The discussion of technical innovation begins with an examination of the conventional dichotomy. Although this requires that a good deal of old material be covered, it seems essential that this matter be put to rest at the outset. This completed, a series of recent refinements to the technical innovation-market structure issue are then considered in Section 2.
However complex the subject matter under consideration may appear to be, there is always a possibility that the underlying structure is fundamentally simple. The trick is to cut through to the heart of the matter. Once this is discovered, all else falls easily into place. What I characterize here as the conventional dichotomy — that, depending on one’s lights, the market structure most conducive to technical progress involves large size and monopoly power or, alternatively, small size and competition — aspires to do this. Unfortunately, the effort in this instance must be judged a failure.
1. The Arguments
Relations between market structure and economic performance are rarely simple, albeit that straightforward applications of price theory sometimes suggest otherwise. This is especially true when the performance measure in question is technological progress. Although Arrow has shown that the incentive to innovate is greater under competition than under monopoly (1962, pp. 619-622), this assumes that the appropriability of the potential gains attributable to the innovation is not significantly greater under monopoly. Also, Demsetz argues that Arrow’s demonstrated incentive differential fails when appropriate standardization techniques are employed (1969, pp. 16-19). Scherer’s more recent study (1967) of the effects of rivalry on R & D, mainly in the context of a duopoly model, concludes that rivalry generally favors R & D—but even this is qualified and, considering the special nature of the model, scarcely dispositive. Kamien and Schwartz (1974) are intrigued by the possibility that there exists an optimum degree of rivalry. They show that an intermediate degree of competition appears often to promote a faster development rate, but the relation between their competition variable and market structure is unclear and the assumptions of the model are rather restrictive in any case.
The upshot is that there does not appear to be a great deal of useful policy advice to be gleaned from existing theoretical studies of the relations linking market structure with technical progress.2 Consider, therefore, the conjectures offered by interested observers.
I count here as potentially relevant those observations relating market structure to technical progress that have occurred since antitrust first became a social issue. Consider in this connection the statement of Marshall who, speaking in 1890, expressed the view that differential progressiveness was “the main reason for regarding with some uneasiness any tendency . . . towards [the] consolidations of business” (Pigou, 1956, p. 279). He went on to observe that while the employment of scientific specialists may occasionally place the large firm at a technical advantage, “these advantages count for little in the long run in comparison with the superior inventive force of a multitude of small undertakers” (Pigou, 1956, p. 280). In 1897 Hadley took a similar position (1897, p. 383):
The tendency of monopoly to retard the introduction of industrial improvement is … a more serious thing than its tendency to allow unfair rates. This aspect of the matter has hardly received proper attention. W e have been so accustomed to think of competition as a regulator of prices that we have lost sight of its equally important function as a stimulus to efficiency. Wherever competition is absent, there is a disposition to rest content with old methods, not to say slack ones. In spite of notable exceptions this is clearly the rule.
Subsequent developments in competitive theory were, as Stigler points out, much concerned with the refinement of static efficiency analysis (1956, pp. 270-271). Attention to the progressiveness dimension correspondingly wanted. Partly as a reaction to this almost exclusive focus on static resource allocation, and partly as a result of his own interest in economic development, Schumpeter countered with what might be characterized as the alternative hypothesis. Like Hadley, he took the position that the kind of competition that counts is attributable to “the new commodity, the new technology, the new source of supply, the new type of organization (the largest-scale unit of control for instance) — competition which commands a decisive cost or quality advantage and which strikes not at the margins of the profits and the outputs of the existing firms but at their foundations and their very lives” (1942, p. 84). Unlike Hadley, however, this emphasis on dynamics led Schumpeter to conclude that it is transient monopoly, not competition, that is a prerequisite to progress (1942, Chap. 8).
Neither Hadley nor Schumpeter was especially precise regarding a specification of the monopolistic condition; Schumpeter was especially ambiguous in this respect. As Stigler notes, Schumpeter regards “every departure from perfect competition in a stationary economy … [as| monopoly, and this is why it is so easy to show that monopoly is necessary to progress” (1956, p. 272). Thus, while restoring attention to the progressiveness dimension in a useful way, Schumpeter left the matter of the optimum organization of industry somewhat unclear.
In what might be characterized as the neo-Schumpeterian position, this structural ambiguity is largely overcome: large size and structural monopoly are said to be most conducive to technical progress. Galbraith expresses it as follows (1952, p. 91; emphasis added);
… a benign Providence*. . . has made the modern industry of a few large firms an almost perfect instrument for inducing technical change. It is admirably equipped for financing technical development. Its organization provides strong incentives for undertaking development and putting it into effect….
There is no more pleasant fiction than that technical change is the product of the matchless ingenuity of the small man forced by competition to employ his wits to better his neighbor. Unhappily, it is a fiction. Technical development has long since become the preserve of the scientist and the engineer. Most of the cheap and simple inventions have, to put it bluntly, been made.
This view, hereinafter referred to as the Galbraithian hypothesis, has been expressed (both before and since) in a variety of ways [see the citations in Jewkes, Sawers, and Stillerman (1959, pp. 29-31)]. It is, in my experience, the one that prevailed among most undergraduates in the 1960’s. One might characterize it as a truth that every schoolboy knows; it may even qualify as the conventional wisdom. But this may be attributable more to the effectiveness of the media (“progress is our most important product”; “better things for better living. . . through chemistry,” and so forth) than to the empirical (nonfictional) accuracy of the argument.
2. The Evidence
I deal here strictly with the matters of how R & D expenditures vary in relation to large size and with the influence of industrial concentration on innovative performance. Large size does not necessarily imply monopoly power, and in this sense might be distinguished from a pure monopoly theory of innovation. There is, however, a tendency for large absolute size to be associated with large relative size, and the Galbraithian position seems to be that large absolute and relative size, individually and in combination, can be expected to yield greater innovative performance. Consider the evidence on absolute size as it bears on the conventional dichotomy first.
Based on data reported by the National Science Foundation, which con- siders three size classes of firms (below 1,000 employees, 1,000-5,000 employees, and over 5,000 employees), there is a clear tendency for firms in the largest of these three-size classes to spend more on R & D as a percentage of sales (Nelson, Peck, andKalachek, 1967, p. 67). This is, however, an insufficiently fine basis for classification: virtually all the firms in the Fortune 500 largest industrials series had over 5,000 employees in 1968; General Motors numbered its employees at 750,000 in that year. A breakdown within the over 5,000 category is thus indicated if the absolute size question is to be dealt with in policy relevant terms. Hamberg’s investigation of the relation between R & D intensity and firm size among 387 firms selected from the Fortune list of the 500 largest industrials in 1960 is useful for this purpose. Taking these corporations as a group and correlating the ratio of R & D employment/total employment against size measured as total employment, a positive, significant correlation was obtained. When total assets are used to measure size, however, a negative, insignificant result was reported (1966, p. 58).
A major problem in interpreting these results nevertheless arises in that the technological potential to innovate differs greatly across industries. Therefore, Hamberg distributed these firms across seventeen two- and three- digit industry groups. He then investigated the elasticity of R & D employment in relation to size (measured as total employment or total assets in each industry) by obtaining least-squares estimates to the regression equation (where the subscripts i, j refer to the firm and industry respectively, Y is R & D employment, X measures size, and the estimated value of Bj is the relevant elasticity). Elasticity values greater than unity favor the Galbraithian hypothesis, while values less than unity favor the competitive version. Hamberg generally obtained elasticity values greater than one, although often these were not statistically significant (1966, pp. 60-63). Overall, he concluded, “a case can be made for the hypothesis that research intensity . . . increases with size among the larger firms in but three industries” (1966, p. 61).
Comanor performed a roughly parallel set of tests using a somewhat finer industry classification than Hamberg (1967, pp. 640-643). Among the twenty- one industries in his sample, six had elasticities that exceeded unity (but none was statistically significant) while fifteen had elasticities less than unity (of which seven were statistically significant). He attributed the differences in his results and Hamberg’s mainly to the broader industry classifications used by Hamberg (1967, p. 641). The true relation, apparently, is sufficiently close on this issue that modest changes in industry definitions yield different results.
Mansfield approached the matter by posing a somewhat different question. He examined R & D expenditures (expressed as a percent of sales) of the very largest firms in relation to their large but not giant-sized rivals in the chemical, petroleum, drug, steel, and glass industries. “Except for the chemical industry, the results provide no evidence that the largest firms in these industries spent more on R & D, relative to sales, than did somewhat smaller firms. In the petroleum, drug, and glass industries, the largest firms spent significantly less; in the steel industry, they spent less but the difference was not statistically significant” (1968, pp. 94-95). Scherer also found that, among a wider class of industries, the R & D intensity of the largest firms did not usually exceed that of their upper middle-sized rivals. He observed that the data do not support the contention that “gigantic scale is … an essential condition for vigorous industrial research and development activity: . . . bigness may indeed be a stifling factor” (1965, p. 265).
Roughly, these results are displayed graphically in Figure 8, where R & D intensity is plotted on the ordinate and relative (not absolute, since the comparisons are strictly intra-industry) firm size on the abscissa. Assuming that R & D expenditures experience constant returns to scale, one could, based on the relations shown in Figure 8, make a case for the proposition that the relative contribution to progressiveness is greatest among upper middle- sized firms. Giant size, however, is not obviously warranted on this criterion and might instead be actively resisted.
Consider now the influence of industrial concentration, expressed gener- ally as a four-firm concentration ratio, on R & D expenditures. Hamberg (1966, Chap. 4) and Horowitz (1962) reported a positive correlation between R & D expenditures and industrial concentration. Scherer discovered a much weaker but slightly positive association (1965, pp. 1119-1121). Kendrick concluded from an examination of Terleckyj’s data that there is no significant correlation between productivity changes and industrial con centration (1961, p. 170). Stigler detected “hints that industries with lower concentration had higher rates of technological progress” (1956, p. 278), while I, using Mansfield’s data, found a negative correlation between the proportion of innovations introduced by the four largest firms and industrial concentration (1965). (The last is strictly a small-numbers result and allows for productivity as well as expenditure effects.)
Mansfield’s is probably the best balanced view of this matter. He con- sidered diffusion (the rate at which an innovation, once introduced, is adopted by other firms in the industry) as well as the proclivity to innovate. He fould that while greater concentration may be associated with a lower rate of diffusion (1968, p. 217), overall — except, possibly, for innovations that require a large amount of capital — “there is no statistically significant relationship between an industry’s concentration and its estimated rate of technological change” (1968, p. 245).
I conclude, as did Nelson, Peck, and Kalachek, that the Galbraithian position “is somewhat exaggerated” (1967, p. 67), or as Mansfield put it, the “available evidence does not seem to support this hypothesis” (1968, p. 245). This does not, however, imply that the competitive alternative is endorsed. Progressiveness performance is too complex to be adequately characterized by either of these polar extremes. At the very least, important exceptions to the proposition that competitive market structures favor progressiveness need to be admitted.
Source: Williamson Oliver E. (1975), Markets and hierarchies: Analysis and antitrust implications, A Study in the Economics of Internal Organization, The Free Press.