email icon Email this citation


Transparency’s Three Paths of Influence

Ronald B. Mitchell

International Studies Association
41st Annual Convention
Los Angeles, CA
March 14-18, 2000


Transparency facilitates cooperation in international regimes in three ways. In a “monitoring” model, information about each state’s actions is collected and given to other states in hopes that treaty supporters will reward compliers and/or punish violators. In a “mobilizing latent responders” model, information regarding the actions of states or non-state actors is given to states or non-state actors whose basic interests (rather than their treaty support) make them likely to respond to that information in ways that make the targeted actor more likely to adopt the behaviors the regime seeks to promote. In a “facilitating rational choices” model, information regarding the state of the world and the likely consequences of certain actions is given to those state and non-state actors likely to engage in such actions in the hope that they will respond to that information by adopting the behaviors the regime seeks to promote. Although most research to date has focused on the first of these, the latter two are both common and effective at facilitating international cooperation.

Draft of 06 March 2000


Scholars increasingly view transparency as crucial to a regime’s ability to influence state behavior. Recent work on compliance and effectiveness has repeatedly demonstrated that the presence or absence of well-designed and well-implemented information systems explains considerable variation in the outcomes that regimes produce (Chayes and Chayes 1995; Finel and Lord 1999; Finel and Lord forthcoming; Mitchell 1998; Victor, Raustiala, and Skolnikoff 1998; Weiss and Jacobson 1998). Yet, the ways in which transparency influences state and nonstate actor behavior in regulatory regimes remain undertheorized. In what follows, I develop a framework for thinking about three distinct mechanisms that regulatory regimes adopt in their efforts to influence behavior.

Many, though not all, regimes seek to induce behavioral change in the behavior of member states or the corporate and individual substate actors under their jurisdiction. Such regulatory regimes have a variety of mechanisms available for increasing a targeted actors propensity to engage in a “desirable” behavior or refrain from an “undesirable” one (Young 1999). They can alter a target’s incentives to engage in these behaviors, their opportunities to engage in them, or their perceptions of the consequences of doing so (Mitchell 1999). Information and transparency play important, but differentiable, roles in these strategies. Three modalities appear to be operative. In a “monitoring” model of transparency, information about targeted actors’ behaviors is collected and made available to member state governments in hopes that treaty supporters will reward compliers and/or punish violators, even if doing so runs against the enforcing states short-term interests. In a “mobilizing latent responders” model, information about targeted actors’ behaviors is made widely available to states and non-state actors whose basic interests (rather than their support for the treaty) make them likely to respond in ways that increase targeted actors’ propensities to engage in regime-consistent behavior. In a “facilitating rational choice” model, information regarding the state of the world and the likely consequences of certain actions is given to the targeted state and non-state actors themselves in the hope that they will respond to that information by adopting the behaviors the regime seeks to promote. Most research to date has focused on the first of these. The latter, however, are common in, and may well be effective at, facilitating international cooperation.



Regimes vary considerably in their basic goals. Young distinguishes regulatory, programmatic, procedural, and generative regimes. Procedural regimes establish procedures for collective choice, programmatic regimes pool resources for projects that would not be carried out unilaterally, and generative regimes develop social practices where none previously existed (Young 1999, 29-31). Transparency and information undoubtedly play roles in the effectiveness of each of these three regime types. I concern myself here with exploring the influence of transparency in regulatory regimes that “prescribe actions that regime members are expected to take or to refrain from in more or less well-defined situations” (Young 1999, 28).

Transparency as used here refers to the availability of regime-relevant information. Although transparency also includes the “openness” of a government’s or international institution’s decision-making procedures to external observers, I focus on a narrower conception of transparency as information related to the behaviors that a regulatory regime seeks to influence. Transparency varies in degree across regimes, within regimes, and over time: some regimes will be more transparent than others; a single regime will be more transparent in certain respects than in others; and regimes will be more transparent during certain periods than others. Transparency is manipulable via the regime’s information system – the actors, rules, and processes by which the regime collects, analyzes, and disseminates information (Mitchell 1998).

Within regulatory regimes, any given actor can be thought of as playing either or both of two roles. Targets are those whose behavior the regime seeks to alter. Responders, or potential responders, are those who may choose to adopt new behaviors once they become aware that efforts to induce behavioral change by the target have either failed or succeed. Thus, the behavior of responders is contingent on knowledge regarding whether and to what extent the target has brought their behavior into conformance with the regime. As we shall see, responders can either reward desirable behavior or sanction undesirable behavior. I refer to desirable and undesirable behavioral change (from a regime supporter’s perspective) to avoid the more constraining terminology of compliance and noncompliance which carries with it an either/or logic. Considerable evidence suggests that states are, and analysts should be, attentive to positive behavioral changes that fall short of compliance (“good faith noncompliers”) and negative behavioral changes that nonetheless leave a target in compliance (“unintentional compliers”).


Power, interests, knowledge, and behavior

Recent contributions to the regimes literature have noted the influence of power, interests, and knowledge on the behavior of states (and presumably non-state actors) in the international sphere. Although transparency and information clearly sit in the arena of knowledge, the ways in which international institutions can use information to influence behavior has much to do with how it influences power and interests. Consider the reasons why a targeted actor might engage in some undesirable behavior. As I have argued elsewhere, the determining factors in an actor’s choice of noncompliance and related undesirable behaviors can arise from either interests or power (Mitchell 1996). As delineated in the following table, undesirable behaviors can result from calculations of independent self-interest, calculations of interdependent self-interest, incapacity, or inadvertence.

If we accept the standard rationality assumption regarding actors – that each pursues their own self-interest within the constraints imposed by their available resources, each of these factors has an “imperfect information component” which present instances in which information might wield influence over the choice of the undesirable behavior. In many cases, of course, targets choose undesirable behaviors based on a calculus of independent and myopic self-interest in which other actors cannot be relied upon to act in anything but their own independent and myopic self-interest. Put differently, actors sometimes choose undesirable behaviors in a context within which doing so is a dominant strategy regardless of the behavior of other actors. Misinformation can produce undesirable behavior in such contexts if the target is misinformed about the costs and benefits of the undesirable behavior relative to those of desirable behaviors. In other cases, actors are making contingent decisions in which the privately optimal behavior depends on how other actors will respond to the behaviors engaged in. Imperfect information plays a role here to the extent that the target may be misinformed about how likely other actors are to respond positively to desirable behaviors or negatively to undesirable ones, and how costly or beneficial those responses will be. A third, and related, problem arises simply when actors engage in inadvertent or accidental behavior that they themselves would have preferred to avoid. Although this often can arise from problems in the ability to exercise complete control over behaviors or outcomes, it also can have an informational dimension. Actors can engage in socially undesirable behavior because they are misinformed about the current state of the world or the actual behavior of others, and thereby misunderstand what the privately optimal behavior is.

Knowledge also influences a target’s engaging in undesirable behavior even when power-related factors are determining. In many cases, as empirical literature has increasingly made clear, undesirable behaviors may arise because the actor lacks the capacity to engage in desirable ones. Low or delayed compliance by developing countries under numerous environmental treaties has been demonstrated to have as much to do with an absence of the necessary financial, technical, or administrative resources as with a lack of incentives (Chayes and Chayes 1995; Weiss and Jacobson 1998). But some of that literature has demonstrated that states and nonstate actors may not be aware of or recognize opportunities that are available or ways in which they could re-deploy the resources they do have so that they can engage in a socially desirable behavior that they also consider privately desirable but in which they believed they were incapable of engaging.

Determining factor in target engaging in undesirable behavior Imperfect information component
Interests Independent self-interest – dominant strategy of engaging in undesirable behavior Misinformed about probability and magnitude of non-strategic consequences of undesirable and desirable behaviors
Interdependent self-interest – engages in undesirable behavior based on expectations regarding others’ actions Misinformed about probability and magnitude of how other actors will respond to undesirable and desirable behaviors
Power Incapacity to engage in desirable behavior Misinformed about opportunities or resources needed to engage in desirable behavior
Inadvertent/accidental – engaging in undesirable behavior contrary to own interests Misinformed about state of the world

Power and interests play important roles in determining why other actors might fail to respond to the undesirable behaviors of those who the regime targets. Here too, knowledge or imperfect information conditions the influence of these factors. Consider the conditions under which a potential responder will want to respond to desirable or undesirable behaviors by targets. An actor’s interest in responding, of course, will reflect its assessment of the costs and benefits of responding to the target’s behavior. As with targets themselves, this self-interest can be contingent on the behavior of other potential respondents or not. In some cases it is not, with actors having dominant strategies to reward targets for desirable behavior or punish targets for undesirable behavior regardless of the behavior of other potential respondents. In this case, the potential respondent may still fail to respond if it misunderstands the consequences it incurs as a result of the target’s behavior. In other cases, the interests of an actor in responding to a target are contingent on whether and how other actors respond. As has been noted, enforcement (whether involving sanctions or rewards) involves a collective action problem involving both free-riding and reassurance problems of its own (Axelrod and Keohane 1986; Chayes and Chayes 1995).

In addition, potential responders may lack the capacity to respond effectively to the behavior of targets. Here, imperfect information plays a role in that it may well prove too difficult or costly for individual actors to collect information regarding whether undesirable acts were committed and what actors committed them. Lacking such information can create the equivalent circumstances in which actors with interests in sanctioning undesirable behaviors or rewarding desirable ones nonetheless fail to bring their resources to bear. Without such information, the potential responder with real interests in responding may fail to respond.

Determining factor in responder failing to respond to undesirable behavior Imperfect information component
Interests Independent self-interest – dominant strategy not to respond Misinformed about probability and magnitude of consequences of undesirable and desirable behaviors by target
Interdependent self-interest – collective action and free-riding problems create benefits from not responding Misinformed about probability and magnitude of how other potential responders are likely to respond to undesirable and desirable behaviors by target
Power Incapacity to respond – may lack the power to respond potently to target behavior Lacking information regarding who committed undesirable or desirable behaviors, when they occurred, and what their consequences were


Transparency as monitoring

Mainstream notions view transparency as influencing behavior by providing the information requisite for actors to sanction those that violate regime rules. For those who argue that significant or “deep” cooperation only occurs in regimes backed by sanctions, transparency primes the sanction pump (Downs, Rocke, and Barsoom 1996). Sanction-based strategies are contingent strategies. They work only if potential responders impose costs only on those engaged in undesirable behaviors and refrain from imposing those costs on those engaged in desirable behaviors. The logic of transparency as monitoring is identical for rewards. Rewards also involve contingent strategies. Their efficacy depends on potential responders providing rewards only to those engaged in desirable behaviors and not undesirable ones. In both contexts, information regarding whether targets engaged in desirable or undesirable behaviors is essential to the functioning of the regime’s implementation system.

The effectiveness of a monitoring model of transparency depends on the characteristics of the monitoring system, characteristics of the responders, and characteristics of the targets. With respect to the monitoring system, of course, the difficulty of detecting whether a target is engaged in desirable or undesirable behavior varies considerably depending on the type of activity being regulated. Therefore, regimes may vary considerably in how complex an information system is required to create transparency adequate to prompt responses to desirable or undesirable behaviors. The nature of some behaviors that states seek to regulate prevents them from being concealed. Consider several factors that influence the difficulty of ensuring that a regulated act is transparent.

Transparency poses few obstacles in cases in which the regulated act involves a transaction involving at least one actor whose interests are coincident with regime goals. Tariffs, for example, are inherently transparent. States cannot impose a tariff that exceeds World Trade Organization (WTO) ceilings without importers knowing. The act of collecting the tariff involves revealing the tariff rate to actors who have incentives to monitor that rate and to ensure that the government complies with WTO restrictions on that rate. Export subsidies may be easier to conceal. Although governments can provide indirect export subsidies, direct export subsidies can only be provided to exporters that know about them. Unlike with tariffs, however, exporters have no incentives to ensure their governments comply with WTO rules.

Ensuring transparency is also easier if the regulated acts leave long-lasting evidence both that the act was committed and who committed it. This depends on much on the definition of the behavior being regulated as in the behavior itself. Thus, initial regulations addressing ocean oil pollution which banned certain discharges of oil at sea proved far less transparent than later, environmentally-equivalent, regulations that required installation of equipment that prevented such discharges (Mitchell 1994). The temporal nature of oil discharges meant that evidence, especially legally adequate evidence, that a tanker had violated discharge regulations proved almost impossible to obtain. By contrast, any tanker that failed to install required equipment could not conceal that fact from any government interested in assessing compliance (although assessing its proper use was, like discharges, far less transparent).

These and other factors dictate that regime information systems vary in the level of certainty they provide with respect to whether a given actor was engaged in desirable or undesirable behavior. Regime information systems, like other such systems, can be described in terms of the extent to which they produce Type I and Type II errors, or false negatives and false positives.

Even a perfectly transparent regime information system may fail to induce behavioral change, either as a regime or in particular cases. The value of transparency in a monitoring model depends on the extent to which responders make their behavior contingent on that information or, more precisely, the extent to which targets believe that responders will make their behavior contingent on that information. The effectiveness of a regime with a given level of transparency will depend on the willingness and ability of potential responders credibly to threaten potent sanctions or convincingly to offer attractive rewards. Sanctions will only work if potential responders are sufficiently likely to impose sanctions that are sufficiently costly to the target that the target views it as less advantageous to engage in the undesirable than the desirable behavior. Rewards will only work if potential responders are sufficiently likely to provide offered rewards that are sufficiently valuable to the target that the target views it as more advantageous to engage in the desirable than the undesirable behavior.

Sanctions are imposed or rewards provided precisely and only because doing so is believed to increase the likelihood that the target will conform to regime dictates. The responder views imposing sanctions or providing rewards as costly behaviors which will be worth engaging in only to the extent that the expected utility of inducing the target to become more conformant outweighs the costs of doing so. A potential responder’s willingness to use sanctions or rewards constitutes a schedule that is a function of the costs of a particular level of sanction or reward, the probability that that level of sanction or reward will induce a change in the target’s behavior, the magnitude of the change that that level of sanction or reward will induce in the target’s behavior, and the value the responder places on that magnitude of change in the target’s behavior. The value of sanctions and rewards stems only from their ability to alter the target’s behavior. As many scholars have noted, “a promise is costly when it succeeds, and a threat is costly when it fails” (Schelling 1960, 177). Schelling’s focus on the difference in the strategies ignores their similarity in that both strategies involve a response that responders consider as costly. A rational responder faced with a target known with certainty to be unsusceptible to the responder’s available sanctions or rewards will prefer not to impose sanctions or offer rewards. Put differently, in a one-shot game in which the target is already engaged in the undesirable behavior, the responder prefers not to impose previously threatened sanctions. Likewise, in a one-shot game in which the target is already engaged in the desirable behavior, the responder prefers to not provide previously promised rewards. These difficulties become even more pronounced in multilateral settings in which sanctions or rewards are plagued by collective action problems that create second-order prisoners’ dilemma dynamics (Axelrod and Keohane 1986). The regime’s ability to surmount these problems depends either on targets believing that responders will carry out promises or threats and therefore not testing the system by engaging in undesirable behaviors or on responders being willing to follow through on promises or threats that run counter to their short-term interests to encourage longer-term adoption of desirable behaviors.

Both targets and responders are aware that the magnitude and likelihood of rewards or sanctions are a function of the responder’s interests in inducing behavioral change by the target and the size and appropriateness of the resources they have available for doing so. Many of the rules of thumb delineated by Hufbauer and Schott in their seminal work on sanctions speak to the fact that sanctions will work only if the target is willing to incur the costs of sanctioning and has at its disposal a resource it can use to sanction that the target would consider costly if imposed (Hufbauer, Schott, and Elliott 1990). Those same rules of thumb, however, could be applied mutatis mutandi to rewards. As recent experience with North Korea demonstrates, a target’s willingness to continue engaging in desirable behavior is expressly contingent on the ability of those that have offered rewards to actually deliver them (Dorn and Fulton 1997; Snyder 1997).

As is obvious, the efficacy of sanctions or rewards depends as much on the characteristics of the target as the characteristics of the responder. A target currently engaged in desirable behavior will decide whether to adopt undesirable behaviors based on a subjective assessment of the probabilities and magnitude of potential sanctions it will face for doing so and a similar assessment of the probabilities and value that existing rewards will continue or new rewards will be provided for maintaining desirable behaviors. A target currently engaged in undesirable behavior will decide whether to adopt desirable behaviors based on a subjective assessment of the probabilities that potential sanctioners will sanction, or will continue sanctioning, them and the costs of those sanctions if carried out as well as a similar assessment of the probabilities and value of potential rewards for adopting desirable behaviors. How the target makes and values these responses as well as how it compares the exogenous or non-manipulated costs and benefits of the desirable and undesirable behaviors determine how the target responds. The target’s decision-making, in turn, depends on such factors as its values and preferences, its risk aversion and whether it views the situation from a losses or gains frame, and the quality of its information about how and how likely responders are to respond. This last point highlights that facilitating transparency for the target related to the expected behavior of the responders may play an important role alongside transparency for the responder of target behavior.


Transparency as mobilizing latent responders

Transparency regarding target behavior plays a related but distinct role in many regimes. In this modality, information regarding target behavior is provided to state or non-state actors whose underlying interests will lead them to respond in ways that alter the target’s consequences of engaging in the desirable and undesirable behaviors. The distinction of transparency that mobilizes latent responders from that of monitoring lies in the responders’ having interests unrelated to regime support that lead them to engage in rewarding or sanctioning targets. Assumptions regarding potential responders’ incentives are quite different under a “mobilizing latent responders” model than under a monitoring model. As noted, a monitoring model assumes that the costs to a responder of imposing sanctions or providing rewards exceed the benefits of doing so. A mobilizing latent responders model assumes that some actors exist who consider themselves better off, independent of the actions of any other actors, rewarding desirable behavior or sanctioning undesirable behavior. What prevents them from doing so is their lack of knowledge regarding targets’ behaviors.

Consider such programs as International Organization of Standards (ISO) guidelines, eco-labeling, or environmental certification. Such programs do not directly regulate the behavior of those they ultimately seek to target. Regimes based on such principles can adopt a set of standards for desirable behavior and then provide information to potential responders regarding the extent to which targeted actors’ behaviors conform to those standards. As with the modality of facilitating rational choice described below, the information on targeted behaviors is provided in ways that merely improve recipients’ information about the consequences of their choices and thereby help them in more rationally pursuing their own self-interest. However, in a mobilizing latent responders modality, the information is provided to those believed to have interests that will lead them to reward targets engaging in desirable behaviors and punish targets engaging in undesirable behaviors. Thus, the information is provided to intermediate actors considered to be potential responders rather than to the actors ultimately targeted by the regime.

The effectiveness of this model depends on the accuracy of the assumption that the undesirable behaviors of targeted actors impose sufficient material or normative harms on actors that would lead those actors, if they were aware of the harms, to respond in ways that would be sufficient to shift the incentives of targeted actors toward engaging in desirable behaviors. The responses can involve either rewards, sanctions, or both. In many instances, such an approach simply improves information flow and removes informational asymmetries in a market. The assumption underlying such an approach, however, is that an externality exists because of imperfect information or misinformation rather than because of power inequities between those imposing the externality and those bearing the costs of the externality. It also assumes that collective action problems do not plague the imposition of sanctions or provision of rewards. Recipients must view it as in their own self-interest to alter their behavior in light of the new information and those new behaviors must impose costs on targets engaged in undesirable behavior and/or provide benefits to those engaged in desirable behaviors. These actors must see adopting these new behaviors as dominant strategies in response to target behaviors rather than strategies the value of which derives from how the target responds.

Consider the eco-forestry certification programs being promoted by the Forestry Stewardship Council (FSC) (Dudley, Elliott, and Stolton 1997; Kiker and Putz 1997; Stevens, Ahmad, and Ruddell 1998; West 1995). The current market for lumber and wood products does not provide consumers with information regarding the forest management practices of wood product producers. Thus, consumers cannot distinguish sustainable producers from unsustainable producers. Certification programs, in which producers allow third parties to evaluate whether they are managing their forests sustainably, allow producers to advertise their lumber as sustainable and have that claim be considered credible by consumers. This information allows environmental consumers to become more discriminate in their purchases, thereby rewarding sustainable producers and punishing unsustainable ones. How significant a change in responder behavior occurs will depend on how many “environmental consumers” there are and their relative sensitivity to both the price and the “sustainability” of the lumber they purchase. In turn, how significant a change in target behavior will depend on both the size of responder behavior and other factors related to the net profits producers believe they can earn by switching to sustainable forestry practices. Ultimately, the FSC seeks to make producers harvest more sustainably. Unable to do that directly, they have developed a scheme that entrains consumers in a process which they hope will lead wood product companies to alter their practices.

Transparency as mobilization need not be solely economic in nature. Some sharing of intelligence information among states may serve similar functions. A weak state threatened by a neighbor may publicize sensitive intelligence information related to that state’s capacities or intentions in the belief that other states, once made aware of that information, will take actions aimed at deterring the potential aggressor. The assumption being made is that the third party states will mobilize not out of solidarity with the weak state but out of a desire to protect their own security interests. Obviously, however, the effectiveness of such a strategy depends on whether the weak state has new information that the third party was not aware of, that convinces the third party that their self-interests are at stake, and that leads the third party to take the actions the weak state desired.

As in a monitoring model, the effectiveness of a mobilizing model requires that the responses induced by the new information in the system are sufficient to shift the balance from favoring undesirable behaviors to favoring desirable ones. The success of a campaign such as that involving “dolphin safe tuna” labels depends upon the number of consumers whose behavior is sensitive how tuna was caught. The impact of such a strategy on producer behavior is a function of the number of actors sufficiently mobilized by the new information to alter their behavior and the aggregate costs those consumers’ behaviors can impose on the target. In order to Often these strategies are

Unlike in a monitoring model, however, a target’s estimate of the costs it will bear for continuing an undesirable behavior or the benefits it will receive for engaging in a desirable behavior are not a function of uncertainty regarding the actors willingness to impose sanctions that run counter to its short-term interests. Rather, the uncertainty arises from the difficulty of estimating how many potential responders there are and the costs to the target of their response. Thus, the credibility problem that inheres in transparency as monitoring approaches is absent in a transparency as mobilization model. Although a range of actors receive information in the latter model, those who do not believe themselves better off imposing sanctions or providing rewards are under no obligation to do so. Thus, problems do not arise because those who consider imposing sanctions or providing rewards as costly fail to make credible commitments to do so anyway. Rather, problems arise because the costs that can be imposed in aggregate by those who view sanctioning “good” or rewarding “bad” targets as a dominant strategy may be insufficient to make targets alter their behavior. Although there is no implication that a mobilization strategy will be more effective than a monitoring strategy, effectiveness is undermined solely by potency, rather than credibility, problems.

A mobilizing model may well target the same behaviors as a monitoring model. To the extent that they rely solely on increasing the amount of transaction-relevant information in the economic or political system they seek to influence, however, they do not bump up against norms of sovereignty or of laissez-faire capitalism that would reject equivalent regulations in a command-and-control mode. Essentially, transparency as mobilization is not considered to be as coercive as transparency as monitoring.


Transparency as facilitating rational choice

If transparency operates by making it in the target’s interest to adopt desirable behaviors in the first two modalities, it can also operate in a third modality by simply demonstrating that desirable behaviors are in the target’s interest. Information can facilitate rational choices by the recipients of the information by making targets aware of behavioral alternatives or the “natural” consequences of their own behavior. It is based on the belief that, were they aware of these options or consequences, they would make different behavioral choices. Such an approach assumes that actors engage in behaviors that are socially undesirable only because they are unaware of consequences which demonstrate that they are privately undesirable as well. In essence, this is transparency as education or transparency as learning. States, like people, can learn in the sense of altering their beliefs in response to new information (Breslauer and Tetlock 1991; Haas 1975; Haas and Haas 1995; Keohane and Nye 1989, 264; Nye 1987).

Transparency in this model works through information provided directly to the targets rather than to potential responders. A learning model assumes that the states and nonstate actors targeted by the regime hold values consistent with those of the regime. They adopt behaviors viewed as undesirable under the regime only because they are misinformed about the optimal strategies for achieving their goals. Targets are assumed to be rational actors who are victims of problems of bounded rationality (March 1978). They may lack complete information, have inaccurate information, fail to recognize the significance of information they do have, or misunderstand how to use the information they do have in choosing privately optimal behaviors. In any of these events, actors may make choices that run counter to their own independent self-interest. The new information provided through the regime is assumed to remedy whatever informational obstacles are preventing actors from pursuing their own individual self-interest which, if pursued, would also further the interests of those supportive of the regime.

Acting in a comprehensively rational manner can involve significant search costs. Numerous obstacles exist to identifying all available behavioral options, their associated costs and benefits, and any probability distributions associated with those costs and benefits. Overcoming those obstacles often proves so difficult and costly that it indeed is rational to not collect information which, if collected, would alter one’s choices. The cost of collecting the information that would allow one to make more optimal choices outweighs the expected increase in benefits from making those choices. International institutions and regimes can alter the decisions of actors in such cases by reducing the costs of acquiring decision relevant information (Keohane and Nye 1998). The nature of bounded rationality dynamics suggest that actors frequently may not seek out information which might help them make more optimal decisions, but they may incorporate such information into their decision-making, if provided with it.

What distinguishes the learning model from the monitoring or mobilizing models is that addresses the problem as an imperfect information problem rather than an externality problem. The point of promoting transparency in the monitoring and mobilizing models is to engage those negatively affected by an existing externality in hopes that they will respond to targets in ways that induce the internalization of the costs they bear. The point of a learning model is to enlighten targets themselves about the effects of their own behavior that may be harmful to their own interests but of which they may be unaware. In a learning model, the behavior in question is assumed to be privately as well as socially suboptimal; in the two other models, the behavior in question is assumed to be socially suboptimal but privately optimal.

The various international regimes based on variants of the principle of prior informed consent provide a common example of this sort of regime. Conventions regulating trade in pesticides, hazardous waste, hazardous chemicals, and, more recently, bio-engineered organisms all rely on a model in which strictures are not placed directly on trade but in which exporters are required to meet certain disclosure requirements such that importers can make informed decisions in light of an assessment of their own interests whether to accept an import (Paarlberg 1993; PIC 1998). Although environmental advocates in each of these arenas might fault these agreements for not establishing reduction targets for activities they consider harmful to the environment, the conventions themselves involve a market-type assumption that those involved in these activities should be allowed to engage in them to whatever level they deem appropriate, so long as they have all information necessary to make a fully informed decision in that regard.

Confidence-building measures in arms control are of the same nature. They involve efforts to establish processes so that each side can provide information that allows the other side to avoid actions that, had they had accurate information, they would have chosen not to take. To the extent that actions depend on accurately understanding what the other side is doing and why, informational imperfections in the system can lead to outcomes that are suboptimal for both parties. Confidence-building measures seek to avert those accidental conflicts in which, after all actions and intentions become clear, the attacker would have preferred not to attack. Thus, transparency of this sort helps actors avoid actions that they themselves would consider privately suboptimal but which they would have taken in the face of uncertainty or misinformation.

Two variants of this model are evident in regimes engaged in scientific research. Improved scientific research can produce changes in behavior in two ways. First, scientific research may better identify the connections between behaviors and their consequences. If such research identifies negative consequences for those engaged in those behaviors and that research is made available to those actors, it may induce changes in behavior. Thus, the extensive environmental research conducted under the Convention on Long-Range Transboundary Air Pollution appears to have produced a change in German behavior largely by demonstrating that much of the forest die-off in Germany was due to domestic rather than foreign emissions of acid precipitants (Levy 1993). Second, scientific research may identify new alternative behaviors that targets view as more attractive than current behaviors. Thus, prompted by the regulatory threat of the Montreal Protocol for protection of the ozone layer, chemical companies conducted research into alternative chemicals and processes that turned out cheaper and better than continued use of CFCs. Although the Protocol itself did impose phase-out deadlines for CFCs, the speed with which CFC use has declined suggests that the availability of cheap alternatives, rather than regulatory enforcement, has been the primary driver of changed behavior (Levy 1995).

The effectiveness of a learning modality approach depends critically on the accuracy of the assumption that the target, if fully informed, would view the behavior in question as counter to its own interests. Regardless of the beliefs of those providing the information, the new information must convince the target that their behavior is privately suboptimal. As with a mobilizing latent enforcers model, the effectiveness of a learning model of transparency depends upon the underlying structure of the preferences of information recipients. The aggregate change in behavior induced by the strategy is determined by the proportion of those receiving the new information who engage in the undesirable behavior because of the imperfect information. In many cases, new information may make targets aware that their current behavior has higher costs or fewer benefits, and that alternative behaviors are less costly and more beneficial, than previously realized. However, behavior change will be induced only if that new information convinces them that the net effect of those costs and benefits makes the alternative behavior superior to the existing behavior. That may frequently not be the case – the information may change the elements of the decision calculus but not the decision itself. This mode of transparency will not alter behavior if, even with the new information, targets still view their current behavior as privately optimal.

Imperfect information problems also are often conflated with externality problems. In such cases, resolving the imperfect information problem alone will not alter behaviors significantly, because even properly-informed targets have incentives to engage in levels of the activity that remain socially suboptimal. Equally important, behavior may be driven as much by capacity and opportunity constraints as by interests. A target may lack the ability to make different choices, even after they recognize that their current choices are privately suboptimal. As considerable recent research has demonstrated, considerable noncompliance with international environmental regimes reflects incapacity problems that new information is unlikely to resolve (Haas, Keohane, and Levy 1993; Victor, Raustiala and Skolnikoff 1998; Weiss and Jacobson 1998). Especially in cases involving investment-type problems, those who cannot afford the immediate costs of switching to a new behavior will not do so, even if doing so would provide huge net benefits over the longer term. In economic terms, the ability to pay may prove to be a more binding constraint than the willingness to pay.

Transparency as learning can also work only if the information is provided to those whose interests are coincident with those of the regime. Thus, criticisms of PIC rules have pointed to the fact that information regarding the problems with various hazardous wastes and pesticides is often received by governments who have incentives to import these materials, regardless of their hazards, but that information may well not be passed on to those people potentially harmed by their use. In many countries, even if those people did receive the information, their input to the domestic policy process short-circuits the pathway by which a learning modality of transparency can work.

The three logics of transparency are summarized in the following table.

Logic of transparency Recipients of information Conditions for success Obstacles to success
Monitoring Regime supporters Regime supporters committed to sanctioning and rewarding targets
  • Collective action problems of sanctions or rewards
Mobilizing latent responders All potential responders Some actors have dominant strategy of sanctioning and rewarding targets
  • Few latent responders
  • Targets insensitive to responders behaviors
Facilitating rational choice All targets Targets engage in undesirable behaviors because of information imperfections
  • Interests dictating behavior
  • Incapacity and opportunities constraining behavior



Transparency is not all of a piece. Transparency has three different logics in the international sphere. International regulatory regimes and institutions can establish information systems that involve “transparency as monitoring,” “transparency as mobilizing latent responders,” or “transparency as facilitating rational choice.” In the first of these, the information systems serve as subsystems within a broader implementation system in which parties to the regime are under some, often implicit, obligation to reward those engaged in desirable behaviors and/or punish those engaged in undesirable behaviors as a means of inducing more of the former. In the second approach, information systems are used to disseminate information to actors who are not actively engaged in or concerned about the operation of the regime. The strategy relies on recipients of the information having interests that will lead them, because it is in their immediate self-interest to do so, to take actions that have the consequence of rewarding targets engaged in desirable behaviors and punishing those engaged in undesirable behaviors. The key difference between this model and a monitoring model is that those expected to respond in a “mobilizing latent responders” are not subject to the collective action problems that can plague transparency as monitoring provisions. In the third model, institutions provide information directly to those actors whose behavior is socially suboptimal. Here the assumption is that these actors engage in socially suboptimal behaviors because of imperfect information. Making them aware of new information will alter their calculus of the behaviors that are privately optimal and those new behaviors will be socially optimal as well.

These differing logics highlight differences in the pathways by which information influences state and nonstate actor behavior. They suggest that the different types of transparency will be adopted in different circumstances. They also point to differing conditions under which we should expect different types of transparency to work effectively at altering behavior. We will need to make more careful distinctions in assessing how transparency is deployed in a regime if we are to accurately assess what role, if any, transparency plays in the behavior of states and nonstate actors in international affairs.