult to deveiop, and it is uncertain how widespread or sustainable they are. It is argued that some of the Japanese corporations are organized in this way (for a discussion, see Lifson 1979), but the interpretation of this is subject to dispute. Certain utopian 50- cieties are organized as relational teams, but these have experienced severe continuity problems as the initial membership, which often was highly committed, retired or expired (see Kanter 1972; Manuel and Manuel
The above described match of internal governance structures with the
27 Thls assumes that output is a joint product and that input dlfferences cannot be easily ascertained.

American Journal ot SocioLogy
Fia. 3.—Tbe governance of internal organization
internal transactional attributes just described is summarized in figure 3. Admittedly, describing internal transactions in bivariate, binary terms simp lifies considerably. The overail framework is nevertheless in place and refinements can be made as needed. (Thus, mixed internal governance structures wiII presumably arise to service transactions that take on int ermediate, rather than extreme, M and H values.)
Despite its simplicity, the four-way classification is instructive in several respects. For one thing, even this simple four-way classification of the employment relation is useful in breaking down what has previously been subsumed under the broad heading of unified governance. Second, and related, merely to recognize that a recurrent transaction involves high asset specificity and hence is appropriately organized under unified gove rnance is not sufficient to assure that the efficiency purposes of transact ion cost analysis wiIl be realized. lt is also necessary to recognize that asset specificity breaks down into site, physical, and human asset categ ories and that these have significantly different internal governance ramif ications. Third, differential meterability also matters. The fact that int ernal transactions dimensionalize along lines similar to those used to des cribe transactions generally (see Sec. II) reinforces confidence in the underlying transaction cost approach.
Some Remarks on Union Organization
The foregoing discussion of internal governance structures refers mainly to staff rather than production-level employees. Since it is among the Iatter that union organization appears, the question arises as to whether transa ction cost reasoning has useful applications to the study of coilective


Transaction Cost Approach
organization. To the extent that it does, further confidence in the power of the approach is presumably warranted.
The general reasons that coliective organization of the work force aff ords efficiency benefits when the human assets in question are firm-specific in significant degree have been set out elsewhere (Williamson, Wachter, and Harris 1975). Rather than repeat them here, 1 merely observe that the transaction cost approach to the study of unionization yields testable implications that do not derive from more familiar theories of unionization that rely on power or politics to drive the analysis (Freeman and Medoff 1979). The principal implications are: (1) the incentive to organize prod uction workers within a coilective governance structure increases with the degree of human asset specificity; and (2) the degree to which an internal governance structure is elaborated will vary directly with the degree of human asset specificity. Transaction cost analysis thus predicts that unions wiII arise early in such industries as railroads, where the skills are highly specific, and will arise late in such industries as migrant farm labor, where skills are nonspecific. lt further predicts that the governance structure (job Iadders, grievance procedures, pay scales) will be more fully elaborated in industries with greater specificity than in those with less (steel vs. autos is an example). The preliminary data appear to support both propositions.28
The transaction cost hypothesis does not deny the possibility that unions will appear in settings where human asset specificity is slight. Where this occurs, however, the presumption is that these outcomes are driven more by power than by efficiency considerations. Employers in these circums tances will thus be more incined to resist unionization; successful efforts to achieve unionization will often require the assistance of the political process; and, since power rather than efficiency is at stake, the resulting governance structure will be relatively primitive.
As noted at the outset, some of the antecedents and the behavioral ass umptions employed in the transaction cost approach have their origins in the organization theory literature. Further connections between transa ction cost economics and that literature are sketched. The transaction cost approach is then contrasted with the “power” approach to the study of organizations.
28 The arguments and the evidence are developed more fully in Scott R. Vil1iamson

Arnerican Journal of Sociology
Some Comparisons
The transaction cost approach is usefuily compared with the population ecology model, with Thompson’s work on organizations, with a recent surv ey of interorganizational linkages, and with the posterior rationality pers pective. Michael Hannan and John Freeman’s influential statement of the population ecology model poses the following provocative question:
“Why are there so many kinds of organizations?” (1977, p. 936). The transaction cost approach affords a partial answer: there are so many kinds of organizations because transactions differ so greatly and efficiency is realized only if governance structures are tailored to the specific needs of each type of transaction.
Hannan and Freeman also observe that “little attention is paid in the organizations literature to issues concerning Lhe proper units of analysis” (1977, p. 933). They argue, however, that choice of the unit of analysis is important and “invoives subtie issues [with] far reaching consequences for research activity” (1977, p. 933). 1 fully concur and argue that Lhe transaction is usefully made Lhe basic unit of analysis. Among other things, this practice shifts attention away from technology (and technological det erminism) and sensitizes analysts to transaction costs and Lhe crucial importance of organizations for economizing on such costs. This brings organization theory to Lhe fore, since choice of an appropriate governance structure is preeminently an organization theory issue.
The population ecology modei emphasizes adaptive fitness (Hannan and Freeman 1977; Aldrich 1979). lt operates at a relatively high levei of abstraction, however, and hence does not offer specific predictions as to which particular organizations wiii have superior properties in which circ umstances. The transaction cost approach has addressed this issue mainly in the context of commercial organizations, in which both product and capital market competition are Lhe sources of natural selection pressures. How broadly it will appiy elsewhere remains to be seen. li is nevertheless interesting that public utilities can be studied in this way (Wiliiamson 1976). More generally, any issue that can be posed, directiy or indirectly, as a contracting probiem can be analyzed to advantage in transactionc ost-economizing terms.
The transaction cost approach has numerous points of contact with Thompson’s work. Thus both he and 1 emphasize that human agents are subject to bounded rationality and that the basic problem with which
29 An iliustration of a problem that 1 once believed to be outside the scope of transa ction cost analysis is the oligopoly issue. Once 1 bad rethought the issue in contracti ng terms, it became clear that a number of useful statements could be made about the llkclihood of successful coilusion among oligopolists. See Wllhiamson (1975, chap.

Transaction Cost Approach
organizations must contend is adapting effectively to uncertainty.8° Both of us are also interested in the problem of efficient boundaries (what Thompson refers to as the “domam” [1967, p. 26]), and we both contend that economizing on “coordination costs” (Thompson 1967, pp. 57—65) is crucial to the definition of the boundary and to the way in which internal relations are ordered. 1 also pick up his notion of the “technical core” (Thompson 1967, p. 11) in my discussion of efficient boundaries, and his discussion of power as a reciprocal condition (Thompson 1967, p. 32) is similar to (though in other respects it goes beyond) mine.
Thompson and 1 differ in that he does not appear to make allowance for trade-offs between production economies and transaction cost econom ies,31 while 1 do. In addition, he does not dimensionalize transactions. Many of his propositions appear to be nontestable for this reason, but at least some of them could be restated to advantage using the dimensionalizat ion of transactions proposed above. His contracting, coopting, and coa lescing arguments (Thompson 1967, pp. 35—37), for example, can be expressed in terms of the frequency, uncertainty, and asset specificity properties of the transactions in question. Thus, assume that the transa ctions in question are recurring and involve an intermediate degree of uncertainty. Then autonomous contracting will be used when assets are nonspecific: obligational contracting (which is akin to co-opting) wiIl be used for assets of an intermediate degree of specificity; and merger (coa lescing) occurs if assets, especially human assets, are highly specific. We also differ somewhat in our treatments of coilective bargaining. 1 contend that the governance structure within which coliective bargaining operates will be specifically attuned to the nature of the human assets in question. This is not inconsistent with Thompson’s discussion (1967, pp. 109—10) but goes beyond it.
lt is also of some interest to relate the transaction cost approach to the recent survey of interorganizational linkages by Laumann, Galaskiew icz, and Marsden (1978). Similarities here include their discussion of
30 See Thompson (1967, pp. 9—13). Thompson’s view that “structure is a fundamental vehicle by which organizations achieve bounded rationality” (1967, p. 54) is dose in spirit lo mine, tbough 1 would express it somewhat differently. The manner iii which the internal affairs of lhe firm are decomposed determines whether the organization is able lo cope effectively within the bounded rationality limits to which lIs mana gement is subject.
31 Thus, Thompson refers repeatedly to minimizing activities without inquiring whether
successive minimizing efforts are independent. If they are not, it is not possible
simultaneously to “minimize the power of task-environment elements” (Thompson
1967, p. 32) and to “group positions lo minimize coordination costs” (Thompson
1967, p. 57). Moreover, the trade-offs between organizing costs and operating costs
need lo be faced. For any given output, lhe object is to minimize lhe sum rather
than cither one.

American Journal of Sociology
modes, relationships, and linkages in an open-systems context. They adopt a relatively microanalytic approach to the study of transactions and cont end that “interorganizational and intraorganizational transactions [mustj be distinguished, which thus implies that the problem of delineating org anizational boundaries be faced” (1978, p. 460). This is precisely the issue addressed in Section III, above. They argue further that the “specific form taken by the total network . . . wiIl aiso be influenced by t1e context of the relationships . . . as well as by the modality or normative context within which network formation occurs” (1978, p. 461). Expressed in my terms, it does not suffice to assign a transaction to one governance struct ure (a firm or a market) or another. It is furthermore necessary to attune the exchange reiationship to the continuity needs of the parties. When these are minimal, autonomous contracting is both efficient and effective. As the needs for contingent cooperation increase, however, autonomous contracting is supplemented by mandated rules or by mutual efforts (inc luding merger) to discourage aggressive suboptimization (Laumann et ai. 1978, p. 468). Within internal organization, moreover, there is a further need to examine the characteristics of the employment relation and to attune it in a discriminating way (see Sec. IV, above).
Whereas Laumann et ai. describe network modalities in terms of comp etitive and cooperative modes, 1 favor a three-way description in which networks are descrihed as autonomous, cooperative, and strategic. Introd uction of this last goes beyond the scope of this paper but makes allowa nce for “interorganizational relations [that] take on a more perduring nature than that of the narrowly defined instrumentalities of procuring necessary inputs and disposing of products, . . . [but include] seeking unfair advantage and subverting the market mechanism” (Laumann et ai. 1978, p. 467). Whether such strategic uses of interorganizational relations are feasibie turns on market structure considerations. A transaction cost interpretation of strategic abuses can be developed and has been set out elsewhere (Wiiiiamson 1979a).
The Laumann et ai. discussion of the “resource-dependency” theory is interesting in two respects (1978, p. 470). For one thing, Laumann et ai. question whether it is sufficient to focus on dyadic exchange. My answer is that dyadic exchange is very powerful and Iess deiimiting than some suggest (though 1 concede that triadic or higher-order analysis is sometimes indicated). They also observe that exchange theory has a tendency “to become tautological” and that specific exchanges, once formed, may be resistant to reassignment thereafter. 1 examine the tautological aspects of exchange theory in the discussion of “power” below. The difficulty of changing trading partners to which they refer is akin to my distinction

Transaction Cost Approach
between ex ante and ex post competition. The issue is this: do the benefits of large-numbers bidding competition (which condition can normaily be presumed at the outset, when ali potential bidders are at a parity in exp erience respects) continue at the contract renewai interval, or are they upset during contract execution? The transaction cost answer is that the initial large numbers bidding competition will be trans formed into one of bilateral exchange at the contract renewal interval if execution entails nont rivial transaction specific investments. Winners will tben enjoy an adv antage over nonwinners, but not otherwise.32
Consider finaily the relation between bounded rationality, as it is used by Simon and employed here, and the concepts of hyperrationality and “posterior rationaiity” (Weick 1969; March 1973, 1978). Bounded rat ionality has been defined as behavior that is “intendedly rationai, but only limitedly so” (Simon 1961, p. xxiv). Insistence that the limited capacities of human agents have important organizationai ramifications distinguishes Simon’s work from that of the hyperrationality genre. But the absence of hyperrationality does not imply irrationality. On the contrary, the human agents with whom Simon is concerned are attempting effectively to cope. This is what intended rationaiity is ali about. To regard organizations as devices by which to economize on bounded rationaiity is thus suggested by this perspective and is central to the transaction cost approach.
Weick’s emphasis is rather different. He argues that decisions made by boundediy rational actors “will be made in terms of localized disturbances to which abbreviated analyses wili be applied, with short-term recomm endations as the result. A search for more stable solutions . . is unl ikely; consequences are not given much attention, and apparently logical solutions may prove faulty as their consequences ramify” (1969, p. 10). Accordingly, Weick treats cognitions as retrospective (1969, p. 30) and contends that environments are “enacted” (1969, p. 64). As March puts it, “Posterior rationaiity modeis maintain the idea that action shouid be consistent with preferences, but they conceive action as being antecedent to goais” (1978, p. 593).
lntended rationality and posterior rationality modeis have different org anizational design ramifications. Thus, whereas Simon recognizes hierarchy as a means by which to effect semidecomposabiiity, thereby to economize on bounded rationaiity and produce order out of organizational chaos (by
32 Another point of contact between transaction cost economics and the Laumann et ai. survey concerns the frequency diniension for describing transactions. As they point out, whether relationships are “episodic or highly recurrent” affects the way they are organired (1978, p. 465). Although recurrent transactions are emphasized throughout this paper frequency is expressly included in my discussion of governance structures elsewhere (1979b, pp. 246—54).

American Journal of Sociology
permitting operating and strategic decisions, for example, to be clearly distinguished), Weick gives little attention to rational organizational des ign. Unable to plan or prepare for contingencies, Weick’s organízations are given to myopic groping. Ex ante planning gives way to ex post rationalization.
Since the study of organizations can usefully be informed by both pers pectives, a forced choice between them is unnecessary and unwise. The question of concentrating research resources nevertheless needs to be faced. Inasmuch as our understanding of organizational anatomy is still primit ive, since the study of anatomy logically precedes pathology, and as transaction cost economizing is central to the design and assessment of governance structures, 1 urge that greater attention to anatomy—viewed through the lens of transaction cost reasoning—is indicated at this juncture.
The resource-dependency model sometimes makes reference to efficiency but more often relies on power in explaining organizational outcomes. Inasmuch as power is very poorly defined and hence can be used to expIam virtually anything, the tautological objection to resource-depend ency analysis is easily understood. Ready access to a power explanation has also had the unfortunate effect of removing efficiency analysis from center stage.
Thus consider Jeffrey Pfeffer’s assertion that if “the chief executive in a corporation always comes from marketing . . - there is a clue about power in the organization” (1978, p. 23). Viewed from a power perspective, the argument evidently is that the marketing people in this corporation have “possession of control over critical resources” (1978, p. 17), have prefe rential access to information (1978, p. 18), and are strategically located to cope with “critical organizational uncertainty” (1978, p. 28). 1 do not disagree with any of this, but would make the more straightforward argum ent that the marketing function in this organization is especially critical to competitive viability.
As Ouchi and 1 have argued elsewhere (1981), those parts of the ent erprise that are most critical to organizational viability will be assigned possession of control over critical resources, will have preferential access to information, and wiIl be dealing with critical organizational uncertaint ies. In some organizations this may be marketing, in others it may be R & D, and in still others it may be production. Indeed, we argue that failure to assign control to that part of the enterprise on which viability turns would contradict the efficiency hypothesis but would presumably be explained as a power outcome.

Transaction Cost Approach
Or consider the transformation of the merchant capitalist described by Glenn Porter and Harold Livesay. They report that during the first two centuries after the initial English settlement on the North American cont inent, “urban merchant capitalists . . . were the wealthiest, best informed, and most powerful segment of early American society” (1972, p. 6). These all-purpose merchants nevertheless gave way to specialized merchants early in the l9th century; such merchants then became “the most important men in the economy” (1972, p. 8). But specialized merchants in turn found their functions sharply cut back by the rise late in the 1800s of integrated manufacturers: “The long reign of the merchant had finally come to a dose. In many industries the manufacturer of goods had also become their distributor. A new economy dominated by the modern, int egrated manufacturing enterprise had arisen” (1972, p. 12).
Power theory must confront two troublesome facts in explaining these changes. First, why would the ali-purpose and later the spedalized merc hants ever permit economic activity to be organized in ways that would remove power from their control? Second, why did power leak out select ive2y—with the merchant role being appropriated extensively by some manufacturers but not by others? As discussed above and developed elsew here (Wifliamson 1980a), the transaction cost approach explains both in terms of efficiency. Perhaps power theory can sometimes add detail. However, until it has been much more carefully delimited—which, 1 subm it, wifl entail dimensionalizing—power theory, as an overall approach to the study of organizational change, is a pied piper whose enticements are better resisted in favor of more mundane efficiency considerations.
Transaction cost analysis is an interdisciplinary approach to the study of organizations that joins economics, organization theory, and aspects of contract law. li provides a unified interpretation for a disparate set of organizational phenomena. Although applications additional to those set out here have been made, the limits of transaction cost analysis have yet to be reached. Indeed, there is reason to believe that the surface has merely been scratched.
Transaction cost reasoning probably has greater relevance for studying commercial than noncommercial enterprise, since natural selection forces operate with greater assurance in the former. Transaction cost economizing
33For applications to organization form, see n. 4 above; for a discus5ion of oligopoly, see n. 29 above; natural monopoly is assessed in transaction cost terms in Williamson

Amer.ican Journai of Socioiogy
is nevertheless important to ali forms of organization. Accordingiy, Lhe foflowing proposition applies quite generally: governance structures that have better transaction cost economizing properties wiil eventualiy disp lace those that have worse, ceteris paribus. The cetera, however, are not always paria, whence the governance implications of transaction cost analys is wiII be incompleteiy realized in noncommercial enterprises in which transaction cost economizing entails the sacrifice of other vaiued objectives (of which power wiIi often be one; Lhe study of these trade-offs is an important topic on the future research agenda).
Certain methodologicai features of Lhe transaction cost approach should perhaps be made more explicit. Three are especially noteworthy. For one thing, the transaction cost approach employs functional analysis in the foiiowing sense: “Institutions are functional if reasonable men might create and maintain them in order to meet social needs or achieve social goais” (Simon 1978, p. 3).34 Second, the approach straddles Lhe methodol ogical dispute that separates maximizers and satisficers. Thus it relies on economizing arguments (which disciplines the anaiysis and appeals to macim izers) but substitutes comparative institutional for optimizing procedures (which is more in the spirit of satisftcing). Inasmuch as the assessment of discrete structurai alternatives can often be performed without “elabor ate mathematicai apparatus or marginal calculation” (Simon 1978, p. 6) and is furthermore entire)y adequate for many purposes, such an unp retentious approach to Lhe study of organizations has much to commend it. Third, as already noted, the transaction cost approach relies—in a somewhat informal, background, and long-run way—on Lhe operation of natural selection forces.
While it is injudicious to claim too much for Lhe transaction cost app roach, neither do 1 want to ciaim too little. At present, it is probably under- rather than overapplied to organization theory. In contrast with the highly microanalytic approach to the study of organizations, in which personalities and detailed organizationai procedures are scrutinized, and the highly aggregative approach to organizations employed in mainline econ omics, Lhe transaction cost approach employs a semimicroanalytic levei of analysis. This appears to be a levei of analysis at which sociologists and other students of organization enjoy a comparative advantage. Facility with the apparatus, however, requires that an irreducibie minimai investment in transaction cost reasoning be made. This paper attempts both to suppiy requisite background and to make substantive headway on some of the governance issues of common interest to economics, iaw, and sociology.
34 The only change that is necessary for my purposes is that “private or social” should be substituted for “social” in the two places where “social” appears lii this quotation.

Transaction Cost Approach
Alchian, Armen A., and Harold Dcmsetz. 1972. “Production, Jnformation Cosis, and Economic Organization.” Amencan Economk Revie,.v 62 (December): 777—95.
Aldrich, Howard E. 1979. Organizations and Environments. Englewood Cliffs, N.J.:
Arrow, Kenneth 3. 1969. “The Organization of Economic Activity.” The Analysis and Evalw.zlion of Ptsblic Expenditure: The PPB System. Joint Economic Comrnittce, 9lst Cong., lst sess., pp. 59—73.
llarnard, Chester 1. 1938. The Functions 01 the Executive. Cambridge, Masa.: Harvard University Press.
Baumol, William. 1959. Business Belsavior, Value, and Growtk. New Vork: Macmillan. Chandier, Alfred D., Jr. 1962. Strategy and Structure. New York: Doubleday.
1977. The Visible lland. Cambridge, Mass.: Harvard University Press.
Coase, Ronald H. (1937) 1952. “The Nature of lhe Firm.” Pp. 386—405 in Readings
in Prke Theory, edited by G. J. Stigier and K. E. Houiding. Homewood, III.: Irwin.
Comnions, John R. 1934. Institutional Economics. Madison: University of Wisconsin Press.
Cyert, Richard M., and James G. March. 1963. A Bekavioral Theory of the Firm. Engiewood Cliffs, N.J.: Prentice-Hail.
Diamond, Peter. 1971. “Political and Economic Evaluation of Social Effects and Ext ernalities: Comment.” Pp. 30—32 in Frontiers of Quantitative Economics, edited by M. Intrilligator. Amsterdam: North-Hoiland.
Doeringer, Peter, and Michaci Piore. 1971. Internal Labor Markets and Manpower Analysi.s. Lexington, Mass.: Heath.
Feiler, David E. 1973. “A General Theory of the Coliective Bargaining Agreement.” California Law Review 61 (May): 663—856.
Freeman, Richard B., and James L. Medoff. 1979. “The Two Faces of Unionisni.” Public Interest (Fali), pp. 69—93.
Fuiler, Lon L. 1964. The Morality oj Law. New Haven, Conn.: Vaie University Presa. Goffman, Erving. 1969. Strategk Interaction. Philadelphia: University of Pennsylv anla Presa.
Hannan, Michaci T., and John Freeman, 1977. “The Population Ecology of Organizat ions.” American Journal 01 Sociology 82 (March): 929—64.
Hayek, Friedrich. 1945. “The Use of Knowledge in Society.” American Economk Review 35 (September) 519—30.
Jensen, Michaei C., and William H. Meckling. 1976. “Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure.” Journai of Financial Econoinics 4 (October): 305—60.
Kanter, Rosabeth Moas. 1972. Community and Commitment. Cambridge, Mass, Harv ard University Presa.
Klein, Benjamin. 1980. “Transaction Cost Dcterminants of ‘Unfair’ Contractual Arr angements.” American Ecanomsc Review 70 (May): 356—62.
Kiein, Benjamin, Robert G. Crawford, and Armen A. Alchian. 1978. “Vertical Integ ration, Appropriabie Quasi-Rents, and the Competitive Contracting Process.” Journ a! of 1.aw and Economics 21 (October): 297—326.
Knight, Frank H. 1965. Risk, Uncertainty and Profit. Ncw York: Harper & Row. Laumann, Edward O., Joseph Galaskiewicz, and Peter V. Marsden. 1978. “Community Structure as Interorganizationa] Linkages.” Pp. 455—84 in Annuai Ret,iew of So. cioiogy. Vol. 4. New York: Annual Reviews.
Lifson, Thomas B. 1979. “An Emergent Administrative Systezn: Znterpersonal Netw orks in a Japanese General Trading Firm.” Working Paper 79-55, Harvard Univ ersity, Graduate School of Business.
Llewellyn, Karl N. 1931. “What Price Contract?—an Essay in Perspective.” Vale Law Journai 40 (May) 704—51.

American Journal of Sociology
Macaulay, Stewart. 1963. “Non-contractual Relations in Business.” American Sociol ogical Re,,iew 28: 55—70.
McKean, Roland. 1971. “The Vertical Integration of Production: Comment.” American Economk Review 61 (May): 124—26.
Macneil, lan R. 1974. “The Many Futures of Contract.” University of Southern Cal!j ornia Law Rcview 67 (May): 691—816.
Manuel, Frank E., and Fritzie P. Manuel. 1979. fltopian l’hought in Lhe Western World. ambridge, Mass.: Harvard Universlty Presa.
March, James G. 1973. “Model Bias in Social Action.” Review of Educational Res earck 42:413—29.
1978. “Bounded Rationality, Ambiguity, and thc Engineering of Choice.” Bel! Journal of Economics 9 (Autumn): 587—608.
March, James G., and Herbert A. Simon. 1958. Organizations. New York: Wiley. Marris, Robin. 1964. The Economic Theory o! “ManagerioJ” Capito.Zism. New York:
Free Presa.
Marschak, Jacob, and Roy Radner. 1972. The Theory 01 l’eoms. New Haven, Conn.:
Yale University Presa.
Mortenson, Dale T. 1978. “Specific Capital and Labor Turnover.” Bel! JournoJ oj Economias 9 (Autumn): 572—86.
Ouchi, William G. 1977. “Review of Markets and Hierarchies.” Administratite Science Quarterly 22 (September): 541—44.
1980a. “Efficient Boundaries.” Mimeographed. Los Angeles: University of Calif ornia, Los Angeles.
1980b. “Markcts, Burcaucracies, and Clans.” Administrative Science Quarterly
25 (March): 129—42.
Pfeffer, Jeffrey. 1978. Organi:ational Design. Northbrook, fli.: AHM. Porter, Glenn, and Harold C. Livesay. 1971. Merchanis and Manujacturers. Baltimore:
Johns Hopkins University Presa.
Radner, Roy. 1968. “Competitive Eguilibrium under Uncertainty.” Econometrica 36 (January): 31—58.
Simon, Herbert A. 1957. Mod.4s of Mau. New York: Wiley.
1961. Ad,ninistrativc Bekovior. 2d cd, New York: Macmillan.
1978. “Rationality as Process and as Product o! Thought.” American Econ omic Review 68 (May): 1—16.
Summers, Clyde. 1969. “Coilective Agrecments and the Law o! Contracts.” Yale Law Journal 78 (March): 525—75.
Teece, David. 1980. “Economies o! Scope and the Scope of the Enterprise.” Journal of Economic Behavior and Organization 1 (September): 223—48.
Thompson, James D. 1967. Orgonizations in Action. New York: McGraw-Hill. Weick, K. E. 1969. The Social Psychology oj Orgonizing. Reading, Masa.: AddisonW esley.
Williamson, Oliver E. 1964. Tke Economias oj Discretionary Behavior. Englewood Cliffs, N.J.: Prentice-Hail.
1970. Corporote Cont rol and Business Behavior. Englewood Cliffs, N.J.:
1971. “The Vertical Integration o! Production: Market Failure Considerat ions.” American Economia Review 61 (May): 112—23.
1975. Morkets and Hierarckies. New York: Free Presa.
1976. “Franchise Bidding for Natural Monopolies—in General and with Res pect to CATV.” Bel! Journol ol Economics 7 (Spring): 73—104.
1979a. “Assessing Vertical Market Restnctions: Antitrust Ramifications of tbe Transaction Cost Approach.” University o! Pennsylvonia Law Review 127 (April):
1979b. “Transaction Cost Economics: Thc Governance of Contractual Relat ions.” Journal oj Law and Economics 22 (October): 233—61.

Transaction Cost Approach
1980. “OrganizaUonal Jnnovation: The Transaction Cost Approach.” Discuss ion Paper no. 83, Center for the Study of Organizational Innovation, Univcrsity of Pennsylvania, September.
1981. “The Modern Corporation: Origins, Evolution, Attributes.” Journal o! Economic Literature, in press.
Williamson, Oliver E., and William G. Ouchi. 1981. “The Markets and Hierarchies Program of Research: Origins, Implications, Prospects.” In Organizational Design, edited by WiUiam Joyce and Andrew Van de Ven. Ncw York: Wiley.
rjl1jamson, Otiver E., Mlchael L. Wachter, and Jeffrey E. Harris. 1975. “Understandi ng the Employment Relation: The Analysis of Idiosyncratic Exchange.” Bel Journal oj Economics 6 (Sprlng): 250—80.
Williamson, Scott R. 1980. “A Selective History of the US. Labor Movcment.” B.A. thesis, Yale University.

Nenhum comentário:

Postar um comentário