Frank Knight made early reference to the limitations to firm size puzzle when, in 1921, he observed that the “diminishing returns to management is a subject often referred to in economic literature, but in regard to which there is a dearth of scientific discussion” (1965, p. 286, n. 1). And in 1933 he elaborated as follows:
The relation between efficiency and size of firm is one of the most serious problems of theory, being, in contrast with the relation for a plant, largely a matter of personality and historical accident rather than of intelligible general principles. But the question is peculiarly vital, because the possibility of monopoly gain offers a powerful incentive to continuous and unlimited expansion of the firm, which force must be offset by some equally powerful one making for decreased efficiency. [1965, p.xxiii; emphasis in original]
Tracy Lewis’s recent remarks that large established firms will always realize greater value from inputs than small potential entrants are apposite:
The reason is that the leader can at least use4he input exactly as the entrant would have used it, and earn the same profits as the entrant. But typically, the leader can improve on this by coordinating production from his new and existing inputs.
Hence the new input will be valued more by the dominant firm. [1983, p.1092 emphasis added]
If the dominant firm can use the input in exactly the same way as the smaller entrant, then the larger firm can do everything the smaller firm could. If it can improve on the input usage, it can do more. Accordingly, industries are not everywhere organized aS monopolies solely because of public policy vigilance and restraints.
A different way of posing the issue is in terms not of horizontal but of vertical integration. Thus Ronald Coase inquired, “Why does the entrepreneur not organize one less transaction or one more?” (1952, p. 339). More generally the issue is, “Why is not all production carried on in one big firm?” (p. 340).
Various answers have been advanced. They all suffer, however, from a failure to adopt and maintain the relevant comparative institutional standard. Thus consider Knight’s response: “The question of diminishing returns from entrepreneurship is really a matter of the amount of uncertainty present. To imagine that a man could adequately manage a business enterprise of indefinite size and complexity is to imagine a situation in which effective uncertainty is entirely absent” (1965, pp. 286-87). In effect, Knight attributes limitations upon entrepreneurship to a condition of bounded rationality. As uncertainty increases, problems of organization become increasingly complex, and bounds on cognitive competence are reached. But he does not address the issues in a genuinely comparative manner.
Thus suppose that two firms are competing. In principle, net gains ought always to be available by merging the two. Economies of scale can be more fully exploited. Certain overhead and rivalry expenses can be curtailed. And product prices may improve—at least temporarily. Joining the two does not increase the aggregate uncertainty. Since the gaming moves and replies of rivalry have been removed, uncertainty has arguably been reduced. Moreover—and this is the really critical point—decisions need not be forced to the top but can always be assigned to the level at which the issues are most appropriately resolved. Specifically, by conferring semiautonomous status on what had previously been fully autonomous firms in the premerger period, the best of both worlds can presumably be realized. If, for example, demand or cost inter-action effects are such that net gains can be had by moving decisions to the top, it will be done. Those decisions, however, that are most efficiently made at operating levels will remain there. Intervention at the top thus always occurs selectively, which is to say only upon a showing of expected net gains. The resulting combined firm can therefore do everything that the two autonomous firms could do previously and more. The same argument applies, moreover, not merely to horizontal mergers but also to vertical and conglomerate mergers. The upshot is that, possible public policy restraints (against monopoly, vertical integration, or aggregate size) aside, a compelling reason to explain why all production is not concentrated in one large firm is not reached by Knight’s argument.
Although other efforts to explain limitations on firm size have since been made, none addresses the issues in the way I have just posed it. And none really disposes of the puzzle. Thus consider my treatment of limits on firm size in terms of the “control loss” phenomenon (Williamson, 1967b). It involved the application to hierarchical organization of what F. C. Bartlett referred to as the serial reproduction effect in transmitting messages or images between individuals. His experiments involved the oral transmission of descriptive and argumentive passages through a chain of serially linked individuals. Bartlett concluded from a number of such studies:
It is now perfectly clear that serial reproduction normally brings about startling and radical alterations in the material dealt with. Epithets are changed into their opposites; incidents and events are transposed; names and numbers rarely survive intact for more than a few reproductions; opinions and conclusions are reversed— nearly every possible variation seems as if it can take place, even in a relatively short series. At the same time the subjects may be very well satisfied with their efforts, believing themselves to have passed on all important features with little or no change, and merely, perhaps, to have omitted unessential matters. [1932, p. 175]
Bartlett illustrated this graphically with a line drawing of an owl which, when redrawn successively by eighteen individuals, each sketch based on its imme- diate predecessor, ended up as a recognizable cat; and the farther from the initial drawing one moved, the greater the distortion experienced (1932, pp. 180-81).
I applied the same argument to the firm size dilemma by invoking bounded rationality and noting that limited spans of control are thereby implied. If any one manager can deal directly with only a limited number of subordinates, then increasing firm size necessarily entails adding hierarchical levels. Transmitting information across these levels experiences the losses to which Bartlett referred, which are cumulative and arguably exponential in form. As firm size increases and successive levels of organization are added, therefore, the effects of control loss eventually exceed the gains. A limit upon radial expansion is thus reached in this way.
Plausible though the argument seemed at the time, it does not permit selective intervention of the kind described above. Rather, the entire firm is managed from the top. All information that has a bearing on decisions is transmitted across successive levels from bottom to top; all directives follow the reverse flow down.
The scenario thus contemplates comprehensive (unselective) linkages between stages. Internal organization need not, however, adopt this structure. Suppose instead that the parent firm deals with each of its parts by exercising forebearance with respect to those activities where no net gains are in prospect (in which event the parent directs the operating part to replicate small firm behavior) and intervenes wherever coordination yields net gains. The puzzle to which I referred at the outset is evidently restored—or at least the serial reproduction loss “solution” does not apply—if such selective intervention is admitted.
The same is true, moreover, of arguments that firm size is limited by growth (Penrose, 1959) or by organization capital (Prescott and Visscher, 1980). Both arguments ignore the possibility that merger may be coupled with selective intervention. Thus if a series of small firms can grow rapidly, or if small firms can acquire valued organizational capital, a merger of those very same firms can selectively do the same and more.
An important recent paper by John Geanakoplos and Paul Milgrom (1984) traces firm size limitations to the “deadlines and delays” that attend hierarchical modes of organization. But the alternative mode of organization with respect to which hierarchy experiences a cost disadvantage is nowhere described. If the set of activities to be organized is held constant, and if internal organization can intervene selectively, the purported disability of hierarchy in relation to a collection of small firms does not withstand scrutiny.
The upshot is that limitations on firm size of a comparative institutional kind have yet to be described.
Source: Williamson Oliver E. (1998), The Economic Institutions of Capitalism, Free Press; Illustrated edition.