The assumption of unbounded rationality in the model of efficient mar- kets is important enough to warrant a separate discussion. Unbounded rationality requires that all decision makers are fully informed and are able to fully process all relevant information so as to optimize their respective positions. They know their options and also the options of competitors (as either sellers or buyers). They therefore can compare alternatives at the margin. Predictions of the model of efficient decision making are predicated on this important assumption of unbounded rationality.
These conditions rarely, if ever, exist in decision situations, however. Shubik (1971) notes that a decision maker ‘lives in an environment about which his information is highly incomplete. Not only does he not know how to evaluate many of the alternatives facing him, he is not even aware of a considerable percentage of them’ (p. 358). This signifi- cantly limits the usefulness of this model as one of efficient decision making. The model yields predictions that at best approximate the actual outcomes and at worst provide incorrect and misleading inputs to applications of the model, for example, to tax policy. At which end of the spectrum the predictions fall depends on the decision situation in question. For simple decision situations where market conditions are clearly specified and information requirements are minimal, and there- fore unbounded rationality may most likely prevail, the model should provide useful predictions of responses to changes in relative prices or opportunity costs. Simon (1987b) states that the case for unbounded rationality can be made ‘when the decision making takes place in situations so transparent that the optimum can be reasonably approxi- mated by an ordinary human mind’ (p. 267). Note that he does not con- cede that an optimum can be achieved, only that it can be reasonably approximated, even in these very best of decision circumstances. Theses very best of circumstances are not likely to exist in decision situations such as occur in complex organization, however. Important issues in the decision process in complex organizations cannot be easily abstracted, so that predictions based on a model that assumes unbounded ration- ality are less useful.
For example, in many organizations the output is difficult to define as well as to measure. This may be less of a problem in manufacturing industries but is particularly true in service industries. This problem is not exclusive to public bureaus and nonprofit organizations; it also applies to for-profit corporations. The output of hospitals and law firms provide two illustrations of this type of difficulty. In hospitals output may be measured by case, or by patient-day, and in law firms output is measured by billable hours, an input measure. In the model of efficient markets, outputs (and inputs) are assumed to be homogeneous across units. When they are not, as in types of hospital cases, how are the relative prices and opportunity costs to be measured? How do decision makers determine the optimal mix of hospital cases or legal cases?
Demsetz (1993) captures this limitation of the neoclassical model as well. He states that
The model sets the maximizing tasks of the firm in a context in which decisions are made with full and free knowledge of production possibilities and prices. The worldly roles of management, being to explore uncertain possibilities and to control resources con- sciously … are not easily analyzed in a model in which knowledge is full and free. (pp. 160–1)
The assumption of bounded rationality captures this and other complexities. As noted in Chapter 2, bounded rationality explicitly rec- ognizes the limited ability of decision makers to obtain accurate and even adequate information and to process the information that they can get. Simon (1987b) emphasizes that bounded rationality is an effi- cient decision process that conserves resources required to choose among alternatives:
[I]t is not reasonable to talk about finding ‘all the alternatives.’ The generation of alternatives is a lengthy and costly process, and one where, in real world situations, even minimal completeness can seldom be guaranteed … [H]uman alternative-generating behavior observed … is usually best described as heuristic search aimed at find- ing satisfactory alternatives, or alternatives that represent an improve- ment over those previously available. (p. 267)
Beckert (1996, pp. 807–8) notes that as early as 1921 both Knight and Keynes ‘focused on uncertainty as a limitation to the rational-actor model’ (p. 813, italics in the original). They each defined uncertainty as a situation where no probabilities could be assigned. Knight attributed uncertainty to lack of available information or ‘prior experience’ (1921, p.229). He proposed specialization through hierarchical structure, essentially providing a rationale for the construct of a firm, as a solution to the problem of uncertainty. Keynes attributed uncertainty to a ‘weak- ness of our reasoning power (1973a, p. 34). Keynes, relating uncertainty to investment decisions, proposed that:
In practice, we have tacitly agreed, as a rule, to fall back on what is, in truth, a convention. The essence of this convention – though it does not, of course, work out quite so simply – lies in assuming that the existing state of affairs will continue indefinitely, except in so far as we have specific reasons to expect a change … [O]ur existing knowledge does not provide a sufficient basis for a calculated math- ematical expectation. (1973b, p. 152, italics in the original)
Beckert states that uncertainty, in the Knight–Keynes sense, is the basis for his concept of ‘intentional rationality,’ similar to bounded rationality: ‘Actors are considered intentionally rational when they want to achieve a goal that optimizes their utility, but do not know the best means to apply for realizing this goal’ (p. 819).
In this context decision making is no longer an act of simple optimiza- tion. Instead decision making becomes a process of framing the issue, developing alternatives, establishing decision targets, and processing information that is limited and filtered through layers of organizational hierarchy (Williamson, 1993 and March, 1997). The decision outcomes yield a solution, or a range of solutions. But the solution is not neces- sarily unique and is likely to change during the decision process. Blyth (2002), commenting on problems with the uniqueness of equilibrium in optimization models of rational choice, notes the importance of perspectives of the decision maker on the decision outcome:
In the economic world … the ideas that agents have about the impacts of their actions, and those of others, shape outcomes them- selves. If agents in the economy hold different ideas about how the economy works, this can lead to such agents taking a variety of actions, thereby producing radically different outcomes in the same circumstances. (p. 33)
Thus any solution under conditions of bounded rationality is not optimal in the usual and mathematical sense of meeting the criteria of decision calculus.
For this reason the concept of bounded rationality has met with some resistance in economics (see, for example, Friedman, 1953, and further comments below). Beckert (1996, p. 813) notes that economists have moved away from the Knight–Keynes concept of uncertainty. For example, Hirshleifer and Riley (1992) remove the distinction between uncertainty and risk by applying subjective probabilities to situations of uncertainty, thus permitting optimization techniques. The importance and relevance of bounded rationality could not be ignored even by economists, however. Using the term ‘rationality’ in the usual economic sense of unbounded optimization, Boulding (1966) acknowledges the significance and efficiency (if not the terminology) of bounded rationality and the potential inefficiency of unbounded rationality:
Decision-making by instinct, gossip, visceral feeling, and political savvy may stand pretty low on the scale of rationality, but it may have the virtue of being able to take in very large systems in a crude and vague way, whereas the rationalized processes can only take in subsystems in their more exact fashion, and being rational about subsystems may be worse than being not very rational about the system as a whole … On the other hand, the economist has a certain mind-set in favor of his own skills, and it is easy for him to leave out essential variables with which he is not familiar. Here, indeed, a little learning may be a dangerous thing … (p. 11)
In decision theory and in organization theory the notion of bounded rationality in economic analysis of managerial decisions has been applied and modified in a variety of ways, effectively broadening or at times altering its meaning. Cyert, Feigenbaum, and March (1959) sug- gested that the problems associated with the distinction between satisficing and maximizing can be avoided through a technique that uses both. To do this they proposed a nine-step decision process that is resolved by choosing the most satisfying solution. That is, they maximize from a set of satisfactory outcomes. Baumol and Quandt (1964) went further and redefined bounded rationality as (additionally) constrained optimization by adding information constraints to a profit maximizing or utility maximizing model of decision behavior.
These interpretations clearly are not the interpretation intended by Simon (1979, p. 504 and pp. 508–9), who conceived of bounded ration- ality as a theory of decision making that is an alternative to optimiza- tion and as a way of being efficient by economizing on information as a scarce resource. Simon (1987a) further clarified his interpretation of bounded rationality in his explanation of behavioral economics:
Behavioural economics is concerned with the empirical validity of these neoclassical assumptions about human behaviour and, where they prove invalid, with discovering the laws that describe behaviour correctly and as accurately as possible. As a second item on its agenda, behavioural economics is concerned with drawing out the implica- tions, for the operation of the economic system and its institutions and for public policy, of departures of actual behaviour from the neo- classical assumptions. A third item on its agenda is to supply empiri- cal evidence about the shape and content of the utility function (or whatever construct will replace it in a [sic] empirically valid behavioural theory) so as to strengthen the predictions that can be made about human economic behaviour. (p. 221)
Furubotn and Richter (2000), referring to Williamson’s (1985) emphasis of the importance of transactions costs, demonstrate that the continued use of the (inaccurate) interpretation of bounded rationality as an addi- tionally constrained form of optimization is still current in economic analysis of organizations. They refer to this as imperfect individual rationality and interpret this as a form of incomplete contracting due to positive transactions costs (Furubotn and Richter, 2000, pp. 3–4, italics in the original). Thus even while explaining the concept of bounded rationality they show that Simon’s interpretation is still being either ignored or misunderstood by economists in their applications.
Even March as late as 1997 reflected the nature of this controversy by indicating on the one hand, its general acceptance, and on the other hand, its acceptance within the context of optimization:
[V]irtually all modern theories of rational choice are theories of limited (or bounded) rationality … The key scarce resource is attention; and theories of limited rationality are, for the most part, theories of the allocation of attention … Search is stimulated by a failure to achieve a goal and continues until it reveals an alternative that is good enough to satisfy existing evoked goals. New alternatives are sought in the neighborhood of old ones … Theories of limited rationality are also theories of slack … When performance falls below the goal, search is stimulated, slack is decreased, and aspirations decreased … (p. 12)
Although March explicitly acknowledges that a solution may be ‘good enough,’ he interprets these as signals that the search process should continue. Just as occurs with optimization, the solution search process (that is, the attention) does not stop at that point of being ‘good enough’ or when slack exists. The implication is that bounded rational decision making can yield only short run outcomes, a point with which Simon specifically disagrees (1979, pp. 509–10).
Radner (1997) also points to this controversy in an essay on bounded rationality. He states that the current economic analysis of decision making relies on optimization and suggests that the alternative behav- ioral approach carries with it a stigma:
This [decision maker’s uncertainty about the logical implications of what he or she knows] confronts the economist with a dilemma. On the one hand, he or she can continue to investigate models of rational decision making that are simple enough to be tractable for the econ- omist but are hopelessly unrealistic (this is the current mainstream approach). On the other hand, at the risk of being branded a ‘behav- ioral economist,’ he or she can abandon the attempt to explain observed behavior as rational and simply record various empirical regularities such as rules of thumb. As a compromise, the economist can try to show that competition will weed out ‘irrational’ or ‘nonoptimal’ behavior in the long run, although individual decision makers and organizations are not capable of deliberately determining what is optimal … (p. 333, italics in the original)
The controversy regarding bounded rationality and its application in economic theories of decision making and organizations has deep roots. As children schooled from an early age by their parents and teachers in ways of viewing the world, economists are schooled in the context of optimization from the earliest course in the principles of economic the- ory, steeped in the Marshallian tradition of marginal analysis. These first exposures to economic modeling assume rational behavior and define that rational behavior to be marginal analysis. This behavioral definition is invoked throughout the economics educational process, and some- where along that road the fact that this defined behavior is assumed behavior falls by the wayside. It becomes a fact of economic life. Indeed, rational behavior defined as optimization has become such a part of the economist’s being that any complex human decision can be analyzed through simplification and marginal analysis: investment decisions (Modigliani and Miller, 1958), altruistic decisions (Rose-Ackerman, 1987), and even choice of spouse (McKenzie and Tullock, 1981).
The same approach is used in economic analysis of decisions in organizations. The illustrations given earlier and related to Figure 3.1 are an example of this type of application. This provides the underlying theory for analysis of decisions in the firm (for example, Coase, 1937; Williamson, 1963; and Alchian and Demsetz, 1972); in nonprofits (for example, Newhouse, 1970; James and Rose-Ackerman, 1986; and Hansmann, 1987b); and in the public sector (for example, Weingast, Shepsle, and Johnsen, 1981 on legislative behavior and Niskanen, 1971 on bureau behavior). Recent game-theoretic models are also based on optimization at each stage of the game, given its rules (Radner, 1997).
Optimization provides a unique theoretical solution; bounded ration- ality does not. As Demsetz (1993) has noted, ‘The cost of maximizing is ignored or implicitly assumed to be zero. De facto, the resources that might be required to make maximizing decisions are treated as if they are not scarce’ (p. 161, italics in the original). Thus the uniqueness comes at the cost of simplifying away those aspects of organizational context and resource issues that make the decision problem interesting and important. Some simplification and abstraction is essential for any model, of course. The issue is what and how much to simplify, and for what purpose. Williamson (1990, p. 179) notes that it is ‘now generally agreed that the satisficing approach has not been broadly applicable,’ which of course is true for models that are based in optimization tech- niques. That the controversy reflects a disciplinary attitude is captured by Douglas (1990), who recognizes Simon’s point that bounded rational behavior is efficient because it conserves resources that are employed in the decision making process. She notes that ‘[w]hereas Simon treated [bounded rationality] as a good thing because it is a form of economiz- ing in cognitive energy, Williamson … treats boundedness as a weak- ness, a source of incompetence’ (pp. 106–7).
Source: Carroll Kathleen A. (2004), Property Rights and Managerial Decisions in For-Profit, Nonprofit, and Public Organizations: Comparative Theory and Policy, Palgrave Macmillan; 2004th edition.