It is a common practice in economic theory to make a series of behavioral assumptions for micro-units and to generate from these assumptions a parallel series of implications for aggregations of such units. Assumptions about the firm, consumer, and investor lead to a series of predictions of the behavior of markets, segments of the economy, and the economy as a whole.
Since these aggregate predictions can be compared with aggregate data gathered from the real world, implicit confirmation of the whole model (including the underlying behavioral assumptions) is obtained by testing aggregate predictions. Although the traditional justification of this methodology is questionable (see Appendix A), the methodology itself is, in principle, unexceptionable.
We have reviewed some of the classic assumptions of firm behavior used in aggregate models. Although most of them seem to us implausible as descriptions of the actual behavioral processes used by firms, they have been useful for certain kinds of aggregate models. In fact, much of macroeconomic theory is built upon such assumptions even if the assumptions are not always necessary to the theory.
The approach pursued in this book and the models outlined here make it possible to develop alternative aggregate models built upon alternative assumptions about business firms. Provided we can solve the computation problems of aggregation, we should be able to elaborate behavioral theories of markets or economies. To a limited extent, we have already shown some of the implications of our models for the simplest forms of oligopoly — the duopoly models of Chapters 5 and 8. Although those efforts are not conclusive, they do indicate both some kinds of aggregate implications and some procedures for evaluating the implications. In this section, we will suggest briefly some general implications for aggregate systems and some approaches to the problem of constructing useful aggregate models.
1. Oligopoly theory and the behavioral theory of the firm
Although we have presented some models that involve interacting firms within an oligopolistic market, we have not attempted to consider oligopoly theory in any detail. Such a consideration must await further empirical work on oligopoly market behavior and its relation to the predictions generated by behavioral models. However, we can suggest a few implications of the theory for work on oligopoly markets. The suggestions include proposals for modification of both the substantive content of theories of firm behavior and the theoretical objectives of oligopoly theory. The substantive modifications were discussed at some length earlier. For the most part, a behavioral model of oligopoly behavior should emphasize adaptation, problem solving, uncertainty avoidance, and incomplete rationalization of decisions. It should tend to de-emphasize explicit omniscience and goal clarity.
These substantive modifications stem from the modifications we have proposed in the theoretical objectives of oligopoly theory. Conventionally, oligopoly theory considers the ways in which the theory of competition must be modified to deal with a market dominated by a few major firms. The shift in “competition” from a large number of essentially anonymous small firms to a small number of readily identifiable large firms leads to an explicit treatment of expectations about the behavior of specific competitors. The objective is to develop a model containing a few variables capable of yielding a determinate result of considerable generality. In order to reduce the decision process to manageable proportions, the classical theory ignores some phenomena and aggregates others into a handful of constructs. As we have observed above, such a strategy assumes that at some meaningful level of generality it is possible to construct a relatively simple model of an oligopolistic market. We are constrained to consider only those phenomena that are susceptible to prediction by such models.
The behavioral theory of the firm relaxes considerably the requirements of simplicity in the models, expands considerably the phenomena that can be considered explicitly, and reduces considerably the need for aggregation of phenomena into summary variables. As a result, the theory gives up some attributes of generality. The distinctions made among firms and industries are finer and the predictions, therefore, tend to be more specific. Since the predictions depend on a larger number of variables, the models require more knowledge in order to generate firm predictions. The reaction of one variable to movement in another variable depends on a number of other contingencies within the model.
To illustrate the differences between the two approaches, consider the treatment of competition within oligopoly theory. If we acknowledge the relevance of specific competitors for decisions, we are constrained to introduce some form of the conjectural variation term — an expectation about behavior by key competitors. For such models we assume that firms have (or form) some expectations about the reaction of other firms to their own behavior. In most forms of the theory, the conjectural variation term is quite simple. It is a crude aggregate construct to summarize the way in which the firm anticipates competitor’s behavior.
No one would seriously argue either that, in fact, a conjectural variation term is explicitly formulated by many business firms or that we can represent all of the processes by which firms attempt to deal with competition in an oligopoly by introducing such a simple concept. However, the conjectural variation term is consistent with the general theoretical strategy of oligopoly theory and allows some kinds of solutions to some problems.
An alternative strategy is to examine the actual processes by which firms deal with the problem of oligopolistic competition. Such an approach — as reflected in the behavioral theory of the firm — identifies two conspicuous techniques used by the business organization. The firm attempts to learn from its experience. That is, rather than “calculate” a conjectural variation term, it “learns” one on the basis of market feedback. Even more important, the firm attempts to avoid the uncertainties of competition by developing information systems that substitute direct knowledge for forecasts. The possibility that business firms might in a general way be adaptive and that they might develop procedures for sharing information is a feature of imperfect competition that few students of oligopoly would deny. What distinguishes the behavioral theory of the firm from most other theories of the firm is the attempt to consider relatively slow adaptation and partial information systems rather than limiting attention to immediately adaptive or complete information systems.
The significance of the difference hinges on the extent to which the firms with which we deal can, in fact, be adequately described in the terms encompassed by more conventional theories. For example, one possible information system for firms in an oligopoly is complete collusion. Normally, it provides the maximum amount of information sharing within the industry. It can also be encompassed rather easily within the theory. Such collusion has, however, two major drawbacks for the typical firm. First, social regulation of market systems ordinarily inhibits the development of collusive information sharing. Various prohibitions of conspiracy and collaboration among competitors (in combination with erratic enforcement) make the costs of manifest collusion highly uncertain. Second, the full effectiveness of collusion depends on assured cooperation among the several firms in the industry, and for at least some firms in the industry the uncertainties of polyarchy within a cartel are at least as forbidding as the uncertainties of oligopolistic markets. Because of the costs involved, the difficulties in enforcing cooperation, and the substantial commitment in the culture to the immorality of complete collusion, such a strategy is generally not considered except when the organization — or some part of it — feels under considerable pressure from actual or impending failure to achieve critical goals. Attempts at some form of collusion are certainly common. The conditions necessary often exist somewhere in a complex organization, and the conventional delegation of authority and decentralization in an organizational coalition lend themselves to decentralized collusion. However, the collusion ordinarily associated with legal action by the government is apparently not a standard operating procedure for most large business firms in the United States.
There are, however, a number of alternative information systems much more commonly in use, some of which were mentioned in earlier chapters. Salesmen exchange information with each other directly. Customers serve as brokers of information gathered from their various alternative suppliers. Information on plans, styles, product mix, and expectations are both systematically leaked and diligently pirated. Trade and professional associations provide for the systematic exchange of information. Widely shared standard operating procedures make explicit information redundant in many cases. Various kinds of external consultants serve as indirect means of pooling information. Some of these information systems are customarily viewed as unilateral intelligence activities, but, in fact, most of them thrive as accepted devices not simply for acquiring information but also for sharing it.The firm or subunit that refuses to offer information is at a disadvantage in gathering it from the information broker involved.
If this description is accurate, the difficulty of oligopoly theory is clear. It deals relatively effectively with complete exchange of information and relatively effectively with no exchange of information but it deals poorly with the intermediate information systems that are apparently typical of most firms in oligopolistic industries. At the same time, it should be clear that the behavioral theory of the firm can — at least thus far — treat intermediate systems precisely only in special situations (e.g., the department store model, the general price and output model). Beyond such special situations the theory provides only a general framework for understanding the operation of oligopolies.
In a similar way, if we examine the possibilities for dealing with organizational learning in an oligopolistic market, we must admit that aside from a few special applications we are not yet in a position to make effective use of the observation that firms adapt to their environments. Some efforts (Chapter 5) to examine adaptive conjectural variation terms have been useful, and some more general adaptive models (Chapter 8) seem to portend more general applicability of detailed learning constructs. 11. However, at this level of theory the main contribution the behavioral theory of the firm can offer is a different set of general considerations for the development of a theory of oligopoly.
2. Computer models of aggregates of firms
The research strategy of expanding the range of phenomena explicitly treated in a theoretical model makes little sense for oligopoly theory unless it is feasible to develop computer models of aggregates of firms. There does not appear to be any alternative method of dealing with the kinds of models described here. Conversely, the step from a predictive computer model of a specific firm to a predictive computer model of a specific industry does not seem in principle overwhelming. In fact, we have moved rather easily back and forth from one to the other (e.g., in the can industry model, the department store model, and the general model).
For comparative statics these models do not seem to have any particular advantage over other models except in special cases. For describing changes over time in a complex system of interrelated decisions and reactions, however they open to study a large class of economic phenomena that have not yielded to previous theoretical efforts. The general model of price and output, for example, predicts time series of prices, profits, market shares, inventory levels, output, general sales strategies, costs, and organizational slack. In order to generate such output, of course, the models require not only the parameters describing attributes of the firms involved but also an adequate characterization of the external environment — as represented particularly by the demand function. The individual firms represented may consist of all of the firms in the market (or segment of the economy) being simulated, or the model may be restricted to a sample of firms from which some prediction can be made. The latter procedure seems more feasible from the point of view of model construction, but it does pose some difficult technical problems in extrapolation. Aggregation to still larger units (e.g., the economy) is also possible through sampling. Although such aggregation may not involve major conceptual difficulties, the problems in obtaining relevant microdata are large enough to make detailed economy-wide aggregation seem somewhat farther in the future.
Source: Skyttner Lars (2006), General Systems Theory: Problems, Perspectives, Practice, Wspc, 2nd Edition.