Scholars typically model the politics of global public goods or common pool resources as difficult collective action problems. Theories of international organization aim to explain how institutions can promote cooperation by solving the free rider problem. Based on an analysis of a quintessential global collective action problem—international climate mitigation—this article challenges both this diagnosis of the problem and the concomitant institutional remedies. Important elements of climate mitigation exhibit three key features that depart from the canonical model: joint goods, preference heterogeneity, and increasing returns. The presence of these features creates the possibility for “catalytic cooperation.” Under such conditions, the chief barrier to cooperation is not the threat of free riding but the lack of incentive to act in the first place. States and other actors seek to solve this problem by creating “catalytic institutions” that work to shift actors’ preferences and strategies toward cooperative outcomes over time. While catalytic institutions can be seen in many areas of world politics, the 2015 Paris Agreement on climate change has put this logic of cooperation at its core, raising the possibility that similar catalytic institutions may facilitate cooperation in other areas of world politics characterized by analogous problem structures.
To the extent collective action and free riding dynamics do not drive climate policy—the premise of this special section—what are the implications for international cooperation? Traditionally, scholars have understood climate change as a tragedy of the commons and international cooperation on mitigation as a kind of global Prisoner’s Dilemma.1 They therefore advance solutions such as international institutions to monitor compliance, sanction defection, facilitate reciprocity, or assign property rights. But if we instead see climate policy as primarily driven by domestic distributional conflicts, are different strategies needed to promote cooperation, or is collective action simply not relevant?
This article aligns with Alkin and Mildenberger (this issue), arguing that the conventional wisdom mischaracterizes the nature of collective action for climate mitigation. However, it also demonstrates that cooperation remains important when certain conditions are met. Specifically, the article identifies features of cooperation problems that make “catalytic cooperation” possible and shows that these conditions apply to many areas of climate mitigation.2 The article also explains how international institutions can promote cooperation in such circumstances not by solving the problems of free riding and credible commitments but by mobilizing incremental action that can shift the strategies and preferences of states and other actors over time.
The article illustrates its theoretical arguments through institutional shifts in the United Nations Framework Convention on Climate Change (UNFCCC) process. The 1997 Kyoto Protocol, following most other multilateral environmental agreements, took a “regulatory” approach, in which states negotiated a set of shared reduction targets. States are bound to report on their emissions and, should they miss the agreed target, face (minimal) sanctions. The 2015 Paris Agreement, instead, requires each country to put forward its own pledge, or nationally determined contribution (NDC). These pledges are then reviewed internationally and ratcheted up every five years. Paris also increases the role of cities, businesses, provinces and regions, and other sub- and nonstate actors, creating institutions to recognize and orchestrate their climate action alongside the national pledges. Both these processes aim toward the long-term goal of limiting climate change to “well below” 2°C in this century, which requires effectively decarbonizing the world economy in the next decades.
Can this “pledge and review and ratchet” system work? Existing theories of international organization see the Paris Agreement as a positive but modest achievement for cooperation because it does little to solve free rider problems and lacks robust enforcement provisions (Bang et al. 2016; Barrett and Dannenberg 2016; Keohane and Oppenheimer 2016; Sachs 2019). This article makes two theoretical arguments that invite a different interpretation.
First, it aims to correct a persistent misdiagnosis in international relations theory, which overemphasizes free riding as the chief barrier to international cooperation around the global commons. Though the multiplayer Prisoner’s Dilemma represents prima facie a compelling interpretation of global climate mitigation, much of the observed political behavior and outcomes over the past two decades depart from its theoretical expectations (Alkin and Mildenberger, this issue). Three generalizable patterns emerge: climate policies provide private benefits as well as collective ones (“joint products”); the value actors place on mitigation, as well as the costs and benefits they face, vary enormously (“preference heterogeneity”); and action in the past lowers the costs and raises the benefits of action in the future (“increasing returns”).
To the extent these conditions apply, the catalytic cooperation model, which draws on other elements of cooperation theory, implies a very different set of barriers to, and strategies for, cooperation. If early movers reach a critical mass, and if past action has a large impact on subsequent action, cooperative action can become progressively self-reinforcing over time. This approach aligns with authors who see decarbonization as a problem of overcoming “lock-in” through innovation and scaling (Bernstein and Hoffmann 2019; Farmer et al. 2019) or who emphasize dynamics between front-runners and followers (Torney 2019; Busby and Urpelainen 2020). This article elaborates the implications for, and role of, international cooperation in this approach.
Second, the article introduces the concept of catalytic institutions, demonstrating how the Paris Agreement and also international institutions in areas such as trade, human rights, and global health attempt to support cooperation not by providing credibility but through catalytic mechanisms that help build action over time by progressively shifting preferences. While it is by no means clear that the Paris Agreement will catalyze sufficient cooperation to avert dangerous changes in the Earth’s climate, the article provides a theory to understand how it might.
The next section demonstrates how climate mitigation departs from conventional theoretical assumptions, identifying three conditions that define a catalytic model of cooperation. The subsequent section examines the implications of this model for cooperation. The fourth section then elaborates the logic of catalytic institutions and shows how the Paris Agreement and other international institutions work to shift state preferences and strategies over time. In conclusion, the article identifies future research questions, implied by the model, that will need to be answered to understand prospects for catalytic cooperation in climate mitigation. It also considers how the problem structure of climate change may shift over time and in what other issue areas catalytic institutions may support cooperation.
Three Conditions That Allow Catalytic Cooperation
Theories of collective action in political science, sociology, economics, and other fields focus on the incentive to contribute to collective action given the actions of others. Canonically, IR scholars focus on states, but because the theoretical setup below is more general, and because cities, businesses, and other such entities play an important role in national and global climate politics, I use the term actors.
In the classic model, the level of collective action provided is a production function P of the sum of all actors’ contributions, A (i.e. the product of the number of contributors and the magnitude of their individual contributions) (Marwell and Oliver 1993). For a given actor i, the benefit of acting gi (ai: A) is the value vi it derives from the collective action P(A) minus the cost ci of its individual contribution ai:
In this framework, the canonical interpretation of climate change is that no actor has an individual incentive to act absent some collective agreement; that is, on average, ci(ai) > vi [P(A)]. This situation creates a familiar and intuitive problem, a large-scale Prisoner’s Dilemma (Barrett 2003; Hardin 1968; Sandler 2004). Bigger actors may have more to gain by acting, because they can do more to affect the problem, but if no actor’s contribution is by itself sufficient to gain an overall benefit that outweighs its cost of acting, collective action will not be forthcoming. Furthermore, even if some action should occur, an actor has little incentive to join in because the actor can more easily free ride on others’ efforts. Absent a credible commitment for all to act, no action is taken, and everyone ends up worse off.
But how accurately does the classic model of collective action capture the politics of climate mitigation or of other global collective action problems (Alkin and Mildenberger, this issue)? A review of mitigation politics over the past decades provides evidence for three features of climate’s problem structure that depart from strong assumptions of the Prisoner’s Dilemma model: joint products (contributions to collective action can yield “private” benefits for those that act inaddition to public benefits shared by all), preference heterogeneity (some actors value collective action highly, others little; similarly, the costs and benefits of a given contribution to collective action vary across actors), and increasing returns (action in the past can reduce the cost and increase the benefit of action in the future, while also changing how actors value collective action). In what follows, I explain the logic of each revised assumption and demonstrate its empirical basis in the realm of climate mitigation. To the extent they apply, they shift the classic model of collective action described herein to a “catalytic” model of collective action, which I define in the forthcoming pages.
The classic model assumes that contributing to a collective good is costly to the contributor. In reality, of course, policy actions typically involve a mixture of costs and benefits to different actors. This means that collective goods are often “joint products” in that actors’ contributions provide both a public good to the community and a private benefit to the contributor. Even seemingly pure public goods (e.g., funding basic research to create scientific knowledge) can come with private benefits of some kind (e.g., jobs for researchers). To the extent private benefits exist, we can expect at least some actors—what Olson (1971) termed a “privileged group”—to take at least some actions irrespective of others’ behavior.
If the cost of a contribution tends to exceed the private benefit it provides the contributor, which we can separate out as a new term bi, then the classic assumption is sensible. But the empirical record shows many cases in climate policy for which this assumption is too strong (Lipscy, Unpublished; Sandler 2004, 43). After all, a county’s net contribution to mitigation is almost never a single policy like a carbon tax but rather a wide range of regulations and actions across nearly every sector, including land-use policy, local zoning laws, transportation initiatives, development assistance, and so on. Indeed, many actions relevant for mitigation may not be perceived primarily as climate policy per se.
Given this diversity, it is not surprising that many mitigation actions entail so-called co-benefits, including reducing local air pollution and improving human health, increasing energy security and reliability, developing new industrial sectors, preserving forests, reducing traffic, and so on. Climate action can also bring narrower private benefits, such as material support for influential stakeholders (e.g., subsidies for farmers to make biofuels) or political benefits for politicians who face pro-climate constituencies. In some cases, these private benefits temper the cost of mitigation actions; in other cases, the other benefits are actors’ primary focus, and mitigation is an ancillary result. Even if it were true that, on average, most mitigation actions came with a net cost, some would not, and these “net positive” actions could allow actors to take significant steps even if they do not hold strong pro-climate views (Green 2018).
The classic model assumes relatively symmetric preferences across actors and actions, typically based in an intuitive but crude political economy model in which actors prioritize overall economic performance. While this simplifying assumption may be accurate on average, again, the empirical literature on climate change provides evidence of significant exceptions.
Actors face very different benefits and costs from mitigation policies. For example, countries with water resources, such as Sweden or Costa Rica, have already nearly decarbonized electricity production through hydropower. Moreover, the “natural” variation actors face is exacerbated by the different decision-making institutions through which they form policy preferences and the relative power of “winners” and “losers” of specific climate actions in decision-making processes. Comparative studies of national climate policy formation show how regime types and political-economic systems shape countries’ climate policies by valuing public goods differently, giving fossil fuel interests groups more or less power or by giving more or less value to outcomes in the future (Bayer and Urpelainen 2016; Dubash et al. 2013; Lachapelle and Paterson 2013; Mildenberger 2020; Tobin 2017).
Expanding our understanding of actors’ preferences to consider nonmaterial costs and benefits creates further possibilities to depart from the symmetric assumptions of the standard model. Even when there are net material costs to mitigation actions, it is not obvious that they always dominate other incentives, including ideational preferences for or against green policy. On average, decision makers certainly prefer to avoid expending money with little immediate return, but many decision makers face pro-climate stakeholders, including citizens and voters, customers, and investors. Under these conditions, decision makers may derive substantial political or financial benefits from acting.
Just as actors face sharply different costs and benefits of acting, so too do they value the benefits of collective action vi differently. Climate impacts vary significantly across space and time, but it is common in the literature to assign a “social cost of carbon” (an estimate of the impacts of future warming on welfare) and discount rate to the future benefits of mitigation. These simplifying assumptions can obscure more than they clarify, since both impacts (Ricke et al. 2018) and time horizons vary significantly across actors. For small island states, arid countries near the equator, and coastal cities, the impacts are already severe and will become existential. For more temperate and inland areas, the changes will likely be slower and more moderate. Moreover, differential levels of economic development and state capacity will also mean that some actors will be able to adapt to climate disruptions much more effectively than others. Political institutions will also create variation in actors’ discount factors, since different political systems value long- and short-term impacts differently (Jacobs 2016). For these reasons, we can expect vi to vary dramatically across actors, with some motivated to act even if their action only generates a small level of collective good provision and others unlikely to act even if they could provide significant public goodas unilaterally.3
In the traditional setup, action by others almost always dissuades an actor from acting. If others contribute, the actor can free ride on their efforts. If others slack, the actor can scarcely hope to solve the problem without them.
Moreover, the model assumes, implicitly, that it is useful to conceptualize actors’ preferences as if they do not change over time.4 In reality, some policy choices reinforce themselves through a variety of feedback loops that generate “increasing returns” to action over time (Pierson 2000). In this way, prior action can shift actors’ preferences toward further action over time. Of course, not all realms of climate policy exhibit increasing returns, and in some areas, increasing returns have been too weak to overcome entrenched opposition (Stokes 2020). It is also possible to imagine that the opposite effect may occur, as first-movers harvest low-hanging fruit, leaving only more costly actions for later-movers.
Ultimately, which dynamic dominates is an empirical question, but the literature highlights many examples of increasing returns at play (Levin et al. 2012; van der Ven et al. 2017; Urpelainen 2011, 2013). As Bernstein and Hoffmann (2018) note, decarbonization in one area can scale and entrench in other areas through normalization, capacity building, and coalition building. Material, informational, and normative mechanisms can be identified, affecting different parts of the model vi, bi, and ci.
First, some climate action can affect the material costs of future action, ci, by changing technology and the economic systems around it. The literature on sociotechnical transitions shows how niche innovations by “pioneers” can diffuse to eventually reconfigure larger systems (Geels et al. 2017; Victor et al. 2019; Zenghelis et al. 2018). As technologies are developed and deployed, their costs descend down a “learning curve,” becoming cheaper as more R&D is conducted and as production and distribution systems “learn by doing” and reach economies of scale (Hale and Urpelainen 2015). Renewable energy technologies fit this pattern well. For example, the cost of photovoltaic cells dropped 75 percent from 2010 to 2015 and of wind turbines 30–45 percent in the same period; both technologies are now at or below the cost of fossil fuel alternatives in many parts of the world, sharply altering the incentives for taking climate mitigation action (IRENA 2017). These cost reductions were only possible because of costly actions by first-movers like California, Denmark, and Germany (Breetz et al. 2018; Sawin 2001).
In addition to learning curves, many new technologies and business models demonstrate network effects; like telephones or email, the more people have them, the more useful they are, raising the private benefits, bi, of such actions. Consider electric cars or solar rooftops. For the first-movers, such products are very inconvenient because there are too few charging stations, maintenance technicians, or electric cables and pricing systems that allow homeowners to send power back to the grid. As market penetration increases, however, the enabling environment shifts and later adopters are well supported. Indeed, once network effects are strong enough, they may help to lock in new technologies as default options.
Second, early action can affect the political processes of preference formation for states and other actors by generating new constituencies for action (Breetz et al. 2018; Meckling et al. 2015; Urpelainen 2013; van der Ven et al. 2017). As new technologies emerge and grow, their producers and consumers develop a distributional interest in their continuance and expansion. At first, these new interest groups are unlikely to be able to overcome established incumbents in political contestation. But in economic sectors or geographic regions where incumbents are relatively weak, the new entrants may thrive and, as action spreads, eventually acquire the size and clout to become politically competitive with incumbents in more and more jurisdictions and industries.
Third, prior action can also generate learning effects. Just as mitigation actions generate new technologies and business models that alter material costs, so too do they produce new knowledge about policy design and implementation through experimentation and diffusion, further reducing costs and enhancing benefits (Hoffmann 2011; Sabel and Victor 2017). This effect is significant because many areas of mitigation involve complex policy instruments. Emissions trading systems are a prominent example, where even highly competent regulators like the European Commission have struggled to make their credit allocation and exchange systems operate smoothly. For this reason, the Chinese government has drawn on significant international expertise in the development of its national emissions trading system and proceeded incrementally by first experimenting at the provincial level. In this way, previous actions generate epistemic resources that improve the ability of followers to emulate leaders (Biedenkopf et al. 2017).
Finally, over time, growing actions may change norms around mitigation, altering not just costs and benefits but also how actors value mitigation itself, vi. IR scholars have shown how “norm entrepreneurs” engage in “strategic social construction” by attempting to shift norms toward their policy goals (Finnemore and Sikkink 1998, 895). In the realm of climate, Bernstein and Hoffmann (2018) argue that actions in one sphere, such as a city setting a climate target, can “normalize” low-carbon preferences in ways that spill over to other actors. These models follow a logic of increasing returns. As Finnemore and Sikkink argue (1998), norms progress through a life cycle from emergence, to a “norm cascade” in which they become widely followed in practice, to internalization, in which they are embedded in the beliefs and preferences of most actors. As more action takes place, more of this self-reinforcing logic applies. While decarbonization norms are not currently widespread, some areas of climate mitigation show evidence of norm cascade dynamics, such as divestment from fossil fuel companies (Green 2018).
To the extent these mechanisms apply, prior actions increase the ratio of benefits, bi, to costs, ci, for subsequent actions, as well as the value actors place on mitigation, vi. Under these conditions, actors’ preferences shift dynamically in ways not accounted for in the classic framework.
A Catalytic Model of Cooperation
The foregoing three modifications change the classic model of collective action into a catalytic model of collective action, which can be expressed as follows:
Implications for Cooperation
The classic model suggests dim prospects for cooperation, as can been seen clearly in Schelling’s classic representation (Figure 1) (Schelling 1978; Snidal 1985). The vertical axis represents the payoff to actor i from collective action as a function of the total amount of action, which is measured along the horizontal axis. Actor i may either cooperate or not cooperate, and the payoff for each choice is shown on a different curve. Because the noncooperation curve is always higher than the cooperation curve, actor i will not act without an inducement, I, at least as large as the distance between the cooperation curve and the noncooperation curve. Inducements could come in the form of side payments, sanctions, issue linkage, or, under repeated iterations, a credible pledge to unlock reciprocal cooperation from others.
An exception occurs if a single actor or coalition of actors can contribute at least k, the point where cooperation and noncooperation yield the same payoff. For example, a hegemonic state or coalition of great powers may have an interest in providing a public good, even if it means other states will free ride on them, because the actions of the “k-group” are themselves sufficient to generate enough of the collective good to be worth the costs (Snidal 1985).
How does the catalytic model described change the prospects for cooperation? Three implications follow from the revised assumptions.
Students of collective action have shown how joint products and preference heterogeneity can facilitate cooperation (Hardin 1982; Marwell and Oliver 1993). In the catalytic model, these characteristics make action strictly preferable to inaction for some actors in two ways: first, the actor may value mitigation so much (vi may be so high) that even a very low level of action A, potentially just the actor’s own contribution ai, could produce a sufficient benefit to outweigh the cost ci(ai) (heterogeneous preferences), and second, the private benefit bi(ai) of acting is higher than the cost ci(ai) (joint products). Combining these conditions, we can expect cooperation to be strictly preferable to noncooperation whenever vi[P(A)] + bi(ai) > ci(ai). In Schelling’s representation, this “unilateral cooperation” can emerge as follows (Figure 2).
First, as costs, ci, fall and private benefits, bi, rise, the cooperation curve shifts upward vertically, potentially falling above the noncooperation curve. Even if costs remain higher than benefits, the catalytic model suggests that heterogeneous preferences and joint products can reduce actors’ inducement costs, I.
Second, when actors value the benefits of mitigation more (say, vi is replaced by a higher value vi), the threshold for unilateral action, k, is lower, say, k′. Valuing mitigation outcomes more increases the slope of both the cooperation and noncooperation curves, rotating them counterclockwise, which reduces the size of the k-group required to make cooperation attractive at a given ratio of costs to benefits.
The inverse situation, when vi and bi are low and ci is high, is also of theoretical interest. Under these conditions, collective action is just not “worth it,” meaning that even if a collective deal could be reached, it would not provide sufficient value to deter free riding. In the realm of climate change, this situation is likely applicable at least in part to a number of significant actors, such as corrupt regimes with large oil reserves or fossil fuel–dependent companies focused on quarterly earnings reports.
Increasing Returns and Tipping
In the catalytic model, increasing returns make vi, bi, and ci partially endogenous to A, with the value of collective action and private benefits rising and costs falling as action expands. This change can be represented as an upward shift in the slope of the cooperation curve as action accumulates. In Figure 3, as increasing returns strengthen, k shrinks to k′ or even k′, and I shrinks to I′.
To the extent increasing returns exist, a different pathway emerges for achieving cooperation: unilateral cooperators can become “first-movers” who reduce the costs and increase the benefits of acting for followers and themselves. As Marwell and Oliver (1993, 88) note, “initial contributions lower the necessary interest for subsequent contributions.”
Similar ideas inform “cascade” or “threshold” models that have been used in political science to describe norm diffusion, protests and revolutions, racial segregation of neighborhoods, treaties that require a minimum critical mass to go into effect, and other phenomena, and have been used in experimental settings to analyze the Paris Agreement (Barrett 2003; Barrett and Dannenberg 2017; Finnemore and Sikkink 1998; Granovetter 1978; Kuran 1989; Schelling 1971). But these applications tend to describe coordination games of various kinds, in which actors are trying to converge on an equilibrium or shift from one equilibrium to another. For example, Schelling’s foundational study of neighborhood racial segregation showed how potentially even a small increase in the number of nonwhite residents could induce first the most anti-integration whites to move away, then the next most racist ones, and so on, potentially setting off a chain reaction that could change the racial composition of the entire neighborhood (Schelling 1972). Similarly, in Downs et al.’s (1998) model of evolutionary multilateralism, a small group of first-movers is able to elicit greater cooperation over time by changing the strategic choices followers face. In all such models, actors’ preferences remain constant (though note that preference heterogeneity plays a key role); their choices change as others’ behavior changes.
In the catalytic model developed here, in contrast, actors’ very preferences alter as a consequence of their and other’s actions. If increasing returns are sufficiently strong, the cooperation curve may intersect and rise above the noncooperation curve at a certain level of A. If this happens, a “tipping point” T occurs at which action becomes strictly preferable and self-reinforcing.
How the Number of Actors Affects Cooperation
In the classic collective action approach, a larger number of actors makes cooperation more difficult. If actors are relatively similar, a larger number of them means that each one’s individual contribution will diminish relative to the whole, making it harder to assemble a k-group. Moreover, the transaction costs of negotiating and implementing a collective agreement to induce cooperation are higher (Olson 1971). For these reasons, a number of scholars have emphasized the value of “mini-lateral” institutions with smaller memberships (Eckersley 2012; Falkner 2016).
The catalytic cooperation model, in turn, helps us identify conditions under which increasing the number of actors can instead facilitate cooperation. Building on Olson and Hardin, Marwell and Oliver (1993, 49) show that “when groups are heterogeneous and a good has high jointness of supply, a larger … group can have a smaller critical mass.” This seemingly heterodox logic is deceptively simple. Assume, conservatively, that preferences are distributed normally across actors. Following the logic of joint products and heterogeneous preferences, raising the number of actors therefore increases the likelihood that at least a few actors will hold very pro-action preferences, because we become more likely to reach the “tails” of the distribution of preferences. Under a collective bargaining framework, this would do little to increase cooperation, because pro-action actors would, on average, be balanced by anti-action actors. But in a catalytic cooperation context where increasing returns apply, having a continuous chain of actors along the full preference spectrum—and especially at the pro-action extreme—is critical. This distribution of preferences increases the likelihood of at least one or more actors having a strong enough preference to act even when it is costly to do so.
Furthermore, having more actors in the system makes it easier for increasing returns to set off a chain reaction. As Granovetter’s threshold model shows, a chain reaction can occur when actors’ thresholds for action (k or T in the preceding figures) are sufficiently close together, such that each new action is enough to reach the next most pro-action actor’s threshold (Granovetter 1978; Macy 1991). Assuming a sufficiently continuous distribution, increasing the number of actors means that it is less likely that a large gap between thresholds will stall the effects of increasing returns.
Each of the three mechanisms described shows how cooperation might emerge, given the assumptions of the catalytic cooperation model. While each individually highlights a potentially sufficient mechanism for collective action, they are also additive and complementary, together composing a catalytic pathway to cooperation. This model can of course also potentially combine with “traditional” approaches to cooperation, such as collective agreements to ensure that the benefits of collective action are realized, sanctions against defectors, provision of exclusive benefits, and so on.
What role can international institutions play when commons problems are characterized by the catalytic cooperation model? The standard institutional solution of collective agreement, monitoring, and sanctions corresponds to the classic model of collective action. These features describe the basic contours of the Kyoto Protocol and many other agreements to regulate the global commons. However, as Schelling wrote following the Bush administration’s “un-signing” of the Kyoto Protocol, such an approach has not proven workable in climate mitigation (Schelling 2002).
Because the Paris Agreement allows actors to set their own targets and does not impose penalties for failing to meet them, the traditional theoretical lens gives reason to interpret Paris as a modest achievement. For example, Keohane and Oppenheimer (2016, 10) argue that Paris uses “discretion and vagueness” to make countries comfortable with making commitments (Keohane and Oppenheimer 2016, 10). Future commitments will therefore be determined through domestic politics. Slightly more positively, Victor (2015) notes that “flexibility offers a way to get started and build confidence that, in time, will beget more confidence and a willingness to do more,” but that “eventually a much more integrated global treaty will be needed to make major cuts in the greenhouse gases.” Bang et al. (2016, 209) are more critical, noting that “Paris does little to restructure states’ incentives so as to avoid free riding,” though they hope actors will become more pro-climate over time. Sachs (2019) instead finds little evidence to support the hope that peer pressure from other states or civil society will consistently shift countries toward greater emissions reductions. Barrett and Dannenberg (2016) present experimental evidence that pledge and review systems can help increase contributions, but not to the threshold required to reach the goal they have set. Following these arguments, we can see Paris is a positive but ultimately modest step.
The catalytic model offers a different interpretation of the Paris architecture. I do not make any claims here regarding why the treaty looks the way it does, nor do I consider it a “perfect” example of a catalytic institution. Instead, this section uses key features of the Paris Agreement to illustrate the logic through which, when the conditions specified by the catalytic model are met, institutions can drive collective action. These are: stimulating first-movers and incremental action through flexibility; iterating commitments; and increasing the effect of prior action on subsequent action via material transfers to alter future preferences and capacities, transferring experiences to shape the informational environment, normative goal setting and benchmarking, and domestic constituency building.
Stimulating First-Movers and Incremental Action Through Flexibility
In the catalytic cooperation model, the most important challenge is to reach a critical mass of action that can begin to generate increasing returns. Catalytic institutions therefore seek to stimulate first-movers to come forward and to encourage small steps even from recalcitrant actors.
By lowering the cost of action to what actors are willing to do, flexible commitments like NDCs can encourage early moves. If the bar for cooperation is too high, that is, a minimum a is required like a binding commitment required by Kyoto, these “small steps” might never occur (Urpelainen 2013; Victor 2011). Moreover, flexible commitments allow the most pro-mitigation actors to put forward ambitious commitments, instead of limiting themselves to a least common denominator negotiated outcome.
Similarly, the Paris Agreement seeks to spur first-movers among cities, businesses, and other actors (Hale 2016). Again, this is facilitated by the flexible nature of the Paris commitments, which allows for a variety of commitments and actions at varying degrees of ambition to be made. These actors add further to the mass of first-movers. Moreover, as I have argued, increasing the number of actors in this way makes it easier to create a chain reaction by shrinking the gaps between actors’ thresholds for action. A weakness of the voluntary approach, however, is that it does not necessarily elicit first-movers in sectors where, for example, the costs of acting remain high, meaning some segments of the economy (e.g., heavy industry) require further policy interventions to catalyze action (Victor et al. 2019).
Iteration of Commitments
Even though flexible commitments capture only what actors are willing to do, with increasing returns, what actors are willing to do changes over time. Catalytic institutions therefore create an ongoing process to record updates in actors’ preferences over time. By requiring new commitments every five years, the Paris Agreement allows increasing returns to be translated into new, stronger actions, which may then generate their own effects. A “one-off” commitment model would lack this ability to harvest the beneficial effects of past actions on subsequent actions over time.
As Macy notes, in “cascade” models of collective action, it is much easier to achieve cooperation when choices are serial rather than parallel (Macy 1991). Indeed, iteration of commitment making is a common feature of international institutions. The global trade regime has developed through progressive trade rounds. More proximately, the “framework and protocol” approach to global commons issues like the ozone regime typically involves a series of increasingly stringent negotiated agreements, as was originally pursued in the climate regime (Sebenius 1991). While these regimes use collective agreements enforced with sanctions, they also have catalytic elements in that past cooperation alters future preferences. In trade, for example, the expansion of multinational companies enabled by early trade rounds reshaped the domestic politics of economic openness in major economies by creating powerful new constituencies for integration (Bailey et al. 1997). Paris replicates this logic, but with individual (for both countries and other actors) as opposed to negotiated commitments, which allows it to capture updates in actors’ preferences more quickly and easily.
Increasing the Effect of Prior Action on Subsequent Action
As discussed, many of the mechanisms through which increasing returns accrue fall outside the realm of international institutions. For example, reductions in the price of clean technologies are transmitted through markets and social norms diffuse through a wide range of processes (Bernstein and Hoffmann 2019). However, international institutions can play a number of complementary roles to augment increasing returns.
Material Transfers to Alter Future Preferences and Capacities
Many international agreements involve resource transfers, for example, payments from one country to another to host a military base. Many such transfers are simply the inducement cost, I, needed to create cooperation. Transfers become catalytic when, above and beyond providing a direct side payment, they increase the capability of the recipient to cooperate by affecting vi, ci, and bi. For example, a grant that helps pay for a renewable energy project reduces the one-off cost of cooperation. In contrast, a training and support package that improves a country’s ability to run effective tendering processes for energy procurement could permanently reduce ci, strengthening the country’s preference for cooperation. Similarly, technical support that gives a country a better sense of what impacts it will suffer from climate change may increase the value vi it places on mitigation efforts.
While the Paris Agreement did not contain any firm commitments to increasing funding for developing countries (arguably, the Kyoto Protocol was superior in this regard), and contributions to the Green Climate Fund have proven small, it did create a Paris Committee on Capacity Building that oversees and supports developing countries’ ability to formulate and implement national climate policies. It remains to be seen, however, whether this process will actually generate resource flows that have a catalytic effect.
Instead, the Paris Agreement’s greatest effect on material transfers may come indirectly. Many of the transnational networks of sub/nonstate actors aim at peer-to-peer capacity building (Bulkeley et al. 2014). In addition, important bilateral donors and multilateral development banks are moving to align their portfolios with the goals of the Paris Agreement (MDB Paris Alignment Working Group 2018). These trends flow in part from the Paris Agreement’s catalytic effect as a normative goal setter (see below).
Sharing Experiences to Shape the Informational Environment
Many international institutions seek to generate and transfer information about state behavior to enhance compliance. As discussed in the third section, such information is catalytic when it creates demonstration and learning effects that boost actors’ ability and willingness to undertake mitigation actions, potentially reducing ci and bosting bi.
Several mechanisms in the Paris Agreement help to generate and diffuse these epistemic resources. First, flexible commitments, in addition to stimulating first-movers and incremental action, can help drive the policy innovation through experimentation (Sabel and Victor 2017). As the difficult work of implementing these heterogeneous plans progresses, actors will gain useful experience of what works and what does not.
Second, catalytic institutions generate and diffuse information via review processes (Abbott 2017; Aldy 2018; Sabel and Victor 2017; Victor et al. 1998). In addition to assessing compliance, review processes allow experts to gather information about state behavior vis-à-vis an international obligation, generating and transmitting knowledge about how best to approach the problem (Chayes and Chayes 1995). Peer-to-peer transgovernmental networks have been shown to employ particularly influential versions of this type of review (Slaughter 2004) and have played an important role in the ozone regime and other global commons issues (Victor et al. 1998). The Paris Agreement includes review of individual countries’ implementation of NDCs (Art. 13), review of aggregate progress toward the long-term goal in a Global Stock Take (Art. 14), and a “non-punitive” enforcement review designed to troubleshoot barriers to NDC implementation (Art. 15).
Third, a major goal of linking sub- and nonstate actors to the intergovernmental regime is to promote learning. The number and diversity of sub- and nonstate actors makes them excellent laboratories for climate policy, and many transnational initiatives explicitly promote information exchange (Roger et al. 2017). The Paris system tries to enhance the epistemic benefits of sub- and nonstate climate action by creating structures to review and extract lessons from nonstate actors, including the NAZCA platform, the climate action events at COPs, and an annual Yearbook of Climate Action. Still, as Abbott (2017) notes, there is potential for the UNFCCC and other actors to play a more active role in enhancing the catalytic effect of these transnational elements of the regime.
Normative Goal Setting and Benchmarking
The second section noted how increasing action can normalize itself, potentially generating norm cascades, and therefore altering actors’ preferences in favor of further action. Catalytic institutions seek to drive such processes through goal setting and benchmarking.
Countries often declare their collective intent to eradicate a disease, uphold human rights, or provide other global public goods, such as the Sustainable Development Goals. Article 2 of the Paris Agreement reaffirms countries’ commitment to limit temperature change to “well below” 2°C, aiming toward 1.5°C, which, Article 4 notes, requires making sure no more GHGs are going into the atmosphere than are coming out by the middle of this century. As many observers have noted, absent concrete plans and accounting, goal setting can be cheap talk or even intentionally dissembling (Downs et al. 1996). Nonetheless, scholars have identified various conditions under which, and mechanisms through which, goal setting can affect political behavior (Biermann et al. 2017; Kanie et al. 2017).
First, a goal provides a focal point around which actors can converge. This mechanism is unlikely to sway actors who do not wish to cooperate, but it can enhance efficiency and enable benchmarking among actors guided by a logic of appropriateness, as first-movers are likely to be (Urpelainen 2013; Young 2017). In this regard, it is notable how the Paris Agreement’s 1.5°C and “net zero” goals have been adopted as a common target by a wide range of countries, businesses, subnational governments, investors, and other actors (UNFCCC 2019).
Second, international goals can enhance the leverage of pro-cooperation constituencies in domestic politics, intrafirm deliberations, or other spheres of political contestation. To the extent states or other actors are sensitive about reputational critiques, explicit goal setting raises the costs of noncompliance and gives pro-compliance groups a “hook” for their arguments (Thomas 2001). For example, the second section noted how decision makers sometimes act to satisfy pro-mitigation constituencies, who, after 2015, would be unlikely to accept climate action that was not seen to be “Paris compatible.”
Once a normative goal has been set, it becomes possible to review actors’ progress toward that goal. Benchmarking systems—which grade states against some standard and compare their compliance to others—are now common tools in world politics (Kelley 2017). While some advocates proposed creating a grading system for NDCs in the Paris Agreement as part of the enhanced transparency framework, countries balked at exposing their “nationally determined” contributions to the collective judgment of their peers or others (suggesting the power of such ranking systems). Still, the Paris transparency framework requires countries to report on their progress toward their NDCs and thus creates the conditions under which third parties can compare and rank national ambition (van Asselt 2016; Climate Action Tracker 2019).
Domestic Constituency Building
Catalytic institutions also shift preferences by reaching into the black box of domestic politics and building new constituencies for greater climate action. While orchestration is widespread across many areas of world politics, the climate regime is perhaps unique in the extent to which policy entrepreneurs engaged in “webcraft” to orchestrate new actors to complement intergovernmental diplomacy (Betsill et al. 2015; Hale 2016; Hale and Roger 2014; Slaughter 2017). While transnational governance of climate change had been building over the course of the regime, in the lead-up to Paris, the UN secretary-general, the UNFCCC Secretariat, and the Peruvian and French hosts of COP20 and COP21, respectively, took a much more purposeful approach, mobilizing dozens of initiatives that ultimately came to include more than 10,000 actors. At Paris, this orchestration function was then institutionalized in the UNFCCC process (Hale 2016). As described, these efforts can have catalytic effects by influencing the process of national preference formation. The more sub- and nonstate actors in a country take ambitious climate action, the more they increase the likelihood of both their peers and the national government adopting strong climate policies as well (Bromley-Trujillo et al. 2016; Cao and Ward 2017; Urpelainen 2009).
Conclusions: Where and When Can Catalytic Institutions Work?
The article has identified three features of climate mitigation—joint products, preference heterogeneity, and increasing returns—that depart from standard models of collective action. Together, these characteristics, to the extent they apply, allow for catalytic cooperation. The argument is not that global climate mitigation has no resemblance to a large-scale Prisoner’s Dilemma but rather that choosing this, and only this, analytic model delivers a very partial understanding of global climate politics. In addition, the article has argued that catalytic institutions like those in the Paris Agreement can help drive international cooperation when conditions for catalytic cooperation obtain. Such institutions seek to initiate and stimulate early action, to iterate commitments over time, and to maximize the positive feedbacks from early action.
While the theory does not allow us to predict how effective the catalytic properties of the Paris Agreement will ultimately be, it points our attention to several factors on which its success will depend, identifying important areas for further research.
First, will enough first-movers appear? To generate catalytic cooperation, we need sufficient “unilateral cooperators” to emerge, either because they value cooperation highly (heterogeneous preferences) or because the benefits of acting outweigh the costs (joint products). In the first years since the Paris Agreement, a number of ambitious pledges have been made by small, progressive economies; vulnerable states; and European nations as well as many significant sub- and nonstate actors. But other large developed and emerging countries have either proceeded incrementally, stalled, or reversed (UNEP 2019). Indeed, the same elements of the model point our attention to the actors least likely to cooperate, those who face high decarbonization costs, who value mitigation little, and whose preferences are least likely to be shifted by increasing returns mechanisms (e.g., fossil-dependent authoritarian regimes with relatively closed economies). To the extent these actors dominate, prospects for catalytic cooperation will remain bounded.
Second, to know whether this initial mass of first-movers is sufficient, we need to know the magnitude of increasing returns—material, informational, and ideational. For example, renewable energy has shown impressive growth in the power sector. Can we expect similar rapid scaling in transportation, land use, or heavy industry (Lafond et al. 2018)? Ideational shifts may be even more difficult to predict. For example, will rising civil society activism reshape climate politics globally, or will it be limited to a relatively small number of Western democracies? Will changes in dietary norms scale up to meaningfully reduce deforestation? Empirical answers to these questions are needed to assess the potential for catalytic cooperation to proceed in climate change.
The model also leads us to ask how its assumptions might change over time. Catalytic cooperation likely has more analytic power during earlier phases of cooperation rather than later ones, since it is premised on partially endogenous shifts in preferences that can change the nature of the cooperation problem. Various outcomes are possible in the medium to long term. Most optimistically, every actor of consequence could “tip” and pro-climate preferences diffuse globally. Under such conditions, the problem structure would change into a coordination game, and we would expect actors to create institutions to provide focal points and facilitate implementation (Barrett 2003).
Alternatively, or in a previous phase, we may see particularly recalcitrant actors fail to shift, even as the majority of others do. This could lead to a problem structure not dissimilar to that of nuclear proliferation, in which a few “rogues” impose externalities on others. In this conflictive scenario, the bulk of states could have an incentive to develop more coercive club institutions to exclude and penalize the holdout emitters, for example, through carbon tariffs. In this way, catalytic cooperation may lead to the conditions needed for the emergence of a climate club that could use negative incentives to enforce cooperation (Nordhaus 2015).
Of course, it may also be the case that the Paris Agreement will fail to catalyze sufficient action to force a critical mass of actors to raise their mitigation ambition (Sachs 2019). Under this scenario, the climate regime will look much as it did after Copenhagen, characterized by a flexible multilateral process and increasing pluralism. As the polycentric logic suggests, this will involve many different approaches across different actors, reflecting their individual preferences and strategies, but may not add up to a global solution (Jordan et al. 2018).
Finally, could such a catalytic approach work in other issue areas? A number of international cooperation dilemmas exhibit joint products, preference heterogeneity, and increasing returns. As the third section discussed, these features imply conditions under which catalytic cooperation becomes possible: first, when there is high variation in actors’ preferences, with at least some willing to act unilaterally (in other words, on average, the costs and benefits of free riding are low); second, when prior action alters the costs, benefits, and preference formation processes around subsequent actions, potentially creating tipping points; and third, when many different actors can affect the problem. Clearly many areas of world politics do not fit this description. For example, the cost/benefit of free riding in the realm of nuclear proliferation is high. If one country agrees to take in a number of refugees, it does not lower the cost for other countries to take in additional refugees in the future. And while many actors are affected by trade policies, only sovereign states can raise or lower tariffs.
Still, many other areas of world politics, and especially the realms of development, environment, and social welfare, satisfy these conditions. For example, of the seventeen Sustainable Development Goals, areas like health, gender equality, water and sanitation, food security, energy, urbanization, and responsible consumption and production all are relatively insulated from free riding concerns and affected by a wide range of actors with highly variable preferences. These tend to be, like climate change, classic “intermestic” issues that have domestic political dimensions but also spill across borders. We may therefore expect catalytic cooperation and catalytic institutions to play an important role in these areas of global governance.
For a discussion, see Alkin and Mildenberger (this issue).
I am grateful to David Victor for suggesting this phrase.
Note that the magnitude of these effects is conditioned by what Sandler calls the “aggregation technology,” which can be interpreted here as the nature of the production function P (Sandler 2004, 68). When, as in mitigation, collective action is “summative,” that is, when every contribution helps create the public good, an actor can always get a bit of public benefit from its own contribution, no matter how small. But when a certain threshold of contributions must be reached before any meaningful collective benefit is created, actors, especially small ones, will not obtain any broader benefit vi from unilateral action.
This article received valuable feedback and critique from Michaël Alkin, Liliana Andonova, Scott Barrett, Steven Bernstein, Joshua Busby, Jarred Finnegan, Fergus Green, Jessica Green, Matthew Hoffmann, Robert O. Keohane, Mareike Kleine, Matto Mildenberger, Michael Oppenheimer, Lauge Poulsen, Charles Roger, Anne-Marie Slaughter, Duncan Snidal, Harro van Asselt, Johannes Urpelainen, David Victor, and the participants of the London School of Economics International Political Economy seminar, the Oxford International Political Economy seminar, the Nuffield College International Relations group, the London School of Economics Grantham Research Institute workshop, and the other authors of this special section, as well as several anonymous reviewers.