Annu.Rev.Polit.Sci.1999.2:297-321 Copyright 1999 by Anmual Reviews.All rights reserved BOUNDED RATIONALITY Ajuo asn Bryan D.Jones Department of Political Science,University of Washington,Seattle,Washington 98195;e-mail:bdjones@u.washington.edu KEY WORDS:decision making,behavioral organization theory,behavioral decision theory, political psychology ABSTRACT Findings from behavioral organization theory,behavioral decision theory, survey research,and experimental economics leave no doubt about the fail- ure of rational choice as a descriptive model of human behavior.But this does not mean that people and their politics are irrational.Bounded rational- ity asserts that decision makers are intendedly rational;that is,they are goal- oriented and adaptive,but because of human cognitive and emotional archi- tecture,they sometimes fail,occasionally in important decisions.Limits on rational adaptation are of two types:procedural limits,which limit how we go about making decisions,and substantive limits,which affect particular choices directly.Rational analysis in institutional contexts can serve as a standard for adaptive,goal-oriented human behavior.In relatively fixed task environments,such as asset markets or elections,we should be able to divide behavior into adaptive,goal-oriented behavior(that is,rational action)and behavior that is a consequence of processing limits,and we should then be able to measure the deviation.The extent of deviation is an empirical issue. These classes are mutually exclusive and exhaustive,and they may be exam- ined empirically in situations in which actors make repeated similar choices. INTRODUCTION Do people make rational decisions in politics and economics?Not if by "rational"we mean that they demonstrate conformity to the classic expected- utility model.There is no longer any doubt about the weight of the scientific evidence;the expected-utility model of economic and political decision mak- 3 ing is not sustainable empirically.From the laboratory comes failure after failure of rational expected utility to account for human behavior.From sys- tematic observation in organizational settings,scant evidence of behavior based on the expected-utility model emerges. 297 1094-2939/99/0616-0297S08.00
1094-2939/99/0616-0297$08.00 297 Annu. Rev. Polit. Sci. 1999. 2:297–321 Copyright © 1999 by Annual Reviews. All rights reserved BOUNDED RATIONALITY Bryan D. Jones Department of Political Science, University of Washington, Seattle, Washington 98195; e-mail: bdjones@u.washington.edu KEY WORDS: decision making, behavioral organization theory, behavioral decision theory, political psychology ABSTRACT Findings from behavioral organization theory, behavioral decision theory, survey research, and experimental economics leave no doubt about the failure of rational choice as a descriptive model of human behavior. But this does not mean that people and their politics are irrational. Bounded rationality asserts that decision makers are intendedly rational; that is, they are goaloriented and adaptive, but because of human cognitive and emotional architecture, they sometimes fail, occasionally in important decisions. Limits on rational adaptation are of two types: procedural limits, which limit how we go about making decisions, and substantive limits, which affect particular choices directly. Rational analysis in institutional contexts can serve as a standard for adaptive, goal-oriented human behavior. In relatively fixed task environments, such as asset markets or elections, we should be able to divide behavior into adaptive, goal-oriented behavior (that is, rational action) and behavior that is a consequence of processing limits, and we should then be able to measure the deviation. The extent of deviation is an empirical issue. These classes are mutually exclusive and exhaustive, and they may be examined empirically in situations in which actors make repeated similar choices. INTRODUCTION Do people make rational decisions in politics and economics? Not if by “rational” we mean that they demonstrate conformity to the classic expectedutility model. There is no longer any doubt about the weight of the scientific evidence; the expected-utility model of economic and political decision making is not sustainable empirically. From the laboratory comes failure after failure of rational expected utility to account for human behavior. From systematic observation in organizational settings, scant evidence of behavior based on the expected-utility model emerges. Annu. Rev. Polit. Sci. 1999.2:297-321. Downloaded from arjournals.annualreviews.org by UNIVERSITY OF WASHINGTON - HEALTH SCIENCES LIBRARIES on 09/25/06. For personal use only
298 JONES Does this mean that people (and therefore their politics)are irrational? Not at all.People making choices are intendedly rational.They want to make rational decisions,but they cannot always do so. The implication for politics is that rational responses to the environment characterize decision making generally,but at points-often important points-rationality fails,and as a consequence there is a mismatch between the decision-making environment and the choices of the decision maker.We refer to this mismatch as "bounded rationality showing through"(Simon 1996b). asn This conception has an important implication.In structured situations,at euosjad least,we may conceive of any decision as having two components:environ- mental demands(seen by the individual as incentives,positive or negative) 豆 and bounds on adaptability in the given decision-making situation.Ideally,an analysis based on rational choice should be able to specify what the environ- 9086000 mental incentives are and to predict decisions based on those incentives.What cannot be explained is either random error (even the most rational of us may make an occasional mistake,but these are not systematic)or bounded rationality showing through.Standard statistical techniques give us the tools to distinguish systematic from random factors,so in principle it should be possi- ble to distinguish the rational,adaptive portion of a decision from bounds on rationality. popeojuo One may think ofany decision as arising from two sources.One is the exter- nal environment-how we respond to the incentives facing us.The other is the internal environment-those parts of our internal make-ups that cause us to 125-76 deviate from the demands of the external environment(Simon 1996b). We are not,however,thrown into a situation in which all residual system- atic deviations from rational choices are treated prima facie as bounded ra- 26661 tionality.A very limited set of facets of human cognitive architecture accounts for a very large proportion of the deviations from adaptation.These may be placed into two classes:procedural limits,which limit how we go about mak- ing decisions,and substantive limits,which affect particular choices directly. Of procedural limits,I cite two as being extraordinarily important in struc- 台 tured,institutional settings (such as voting in mass publics or in legislative bodies),attention and emotion.Of substantive limits,I cite but one-the ten- dency of humans to "overcooperate,"that is,to cooperate more than strict adherence to rationality would dictate. The primary argument in this essay is that most behavior in politics is B adaptive and intendedly rational but that limits on adaptive behavior,imposed by human cognitive/emotional architecture,may be detected in even the most stable of environments.I advocate a research strategy that explicitly divides political action into the two categories of intended rationality and deviations from(or bounds on)intended rationality and explores empirically the implica- tions for the outputs of institutions and the institutional processes responsible
Does this mean that people (and therefore their politics) are irrational? Not at all. People making choices are intendedly rational. They want to make rational decisions, but they cannot always do so. The implication for politics is that rational responses to the environment characterize decision making generally, but at points—often important points—rationality fails, and as a consequence there is a mismatch between the decision-making environment and the choices of the decision maker. We refer to this mismatch as “bounded rationality showing through” (Simon 1996b). This conception has an important implication. In structured situations, at least, we may conceive of any decision as having two components: environmental demands (seen by the individual as incentives, positive or negative) and bounds on adaptability in the given decision-making situation. Ideally, an analysis based on rational choice should be able to specify what the environmental incentives are and to predict decisions based on those incentives. What cannot be explained is either random error (even the most rational of us may make an occasional mistake, but these are not systematic) or bounded rationality showing through. Standard statistical techniques give us the tools to distinguish systematic from random factors, so in principle it should be possible to distinguish the rational, adaptive portion of a decision from bounds on rationality. One may think of any decision as arising from two sources. One is the external environment—how we respond to the incentives facing us. The other is the internal environment—those parts of our internal make-ups that cause us to deviate from the demands of the external environment (Simon 1996b). We are not, however, thrown into a situation in which all residual systematic deviations from rational choices are treated prima facie as bounded rationality. A very limited set of facets of human cognitive architecture accounts for a very large proportion of the deviations from adaptation. These may be placed into two classes: procedural limits, which limit how we go about making decisions, and substantive limits, which affect particular choices directly. Of procedural limits, I cite two as being extraordinarily important in structured, institutional settings (such as voting in mass publics or in legislative bodies), attention and emotion. Of substantive limits, I cite but one—the tendency of humans to “overcooperate,” that is, to cooperate more than strict adherence to rationality would dictate. The primary argument in this essay is that most behavior in politics is adaptive and intendedly rational but that limits on adaptive behavior, imposed by human cognitive/emotional architecture, may be detected in even the most stable of environments. I advocate a research strategy that explicitly divides political action into the two categories of intended rationality and deviations from (or bounds on) intended rationality and explores empirically the implications for the outputs of institutions and the institutional processes responsible 298 JONES Annu. Rev. Polit. Sci. 1999.2:297-321. Downloaded from arjournals.annualreviews.org by UNIVERSITY OF WASHINGTON - HEALTH SCIENCES LIBRARIES on 09/25/06. For personal use only
BOUNDED RATIONALITY 299 for those outcomes.[The analysis presented here is further developed in my Traces of Eve:Adaptive Behavior and Its Limits in Political and Economic In- stitutions (manuscript in preparation).] BOUNDED RATIONALITY:BIRTH AND DEVELOPMENT Ajuo Bounded rationality is a school of thought about decision making that de- asn veloped from dissatisfaction with the "comprehensively rational"economic and decision theory models of choice.Those models assume that preferences JO'SMOl euosjad are defined over outcomes,that those outcomes are known and fixed,and that decision makers maximize their net benefits,or utilities,by choosing the 2 alternative that yields the highest level of benefits(discounted by costs).The subjective expected-utility variant of rational choice integrates risk and uncer- Aorenuue's]eumofre woy 9086000 tainty into the model by associating a probability distribution,estimated by the decision maker,with outcomes.The decision maker maximizes expected utility.Choices among competing goals are handled by indifference curves- generally postulated to be smooth(twice differentiable)-that specify substi- tutability among goals. A major implication of the approach is that behavior is determined by the popeojuo mix of incentives facing the decision maker.A second implication is that adjustment to these incentives is instantaneous;true maximizers have no learn- ing curves. Like comprehensive rationality,bounded rationality assumes that actors 125-76 are goal-oriented,but bounded rationality takes into account the cognitive limitations of decision makers in attempting to achieve those goals.Its scien- 26661 tific approach is different;rather than making assumptions about decision making and modeling the implications mathematically for aggregate behavior (as in markets or legislatures),bounded rationality adopts an explicitly behav- ioral stance.The behavior of decision makers must be examined,whether in the laboratory or in the field. The Birth of Bounded Rationality ALISHAINO Simon(1999;see also Simon 1996a)reminds political scientists that the no- tion of bounded rationality and many ofits ramifications originated in political science.Over his long career,Simon made major contributions not only to 3 political science (as the founder of the behavioral study of organizations)but also to economics(as a Nobelist),psychology(as a founding father of cogni- tive psychology),and computer science(as an initiator of the field of artificial intelligence). In the 1940s and 1950s,Simon developed a model of choice intended as a challenge to the comprehensive rationality assumptions used in economics
for those outcomes. [The analysis presented here is further developed in my Traces of Eve: Adaptive Behavior and Its Limits in Political and Economic Institutions (manuscript in preparation).] BOUNDED RATIONALITY: BIRTH AND DEVELOPMENT Bounded rationality is a school of thought about decision making that developed from dissatisfaction with the “comprehensively rational” economic and decision theory models of choice. Those models assume that preferences are defined over outcomes, that those outcomes are known and fixed, and that decision makers maximize their net benefits, or utilities, by choosing the alternative that yields the highest level of benefits (discounted by costs). The subjective expected-utility variant of rational choice integrates risk and uncertainty into the model by associating a probability distribution, estimated by the decision maker, with outcomes. The decision maker maximizes expected utility. Choices among competing goals are handled by indifference curves— generally postulated to be smooth (twice differentiable)—that specify substitutability among goals. A major implication of the approach is that behavior is determined by the mix of incentives facing the decision maker. A second implication is that adjustment to these incentives is instantaneous; true maximizers have no learning curves. Like comprehensive rationality, bounded rationality assumes that actors are goal-oriented, but bounded rationality takes into account the cognitive limitations of decision makers in attempting to achieve those goals. Its scientific approach is different; rather than making assumptions about decision making and modeling the implications mathematically for aggregate behavior (as in markets or legislatures), bounded rationality adopts an explicitly behavioral stance. The behavior of decision makers must be examined, whether in the laboratory or in the field. The Birth of Bounded Rationality Simon (1999; see also Simon 1996a) reminds political scientists that the notion of bounded rationality and many of its ramifications originated in political science. Over his long career, Simon made major contributions not only to political science (as the founder of the behavioral study of organizations) but also to economics (as a Nobelist), psychology (as a founding father of cognitive psychology), and computer science (as an initiator of the field of artificial intelligence). In the 1940s and 1950s, Simon developed a model of choice intended as a challenge to the comprehensive rationality assumptions used in economics. BOUNDED RATIONALITY 299 Annu. Rev. Polit. Sci. 1999.2:297-321. Downloaded from arjournals.annualreviews.org by UNIVERSITY OF WASHINGTON - HEALTH SCIENCES LIBRARIES on 09/25/06. For personal use only
300 JONES The model first appeared in print in Administrative Behavior (1947),which critiqued existing theories of public administration and proposed a new ap- proach for the study of organizational decision making.Simon gave great credit for the initiation of his innovative work to the behavioral revolution in political science at the University of Chicago,where he studied for all of his academic degrees.Although most political scientists are aware of Simon's contributions,many fail to appreciate that bounded rationality was the first, and because of its ripple effects in so many disciplines,the most important idea asn (even academic school of thought)that political science has ever exported.I A briefretelling of the tale is in order.As an undergraduate at the University of Chicago,Simon returned to his native Milwaukee in 1935 to observe budg- eting in the city's recreation department.He wrote: I came as a gift-bearing Greek,fresh from an intermediate price theory course taught by the grandfather of Chicago-School neoclassical laissez- faire economics,Henry Simons....My economics training showed me how to budget rationally.Simply compare the marginal utility of a proposed expen- diture with its marginal cost,and approve it only if the utility exceeds the cost.However,what I saw in Milwaukee didn't seem to be an application of this rule.I saw a lot of bargaining,of reference back to last year's budget,and incremental changes in it.If the word"marginal"was ever spoken,I missed it.Moreover,which participants would support which items was quite predictable....I could see a clear connection between people's positions on budget matters and the values and beliefs that prevailed in their sub- organizations. I brought back to my friends and teachers in economics two gifts,which I 129-562:c6661 ultimately called "organizational identification"and "bounded rationality." (Simon 1999) In his autobiography,Simon noted the importance of these two notions for his later contributions to organization theory,economics,psychology,and computer science."I would not object to having my whole scientific output described as largely a gloss-a rather elaborate gloss,to be sure-[on these two ideas]"(Simon 1996a:88). Bounded rationality and organizational identification (now considered a consequence of bounded rationality)won ready acceptance in political sci- ence,with its emerging empiricist orientation,but they were largely ignored in the more theoretical discipline of economics.Or,as Simon (1999)puts it, ITwo recent incidents convinced me of the need to remind political scientists that Simon's "tribal allegiance"(1999)is to our discipline.A well-regarded political scientist recently commented,"I didn't know that Simon was a political scientist."In a written review,a cognitive psychologist somewhat haughtily informed me that Simon's work on organizations,and in particular March Simon's Organizations(1958),was intended to extend his work on problem solving to organizational behavior.Of course,the intellectual path was the other way around
The model first appeared in print in Administrative Behavior (1947), which critiqued existing theories of public administration and proposed a new approach for the study of organizational decision making. Simon gave great credit for the initiation of his innovative work to the behavioral revolution in political science at the University of Chicago, where he studied for all of his academic degrees. Although most political scientists are aware of Simon’s contributions, many fail to appreciate that bounded rationality was the first, and because of its ripple effects in so many disciplines, the most important idea (even academic school of thought) that political science has ever exported.1 A brief retelling of the tale is in order. As an undergraduate at the University of Chicago, Simon returned to his native Milwaukee in 1935 to observe budgeting in the city’s recreation department. He wrote: I came as a gift-bearing Greek, fresh from an intermediate price theory course taught by the grandfather of Chicago-School neoclassical laissezfaire economics, Henry Simons.... My economics training showed me how to budget rationally. Simply compare the marginal utility of a proposed expenditure with its marginal cost, and approve it only if the utility exceeds the cost. However, what I saw in Milwaukee didn’t seem to be an application of this rule. I saw a lot of bargaining, of reference back to last year’s budget, and incremental changes in it. If the word “marginal” was ever spoken, I missed it. Moreover, which participants would support which items was quite predictable.... I could see a clear connection between people’s positions on budget matters and the values and beliefs that prevailed in their suborganizations. I brought back to my friends and teachers in economics two gifts, which I ultimately called “organizational identification” and “bounded rationality.” (Simon 1999) In his autobiography, Simon noted the importance of these two notions for his later contributions to organization theory, economics, psychology, and computer science. “I would not object to having my whole scientific output described as largely a gloss—a rather elaborate gloss, to be sure—[on these two ideas]” (Simon 1996a:88). Bounded rationality and organizational identification (now considered a consequence of bounded rationality) won ready acceptance in political science, with its emerging empiricist orientation, but they were largely ignored in the more theoretical discipline of economics. Or, as Simon (1999) puts it, 300 JONES 1 1Two recent incidents convinced me of the need to remind political scientists that Simon’s “tribal allegiance” (1999) is to our discipline. A well-regarded political scientist recently commented, “I didn’t know that Simon was a political scientist.” In a written review, a cognitive psychologist somewhat haughtily informed me that Simon’s work on organizations, and in particular March & Simon’s Organizations (1958), was intended to extend his work on problem solving to organizational behavior. Of course, the intellectual path was the other way around. Annu. Rev. Polit. Sci. 1999.2:297-321. Downloaded from arjournals.annualreviews.org by UNIVERSITY OF WASHINGTON - HEALTH SCIENCES LIBRARIES on 09/25/06. For personal use only
BOUNDED RATIONALITY 301 economists "mostly ignored [bounded rationality]and went on counting the angels on the heads of neoclassical pins." Procedural Rationality Simon spent a great deal of time and energy attacking the abstract and rarefied economic decision-making models.Much of his attack was negative-show- ing how the model did not comport with how people really made decisions. asn But Simon also developed what he termed a procedural model of rationality, based on the psychological process ofreasoning-in particular his explanation of how people conduct incomplete searches and make tradeoffs between values. Since the organism,like those of the real world,has neither the senses nor the wits to discover an"optimal"path-even assuming the concept of optimal to be clearly defined-we are concerned only with finding a choice mechanism that will lead it to pursue a"satisficing"path that will permit satisfaction at some specified level of all of its needs.(Simon 1957:270-71) Simon elaborated on his "satisficing"organism over the years,but its fundamental characteristics did not change.They include the following: 1.Limitation on the organism's ability to plan long behavior sequences,a limitation imposed by the bounded cognitive ability of the organism as well as the complexity of the environment in which it operates. -129-562:26661 2.The tendency to set aspiration levels for each of the multiple goals that the organism faces. 3.The tendency to operate on goals sequentially rather than simultaneously because of the"bottleneck of short-term memory." 4.Satisficing rather than optimizing search behavior. An alternative satisfices if it meets aspirations along all dimensions (attrib- utes).If no such alternative is found,a search is undertaken for new alterna- tives. ALISHAINn Meanwhile,aspirations along one or more dimensions drift down gradually until a satisfactory new alternative is found or some existing alternative satis- fices(Simon 1996b:30). B In detailing the general requirements of an organism operating under bounded(as contrasted with comprehensive)rationality,Simon(1983:20-22; see also Simon 1995)notes the following requisites:(a)"Some way of focus- ing attention,”(b)“a mechanism for generating alternatives,”(c)“a capacity for acquiring facts about the environment,"and (d)"a modest capacity for drawing inferences from these facts
economists “mostly ignored [bounded rationality] and went on counting the angels on the heads of neoclassical pins.” Procedural Rationality Simon spent a great deal of time and energy attacking the abstract and rarefied economic decision-making models. Much of his attack was negative—showing how the model did not comport with how people really made decisions. But Simon also developed what he termed a procedural model of rationality, based on the psychological process of reasoning—in particular his explanation of how people conduct incomplete searches and make tradeoffs between values. Since the organism, like those of the real world, has neither the senses nor the wits to discover an “optimal” path—even assuming the concept of optimal to be clearly defined—we are concerned only with finding a choice mechanism that will lead it to pursue a “satisficing” path that will permit satisfaction at some specified level of all of its needs. (Simon 1957:270–71) Simon elaborated on his “satisficing” organism over the years, but its fundamental characteristics did not change. They include the following: 1. Limitation on the organism’s ability to plan long behavior sequences, a limitation imposed by the bounded cognitive ability of the organism as well as the complexity of the environment in which it operates. 2. The tendency to set aspiration levels for each of the multiple goals that the organism faces. 3. The tendency to operate on goals sequentially rather than simultaneously because of the “bottleneck of short-term memory.” 4. Satisficing rather than optimizing search behavior. An alternative satisfices if it meets aspirations along all dimensions (attributes). If no such alternative is found, a search is undertaken for new alternatives. Meanwhile, aspirations along one or more dimensions drift down gradually until a satisfactory new alternative is found or some existing alternative satisfices (Simon 1996b:30). In detailing the general requirements of an organism operating under bounded (as contrasted with comprehensive) rationality, Simon (1983:20–22; see also Simon 1995) notes the following requisites: (a) “Some way of focusing attention,” (b) “a mechanism for generating alternatives,” (c) “a capacity for acquiring facts about the environment,” and (d) “a modest capacity for drawing inferences from these facts.” BOUNDED RATIONALITY 301 Annu. Rev. Polit. Sci. 1999.2:297-321. Downloaded from arjournals.annualreviews.org by UNIVERSITY OF WASHINGTON - HEALTH SCIENCES LIBRARIES on 09/25/06. For personal use only
302 JONES I cannot do justice to the importance for other disciplines of Simon's "gloss"on bounded rationality.Just one note:The study of problem solving is grounded in the intended rationality of problem solvers,as is the study of judg- ment (Newell 1968,1990).By imposing a task environment,experimenters can examine that part of the problem solver's behavior that may be explained objectively,via the nature of the task environment,and compare it with that part that can be explained only with reference to failures to overcome system- atic internal limitations-bounded rationality showing through (Newell asn Simon 1972,Simon 1996b). The principle that rationality is intended but not always achieved,that what JO'SMOl "shows through"from the inner environment of the problem solver can be 2 systematically studied,is a principle that I consider extraordinarily useful in the study of human behavior in relatively set institutional task environments. BOUNDED RATIONALITY IN POLITICAL SCIENCE Bounded rationality has been a key component since the 1950s in public- administration and public-policy studies.In more recent times,partly in reac- tion to the attitudinal model of voting behavior,the approach has been used to understand political reasoning(Iyengar 1990,Sniderman et al 1991,Marcus McKuen 1993).Nevertheless,bounded rationality,born in organization the- ory(Simon 1947),has had its greatest impact in political science in the study of governmental organizations. 会 125-76 The fundamental premise underlying organizational studies in political science is that the behavior of organizations mimics the bounded rationality of the actors that inhabit them(March 1994).This correspondence is not simply 26661 an analogy among phenomena at different levels;the relationship is causal. This premise characterized behavioral organization theory generally,along with the insistence that organizational science be grounded in the observation of behavior in (and analysis of data from)organizational settings.The most important components of the political theory of organizations were the con- cepts of limited attention spans,habituation and routine,and organizational identification.Behavioral organization theory,unlike the subjective expected- utility approach,viewed uncertainty not as simple probabilities attached to specified outcomes,but as infecting the very specification of outcomes them- selves. B Over and over again,students of the behavior of public organizations reported findings that did not comport with the demands of"objective ration- ality"(Simon 1985:294).Search was incomplete,selective,and nonoptimal (Simon 1985,Jones Bachelor 1994).Decision makers did not need simply to choose among alternatives;they had to generate the alternatives in the first place(Simon 1983,1996b;Chisholm 1995).Problems were not givens;they
I cannot do justice to the importance for other disciplines of Simon’s “gloss” on bounded rationality. Just one note: The study of problem solving is grounded in the intended rationality of problem solvers, as is the study of judgment (Newell 1968, 1990). By imposing a task environment, experimenters can examine that part of the problem solver’s behavior that may be explained objectively, via the nature of the task environment, and compare it with that part that can be explained only with reference to failures to overcome systematic internal limitations—bounded rationality showing through (Newell & Simon 1972, Simon 1996b). The principle that rationality is intended but not always achieved, that what “shows through” from the inner environment of the problem solver can be systematically studied, is a principle that I consider extraordinarily useful in the study of human behavior in relatively set institutional task environments. BOUNDED RATIONALITY IN POLITICAL SCIENCE Bounded rationality has been a key component since the 1950s in publicadministration and public-policy studies. In more recent times, partly in reaction to the attitudinal model of voting behavior, the approach has been used to understand political reasoning (Iyengar 1990, Sniderman et al 1991, Marcus & McKuen 1993). Nevertheless, bounded rationality, born in organization theory (Simon 1947), has had its greatest impact in political science in the study of governmental organizations. The fundamental premise underlying organizational studies in political science is that the behavior of organizations mimics the bounded rationality of the actors that inhabit them (March 1994). This correspondence is not simply an analogy among phenomena at different levels; the relationship is causal. This premise characterized behavioral organization theory generally, along with the insistence that organizational science be grounded in the observation of behavior in (and analysis of data from) organizational settings. The most important components of the political theory of organizations were the concepts of limited attention spans, habituation and routine, and organizational identification. Behavioral organization theory, unlike the subjective expectedutility approach, viewed uncertainty not as simple probabilities attached to specified outcomes, but as infecting the very specification of outcomes themselves. Over and over again, students of the behavior of public organizations reported findings that did not comport with the demands of “objective rationality” (Simon 1985:294). Search was incomplete, selective, and nonoptimal (Simon 1985, Jones & Bachelor 1994). Decision makers did not need simply to choose among alternatives; they had to generate the alternatives in the first place (Simon 1983, 1996b; Chisholm 1995). Problems were not givens; they 302 JONES Annu. Rev. Polit. Sci. 1999.2:297-321. Downloaded from arjournals.annualreviews.org by UNIVERSITY OF WASHINGTON - HEALTH SCIENCES LIBRARIES on 09/25/06. For personal use only
BOUNDED RATIONALITY 303 had to be defined(Rochefort Cobb 1994).Solutions did not automatically follow problems;sometimes actors had set solutions ready to apply to prob- lems that could occur (Cohen et al 1972,Kingdon 1996,Jones Bachelor 1994).Choice was based on incommensurate goals,which were ill-integrated (March 1978;Simon 1983,1995;Jones 1994).Organizations seemed to have limited attention spans and,at least in major policy changes,serial processing capacity (Simon 1983,Jones 1994,Cobb Elder 1972,Kingdon 1996). The three most important strands of research stemming from behavioral or- asn ganizational theory in political science focused on incremental budgeting,on the impacts oforganizational routine on policy outputs,and on policy agendas. JO'SMOl euosjad 2 Incremental Budgeting Incremental decision making was developed not only as a descriptive model of decisions by bounded actors but as a normative mechanism for use in an uncer- ueunoeo popeojuo 9008.6000 tain world (Lindblom 1959).If people are handicapped by limited cognition, and if the world is fundamentally complex and ambiguous,then it made sense for a decision maker to (a)move away from problems,rather than toward solu- tions;(b)make only small moves away from the problem;and (c)be willing to reverse direction based on feedback from the environment.Wildavsky (1964; see also Fenno 1966,Meltsner 1971),in his classic observational studies of federal budgeting,noted that such incremental budgeting was governed by de- cision rules based on two norms:base and fair share.What was the agency's base,and what was a fair share given changes in the agency's environment since last year's budget meeting?Incrementalism was even criticized as too N rational a characterization of budget processes,because of the adoption of roles by budget decision makers(Anton 1966,Crecine 1969).Incrementalism, 6l in effect a small-step hill-climbing algorithm,implied adjustment to local op- tima rather than global ones. Incrementalism in decision making implied incrementalism in organiza- tional outcomes-so long as one also modeled exogenous"shocks"(Davis et al 1966,1974).Students of the budgetary process concluded that incremental- 台 ism did not fit even endogenous decision processes(Wanat 1974,Gist 1982). Pure incrementalism did not seem to characterize governing organizations.In essence,there were too many large changes in budget processes.But it was realized that attentional processes are selective (as the incremental model rec- ognized)and subject to occasional radical shifts.Incorporating this aspect of 3 attentional processes better accounts for the distribution of budget outcomes Padgett1980,1981;Jones et al1996,1997,1998). Organizational Habits and Routines Cognitive limits of human decision makers imposed limits on the ability of the organization to adjust to its environment.Rather than maximizing,organiza-
had to be defined (Rochefort & Cobb 1994). Solutions did not automatically follow problems; sometimes actors had set solutions ready to apply to problems that could occur (Cohen et al 1972, Kingdon 1996, Jones & Bachelor 1994). Choice was based on incommensurate goals, which were ill-integrated (March 1978; Simon 1983, 1995; Jones 1994). Organizations seemed to have limited attention spans and, at least in major policy changes, serial processing capacity (Simon 1983, Jones 1994, Cobb & Elder 1972, Kingdon 1996). The three most important strands of research stemming from behavioral organizational theory in political science focused on incremental budgeting, on the impacts of organizational routine on policy outputs, and on policy agendas. Incremental Budgeting Incremental decision making was developed not only as a descriptive model of decisions by bounded actors but as a normative mechanism for use in an uncertain world (Lindblom 1959). If people are handicapped by limited cognition, and if the world is fundamentally complex and ambiguous, then it made sense for a decision maker to (a) move away from problems, rather than toward solutions; (b) make only small moves away from the problem; and (c) be willing to reverse direction based on feedback from the environment. Wildavsky (1964; see also Fenno 1966, Meltsner 1971), in his classic observational studies of federal budgeting, noted that such incremental budgeting was governed by decision rules based on two norms: base and fair share. What was the agency’s base, and what was a fair share given changes in the agency’s environment since last year’s budget meeting? Incrementalism was even criticized as too rational a characterization of budget processes, because of the adoption of roles by budget decision makers (Anton 1966, Crecine 1969). Incrementalism, in effect a small-step hill-climbing algorithm, implied adjustment to local optima rather than global ones. Incrementalism in decision making implied incrementalism in organizational outcomes—so long as one also modeled exogenous “shocks” (Davis et al 1966, 1974). Students of the budgetary process concluded that incrementalism did not fit even endogenous decision processes (Wanat 1974, Gist 1982). Pure incrementalism did not seem to characterize governing organizations. In essence, there were too many large changes in budget processes. But it was realized that attentional processes are selective (as the incremental model recognized) and subject to occasional radical shifts. Incorporating this aspect of attentional processes better accounts for the distribution of budget outcomes (Padgett 1980, 1981; Jones et al 1996, 1997, 1998). Organizational Habits and Routines Cognitive limits of human decision makers imposed limits on the ability of the organization to adjust to its environment. Rather than maximizing, organizaBOUNDED RATIONALITY 303 Annu. Rev. Polit. Sci. 1999.2:297-321. Downloaded from arjournals.annualreviews.org by UNIVERSITY OF WASHINGTON - HEALTH SCIENCES LIBRARIES on 09/25/06. For personal use only
304 JONES tions tended to adopt task performance rules,which routinized even the most important decisions of the organization (March Simon 1958).Firms routi- nized price and output decisions(Cyert March 1963).Learning in organiza- tions seemed to be a slow,evolutionary,conflictual process (Sabatier Jenkins-Smith 1993,Lounamaa March 1985,Ostrom 1990)rather than the instantaneous adjustment process that rational organization theory would im- ply.Participants identified with the rules of the organization,adhering to them asn even in the face of evidence of problems (Jones 1980,1985).This could cause disjoint"lurches"as organizations were finally forced to adjust to changes in their environments (Dodd 1994). JO'SMOl Routines in service organizations invariably generated unintended conse- 2 quences,many of which went unrecognized or unaddressed.For example, distributional consequences of supposedly neutral rules were often ignored 9086000 (Levy 1974,Mladenka 1978). In other cases,an organization might have contradictory demands on it. Such contradictory demands are handled in economics via indifference curves, which specify a decision maker's preferences under all combinations of the de- mands.Instead of a rational process for handling tradeoffs,public service or- ganizations tended to develop task performance rules for each demand.The re- sponse of the organization depended on which set of rules was activated.A study of Chicago's Building Department revealed that two sets of task per- formance rules were in effect.One set directed resources in accordance with the severity of the problem.These rules embodied the classic administrative norm of neutral competence.A second set of rules,less explicit but just as im- 世w沿 portant,directed resources based on responsiveness to political forces.The distribution of organizational outputs to neighborhoods depended on an atten- 666 tion rule,activated by middle management,that governed which set of rules was to be put in force.Neutral competence was the default;response to politi- cal forces required an override of standard operating procedures,but the atten- 美 tion rule override happened so often that it could easily be detected in organ- izational outputs (Jones 1985). Policy Agendas If individuals have limited attention spans,so must organizations.The notion of policy agendas recognizes the "bottleneck"that exists in the agenda that any policy-making body addresses (Cobb Elder 1972).These attention pro- B cesses are not simply related to task environments-problems can go for long periods of time without attracting the attention of policy makers(Rochefort Cobb 1994).A whole style of politics emerges as actors must strive to cope with the limits in the attentiveness ofpolicy makers-basically trying to attract allies to their favored problems and solutions.This style of politics depends on connections driven by time-dependent and often emotional attention processes
tions tended to adopt task performance rules, which routinized even the most important decisions of the organization (March & Simon 1958). Firms routinized price and output decisions (Cyert & March 1963). Learning in organizations seemed to be a slow, evolutionary, conflictual process (Sabatier & Jenkins-Smith 1993, Lounamaa & March 1985, Ostrom 1990) rather than the instantaneous adjustment process that rational organization theory would imply. Participants identified with the rules of the organization, adhering to them even in the face of evidence of problems (Jones 1980, 1985). This could cause disjoint “lurches” as organizations were finally forced to adjust to changes in their environments (Dodd 1994). Routines in service organizations invariably generated unintended consequences, many of which went unrecognized or unaddressed. For example, distributional consequences of supposedly neutral rules were often ignored (Levy 1974, Mladenka 1978). In other cases, an organization might have contradictory demands on it. Such contradictory demands are handled in economics via indifference curves, which specify a decision maker’s preferences under all combinations of the demands. Instead of a rational process for handling tradeoffs, public service organizations tended to develop task performance rules for each demand. The response of the organization depended on which set of rules was activated. A study of Chicago’s Building Department revealed that two sets of task performance rules were in effect. One set directed resources in accordance with the severity of the problem. These rules embodied the classic administrative norm of neutral competence. A second set of rules, less explicit but just as important, directed resources based on responsiveness to political forces. The distribution of organizational outputs to neighborhoods depended on an attention rule, activated by middle management, that governed which set of rules was to be put in force. Neutral competence was the default; response to political forces required an override of standard operating procedures, but the attention rule override happened so often that it could easily be detected in organizational outputs (Jones 1985). Policy Agendas If individuals have limited attention spans, so must organizations. The notion of policy agendas recognizes the “bottleneck” that exists in the agenda that any policy-making body addresses (Cobb & Elder 1972). These attention processes are not simply related to task environments—problems can go for long periods of time without attracting the attention of policy makers (Rochefort & Cobb 1994). A whole style of politics emerges as actors must strive to cope with the limits in the attentiveness of policy makers—basically trying to attract allies to their favored problems and solutions. This style of politics depends on connections driven by time-dependent and often emotional attention processes 304 JONES Annu. Rev. Polit. Sci. 1999.2:297-321. Downloaded from arjournals.annualreviews.org by UNIVERSITY OF WASHINGTON - HEALTH SCIENCES LIBRARIES on 09/25/06. For personal use only
BOUNDED RATIONALITY 305 rather than a deliberate search for solutions(Cohen et al 1972,March Olsen 1989,Kingdon 1996,Baumgartner Jones 1993). Because attention processes are time dependent and policy contexts change temporally,connections between problems and solutions have time depend- ency built into them.As an important consequence,policy systems dominated by boundedly rational decision makers will at best reach local rather than global optima.Because of the time dependence of attentional processes,all policy processes will display considerable path dependence (March 1994) asn OBJECTIONS TO THE EXPECTED-UTILITY MODEL: BEHAVIORAL DECISION THEORY 豆 ou'seuofe wo popeojumo The expected-utility model incorporates risk and uncertainty into models of 9008960210 rational choice.Instead of maximizing utility,decision makers maximize expected utility in choice situations in which the consequences of choice are risky(may be characterized by known probabilities)or uncertain (are charac- terized by unspecified probabilities). Numerous empirical studies of human decision making,from experiments in the laboratory to large-scale social surveys to observational studies in the field,have demonstrated that humans often do not conform to the strictures of choice theory (Slovak 1990).This study of how people actually behave in choice situations is known as behavioral decision theory.Even defenders of choice theory have retreated in the face of the onslaught of empirical findings. Expected-utility theory is no longer seriously entertained as an accurate de- 世 scriptive theory (Halpern Stern 1998b). Again,this does not imply that people are irrational,nor that people inter- 666 acting in large-scale institutions make large-scale mistakes.Intendedly ra- tional actors in large-scale institutions may respond collectively to the tasks they face adaptively.Wittman(1995:16)notes that"even if some individuals make incorrect choices,the law of large numbers is likely to yield the correct majority choice.” 台 Many of these objections are quite fundamental-so much so that it seems impossible to develop a serious empirical theory of choice without taking them into consideration.They address both(a)the limitations of humans to com- prehend and act on inputs from the environment and (b)the fundamental complexity of the environment,which is vastly underestimated in standard B rational choice theories. The Nature of the Decision Maker Empirical objections to rational choice are so voluminous that they are,in ef- fect,a laundry list of problems.The first set has to do with the nature of the de- cision maker
rather than a deliberate search for solutions (Cohen et al 1972, March & Olsen 1989, Kingdon 1996, Baumgartner & Jones 1993). Because attention processes are time dependent and policy contexts change temporally, connections between problems and solutions have time dependency built into them. As an important consequence, policy systems dominated by boundedly rational decision makers will at best reach local rather than global optima. Because of the time dependence of attentional processes, all policy processes will display considerable path dependence (March 1994). OBJECTIONS TO THE EXPECTED-UTILITY MODEL: BEHAVIORAL DECISION THEORY The expected-utility model incorporates risk and uncertainty into models of rational choice. Instead of maximizing utility, decision makers maximize expected utility in choice situations in which the consequences of choice are risky (may be characterized by known probabilities) or uncertain (are characterized by unspecified probabilities). Numerous empirical studies of human decision making, from experiments in the laboratory to large-scale social surveys to observational studies in the field, have demonstrated that humans often do not conform to the strictures of choice theory (Slovak 1990). This study of how people actually behave in choice situations is known as behavioral decision theory. Even defenders of choice theory have retreated in the face of the onslaught of empirical findings. Expected-utility theory is no longer seriously entertained as an accurate descriptive theory (Halpern & Stern 1998b). Again, this does not imply that people are irrational, nor that people interacting in large-scale institutions make large-scale mistakes. Intendedly rational actors in large-scale institutions may respond collectively to the tasks they face adaptively. Wittman (1995:16) notes that “even if some individuals make incorrect choices, the law of large numbers is likely to yield the correct majority choice.” Many of these objections are quite fundamental—so much so that it seems impossible to develop a serious empirical theory of choice without taking them into consideration. They address both (a) the limitations of humans to comprehend and act on inputs from the environment and (b) the fundamental complexity of the environment, which is vastly underestimated in standard rational choice theories. The Nature of the Decision Maker Empirical objections to rational choice are so voluminous that they are, in effect, a laundry list of problems. The first set has to do with the nature of the decision maker. BOUNDED RATIONALITY 305 Annu. Rev. Polit. Sci. 1999.2:297-321. Downloaded from arjournals.annualreviews.org by UNIVERSITY OF WASHINGTON - HEALTH SCIENCES LIBRARIES on 09/25/06. For personal use only
306 JONES SEARCH BEHAVIOR In general,people do not consider all aspects of a deci- sion facing them.They must factor the decision to make it manageable,ex- amining only relevant aspects.They do not undertake complete searches for information,and they ignore available information-especially if it is not relevant to the factors they have determined to characterize the structure of the problem. SEARCH MUST INCLUDE BOTH ALTERNATIVES AND ATTRIBUTES Different asn physiological and psychological mechanisms probably underlie the search for euosjad attributes(which is equated in ordinary language with understanding a prob- JO'SMOl lem)and the search for alternatives (which involves the choice under a given 豆 decisional structure,design,or understanding)(Jones 1996). CALCULATIONS People generally cannot perform the calculations necessary useuofeo popeojuMo 9000600 even for a reduced set of options in a decision-making situation.This is actu- ally the least problematic limitation in decision making.They can,given time, write down and manipulate the numbers. COGNITIVE ILLUSIONS AND FRAMING When identical options are described in different terms,people often shift their choices.For example,if a choice is described in terms of gains,it is often treated differently than if it is described in terms of losses.This shift demonstrates the concept of framing,developed by psychologists Daniel Kahneman and Amos Tversky.They claim that this tendency violates a major,if often unstated,assumption of rational choice- namely the axiom of invariance,which states that the "preference order be- tween prospects should not depend on the manner in which they are described" (Kahneman Tversky 1983:343).They bolster their claim with numerous 666l convincing experiments indicating that decision makers tend to choose differ- ent alternatives when they are described in positive terms (for example,in terms of the number of lives saved with a vaccine)than when they are de- scribed in negative terms (the number of people who will die).Kahneman Tversky(1983:343)state,"In their stubborn appeal,framing effects resemble perceptual illusions more than computational errors." SELF-CONTROL People often seem to need to bind themselves in some way to establish self-control over their behavior in the future.A major mechanism for dealing with likely future lapses in self-control is to establish binding rules that 3 prohibit the unwanted behavior.For example,Thaler(1991)has developed the notion of mental accounting to explain the tendency of people to separate cate- gories of income and impose more constraints on some (investment income) than on others (a Christmas bonus).People also tend to treat gains differently from losses,applying different risk functions to them,essentially being more risk-adverse for gains than for losses(Kahneman Tversky 1983,1985)
SEARCH BEHAVIOR In general, people do not consider all aspects of a decision facing them. They must factor the decision to make it manageable, examining only relevant aspects. They do not undertake complete searches for information, and they ignore available information—especially if it is not relevant to the factors they have determined to characterize the structure of the problem. SEARCH MUST INCLUDE BOTH ALTERNATIVES AND ATTRIBUTES Different physiological and psychological mechanisms probably underlie the search for attributes (which is equated in ordinary language with understanding a problem) and the search for alternatives (which involves the choice under a given decisional structure, design, or understanding) (Jones 1996). CALCULATIONS People generally cannot perform the calculations necessary even for a reduced set of options in a decision-making situation. This is actually the least problematic limitation in decision making. They can, given time, write down and manipulate the numbers. COGNITIVE ILLUSIONS AND FRAMING When identical options are described in different terms, people often shift their choices. For example, if a choice is described in terms of gains, it is often treated differently than if it is described in terms of losses. This shift demonstrates the concept of framing, developed by psychologists Daniel Kahneman and Amos Tversky. They claim that this tendency violates a major, if often unstated, assumption of rational choice— namely the axiom of invariance, which states that the “preference order between prospects should not depend on the manner in which they are described” (Kahneman & Tversky 1983:343). They bolster their claim with numerous convincing experiments indicating that decision makers tend to choose different alternatives when they are described in positive terms (for example, in terms of the number of lives saved with a vaccine) than when they are described in negative terms (the number of people who will die). Kahneman & Tversky (1983:343) state, “In their stubborn appeal, framing effects resemble perceptual illusions more than computational errors.” SELF-CONTROL People often seem to need to bind themselves in some way to establish self-control over their behavior in the future. A major mechanism for dealing with likely future lapses in self-control is to establish binding rules that prohibit the unwanted behavior. For example, Thaler (1991) has developed the notion of mental accounting to explain the tendency of people to separate categories of income and impose more constraints on some (investment income) than on others (a Christmas bonus). People also tend to treat gains differently from losses, applying different risk functions to them, essentially being more risk-adverse for gains than for losses (Kahneman & Tversky 1983, 1985). 306 JONES Annu. Rev. Polit. Sci. 1999.2:297-321. Downloaded from arjournals.annualreviews.org by UNIVERSITY OF WASHINGTON - HEALTH SCIENCES LIBRARIES on 09/25/06. For personal use only