Ambiguity and Multiple Streams
When leaders in ancient Greek times confronted a critical juncture in political life, they often sought Apollo’s advice in Delphi. The god’s words, delivered obscurely through his priestess, Pythia, and construed shrewdly by priests, offered a window into a possible future. The process was ritualistic and mystical, and the results were always subject to interpretation. Not only did priests interpret Pythia’s utterings as they saw fit, but recipients also interpreted the oracle handed down by the priests in ways that suited them. The ambiguity inherent in the process enabled those involved to pursue self-interest and to infuse meaning into a partially comprehensible world. Modern democratic politics resembles Pythia’s cave quite well. The cave is an occasion to mix people, meaning, resources, self-interest, and institutions. It obfuscates and clarifies at the same time. Individuals play a key role in a process full of norms, rituals, and traditions. Ambiguity facilitates taking appropriate action and shaping preferences without a priori estimating the consequences.
The Multiple Streams Approach (MSA) is a lens or framework—I use the terms interchangeably—that explains how policies are made by government under conditions of ambiguity. It illuminates Pythia’s cave. As a framework of choice, it has proven useful in explaining national, supranational, and subnational policies in a variety of cross-national settings. Google Scholar lists over 10,000 citations since Kingdon’s seminal work in 1984. Some 1,900 peer-reviewed journal articles have mentioned or applied the framework since 2000. Deriving inspiration from organizational theory, it yields insight into the dynamics of the entire policy process—agenda setting, decision-making, and implementation. Three streams are identified as flowing through the policy system: problems, policies, and politics. Each is conceptualized as largely separate from the others, with its own dynamics and rules. At critical points in time, termed “policy windows,” the streams are coupled by policy entrepreneurs using a variety of strategies. The combination of all three streams into a single package dramatically enhances the chances that policymakers will adopt a specific policy.
The first section provides a panoramic view of the approach by presenting its assumptions and guiding logic. The second outlines the main structural elements of the framework. The third discusses the various processes by which elements come together to provide answers to the puzzle of choice. The fourth addresses limitations of the lens, and the conclusion proposes an agenda for future research.
ASSUMPTIONS AND LOGIC
The basic outline of MSA was put forth by Kingdon (1995) in the tradition of Cohen, March, and Olsen’s 1972 garbage can model of organizational choice. Collective choice is not merely the derivative of individual efforts aggregated in some fashion but rather the combined result of structural forces and cognitive and affective processes that are highly context dependent.
Level and Unit of Analysis
Pitched at the systemic level, MSA incorporates the entire system or a separate decision as the unit of analysis. Much like systems theory, it views choice as collective output formulated by the push and pull of several factors. It shares common ground with complexity theories in being attentive to dynamic interactions, in assuming a considerable amount of residual randomness, and in viewing systems as constantly evolving and not necessarily settling into equilibrium (Kingdon 1994, 219).
MSA deals with policymaking under conditions of ambiguity. Ambiguity refers to “a state of having many ways of thinking about the same circumstances or phenomena” (Feldman 1989, 5). These ways may not be reconcilable, creating vagueness, confusion, and stress. It is different from uncertainty, a related concept, in that the latter refers to the inability to accurately predict an event. Ambiguity may be thought of as ambivalence, whereas uncertainty may be referred to as ignorance or imprecision (March 1994, 178–179). Although more information may reduce uncertainty (Wilson 1989, 228), more information does not reduce ambiguity. For example, more information can tell us how AIDS is spread, but it still won’t tell us whether AIDS is a health, educational, political, or moral issue.
At the heart of the lens lies the garbage can model of choice (Cohen, March, and Olsen 1972). Choice is conceptualized as a garbage can into which participants, who drift in and out of decisions, dump largely unrelated problems and solutions. No one person controls the process, while fluctuating attendance, opportunities, and attention give choice highly interactive qualities. The point is to stress the dynamic, complex, and chaotic nature of political life. MSA was developed to explain policy in situations that more closely approximate empirical reality and under conditions that other frameworks simply assume away.
Kingdon (1995) adapts this model to policy output by the U.S. federal government while adding a few structural features of his own. Ambiguity is operationalized contextually. It exists in organizations or governments, which are called organized anarchies, and is measured by three indicators: fluid participation, problematic preferences, and unclear technology. First, participation in such organizations is fluid. Turnover is high, and participants drift from one decision to the next. Legislators come and go, and bureaucrats, especially high-level civil servants, often move from public service to private practice. Moreover, nongovernmental actors, such as employer associations, trade unions, and consumer groups, exercise a significant influence over the form certain decisions will take. The time and effort that participants devote to any one decision varies considerably.
Second, people often don’t know what they want. To say that policymakers almost never make their objectives crystal clear is hardly novel, but it is true that quite often time constraints force politicians to make decisions without having formulated precise preferences. Decisions are made as the process unfolds, and they may even be facilitated by opaqueness (Sharkansky 2002). This situation stands in stark contrast to decision-making in most business firms, where the ultimate goal is clear—to make a profit. As Cohen, March, and Olsen aptly put it, organized anarchies “can be described better as a collection of ideas than as a coherent structure” (1972, 1).
Third, technology—that is, an organization’s processes that turn inputs into products—is unclear. Members of an organized anarchy, such as a university or national government, may be aware of their individual responsibilities, but they exhibit only rudimentary knowledge of how their job fits into the overall mission of the organization. Jurisdictional boundaries are unclear, and turf battles between different departments or agencies are common. Members of the legislature often complain of unaccountable officials, who, in turn, frequently express their frustration with overburdening reporting rules and independent-minded public managers. Past experience often guides their actions, making trial-and-error procedures indispensable learning tools.
Under such extreme conditions, theories based on rational behavior are of limited utility. Because problems and preferences are not well known, selecting the alternative that yields the most net benefits is difficult. The problem under conditions of ambiguity is that we often don’t know what the problem is; its definition is vague and shifting. Distinguishing between relevant and irrelevant information is problematic, which can lead to false and misleading interpretations of facts. Choice becomes less an exercise in solving problems and more an attempt to make sense of a partially comprehensible world (Weick 1995). Contradictions and paradoxes appear: state agencies are told to strengthen their oversight functions; at the same time their budgets are slashed. Information is requested and produced but not used in any decisions (Feldman 1989).
Decisions are made by allocating attention through activating or overcoming temporal constraints and biases. Most decisions in garbage cans are made not by problem resolution but by flight (when problems leave the choice arena) or oversight (by action before the activation of problems) (Cohen, March, and Olsen 1972). The process is generally sensitive to energy load (the resources needed to make a decision) and problem load (the number of problems under consideration). In this way, the content of problems or solutions is as significant as—or even less so than (March 1994, 218)—the process and timing of decisions.
Who pays attention to what and when is critical. Time is a unique, irreplaceable resource whose supply is totally inelastic. As Drucker categorically asserts, “No matter how high the demand, the supply will not go up. There is no price for it and no marginal utility curve for it” (1985, 26). Because the primary concern of decision-makers—be they policymakers, business executives, or top civil servants—is to manage time effectively rather than to manage tasks (Drucker 1985; Mackenzie 1997), it is reasonable to pursue a lens that accords significance to time rather than rationality.
Three assumptions guide the framework. It is very important to make them explicit because, as Ruggie perceptively observes, “it’s not enough to be right; in the policy sciences, we also want to be right for the right reasons” (1998, 13).
Assumption 1: Individual attention or processing is serial; systemic attention or processing is parallel. On the one hand, because of biological and cognitive limitations, individuals can only attend to one issue at a time. This means that the number of issues under the active consideration of policymakers is relatively small. In addition, the number of pet projects that any one entrepreneur will push to adopt will be quite limited. On the other hand, division of labor in organizations or governments enables them to attend to many issues simultaneously (March and Simon 1958; Jones 2001). This capacity is, of course, not infinite, but government can simultaneously put out fires in California, conduct trade negotiations with the European Union, investigate mail fraud, and mourn the loss of soldiers killed in action.
Concern with processing capacity in decision-making was first introduced by Herbert Simon (1957). Multiple Streams shares with Simon’s concept of bounded rationality a focus on attention and search activities particularly as they relate to the order by which alternatives are considered. Both argue that the sequence by which solutions are considered strongly affects the decision outcome. They differ in the level of theorizing and the problem-solution sequence. Simon theorizes at the individual level and argues that individuals possess only serial-processing capacities. Computational and cognitive limitations substantially bias the search and selection processes, while environmental cues and organizational slack (excess supply of resources relative to demand) favor some solutions over others. Political systems, however, contain many subsystems that facilitate attention to many issues simultaneously, a phenomenon known as parallel processing. Consequently, attention and search can be quite abrupt and disorderly from the system’s point of view (Jones 1994). Whereas Simon in general imposes a boundedly rational order on the process of policymaking, theorizing from the micro to the macro level, MSA attempts to uncover rationality, theorizing from the macro to the micro.
Assumption 2: Policymakers operate under significant time constraints. These people often do not have the luxury of taking their time to make a decision. While this does not imply that all decisions are crises, it suggests that there is a sense of urgency in addressing them. Because many issues vie for attention, policymakers need to “strike while the iron is hot.” In effect, time constraints limit the range and number of alternatives to which attention is given.
Assumption 3: The streams flowing through the system are independent. This assumption is related to the first one in that if systems can do things in parallel, then each element or stream may be conceived as having a life of its own. The stream of problems includes concerns that individuals inside and outside the policy system have. Policies (solutions) are people’s products usually generated in narrow policy communities; they are answers to questions that may not be produced only when needed. Politics is a stream that refers to the broader political discourse within which policy is made. It includes legislators and parties, the national mood or climate of opinion, and so forth.
The Logic of Political Manipulation
If ambiguity is pervasive and central to politics, manipulation is the effort to manage ambiguity. It is a political struggle to create winners and losers, to provide meaning and identity, and to pursue self-interest. A central concept is information, which is not value neutral (Jones and Baumgartner 2005). Because ambiguity affords discretion in interpretation, it is strategically manipulated to serve diverse aims for different elements of the policy process. Although manipulation from the point of view of the entrepreneur might involve pursuing self-interest, it serves a different purpose from the point of view of the system. Political manipulation aims primarily to provide meaning, clarification, and identity. In a world replete with ambiguity, the most important aspect of entrepreneurial activity is not to pursue self-interest but to clarify or create meaning for those policymakers, and others, who have problematic preferences. As March puts it, “Decision making may, in many ways, be better conceived as a meaning factory than as an action factory” (1997, 23). It is precisely the inability to formulate preferences on the part of policymakers that makes entrepreneurs rationalists in their narrow pursuit of their pet proposal but meaning suppliers and identity providers in their coupling efforts.
The logic of political manipulation sets MSA apart from other lenses that employ rationality (rational choice) or persuasion (constructivism). Rationalists assume that individuals are utility maximizers. They behave opportunistically in the sense that they engage in deceitful behavior to exploit discrepancies in transaction costs during voluntary exchange (Williamson 1985). Although information feedback may be inadequate to allow every individual to devise optimal strategies, proponents of rationality assume that individuals have clear and consistent ways of arriving at the final decision. Constructivists conceive of policymaking as driven by persuasion and the social construction of identity and meaning. It is a process of deliberation between competing groups, each crafting a reasonable argument in ways that aim to persuade the other side(s). As Majone boldly states, “Argumentation is the key process through which citizens and policy makers arrive at moral judgments and policy choices” (1989, 2).
Despite some similarities, MSA differs from both rational choice and constructivism in central ways. MSA does not reject but complements rationality. Individuals sometimes behave rationally, but the process of making systemic decisions often does not exhibit rational properties. The lens differentiates between two groups of individuals: those who manipulate and those who get manipulated. Policy entrepreneurs are goal-intending manipulators. Policymakers are assumed to have problematic preferences and are subject to manipulation. Whereas rationalists assume that satisficing individuals choose the best option under certain conditions, MSA points out that whether a solution is “good enough” is determined politically by policymakers, not entrepreneurs. The problem-solution sequence and the politics of choice are affected by the degree of fragmentation in the politics and policy streams and the type of policy window. Moreover, policymakers and entrepreneurs utilize labels and symbols that affect meaning. Using these elements strategically alters the dynamics of choice by highlighting one dimension of the problem over others. The process of interpretation inherent in ambiguous situations and the power of discretion enable policymakers to legitimately deviate from established norms. They may overlook negative experiential learning that contradicts preferred policy (Moynihan 2006) or create fantasy documents that purport to cope with high-risk situations (Birkland 2009; Clarke 1999). It’s the strategic use of information in combination with institutions and policy windows that changes context, meaning, and policy over time.
The framework contains five structural elements: problems, policies, politics, policy windows, and policy entrepreneurs.
The problem stream consists of various conditions that policymakers and citizens want addressed. Examples are government budget deficits, environmental disasters, inflation, rising medical costs, and so on. Policymakers find out about these conditions by way of indicators, focusing events, and feedback. Indicators may be used to assess the existence and magnitude of a condition and the scope of change—for example, program costs, infant mortality rates, or highway deaths. Indicators can be monitored either routinely or through special studies. For example, special studies occasionally seek to estimate the number of Americans without health insurance. The drafted indicators can then be used politically to measure the magnitude of change in the hope of catching official attention (Stone 2011). British Conservatives used the Public Sector Borrowing Requirement to point attention to large fiscal imbalances and create the requisite conditions for action. Of course, not all conditions become problems. As Kingdon categorically asserts, problems contain a “perceptual, interpretive element” (1995, 110). Some conditions come to be defined as problems and consequently receive more attention than others (Rochefort and Cobb 1994). How is this done? A range of values is normally associated with a particular issue. Changes in specific conditions may violate those values and therefore activate interest and attention. People define conditions as problems by letting their values and beliefs guide their decisions, by placing subjects under one category rather than another, by comparing current to past performance, and by comparing conditions in different countries.
Focusing events also draw attention to problematic conditions (Birkland 1997). Conditions and focusing events direct attention to specific evaluative dimensions of particular problems; attention is fixed by the media or policy entrepreneurs (Jones 1994). There are several types of focusing events. Zahariadis (2003) offers the examples of prolonged rail strikes and two train accidents; Birkland (2004) offers the example of the 9/11 attacks; Farley et al. (2006) suggest Hurricane Katrina.
Feedback from previous programs is important in that it helps highlight what works and what may not. In this context, successfully implementing a solution in one area may facilitate the adoption of the same solution in a seemingly unrelated area. The cases of spillover of privatization from the area of oil to telecommunications in Britain in the early 1980s and across different countries in later years are good examples.
Attention is, to an extent, a function of what else preoccupies the minds of policymakers. Problems tend to appear to be more intractable when more of them crowd the agenda. Problem load—that is, the number of difficult problems occupying the attention of policymakers—has a significant negative effect on the efficient utilization of information and a strong positive effect on the ability to predict the issue’s place on the agenda (Zahariadis 2003, ch. 6). This point is particularly relevant in times of crisis. When many difficult problems crowd the agenda, policymakers tend to centralize decision-making to conserve resources and control the message (Peters 2011). However, information overload often paralyzes the process or compartmentalizes attention, forcing policymakers at the top to address problems that are often different from those experienced at the street level. Mindful of the fact they are more likely to be vilified when things go wrong than praised when things go right (Hood 2011), policymakers facing difficult problems will likely focus on those they may actually be able to solve rather than those that must be solved (Zahariadis 2013b).
The policy stream includes a “primeval soup” of ideas that compete to win acceptance in policy networks. Ideas are generated by specialists in policy communities—networks that include bureaucrats, congressional staff members, academics, and researchers in think tanks who share a common concern in a single policy area, such as health or environmental policy—and are assessed in various forums and forms, such as hearings, papers, and conversations. Some ideas survive this initial period basically unchanged, others are combined into new proposals, and still others just disappear. Although the number of ideas floating around is quite large, only a few ever receive serious consideration. Selection criteria include technical feasibility, value acceptability, and resource adequacy. Proposals that are or appear to be difficult to implement have fewer chances of surviving this process. Moreover, alternatives that do not conform to prevailing norms or the values of policymakers are less likely to be considered for adoption. Costlier proposals have a higher rate of failure.
Not all policy networks are created equal. In the original conception of MSA, ideas are recombined and rise to the top only incrementally (Kingdon 1995). Zahariadis (2003) has shown that this need not be the case. Institutional configurations—or, put differently, the level of integration—differ across countries, affecting the mode and tempo of ideas, that is, how ideas germinate in the policy stream and how fast they rise to prominence. Integration refers to linkages among participants and is distinguished by variations in four dimensions: size, mode, capacity, and access. Based on these dimensions, networks can be classified as more or less integrated. Less integrated networks are larger in size and have a competitive mode, lower administrative capacity, and less restricted access. Conversely, more integrated networks are smaller in size and have a consensual mode, higher capacity, and more restricted access. Further research will employ more refined network methods and concepts, such as centrality, reciprocity, or density, to add nuance to network interactions.
The politics stream consists of three elements: the national mood, pressure-group campaigns, and administrative or legislative turnover. The national mood refers to the notion that a fairly large number of individuals in a given country tend to think along common lines and that the mood swings from time to time. Government officials sensing changes in this mood, through, say, monitoring public opinion polls, act to promote certain items on the agenda or, conversely, to dim the hopes of others. In addition, politicians often view the support or opposition of interest groups as indicators of consensus or dissent in the broader political arena. For example, if many interest groups voice their support for deregulation, it is likely that government officials will hasten to include the item on the agenda. In the frequent case of conflicting views, politicians formulate an image of the balance of support and opposition. The perception that the balance is tilting one way or another directly affects the chances of the issue’s rising to prominence or falling to obscurity.
In addition to the aforementioned factors, legislative or administrative turnover frequently affects choice in quite dramatic ways. A sudden influx of new members of Congress ideologically predisposed against “big government” is likely to propel the issue of deregulation into high prominence. Moreover, turnover of key personnel in the administration has a significant influence on politics. The advent of a new president or new secretary of defense signifies potential changes. Certain issues, such as proposals to cut the budget, may receive more attention, whereas others, such as comprehensive national health insurance, may simply be pushed into obscurity. For example, the election of Bill Clinton in 1992 elevated national health care on the agenda to an extent not seen since the Richard Nixon administration; the Clinton administration’s failure to enact a health care bill then closed the window of opportunity for change. When President Barack Obama was elected, health care was again catapulted into prominence. Of the three elements in the political stream, the combination of the national mood and turnover in government exerts the most powerful effect on agendas.
Choices are made when the three streams are coupled or joined together at critical moments. Kingdon labels these moments “policy windows” and defines them as fleeting “opportunit[ies] for advocates of proposals to push their pet solutions, or to push attention to their special problems” (1995, 165). Windows define the context within which policy is made. Moreover, they act as catalysts for the adoption of policies, which may often be quite irrelevant to the issue at hand (e.g., Birkland 2004). Troubles arise when policy entrepreneurs use the wrong window to pursue their goals. For example, “by defining bioterrorism as a security rather than a public health issue, policy entrepreneurs [have] squander[ed] the opportunity to institute broad-based reforms that would improve not only the ability to manage a terrorist incident, but also meet other public health needs” (Avery 2004, 275).
Windows are opened by compelling problems or by events in the political stream. The crash of an airplane, for example, brings attention to air-safety issues (Cobb and Primo 2003). In the political stream, a new administration may be ideologically committed to deregulation. Policy windows are of short duration—although Sharp (1994) finds the opposite in relation to drug policy—and may be as predictable as annual budget allocations or as unpredictable as earthquakes. Predictable windows tend to be institutionalized, facilitating spillovers and issue-linkage attempts (Howlett 1998). Once open, some windows cast long shadows in that they leave a legacy that influences later events through the construction of path-dependent narratives (Dudley 2013). Sometimes open windows overlap (Copeland and James 2014), although MSA assumes windows open one at a time. For example, there may simultaneously and independently be a jet crash and the announcement of mechanical problems in certain aircraft (the aircraft need not be the same). Is there a hierarchy of problem (or politics) windows? If one assumes that windows in the politics stream lead to different outcomes than windows in the problem stream (more on that later), are politics windows more “important” than problem windows? Can windows be linked in the policy process, activating different participants and issues along a sequential trajectory (Ridde 2009)?
Policy entrepreneurs are individuals or corporate actors who attempt to couple the three streams. While individuals are mostly conceptualized as entrepreneurs, it is quite often the case that particular organizations, not just their individual representatives, are behind the push for certain policies. They are more than mere advocates of particular solutions; they are power brokers, coalition enablers, and manipulators of problematic preferences and unclear technology (Mintrom and Norman 2009). When windows open, policy entrepreneurs must immediately seize the opportunity to initiate action. Otherwise, the opportunity is lost, and the policy entrepreneurs must wait for the next one to come along. Entrepreneurs must be not only persistent but also skilled at coupling. They must be able to attach problems to their solutions and find politicians receptive to their ideas. A policy’s chances of being adopted dramatically increase when all three streams—problems, policies, and politics—are coupled in a single package.
Not all entrepreneurs are successful at all times. The more successful entrepreneurs are those who have greater access to policymakers. For example, the Adam Smith Institute had greater access to the government during Margaret Thatcher’s tenure in power in Britain because its ideologies matched hers more closely than those of other groups. Hence options put forth by individuals associated with the institute achieved greater receptivity among policymakers. Entrepreneurs with more resources (i.e., the ability to spend more time, money, and energy) to push their proposals have greater rates of success. Finally, entrepreneurs must also employ manipulating strategies to accomplish their goal of coupling the three streams.
How do the elements combine to produce choice? MSA offers answers to three questions of choice (Simon 1983): how is attention rationed, how is search conducted, and how is selection biased?
Attention is scarce. Not everything can be attended to at once. Policymakers need to ration their attention, dividing it among a limited number of issues. Limitations of this sort pose dilemmas for policymakers. MSA argues they are resolved by institutional structure, the type of policy window that opens, and the symbols used to attract attention. Attention to a particular issue is a function of opportunity, bias, formal position in an organization or government, and the number of issues competing for policymaker attention (March and Romelaer 1976). Policy entrepreneurs play a crucial role in capturing the attention of policymakers and manipulating it to their advantage.
Institutional structure strongly affects attention. Because policymakers at the top are frequently overwhelmed by the number and complexity of problems they encounter, they have designed institutions to ease overload. The entire system has been organized into policy communities or subsystems that act as filters in that problems and solutions usually incubate in those communities first before they are taken up by national-level politicians. Hierarchical structure reduces the number of issues to a manageable few and acts as a first step in sorting out available solutions. Moreover, institutional overlap facilitates attention and spillover. As Ackrill and Kay assert, “If a policy issue occupies multiple institutionally connected policy arenas, a policy decision taken in one arena may impact directly on policy decisions in others, even forcing a decision when none otherwise would be made” (2011, 73).
What people pay attention to depends partially on the structure of opportunities that evoke such focus. Choice often involves a problem-solution sequence. Rational choice theory, for instance, assumes that policymakers attend to problems first and then develop policies to solve them. MSA amends this argument by suggesting that opportunities ration attention. In cases when the window opens in the problem stream, the process is consequential—that is, solutions are developed in response to specific problems. For example, a flood or a hurricane (problem) points attention to and seeks to redress possible emergency management deficiencies (solution). If windows open in the politics stream, however, attention is focused on solutions first before problems can be clearly defined. In such cases the process is ideological—that is, policies are made in search of a rationale. What matters more is the solution to be adopted rather than the problem to be solved. Privatization in the United Kingdom is a good example of a policy in search of a rationale (Zahariadis 2003).
Attention is also influenced by the symbols used to attract it. Symbols have both emotive and cognitive functions. They transmit a simple message and arouse emotion. As Simon says, political symbols are particularly influential “in large part because [of their] evocative power, the ability to arouse and fix attention” (1983, 29). Higher-order symbols—that is, symbols that apply to the entire community—have more potency of affect, more uniformity of meaning across individuals, and greater durability of attention. Conditions of ambiguity facilitate political manipulation by way of symbolic politics. The chances of successfully coupling the problem, policy, and politics streams are greater when entrepreneurs attach higher-order symbols to their pet proposals. In this way policy entrepreneurs reach more people, evoke a stronger emotional reaction, convey gains and losses, and spend the least effort explaining exactly what their proposal is about. For example, symbols that derive from the core of a nation’s identity are more likely to facilitate adoption of the policy associated with them. They are also far more likely to make political discourse emotive rather than rational. Emotive arousal leads to the adoption of more confrontational policy (Zahariadis 2005).
The search for solutions and their availability (i.e., their evolution and diffusion) is closely linked to the concept of slack (Cyert and March 1963). Organizations consciously or unconsciously set aside time and resources for use in searching for innovative ideas. High-tech companies, for example, routinely give engineers free time and capacity to think and experiment with ideas not tightly linked to their projects at hand. The aim is to anticipate rather than predetermine the next big idea in rapidly changing environments. The policy process, especially in democratic systems, similarly contains slack. Governments maintain a range of instruments that they constantly decompose and reconstruct to address changing problems. Public and private actors routinely generate and advocate novel ideas. But search in the presence of ambiguity is a costly and politically contentious process. Resources may be spent looking for solutions to nonexistent problems. Government programs may remain idle for ideological or clientelist reasons. Under certain conditions ambiguity reduces political conflict by fostering the generation of solutions that appeal to different target audiences. In the absence of common goals, however, ambiguity dramatically increases the cost of searching for appropriate solutions precisely because each is molded to appeal to different audiences. Public money may be used in pursuit of private objectives but often in costly and unintelligible ways. While policy participants frequently promote solutions in the name of efficiency, they ironically use a very inefficient search process. Talk in policy communities is an innovative and expensive paradox.
Search is heavily influenced by the structure of policy networks. Where policymakers search for solutions and how ideas germinate in the “primeval soup,” to use Kingdon’s metaphor, depends on the degree of integration of the policy communities (or networks). The gestation period of ideas in the policy stream varies from rapid to gradual. The content ranges from totally new to a minor extension of the old. The typology that emerges from these criteria yields four types: quantum (rapid propulsion of new ideas), emergent (gradual gestation of new ideas), convergent (rapid gestation of old ideas), and gradualist (slow gestation of marginal extensions of existing policies) (Durant and Diehl 1989). Integration encourages one type of evolution rather than another. Less integrated networks are more likely to facilitate a quantum to gradualist evolution of ideas, and more integrated networks are likely to follow an emergent to convergent pattern. This is not to say that other combinations are not possible; rather, integration renders such evolutionary trajectories more likely. This hypothesis helps explain the ease with which ideas such as privatization have been gaining prominence among specialists in the United Kingdom but have relative difficulty doing the same in Germany (Zahariadis 2003).
Selection is biased by the manipulating strategies and skills of policy entrepreneurs, who couple problems, policies, and politics into a single package. Strategies include framing, affect priming, “salami tactics,” and the use of symbols. Prospect theory and affect priming theory impart the underlying logic of political manipulation by explaining how information is processed.
Problem representation (framing) makes a difference in what people perceive as losses or gains (Tversky and Kahneman 1981; Quattrone and Tversky 1988). The presentation of an option as a loss relative to the status quo tends to bias choice. People are generally loss averse in the sense that losses loom larger than gains. Jervis (1992) adds to this prediction that sensitivity is tied to the fact rather than the magnitude of gains or losses. Individuals are also likely to engage in risk-seeking behavior when trying to recoup losses (Tversky and Kahneman 1981). For example, politicians are more likely to take drastic, risky measures, such as mobilizing troops or going to war, if they think such action will reverse perceived losses in prestige or credibility.
Manipulation involves not only language but also emotion. Emotional states drive social processes. As Barbalet boldly claims, “Emotion is central to social processes not only in being central to identity and affiliation, in which its role is frequently acknowledged, but also in being the necessary basis of social action and in being responsible for the form action takes” (1998, 65). Using affect priming theory (Bower and Forgas 2000), Zahariadis (2005) hypothesizes that the national mood vitally affects a government’s behavior. The reasoning is that negative mood biases appraisals of the current situation by highlighting negative expectations of one’s reaction to others. A negative mood is likely to lead to more confrontational policy.
Symbols have affective and cognitive dimensions in that they evoke emotions and also convey relatively clear but highly simplified messages (Elder and Cobb 1983). They focus the debate on specific aspects of an issue and bias selection by raising emotive attachment to certain options and by highlighting the cognitive distance from the status quo. The burning of the flag is a good example. Not only does it raise an emotional response in most people, but it also conveys the message that the flag’s desecration constitutes a loss. The implication of this point for MSA is that coupling is more likely to be successful if the proposed solution is presented as a large deviation from the status quo and the problem is represented as a loss. If options represent smaller deviations or a preservation of the status quo, coupling success is more likely when problems are framed as gains. Because policymakers are loss averse, policy entrepreneurs will have greater success if they present them with options that recoup perceived losses. Given that prospect theory argues that individuals are also risk seekers when confronted with losses, efforts to couple the three streams will intensify, and hence be more successful, when problems are defined as losses. This helps to explain Greece’s persistent—and, some might argue, blind—policy of supporting the status quo in Yugoslavia during the Bosnian carnage at a significant diplomatic and economic cost (Zahariadis 2005). The greater the perceived losses, the more stubborn Greece’s defense of the “old ways” became, even when that meant a serious rift with its more powerful European and American allies.
Entrepreneurs who are placed at a high level in government, operate under crisis conditions, and pursue “salami tactics” are more likely to succeed at coupling. A “salami tactic” involves the strategic manipulation of sequential decision-making. Entrepreneurs are assumed to have a grand design for their desired outcome. However, because they are reasonably certain their desired solution will not be adopted because it’s too risky, they cut the process into distinct stages that are presented sequentially to policymakers. This promotes agreement in steps.
But selection is not merely a function of perception. It is also a question of skill. Policy entrepreneurs must be skilled at coupling. Two variables are important: resources and access. Entrepreneurs who are more willing to spend time and energy lobbying politicians and generally pushing their pet projects forward are more likely to experience success. In addition, those with access to the centers of power have an even greater chance of succeeding. Privatization of British Rail, for example, was a better candidate for adoption because think tanks with very strong connections to the governing party pushed for it.
Why do policymakers adopt some policies but not others? The MSA answer can be summarized as follows. During open policy windows, persistent policy entrepreneurs, who constantly search for solutions to important problems, attempt to couple the three streams. Success is more likely when all three streams are coupled, conditional on the type of window that opens and the skills, resources, and strategies entrepreneurs use to focus attention and bias choice.
Despite its wide appeal among policy analysts, MSA has also generated a number of detractors. Analysts have criticized MSA for making a number of unrealistic assumptions and for underspecifying certain processes. As a result, Sabatier (2007, 327) reminds us, MSA has not generated enough clear, falsifiable hypotheses. I have addressed this issue in the two preceding sections. The plethora of applications detailed in Table 2.1 in the appendix further dispels this criticism. I confine my comments here to addressing other criticisms.
MSA has definitely generated a lot of movement in the policy field, but has there been much movement forward? In a strongly worded article, Bendor, Moe, and Shott (2001) criticize the logic and conclusions of the original garbage can simulation (Cohen, March, and Olsen 1972). Because that conceptualization served as an inspiration for MSA, undermining the former adversely affects the latter. I will not summarize Olsen’s (2001) response here, but the question still lingers: Are the conclusions of MSA empirically based rather than assumption driven?
The major criticism of the garbage can model is that the verbal model differs from the computer simulation that accompanied it in the original article. Moreover, the results seem to flow directly out of the assumptions of the model. For example, decisions made by flight or oversight in the garbage can model are attributed directly to the structure of the model; they are not findings but assumptions (Bendor, Moe, and Shott 2001). MSA begins from a different point. It draws inspiration from the work of Cohen and his colleagues, but it also contains structural features of its own. For example, the garbage can model conceives of decisions as being the result of energy fluctuations in each of the streams, more or less fortuitously combined. In contrast, coupling in MSA is purposefully done by policy entrepreneurs (Zahariadis 2003). In addition, MSA is empirically oriented. Kingdon supplies considerable evidence from the fields of transportation and health in the United States to make his case. Additional analyses and extensions across different countries and policy domains have been similarly empirically based (Zahariadis 1995, 2003, 2005; Birkland 1997). Perhaps the critics provide a clue to the answer. Whereas Bendor, Moe, and Shott sharply criticize Cohen, March, and Olsen for having no empirical verification, they praise Kingdon (1995), “whose work is distinguished by careful empiricism tied to theoretical concerns” (2001, 186n28). In other words, whatever the flaws of the garbage can model, MSA is theoretically driven and empirically validated.
Critics also point to more specific problems. I reflect on three of them in the sections below.
Are the streams really independent? MSA argues that although the streams are not completely independent of one another, each can be viewed as having a life of its own. Participants drift in and out of decisions, making some choices more likely than others. Problems rise and fall on the government’s agenda regardless of whether they are solvable or have been solved. Similarly, people generate solutions not necessarily because they have identified a particular problem but because the solution happens to answer a problem that fits their values, beliefs, or material well-being. Changes in the political stream take place whether or not problems facing the nation have changed. Thus, each stream seems to obey its own rules and flows largely independently of the others (Sager and Rielle 2013).
Critics disagree. Mucciaroni (1992) and Robinson and Eller (2010) question the appropriateness of conceptualizing independent streams. The streams can be more fruitfully viewed as interdependent, Mucciaroni maintains, and changes in one stream can trigger or reinforce changes in another, making coupling much less fortuitous and the process more purposive and strategic. For example, the problem of U.S. tax reform was tied to the supply-side tax cuts proposed by conservatives in symbolic and substantive ways long before Ronald Reagan’s rise to power opened a policy window. Sabatier (2007, 332n5) goes further. He views stream independence solely as a contingent relationship subject to empirical verification. Politics does not necessarily operate at the systemic level, policies are not always developed in policy communities, and solutions are not developed independently of problems. Kingdon (1995, 228) himself opens the possibility that coupling (i.e., interaction) may take place in the absence of an open window.
Stream independence is a conceptual device. It has the advantage of enabling researchers to uncover rather than assume rationality (i.e., that solutions are always developed in response to clearly defined problems). Sometimes policies are in search of a rationale or solve no problem (Stone 2011; Zahariadis 2003). Edelman (1988) goes as far as to argue that solutions create problems. Consider, for example, the decision by the George W. Bush administration in 2003 to go to war in Iraq. Whereas the initial rationale had to do with what was claimed to be the clear and imminent danger posed by Saddam Hussein’s possession of weapons of mass destruction, subsequent rationalizations emphasized connections with terrorists, the liberation of Iraq, or democratization and nation building. The solution remained the same—depose Saddam—while the problem constantly drifted in search of an anchor. As insiders, such as former counterterrorism “czar” Richard Clarke (2004), pointed out later, the administration was fixated on Saddam long before the attack. The question was not whether but when and how to launch it.
It is impossible to make the above argument in the absence of stream independence. The key is to specify when policy may be in search of a rationale, but one cannot logically make this statement unless one differentiates between the development of problems and solutions. Besides, assumptions are simplifications of reality. If many policy analysts readily accept the assumption that people don’t have to be rational, that they only need act as if they are rational, then they can also accept the assumption that streams don’t have to be independent and only need flow as if they are independent.
Can hypotheses generated by MSA be quantitatively examined? Methodological pluralism may be a virtue, but statistical analysis adds weight to a lens’s predictions in ways that case studies do not. This is not the place to rehash the old debate over the benefits and drawbacks of quantitative versus qualitative analysis, but it is no secret that most applications of the garbage can stream of research as well as MSA have been qualitative case studies. This has led to significant discontent among critics who charge that “many applications of the approach do little more than describe some parts of an organization as garbage cans, offering descriptions—usually ethnographic accounts that emphasize G[arbage] C[an] T[heory]’s central themes” (Bendor, Moe, and Shott 2001, 186). Apart from utilizing computer simulations that tighten and formalize the verbal lens in order to draw further implications (Zahariadis 2003), it would be useful to test MSA statistically. Can this be done, and if so, how?
Travis and Zahariadis (2002) provide a test. The authors adopt a cybernetic version of the model and make two assumptions. First, they drop the notion of entrepreneurs who manipulate the process using cognitive or affective strategies. Second, they explicitly conceptualize inertia built into the model in the form of baseline funding. Examining U.S. foreign aid allocations, they argue foreign aid is the result of interactions between problems, policies, and politics. The model follows an anchor-and-adjust process whereby policy outputs or funding levels are anchored around a specific level that is subject to periodic adjustments caused by prespecified factors. Policymakers anchor allocations to the previous year’s allocation level, which represents the point of agreement in the foreign aid policy community. Adjustments to the anchor can be made in response to external problems or domestic politics, and they will be made during open policy windows, which are changes in either the problem or the politics stream. Using the idea of negative and positive feedback, the authors view external problems as measured by security concerns, economic activities, and the recipient’s needs. Similarly, domestic politics is conceptualized as being measured by control of the executive branch and control of either or both chambers of Congress. Domestic preferences interact with external stimuli to produce dramatic policy shifts under certain conditions. Reconceptualizing foreign policymaking, the findings show that “foreign aid decisions are not simply made with an eye toward the domestic scene but because of the domestic scene” (Travis 2010, 818).
Following a different path, Sager and Rielle (2013) use Qualitative Comparative Analysis (QCA) to explain the adoption of new alcohol policy programs in the Swiss cantons. QCA employs a set of comparative methods that focuses on constellations of factors rather than examination of a single factor independently. Dichotomizing the dependent variable as program adoption (or rejection), the authors develop new theoretical insights into the way MSA operates. They find, among other things, that centralization of authority coupled with a strong bureaucratic apparatus, which they call cooperative bureaucracy, increases the likelihood of program adoption. In contrast, decentralized structures coupled with low coordination mechanisms or high outsourcing capabilities do not encourage program adoption. While stressing the need to include a more explicit specification of institutional settings, the authors show that MSA can be explored quantitatively to explain programmatic and policy choices.
Entrepreneurs or entrepreneurship? One of MSA’s greatest strengths, its ability to link agency and structure in a single framework, is also a major liability. Critics argue MSA pays limited attention to institutional arrangements (Mucciaroni 1992). Schlager (2007, 307) perceptively adds that specification of institutional structures will more clearly identify different coupling processes within and across government systems. Richardson (2006) sings the praises of MSA’s actor-centered theorizing, but Ackrill and Kay (2011) question the utility of stressing entrepreneurs as individuals. Instead, they argue for developing a theory of entrepreneurship.
The lack of institutional language is indeed puzzling. Barzelay and Gallego (2006, 539) label MSA as “institutionally processual” to emphasize ideational interactions among policy entrepreneurs, experts, media, and decision-makers. Cairney (2009) adds the timing of policy transfer across national boundaries to this mix of factors. Kingdon (1995) addresses the criticism more directly by arguing MSA incorporates institutional arrangements; it just does not use familiar terminology. For example, he notes the impact of value acceptability in the policy stream or the balance of interests in the political stream as evidence of institutional effects. But that’s not enough. Brunner (2008) observes the impact of multilevel governance structures and organizational learning processes on MSA explanations. The presence of different institutional venues, often hierarchically arranged, suggests that the dynamics of each venue strongly affect the final outcome. Do entrepreneurs engage in venue shopping? If so, how? The original garbage can model paid significant attention to institutions through specification of access and decision structures (Cohen, March, and Olsen 1972). The emphasis was lost in MSA and needs to be rediscovered. It is an area of considerable interest that remains underexplored.
The process of coupling needs further refinement and elaboration. Most MSA scholarship stresses the work of individual entrepreneurs. Although fruitful, this line of research risks making the process of coupling too idiosyncratic. For example, specifying the skills and characteristics of individual entrepreneurs reduces the ability to generalize in cross-national environments. Too many contextual factors, institutions, or cultural particulars may affect the outcome. MSA needs an integrated theory of entrepreneurship. Mintrom and Norman (2009) and Brouwer and Biermann (2011) chart possible directions, but more theoretical and empirical research is needed to specify scope conditions and highlight the strategies of success and failure.
Implications for Theory
The MSA is useful in linking the various stages of the policymaking process under the umbrella of a single lens. Politics (policy formation) and administration (implementation) are not so rigidly divorced (March 1994, 109; Olsen 1988). Quite the contrary: difficulties with implementation cannot be divorced from the confusions of policymaking; policy ambiguity encourages administrative autonomy, which in turn facilitates more ambiguity (Baier, March, and Saetren 1986). MSA may be able to address this issue with appropriate revisions and qualifications. Zahariadis (2008b) theorizes that higher levels of ambiguity in EU implementation systems dramatically reduce efficiency of delivery but greatly enhance democratic accountability. Ridde (2009) applies the framework to health policy in Burkina Faso. He finds successful implementation (or lack thereof) depends on coupling the three streams, but unlike other researchers, he also finds an increasing number of policy windows in implementation especially when international actors are involved. As policies develop into concrete plans of action, political conflict becomes increasingly intrabureaucratic conflict. Entrepreneurs and windows are still essential, but the importance of the policy stream is magnified because hierarchically placed agencies vie for resources, control, and meaning.
The lens addresses the issue of ideas in public policy. Although it does not deny the importance of self-interest, it points to the significance of ideas in two ways. First, solutions are developed, Kingdon argues, on the basis not simply of efficiency or power but also of equity. Second, political ideology is a good heuristic in an ambiguous and rapidly changing world (Kingdon 1993, 79). It provides meaning to action or cues for floor voting or serves as an (imprecise) guide to what issues are important. Ideas may be used by politicians not only to define others but to define themselves. People, however, need not be motivated exclusively by ideas. Entrepreneurs whose purpose is to couple the three streams will occasionally bend ideological proclivities in order to take advantage of fleeting opportunities (Zahariadis 2003). The lens is a good way of exploring the impact of ideas without necessarily denying the importance of self-interest.
Finally, MSA has important implications for claims concerning the role of individuals and institutions in policymaking. MSA subscribes to the notion that institutions make things possible, but people make things happen. It points to the importance of policy entrepreneurs and human cognition and emotion as the bases of political manipulation. Moreover, institutions matter (Weaver and Rockman 1993), but their importance is tempered considerably by individuals, timing, and context. Even in the case of smaller (relative to the United States) systems with strong executive control and partisan discipline, such as those of the United Kingdom and Greece, there must still be coupling of three independent streams before new policies are adopted.
Recommendations for Further Research
Future research may proceed along the following lines, drafted in the form of advice to aid theoretical development.
Probe applicability under different conditions. Why do some decisions tend to become garbage cans? Are there characteristics of issues that make some more likely candidates than others? Answers will circumscribe even more carefully the limits of MSA and lead to a better understanding of the policymaking process. While Kingdon (1995) originally implied that the entire process of policy formation constituted a giant receptacle because of characteristics of policymaking at the national level, others have explored the importance of specific issue properties. Rommetveit (1976) suggests that prime candidates for garbage cans are those issues that involve changes in normative structures—basic value priorities in a polity—when no active participant dominates the policy process. When a society is in the process of reordering its values, established norms that underlie state-society relations are challenged. As a result, conventional wisdom is questioned, bringing dissenting groups to the forefront of change. The activation of new groups and disagreement as to the relevant values on which to base the policy decision in turn increase ambiguity and permit the evocation or appearance of new problems and solutions. Such desegmentation of previously established links between windows, problems, and politics complicates the process as new and perhaps unrelated elements are dumped into the can. In this light, the issue of privatization or government reform is a good candidate for applying MSA (Brunsson and Olsen 1993; March and Olsen 1983). Other characteristics of issues, Mucciaroni (1992) maintains, alter the behavior of the system in predictable ways. Zahariadis (2003, 2005) addresses the point, concluding that issue salience, rather than inherent issue characteristics, makes the difference. More empirical work is needed to settle the question.
Probe applicability in different domains. Most of the work utilizing MSA has investigated the politics of making domestic policies at the national level. Adhering to rigid and obsolete disciplinary boundaries drawn many decades ago, policy analysts systematically neglect the area of foreign policy. Yet Allison (1971) showed that differences between domestic and foreign policy are more imagined than real. Can lenses developed in one context be extended to provide credible explanations in the other?
Zahariadis (2005), Mazzar (2007), Travis (2010), Travis and Zahariadis (2002), and Durant and Diehl (1989) find that MSA is a good candidate to bridge the divide between domestic and foreign policy. The key problem is to link domestic and external variables. Despite differences regarding the ability of interest groups and corporate actors to access the foreign policy establishment of a particular country, particularly those representing or having extensive ties to foreign interests, domestic concerns and actors assess and filter external threats while pursuing their own domestic pet projects. Ultimately, foreign policy outcomes need to be acceptable to domestic audiences who will ratify the solutions. The external environment plays a role, but externally generated problems or solutions still need to be domestically interpreted. Policy entrepreneurs play a major part in coupling, just like in the case of domestic policies. Having started as an explanation of domestic policy in a “disorderly” presidential democracy, MSA proves useful even in small, parliamentary democracies, such as Greece, and in foreign policy where participation is less fluid. In contrast to conventional wisdom, theories of domestic policymaking, such as MSA, offer solid and theoretically informed explanations of foreign policy.
On a slightly different note, analysis may further inquire into the applicability of MSA as a lens for explaining the policy process in areas other than educational organizations or national governments. On the domestic subnational level, McLendon (2003) uses the framework to study agenda formation for the decentralization of higher education in three states: Arkansas, Illinois, and Hawaii. Laird (2008) examines the workings of the policy stream in making Colorado’s energy policy, while Protopsaltis (2008) uses it to examine agenda setting in Colorado’s educational policy. Dudley (2013) examines the utility of MSA i