Bounded rationality refers to human behavior that is “intendedly rational, but only limitedly so” (Simon, 1961, p. xxiv). Although it is widely appreciated that human decision makers are not lightning calculators, and occasionally this fact is explictly taken into account in abstract models of market processes ( Radner. 1968), the implications for economic organization have only been scratched.
Bounded rationality involves neurophysiological limits on the one hand and language limits on the other. The physical limits take the form of rate and storage limits on the powers of individuals to receive, store, retrieve, and process information without error. Simon observes in this connection that “it is only because individual human beings are limited in knowledge, foresight, skill, and time that organizations are useful instruments for the achievement of human purpose” ( 1957. p. 199). Inasmuch as the literature on contingent claim markets was at a very early stage of its development in 1957 and since Simon is inexplicit on what alternative modes were subject to displacement by “organizations” (markets, after all, are usefully regarded as organizations also), it is not clear that Simon had the substitution of internal organization for imperfect contingent claims markets in mind. The reasons why the absence of unlimited computational capacity prevents comprehensive contracting of the sort required for the standard theorems on the existence and optimality of a competitive equilibrium to go through have, however, since been examined by Radner ( 1968).
Whereas the absence of a full set of contingent claims markets was of interest to Radner mainly because this permitted him to explain the existence of spot markets and demands for liquidity, which do not appear in the Arrow-Debreu version of the contingent claims model, it is of interest here because incomplete (or costly) contingent claims markets also occasion the development of internal organization. To the extent that internal organization serves to economize on scarce computational capacity and does not experience offsetting disabilities, internal organization is presumably favored. This turns out to be especially significant in relation to adaptive, sequential decision-making.
Language limits refer to [he inability of individuals to articulate their knowledge or feelings by the use of words, numbers, or graphics in ways which permit them to be understood by others. Despite their best efforts parties may find that language fails them ( possibly because they do not possess the requisite vocabulary or the necessary vocabulary has not been evise . and they resort to other means of communication instead. Dem- onstrations, learning-by-doing. and the like may be the only means of achieving understanding when such language difficulties develop.
Bounds on rationality are interesting, of course, only to the extent that e units of rationality are reached-which is to say. under conditions of uncertainty and or complexity. In the absence of either of these conditions the appropriate set of contingent actions can be fully specified at the outset. Thus, it is bounded rationality in relation to the condition of the environment that occasions the economic problem. Given unbounded rationality, contingent claims contracting goes through, whatever the degree of complexity to be dealt with. Similarly, given a sufficiently simple environment, bounded rationality constraints are never reached and com-parative institutional choices between firm and market are not posed not in any interesting way at least. When, however, transactions are conducted under conditions of uncertainty / complexity, in which event it is very costly, perhaps impossible, to describe the complete decision tree, the bounded rationality constraint is binding and an assessment of alternative organizational modes, in efficiency respects, becomes necessary. Feldman and Kanter’s discussion of the decision process is particularly relevant (1965, p. 615):
For even moderately complex problems . . . the entire decision tree cannot be generated. There are several reasons why this is so: one is the size of the tree. The number of alternative paths in complex decision problems is very large A second reason is that in most decision situations, unlike chess, neither the alternative paths nor a rule for generating them is available…. A third reason is the problem of estimating consequences … For many problems, consequences of alternatives are difficult, if not impossible, to estimate. The comprehensive decision model is not feasible lor most interesting decision problems.
As they point out, most decision problems, unlike board games such as chess, are not deterministic but involve decision-making under uncertainty. For these, the comprehensive decision tree is not apt even to be feasible. As Simon indicates, however, and as is maintained here, the distinction between deterministic complexity and uncertainty is inessential. What may be referred to as “uncertainty” in chess is “uncertainty introduced into a perfectly certain environment by inability —computational inability —to ascertain the structure of the environment. But the result of the uncertainty, whatever its source, is the same: approximation must replace exactness in reaching a decision” (Simon, 1972, p. 170). As long as either uncertainty or complexity is present in requisite degree, the bounded rationality problem arises and an interesting comparative institutional choice is often posed.12
2. Some Examples
Chess, as von Neumann and Morgenstern have observed, is a trivial game:“ if the theory of Chess (i.e., the complete tree of possible games) were really fully known there would be nothing left to play” (1953, p. 125). As they subsequently point out, however, and as Simon elaborates, the complete decision tree is prohibitively complex to develop. Assuming, at any given stage in a game of chess, that there are about thirty legal moves, there will be, for a move and its replies, about 103 possibilities to consider.
If the average length of game is forty moves. 101 possibilities must be con-stdered (Simon, 1972, p. 166)-which is unimaginably vast.
Meade’s discussion of the limitations of contingent claims contracting affords a striking example of bounded rationality. After first describing a single gigantic once-for-all forward ‘higgle-haggle’ in which all contingent goods and services (i.e., all goods and services at each possible time- cum- environmental condition) are bought and sold once and for all now for money payments made now” (1971. p .166) he then goes on to con-cede (1971, p.183):
When environmental uncertainties are so numerous that they cannot all be considered . . . or. what comes perhaps to much the same thing. when any particular environmental risks are so hard to define and to distinguish from each other that it is impossible to base a firm betting or insurance contract upon the occurrence or non-occurrence of any one of them, then for this reason alone it is impossible to have a system of contingency or of conditional forward markets.
Although a full explanation for the impossibility conditions described by Meade also involves opportunism (see Section 2. below), a major contributing factor ,s the existence of bounded rationality. Environmental uncertainties, when these become so numerous that they cannot all be considered, presumably exceed the data processing capabilities of the parties. The complete decision tree simply cannot be generated-in which event “the bulk of meaningful future transactions cannot be carried out on any existing present market” (Arrow, 1969, p. 51).
A final example of bounded rationality is afforded by Commons in his description of the political leader as one who “can formulate in language what others feel but could not tell” (1934, p. 750). The bounds on rationality here take the form of language rather than computational limits and evidently vary among individuals. If the specialization of labor is feasible, those whose rationality limits are less severely constrained than others are natural candidates to assume technical, administrative, or political leader- s ip positions – which is to say that a hierarchy can emerge on this account.
3. Internal Organization
As Hayek emphasized (see Chapter 1), the price system has advantages over central planning in circumstances where the relevant information is summarized by price signals. The price system relieves the need for parties who are remotely related to the underlying change in the data to be apprised of the details associated with changing market conditions. Demands on scarce rationality capabilities are correspondingly limited.
Where competitive supply conditions are not satisfied, reliance on summary statistics can be hazardous. For the reasons that will be developed in Section 2 and in the following chapters, internal organization often has attractive properties in that it permits the parties to deal with uncertainty/ complexity in an adaptive, sequential fashion without incurring the same types of opportunism hazards that market contracting would pose. Such adaptive, sequential decision processes economize greatly on bounded rationality. Rather than specifying the decision tree exhaustively in advance, and deriving the corresponding contingent prices, events are permitte to unfold and attention is restricted to only the actual rather than all possible
A further advantage of internal organization is that, as compared to recurrent market exchange, efficient codes are more apt to evolve and be employed with confidence by the parties. Such coding also economizes on bounded rationality. Complex events are summarized in an informal way by using what may be an idiosyncratic language.10 Although, in principle, the parties to recurrent market contracts could devise the same language, thereby realizing the same economies, such exchanges are more subject to risks of opportunism—hence, are less apt to be developed as fully.
An additional advantage of internal organization is that it promotes convergent expectations, serving in this way to attenuate uncertainties that are generated when interdependent parties make independent decisions with respect to changing market circumstances (Malmgren, 1961). It each of the parties to a related set of transactions takes his own observations on how events are changing, infers probable consequences (including how other parties can be expected to adapt), and acts accordingly, there is a risk that the resulting set of decisions will be made in a jointly incompatible manner.
It is not however, that markets are perverse or inherently “predisposed” to to through an extended period of disequilibrium during which a common set of expectations among the parties gradually evolves. In principle this adjustment process could be shortened by having one of the autonomous parties to the exchange temporarily serve as the decision maker for all: he would declare who is to adapt and how. The risk of giving such an assigm ment to an autonomous party is that he will refer to his private gains rather han to a collective calculus in making a recommendation. To the extent that internal organization mitigates such opportunistic behavior, without incurrmg offsetting costs, a shift from market to hierarchy will promote efficient adaptation.
Source: Williamson Oliver E. (1975), Markets and hierarchies: Analysis and antitrust implications, A Study in the Economics of Internal Organization, The Free Press.