Market structure: A Systems Approach

While there are important exceptions, the literature relating technical innovation to firm size and industry structure yields the following average tendencies: R & D expenditures (expressed in relation to firm size) are not usually greater and are often lower for the very largest firms in an industry by comparison with its large but somewhat smaller rivals; the productivity of research expenditures follows roughly the same pattern; industrial R & D is, preponderately, applied in nature, with a few large firms and a few industries offering partial exceptions; research conducted in most large industrial laboratories favors minor improvement inventions rather than major new inventions; the short-reach nature of most industrial research (military and space research and some chemical or other continuous process research con- stituting principal exceptions) does not require large-scale support.

Suppose, as a policy matter, that one is instructed to select a program that will enhance technical progressiveness and that the alternatives are either to encourage or discourage increases in the absolute and relative size of the largest firms in every industry. If the foregoing review is an accurate assessment of the situation (which assumes that the studies are unbiased), a preference for the giant-size discouragement program would not seem injudicious.

One might object that the policy options postulated are unnecessarily constrained. The survey of the evidence in Sections 1 and 2 describes central tendencies, but there are notable exceptions. In particular, the science based industries may warrant individual treatment. Considering that half of the research and development in the United States is publicly financed and that much of it goes to this science based subset, using budget money rather than antitrust policy as a means by which to achieve technical objectives in science based firms is both more direct and, arguably, more effective.

Suppose that science based exceptions were made. One might then object that the policy options are underspecified with respect to other industries. What kinds of size encouragement and discouragement programs are contemplated? Since radical changes can be seriously disruptive, assume that only gradualist measures are intended; side effects attributable to extreme changes are thereby avoided. A more subtle objection might then be registered: the above review tends to regard the technical innovation question as a matter of optimum firm size, while in fact the innovation process ought to be more broadly conceived.

The last point is developed below with the assistance of a rudimentary statement of a systems approach to technical innovation. Possible breakdowns in the system are also considered. The argument is an application of market and organizational failures analysis to the special problems that technical innovation poses.

1. General 

An efficient innovation system (for devising, developing, producing, and distributing new products and services) is the assigned objective. Although it is possible that innovation is an indecomposable process, it seems not implausible that a piecewise development by stages is feasible. The tendency in the technical innovation literature to focus on inputs and early stages of the innovation process has perhaps discouraged the development of a broader viewpoint. As Stigler has emphasized, however, invention is merely the first in a series of uncertain stages in the innovation process,132 and the necessary expertise in bringing the process to completion (which includes not merely invention but also experimental development, market testing, and commercial production and distribution) may differ between stages (Stigler, 1956, pp. 281-82). The necessary judgment, frequently, is not how to proceed but whether to proceed. A critical rejection mechanism for developments that lack essential marketability is needed as well as a capacity to bring feasible projects efficiently to completion.

The innovation process can conveniently, if arbitrarily, be divided into three stages: invention, development, and final supply.133 Corresponding to each is a decision process that may be designated, respectively, as proposal, selection, and a composite process involving coordinated production and distribution. Assuming that innovation can meaningfully be decomposed in this way, the question to be addressed is whether, despite decomposability of the process, a single size or form of organization has optimum properties with respect to all stages. Alternatively, is there a transfer process that permits individual firms to specialize according to characteristic strengths, thereby to realize a larger total yield than would be possible were each required to take the process unassisted to completion? More generally, are there circumstances in which transfer is both feasible and advantageous but others which favor internal—that is, administrative (intrafirm) rather than market (interfirm) — processes instead?

The frequency with which transfer is observed to occur (Hamberg, 1966, p. 162, Jewkesetal., 1959, pp. 168,186, Mansfield, 1968, p. 58) is, presumably, evidence favorable to the interfirm specialization argument. This raises the possibility that an optimum system need not include among its part any firms which, when the R & D process is regarded strictly in intrafirm terms (as is common, if not prevailing), would be considered individually optimal. Yet, many innovations do not go the transfer process route. This raises the question of what factors explain the coexistence of these two processes.

Among the conventional explanations for internal development are the following:

  1. Efficient final supply for many innovations does not require great size. If small firms can easily bring innovations to completion unassisted, transfer is unnecessary.
  2. The need for improvement, especially process improvement, in- novations is apt to be more evident in firms that are already engaged in final supply. In many cases these will be the larger firms. To the extent that close cooperation with production and/or marketing personnel facilitates the accomplishment of such innovation activity, large firms will tend naturally to perform in-house research and development on these items.
  3. Large firms may be induced for defensive purposes to support some minimum internal R & D effort; reliance on transfer is too risky.

Often, it is alleged, these defensive incentives are greatest among large firms in oligopolistic industries.

So much for internal innovation. What explains transfer? If large firms can do everything that small firms can do and more, why should any inno- vations go the transfer process route? Probably the simplest explanation consistent with the “replication plus” argument is that small firms may, occasionally, strictly as a matter of chance, stumble upon an innovation for which they lack the capability to bring to completion. Transfer is dictated on this account.

But while such considerations permit the coexistence of internal and transfer processes to be qualitatively reconciled, the reported volume of transfer activity is too large to be accounted for so easily. Also, other reasons can be advanced for expecting innovation to follow the internal development route. Both organizational and market failures need to be examined in these connections.

Consider the three stage innovation process described (invention, development, efficient final supply) and isolate the stage at which the greatest size requirements are experienced. Assume, as is usually the case, that efficient final supply is the critical stage for determining minimum efficient scale. Does a firm of requisite final supply-size experience any disabilities in attempting itself to perform early stage functions as well ?

Approaching the problem in this way focuses on what may be charac- terized as organizational failures. If these exist, additional reasons to expect the transfer process to operate may be discovered. A symmetrical treatment of these matters would also recognize the possibility of market failures. Are there factors that inhibit the transfer process from operating efficiently ?

2. Organizational Limits of the Large Firm 

As noted, economies of scale in production and distribution will, fre- quently, make large (but not often giant) size the preferred structure for efficient final supply. Large size may not, however, be necessary at either of the two earlier stages. The question that is addressed here is whether there are diseconomies attributable to defective decision-making and/or incentive processes that operate at the proposal and selection stages. Put differently, what prevents the large firm from replicating everything a series of small firms could perform and more ?

The organizational limits of the large firm that are described below serve as affirmative incentives for the transfer process to operate. The argument assumes that there are no intermediate forms of organization in which large irms become part owners of small firms and in other respects provide management or technical assistance to small ventures. This not only simpli- fies the exposition but, for the most part, seems roughly to correspond with prevailing practices. (A recent organizational innovation at General Elec- tric, called the “Technical Ventures Operation,” represents a radical departure from this norm. It is discussed separately in Section 4.5, below. For the purposes of this section and the two sections which follow, I restrict the discussion to autonomous large- and small-firm operations.)


A general discussion of size-related and intrinsic limitational factors of internal organization is given in Chapter 7 and need not be repeated here. There are, however, particular limitations of internal organization that bear on technical innovation that are not included in the earlier discussion. Of special interest is the history dependency of a firm, which constrains the options available to it at any point in time. Among the “life-cycle” attributes of large organizations noted by Downs in his study of bureaucracy134 are the following aging phenomena (1967, pp. 18-20, 96-101) :

  1. As bureaus grow older they encounter more situations and develop institutionalized rules for dealing with them when these Al- though this may have desirable efficiency consequences in many circumstances, in others it serves to limit the bureau’s range of response and, hence, its innovativeness.
  2. When a bureau is first created, it is usually dominated by Zealots, however, tend not to be good administrators. As the bureau grows older and the need for efficient administration increases re- latively, the zealots (who tend to be innovators) are displaced by administrators with more conservative orientations.
  3. Rapid rates of growth are difficult to sustain as absolute size becomes Thus, while successful small organizations in which growth is rapid offer numerous opportunities for individual advancement, promotion is less easy in the large bureaucracy. The more enterprising members of the large bureaucracy may thus be induced to locate elsewhere.

These effects are summarized in Downs’ Law of Increasing Conservatism: “All organizations tend to become more conservative as they get older, unless they experience periods of very rapid growth or internal turnover” (1967, p. 20). Since innovation, in relation to production, tends to be untidy, innovation — which is a poorly structured, high-risk activity — may not be an activity which the large, mature bureaucracy is constitutionally well-suited to handle.


The limitations of the large firm with respect to venture capital noted in Section 1, above might also be mentioned in this respect. Although these limitations may be more than offset by capital access advantages from more conventional sources, the point is that outside venture capital may be spe- cifically earmarked for investment in high-risk inventive activities for which investor appropriability is substantial. Large firms, as usually constituted, are not calculated to attract such sources of funds. In circumstances where high mean-variance outcomes are favored, the risk pooling which the large (and particularly the diversified) corporation provides can simultaneously impair the investment attractiveness of these firms to venture capital suppliers.


In addition to the reasons given above, which relate mainly to the limit- ations of the large firm in relation to the proposal process, the large firm may also experience shortcomings with respect to its selection properties. Its error admission characteristics may be defective. From the standpoint of the career prospects of a bureaucrat, fixed costs are not sunk but need to be justified. As noted in Chapter 7, the decision to proceed frequently becomes a commitment to succeed, whatever the costs. Sequential decisionmaking procedures designed to permit project review on the merits may be overwhelmed by partisan appeals. A tendency to persist beyond judicious cutoff limits easily results.


Bureaucracies generally maintain close correspondence between hierarchical position and compensation. Such a relation can be predicted on sociological grounds (Simon, 1957) and, where the management of proven resources is involved, on economic grounds as well (Mayer, 1960). This may even suffice for some types of innovation, as the views of Daniel P. Barnard (research coordinator at Standard Oil of Indiana) cited in Section 2.2, above appear to suggest. But such incentives are poorly suited to satisfy the entrepreneurial appetites of individuals who are prepared to risk their personal savings and careers in pursuit of big stakes.

Were it that large firms could compensate internal entrepreneurial activity in ways approximating that of the market, the large firm need experience no disadvantage in entrepreneurial respects. Violating the congruency between hierarchical position and compensation appears to generate bureaucratic strains,136 however, and is greatly complicated by the problem of accurately imputing causality. Where did the idea originate who discerned its commercial importance, who refined it, and so forth? Since sorting this out is inordinately difficult,137 distributing rewards among those who participated will have highly arbitrary and possibly counterproductive qualities (in that it may elicit resentment and noncooperation on future ventures). Where, instead, innovations are successfully developed by small firms, the rewards accrue to those who have expressly assumed monetary risks and who have removed their careers from bureaucratic promotion ladders. Although the resulting imputations may sometimes be arbitrary,138 they do not occasion bureaucratic dysfunctions of the types described above.


In principle, the large firm should constitute a superior instrument with which to conduct parallel R &D in comparison with a series of small firms. The reason for this is that it affords an opportunity to realize efficient information exchange and thereby eliminate inferior projects and wasteful duplication at an early point in time (Arrow, 1962b, p. 356). Markets would be vulnerable to opportunism were they to be used for such information exchange purposes. The difficulty is this: The value of information to a purchaser (or group of purchasers) isnot known until he has the information, but once it is disclosed he has acquired it without cost (Arrow, 1962a, p. 615). Were it possible to extract self-enforcing commitments to the effect that “I promise to pay full valuation for all information disclosed,” no problem would exist. But such promises are idly made. The seller, by reason of information impactedness (he does not know the true value to the buyer and can develop this only at great expense, if at all), is unable to enforce such terms.

Often, however, and perhaps usually, the large firm appears not to be able to support a genuine parallel R & D effort (Peck, 1962, p. 294). Its planning propensities may preclude this; however attractive adaptive, sequential decision-making may be in principle, bureaucrats frequently display a preference for comprehensive planning instead (Schlesinger, 1967, p. 189). Partly, this may reflect a lack of appreciation for the experimental approach. But it may also be that it is difficult to assess internal competition on the mertis in a fully dispassionate way. Groups that are associaied with projects that are to be phased out in favor of others are prone to press for project continuation and marshall political support on their behalf. (The discussions of project persistence biases and internal reciprocity arrangements in Chapter 7 are relevant in these connections.)

An organizational dilemma is thereby posed. The internal distortion biases of large firms which themselves sponsor parallel R & D projects are avoided by going to the market, but efficient information-sharing can hardly be expected among a series of small autonomous firms, each pushing its own R & D approach. An uncertain verdict is thus reached in assessing alternative modes of organization in parallel R & D respects.

3. Market Impediments to Transfer 

Some of the leading reasons why large firms are poorly suited to sponsor early stage entrepreneurial activity have been sketched above. Some of the principal market impediments to transfer are considered here.


Stigler has observed: “It is only to be expected that, when a new kind of research develops, at first it will be conducted chiefly as an ancillary activity by existing firms . . . [but eventually] we may expect the rapid expansion of the specialized research laboratory, selling its services generally. These specialized laboratories need not be in the least inferior to captive laboratories (1956, p. 281). But parity in the technical sense to which Stigler refers does not assure parity in a contractual sense. Assigning R & D to outside research organizations may, because of inherent task uncertainty, pose risk-sharing problems which, contractually, are not easily resolved. The cost-plus contract is an example.

The moral hazard problems posed by cost-plus contracts are akin to those discussed in conjuction with the insurance illustration in Chapter 1; the pairing of opportunism with information impactedness again presents the difficulty. Suffice it to observe that where such problems exist there is no clean market solution: a cost-plus contract assigns the risk to the sponsoring agency (which may be well-suited to bear it) but undermines least-cost incentives, while a fixed price contract maintains incentives but shifts the risk in what may be an inefficient (high cost) way.

At least occasionally, vertical integration backward into research is the most attractive way to overcome the dilemma posed when high-risk programs are to be performed: the sponsoring firm (agency) assumes the risk itself and assigns the task to an internal research group. It essentially writes a cost-plus contract for internal development. That this does not have the debilitating incentive consequences that often result when similar contracts are given to outside developers is attributable to differences in the incentive and compliance machinery: managers are employees, rather than “inside contractors” (of the sort described in Chapter 6), and thus are unable to appropriate individual profit streams; also the internal compliance machinery to which the firm (agency) has access is vastly superior to and more delicately conceived than the policing machinery that prevails between organizations. Internal organization thus arises in part because of its superior properties in moral hazard respects.


Considerations of market thinness reinforce the need to be sensitive to tradeoff considerations. If the market to which innovations can be sold is, by reason of high concentration, thin, it may be essential, in order to maintain the bargaining position and hence the incentives of the innovators, that new entry and subsequent expansion be a feasible route by which to bring the process to completion. Lacking the prospect of entering independently into the industry to which the innovation applies, the innovator runs the risk that bids will fail to reflect full valuations — perhaps significantly.

Established firms face a tradeoff between the gains of inhibiting entry, in order to increase current profits and maintain their respective market positions, and the advantages of participating (in a systems sense) in maintaining the flow of new product developments. Inasmuch as the latter involves both uncertainty and future rather than immediate returns, established firms may sometimes opt for an entry inhibiting policy. The social interest, however, would appear generally to be served by preserving easy access.

Access difficulties into industries such as automobiles and basic metals illustrate these difficulties. In the former case, refusal by the assemblers to adopt an innovation can, whatever its merits, stand as a bar to acceptance. In the latter, innovations are apt to involve process changes, the adoption of which requires incorporation into a continuous process operation. If entry is not a feasible route by which to achieve adoption and if resistance to change is contemplated, outside parties will have less incentive to support as much inventive activity as they otherwise would. Relevant in this connection is Arrow’s expressed concern over why patent royalties in the aluminum, petroleum, and chemical industries are so low. “It really calls for some explanation, why the firm that has developed the knowledge cannot demand a greater share of the resulting profits — ideally all except a competitive return on the capital invested” (1962b, p. 355).

I submit that first-mover advantages, including an established market position, in conjunction with market thinness provides a partial explanation for this condition. Scherer notes that “speed of penetration … is perhaps the most important single advantage enjoyed by large firms in developing new products” (1970, p. 353). The innovating firm that does not already have an established market position is thus at a relative disadvantage in this respect. Its fallback, and hence its bargaining position, is thereby weakened; a refusal to sell to less credible. If, in addition, the industry to which it would sell the innovation is one of small numbers, elicited bids may fail to reflect full valuation.140 As a consequence, the inventive incentives among independents and small firms are predictably impaired, with the result that more of the burden of conducting research is shifted onto the large established firms — which, for reasons given earlier, may be constitutionally less well-suited to perform inventive activity.


Consider finally the possibility that certain high-talent, scientific personnel possess a degree of monopoly power in bargaining over their employment conditions. As Marschak has forcefully observed, such uniqueness conditions plainly exist and sometimes have interesting economic consequences (1968, p. 14). Its relevance here is that such researchers may sometimes insist, as part of the inducements-contribution package, that some fraction of their time be available for “undirected” research activity. The large firm that accedes to such stipulations will occasionally find itself originating projects that it would otherwise leave for others to develop.

None of this is to deny that the prospect of appropriating the monopoly gains associated with a patentable invention provides an incentive for firms of any size to sponsor early stage R & D. Presumably, however, the risk adjusted rates of return to R & D are brought into correspondence at the margin with other types of investments, and the question is what firms in what market circumstances enjoy a relative advantage and why. An evaluation of market failures on the one hand and organizational failures on the other is necessary for this purpose.

4. Systems Solution by Classical Specialization

I am inclined to regard the early stage innovative disabilities of large size as serious and propose the following hypothesis: An efficient procedure by which to introduce new products is for the initial development and market testing to be performed by independent inventors and small firms (perhaps new entrants) in an industry, the successful developments then to be acquired, possibly through licensing or merger, for subsequent marketing by a large multidivision enterprise. I do not, however, mean to suggest that this is the only efficient process. As indicated above (perhaps especially with respect to improvement innovations), there are others; also, market failures, in particular instances, discourage stagewise specialization.

The argument is a simple extension of classical specialization arguments to include technical progress within the framework of systems analysis. In consideration of the disabilities that different forms of organization experience at different stages of the innovation process, and in view of the empirical results regarding research proclivities of large firms reported in Sections 1 and 2, the established multidivision form organization that follows a conscious policy of imitation or acquisition as an important part of its new product strategy (supplemented, perhaps, by a set of internal operating rules designed to check its worst tendencies to persist with own-projects for which merit is objectively lacking) may be judged to display rationality of a high order in its allocation of resources to R & D. Put differently, a division of effort between the new product innovation process on the one hand, and the management of proven resources on the other may well be efficient. The frequency with which transfer is observed to occur (Hamberg, 1966, p. 21; Jewkes et al., 1959, pp. 168, 186; Mansfield, 1968, p. 58) plainly adds credence to the hypothesis.

5. Systems Solution by Organizational Innovation

Conceivably the perceptive large firm can attempt to debureaucratize itself and overcome certain of its worst tendencies. Replicating the characteristics of the small firm is not, however, apt to be easy. (The task force approach to projects may have distinct advantages for this purpose, but it is unlikely to be a perfect substitute. Truly entrepreneurial types find neither the latitude nor the rewards of this approach to innovation sufficient.) In comparison with the costs of going to the market and utilizing the transfer process, intensive efforts at debureaucratization are apt to be relatively expensive. If, naturally, organizations experience certain (mainly irreversible) life-cycle changes, it may be more economical simply to acknowledge these and permit specialization by stages to occur rather than force innovation through the internal development route.

The recent development by General Electric of what it refers to as its “Technical Ventures Operation” (TVO) is sufficiently interesting to warrant separate attention. This organizational innovation was originated in 1970 and can hardly be said to be proven. It is furthermore limited to a small subset of G.E.’s new project ventures. Nevertheless, TVO represents an imaginative effort to join the advantage of large and small firms.

The problem for which the TVO was devised to handle is that G.E. found itself developing technologies that were potentially viable but the markets for which were too small in relation to G.E.’s “immense size and method of operating (Sabin, 1973, p. 145). G.E. sometimes attempted to sell the unwanted technologies, but frequently found no “suitable” buyers, while the sale of other developments prevented it from “keeping abreast of a business in which it | had] a scientific interest” (Sabin, 1973. p. 145) and for which knowhow was evidently important.

The possibility of setting up separate but wholly owned new ventures was rejected in favor of a scheme in which G.E. sets up a new entity to which the physical assets as well as the patents or a license to use the technology are transferred. G.E. then “takes about a one-third interest in the new business, and the ramaining two-thirds are split between the scientists and managers who help launch the new company, on the one hand, and outside investors on the other” (Sabin, 1973. p. 147). The rationale for this is, especially in view of the discussion in earlier sections of this chapter, altogether fascinating. According to David J. BenDaniel, who heads the TVO (Sabin 1973, p. 147):

A subsidiary arrangement would still leave the managers bogged down in the hierarchical reporting procedures required by large corporations. Even more important, by giving the new company’s managers a direct ownership stake, G.E. supplies an incentive for performance that can’t be equaled within a large corporation. The managers must invest in the venture themselves. . . .

The rationale for raising outside capital has several aspects. For one thing, any such venture generally needs more financial backing than the entrepreneurs can give it (and G.E. doesn’t want to pour more capital into it). BenDaniel also considers the raising of outside capital a test of the managers’ entrepreneurial mettle (though he usually helps them). And. finally, the whole capital-raising procedure subjects the venture to independent scrutiny by outside businessmen who are unlikely to have their judgment swayed by sentimental attachment to the technology.

The advantages of TVO in relation to G.E. for the projects in question thus reduce to the following: hierarchical communications are reduced; entrepreneurial incentives unavailable within the large corporation are introduced; and commitment biases which would otherwise appear are checked. Plainly TVO and related organizational innovations that are designed to mitigate the bureaucratic distortions to which large firms are subject bears watching.

Source: Williamson Oliver E. (1975), Markets and hierarchies: Analysis and antitrust implications, A Study in the Economics of Internal Organization, The Free Press.

Leave a Reply

Your email address will not be published. Required fields are marked *