Abstract
How can declining states reliably infer the intentions of rising states? One prominent line of argument maintains that because declining states face intractable uncertainty about rising states’ future intentions, preventive war is often unavoidable even between states with truly compatible goals. This article presents a dynamic model of reassurance in which actors are uncertain whether or not their interests conflict. The model shows that by adopting a hedging strategy of limited containment short of war, declining states can reduce risers’ incentives to send dishonest cooperative signals. This, in turn, makes cooperation more credible as a signal of risers’ benign intentions. Moreover, these signals are sufficiently informative to dissuade the decliner from escalating to preventive war even under large power shifts. Thus, although power shifts promote limited competition among states with compatible goals, preventive war rationally occurs only in a bargaining context when the riser’s goals are known to be incompatible.
Keywords
What kind of foreign policy strategies should states adopt in response to relative decline? This longstanding question is of particular importance today as the United States faces numerous rising states in the contemporary international system, most prominently China. Whereas proponents of “engagement” argue that the United States should integrate China into the global economy and international institutions, proponents of “containment” argue that the United States should adopt preventive measures while it is still powerful—such as economic competition or balancing alliances—in order to stem China’s rise and reduce its capacity to revise the international order in the future. 1
Which of these broad strategies a declining state adopts depends largely on its beliefs about rising states’ future intentions. While the decliner would prefer to accommodate benign risers that will maintain its preferred international order in the future, it has strong incentives to oppose hostile risers that intend to revise the status quo.
This article examines how, and under what conditions, declining states can accurately infer a rising state’s future intentions from its behavior in the present. It presents a model of a power shift in a particular signaling context called reassurance in which a declining state is uncertain whether its preferences converge or diverge with a rising state’s. The model shows that under this kind of incomplete information, declining states can elicit relatively credible signals of rising states’ future intentions by adopting a “hedging” strategy of limited containment. 2 Moreover, the information from cooperative signals eliminates the declining state’s incentive to escalate from limited hedging to full-scale preventive war, even under large power shifts and high initial uncertainty. 3
These findings help resolve a longstanding debate regarding the efficacy of reassurance in a dynamic international system. Reassurance optimists contend that states can always credibly signal benign intentions through cooperative behavior, thereby alleviating uncertainty and eliminating “tragic” conflicts between rising and declining states with compatible goals (Glaser 1994, 1997, 2010; Kydd 1997a, 2000b, 2005). 4 However, there is a large literature which suggests that cooperative signals are far less credible during power shifts (Levy 1987; Copeland 2000; Mearsheimer 2001; Rosato 2014/2015). These reassurance pessimists argue that “hostile” rising states—those whose goals are largely incompatible with the declining state’s—have strong incentives to misrepresent their intentions by behaving cooperatively in the present, while they are relatively weak. This makes rising states’ cooperative signals noncredible, leaving the declining state intractably uncertain about risers’ future intentions. Pessimists argue that this uncertainty, combined with the decliner’s increasing future vulnerability, engenders conflict and possibly full-scale preventive war even between rising and declining states with compatible goals.
The model presented below—the dynamic reassurance game—synthesizes the optimist and pessimist positions while identifying logical flaws in each. It shows that a declining state’s containment itself serves as a “screening mechanism” that helps the decliner distinguish “benign” rising states from hostile ones. Consequently, the higher the decliner’s degree of containment, the more credible a riser’s cooperative behavior is as a signal that it is truly benign—that is, that it shares the decliner’s goals for the international order. The information from these credible signals then allows the decliner to maintain a moderate hedging strategy toward benign risers (which always cooperate) and escalate to full-scale war only with those that have revealed their hostile intentions through noncooperative behavior. 5
Thus, contrary to pessimists, the dynamic reassurance game implies that even under large power shifts, rising states’ cooperative signal retains significant credibility. Consequently, only limited competition should ever occur between rising and declining states with truly compatible goals, never full-scale war. However, contrary to optimists, the model also shows that cooperative signals are not always sufficient to avoid competition altogether and that the declining state’s optimal degree of containment in response to cooperation increases with the size of the projected power shift (PPS). Indeed, the decliner’s limited containment is the very mechanism that sustains the credibility of a riser’s cooperative signals under large power shifts. Thus, although the model refutes the pessimist claim that power shifts can lead to full-scale preventive war between states with compatible goals, it supports the weaker claim that power shifts increase competition between compatible states short of war.
How does hedging work as a screening mechanism? In brief, limited containment increases the credibility of a riser’s cooperative signals by reducing incentives for hostile risers to misrepresent. To the extent that a hostile riser anticipates that it will incur some degree of containment even if it cooperates, it has less incentive to send cooperative signals. Instead, the decliner’s containment makes hostile risers more likely to reveal their true preferences by attempting immediate revision, in order to enjoy their preferred international order in the present. As such, cooperation in the face of limited hedging becomes a more credible signal that a riser is truly benign, since hostile types become less likely to exhibit such behavior. 6
The dynamic reassurance game departs from existing models of both reassurance and power shifts. Whereas previous models of reassurance have assumed a static distribution of power, the formal literature on power shifts has analyzed a different strategic setting called coercive bargaining. In contrast to reassurance, actors in a bargaining context are completely informed of the extent to which their ideal outcomes diverge, that is, the size of the asset at stake. Uncertainty in bargaining instead concerns another actor’s resolve: how badly it wants a known asset, rather than how much of an undefined asset it wants in the first place. As shown below, this standard bargaining specification does not adequately capture reassurance. Importantly, however, the dynamic reassurance game does not refute existing models but merely circumscribes their findings.
This article makes two additional contributions. First, it identifies a novel informational benefit of hedging strategies. Although hedging has featured prominently in the literatures on balancing and on US foreign policy toward China, these works have characterized it solely as a strategy for minimizing future vulnerability. Secondly, the article reintroduces hedging into the formal literature on power shifts, which has typically limited the declining state to a dichotomous choice between preventive war and full acquiescence to the riser’s power gains. In contrast, the dynamic reassurance game identifies conditions under which declining states might instead choose an intermediate degree of containment.
The findings of this article suggest that how American policy makers interpret the actions of a rising China depends greatly on US policy itself: contemporary US policy makers may face difficult trade-offs between maximizing cooperation with a rising China in the present, which will obscure its future intentions, or sacrificing some immediate gains from cooperation in order to elicit more credible signals. Although concrete policy prescriptions are beyond the scope of this article, it does help scholars and policy makers to recognize these trade-offs as well as the conditions under which China’s cooperative signals are credible.
The Reassurance Debate
This article addresses interstate reassurance in the context of shifting power. Reassurance occurs when states are incompletely informed about the compatibility of each other’s goals—that is, whether they prefer the same outcome or different outcomes regarding an issue or set of issues. This type of uncertainty is particularly important for states in relative decline. As the distribution of power shifts, a declining state will become increasingly vulnerable to exploitation or revision by hostile rising states, that is, those whose goals for the international order are relatively incompatible with the decliner’s. 7 This gives the decliner a strong incentive to oppose or contain hostile risers in the present, while it is still strong in order to forestall their power gains and mitigate its own future vulnerability (Levy 1987; Copeland 2000). Yet the decliner obviously wants to avoid costly competition with benign risers that have similar inherent goals and will largely maintain its preferred order as they become more powerful (Glaser 2010). Thus, uncertainty about a rising state’s goals presents the decliner with a strategic conundrum regarding whether and to what degree it should cooperate with the riser versus attempting to suppress its power gains.
The literature on interstate reassurance can be grouped roughly into two camps: “reassurance optimists” and “reassurance pessimists.” Optimists assert that credible cooperative signals are widely available, such that states with compatible goals can readily identify each other. A benign state that does not intend to revise the international order can refrain from offensive military investments (Jervis 1978; Glaser 1994, 1997; Kydd 2005), forgo low-cost opportunities for revision (Kydd 1997a), join institutions that raise its opportunity costs of revision (e.g., Ikenberry 2001; Weinberger 2003), or bear costs to support existing international rules and norms (e.g., Johnston 2003; Kydd 1997a). These signals are credible because they are “costly”—they carry high costs for hostile states that prefer an alternative international order but are low cost or even beneficial to benign types that inherently prefer the status quo (Fearon 1997). As a result, optimists conclude that states’ goals must truly be incompatible for conflict to occur (Kydd 1997a, 2005; Schweller 1994).
Problematically, however, reassurance optimists have generally assumed a static distribution of power (Osgood 1962; Jervis 1978; Kydd 1997b, 2000a, 2000b, 2005; Glaser 2010). Although optimists often claim that cooperative signals remain equally credible during power shifts, this claim is outside the scope of existing models of reassurance. 8 Moreover, other scholars—reassurance pessimists—have compellingly argued that large projected power shifts engender formidable barriers to the credibility of rising states’ cooperative signals (Levy 1987; Layne 1993; Copeland 2000; Mearsheimer 2001; Edelstein 2002; Montgomery 2006; Rosato 2014/2015). Hostile rising states have strong incentives to misrepresent their intentions while they are still relatively weak by mimicking the cooperative behaviors of benign risers and refraining from attempting to revise the international order in accordance with their true goals. Although foregoing immediate revision is inherently costly to hostile types, pessimists claim that for hostile rising states, these costs are outweighed by the prospects of avoiding opposition from the decliner and attempting revision under a more favorable distribution of power in the future.
Pessimists conclude that because cooperative signals are likely to be sent by both benign and hostile risers alike, such signals are noncredible, and declining states remain highly uncertain about any cooperative riser’s future intentions. In addition, declining states will become increasingly vulnerable to revision by hostile risers in the future. This combination of uncertainty and future vulnerability gives declining states strong incentives to take preventive action even against risers that have exhibited cooperative behavior. Moreover, pessimists claim that sufficiently large power shifts can produce full-scale preventive wars between mutually benign rising and declining states, due in part to the decliner’s intractable uncertainty about the compatibility of a rising state’s goals. 9
However, this pessimist claim is logically flawed. Pessimists argue that hostile rising states send dishonest cooperative signals in order to avoid preventive opposition from the decliner. Yet they simultaneously argue that these signals will be ineffective for achieving that end—vulnerable decliners will see cooperative signals as noncredible and therefore take preventive action anyway. This begs the question: what incentive do hostile rising states have to misrepresent their goals, if doing so will not allow them to avoid opposition? If (as pessimists claim) risers will incur opposition regardless of whether they cooperate, hostile types have less incentive to forego their goals in the short term by sending cooperative signals. 10 They should instead be inclined to reveal their true types by attempting immediate revision. Conversely, because hostile risers would reveal themselves, continued cooperation in the face of preventive opposition should constitute a credible signal that a riser’s goals are compatible with the decliner’s. Thus, the dual pessimist claims that decliners adopt unconditional preventive strategies and that risers’ cooperative signals are noncredible are at odds.
The model below resolves this contradiction in the pessimist argument. It shows that under large projected power shifts, uncertain decliners adopt an intermediate degree of containment toward cooperative rising states—a hedging strategy. 11 A decliner’s limited containment, in turn, reduces the incentive for hostile risers to misrepresent, thereby increasing the credibility of risers’ cooperative signals. No matter how large the power shift, the decliner’s hedging strategy elicits sufficiently credible signals that the decliner prefers not to escalate to full-scale preventive war with cooperative risers. Indeed, the model shows that the credibility of cooperative signals increases with the size of the power shift, as the decliner hedges more vigorously to offset its increasing vulnerability. Thus, optimists are correct that cooperative reassurance signals remain credible enough to avert full-scale war between truly compatible states even under large power shifts. Yet contrary to optimists, the very hedging mechanism that underpins these credible signals ensures that some degree of competition short of full-scale war is unavoidable under large power shifts and high uncertainty.
Previous Models of Power Shifts and Interstate Signaling
The existing formal literature has not addressed this particular debate regarding the efficacy of reassurance during power shifts and the likelihood of preventive war among states with compatible goals. Although there are a few existing works that explore the interaction of power shifts and incomplete information, these are models of coercive bargaining rather than reassurance (Powell 1996, 1999; Wolford, Reiter, and Carrubba 2011; Debs and Monteiro 2014). 12 As defined by Andrew Kydd, bargaining “focuses on conflict resolution—how conflicts of interest are resolved” short of war, given that they are known to exist, whereas reassurance focuses “on conflict avoidance—how states [signal] their motivations and thereby avoid conflicts of interest in the first place” (Kydd 1997a, 119). In reassurance, actors’ uncertainty pertains to the compatibility of each other’s goals, that is, whether they want the same outcome or divergent outcomes regarding the shape of the international order. This uncertainty is represented in models of reassurance as the size of the asset at stake, defined by the distance between the actors’ ideal outcomes on a particular issue or set of issues (Kydd 1997b, 2000a, 2000b, 2005).
In contrast, in coercive bargaining, the actors are completely informed of their degree of (in)compatibility—they are already aware of the extent to which their ideal outcomes diverge. Uncertainty in bargaining interactions instead pertains to the actors’ resolve to fight over this known asset. 13 In other words, whereas compatibility in reassurance captures how much the actors want at each other’s expense, resolve in bargaining captures how badly they want it. An actor’s resolve is represented in bargaining models by c, its costs of going to war relative to the value it places on the asset (thus, c varies inversely with resolve). In contrast, the source of uncertainty in reassurance—the size of the asset at stake—is known to both actors and normalized to 1 in bargaining models (Fearon 1995).
It is important to note that coercive bargaining and reassurance can occur simultaneously within a particular dyad across different issue areas. A rising and declining state could bargain over a particular issue where their goals are known to conflict, while the riser simultaneously attempts to reassure the decliner that it holds compatible goals for the broader international order. For example, the United States has had bargaining interactions with China over several specific economic disputes such as China’s tariff levels and currency valuation but remains significantly uncertain about whether China will use its growing economic and institutional power to uphold or challenge the existing rules and norms of the liberal economic order (Christensen 2015, 57). The findings of this article regarding reassurance are broadly applicable to any cases in which a declining state is uncertain about the compatibility of a riser’s goals on at least some important issues.
Standard bargaining assumptions are inadequate to capture such cases of reassurance. Indeed, power shifts have very different effects on a declining state’s incentives when uncertainty concerns a riser’s compatibility rather than its resolve. A defining characteristic of reassurance is that it is costless for the declining state to allow a highly compatible (benign) rising state to gain power. This is because a benign riser would use its increased capabilities to support and maintain an international order very similar to the one the decliner most prefers. When the asset at stake is very small (high compatibility), the riser’s demands will not increase beyond what the decliner would consider acceptable, no matter how powerful the riser becomes. In other words, compatibility defines the ceiling on a rising state’s future revision, such that the riser’s maximum demand decreases with its degree of compatibility. Thus, the commitment problem does not obtain when the rising state is known to be highly compatible but does when it is known to be highly incompatible.
Bargaining assumptions that focus on resolve omit these essential features of reassurance. In coercive bargaining, it is costly for the declining state to allow even weakly resolved rising states to gain power, such that larger power shifts increase the declining state’s incentive for preventive action against both highly resolved and weakly resolved risers. This is because both types of riser want all of the disputed asset—the ceilings on their demands given unlimited power are identical. Thus, under standard bargaining specifications, as the distribution of power shifts, a weak riser’s demands increase at the same rate and to the same maximum extent as a strong type’s (this is true of the models in Powell [1996, 1999]; Wolford, Reiter, and Carrubba [2011]; and Debs and Monteiro [2014]). As a result, commitment problems and preventive war can occur regardless of the decliner’s beliefs about which type it is facing. 14 This contrasts with reassurance in which a highly compatible riser’s future demands stop increasing with the PPS beyond a low ceiling that is still acceptable to the decliner, whereas a highly incompatible riser’s future demands continue to increase beyond what the decliner would accept in the present. Thus, the standard bargaining assumptions used in existing models of power shifts under incomplete information fail to capture the key strategic features of reassurance being analyzed in this article.
The model below formalizes reassurance under shifting power. It shows that when the decliner’s uncertainty concerns the rising state’s compatibility, full-scale preventive war does not occur. Rather, in response to this type of uncertainty, rational declining states adopt an intermediate hedging strategy and escalate to general war only when they are confident that a rising state’s goals are above some threshold of incompatibility. In other words, full-scale preventive war occurs only in a bargaining context, where actors are already confident that their goals are significantly incompatible. Importantly, then, this article corroborates a central finding of previous models of power shifts under incomplete information—that commitment problems can cause preventive war—but circumscribes this finding to a coercive bargaining context.
The Dynamic Reassurance Game
The dynamic reassurance game has two players, a declining state, DS, and a rising state, RS, and proceeds in two rounds. The first round corresponds to an early phase of the power shift in which the distribution of power favors DS. This is expressed as
Figure 1 shows the game in extensive form. It begins with nature choosing the compatibility of the actors’ goals, which is represented by the distance between their ideal points for the international order. The decliner’s ideal point is the status quo, normalized to 0. The riser’s ideal point is represented as g, the gain RS gets from revising the international order from 0 to g. Revision toward g gives RS linearly increasing positive utility and gives DS equal and opposite negative utility. Thus, the larger g is, the less compatible the actors are.

Extensive form of the power shift game (expected payoffs written with RS above DS).
Nature draws g from the uniform distribution [0, B0], such that every value of g between 0 and B0 is equally likely. RS observes the exact value of g, but DS is informed only of the range of possible values that g might take.
15
DS’s prior beliefs are therefore expressed in terms of B0. DS’s posterior beliefs are expressed in terms of
Once nature draws RS’s type, RS attempts some degree of revision,
In the second round, after the power shift is complete, the actors automatically reach a peaceful bargain regarding the shape of the international order according to the second-round distribution of power, P2.
20
The expected second-round payoffs are
Results
Equilibria with Complete Information
The complete-information game contains three subgame-perfect equilibria. The first equilibrium occurs when RS’s goals are relatively compatible, expressed as the following condition:
Under this condition, RS can revise to its ideal point and DS will acquiesce, that is, k = 0. Thus, gc represents the most incompatible riser that a completely informed DS would allow to revise to its ideal point without any containment. Let RS be defined as a benign type when
However, when
Note that hedging does not occur under complete information. This is because, as discussed below, hedging is a response to the informational problems that a declining state faces given the riser’s incentive to misrepresent—it allows the decliner to gain information and avoid undesirable preventive war with relatively compatible risers. Yet under complete information, the decliner already knows whether the riser is a benign type it prefers to accommodate or a hostile type it prefers to fight, and therefore has no use for an intermediate strategy.
Equilibria with Incomplete Information
The incomplete-information game has two perfect Bayesian equilibria, named for DS’s response to cooperative signals and depicted in Figure 2. The “appeasement equilibrium” occurs when the PPS is relatively small, while the “hedging equilibrium” occurs when the PPS is relatively large. Proofs are presented in the Online Appendix.

Equilibria of the power shift game under incomplete information (equilibrium strategies written with riser's preceding decliner's).
One note before proceeding: in the equilibria presented below, DS’s prior belief about the maximum value of g,
Appeasement equilibrium
The appeasement equilibrium occurs when
When
On the other hand, when
Conversely, when
Hedging equilibrium
Under a large PPS
As in the appeasement equilibrium, there is a threshold of g, g*, below which RS is compatible enough that it prefers to play r = 0 and incur limited containment, and above which RS attempts immediate revision to its ideal point (r = g) and incurs full containment. In the hedging equilibrium,
Since
Why does DS adopt a hedging strategy in response to cooperative signals under large power shifts instead of fully opposing RS? As elaborated below, DS’s hedging strategy increases the credibility of RS’s cooperative signals. Being completely informed, RS anticipates DS’s limited containment in response to cooperation, which makes hostile risers less inclined to misrepresent. Thus, there is some level of containment, k*, at which the riser’s cooperative signals become sufficiently credible that DS prefers not to escalate further. In other words, DS’s equilibrium degree of containment generates beliefs that are sufficiently optimistic to support that precise degree of containment. DS never forms beliefs in equilibrium that support escalation to full-scale preventive war in response to RS’s cooperation, no matter how large the power shift.
Comparative Statics
The article now returns to its central question: how does the size of the PPS affect the credibility of rising states’ signals and the likelihood and severity of preventive conflict? This section presents the comparative statics of the dynamic reassurance game regarding these two relationships, which differ depending on whether the PPS is small
Effects of small projected power shifts
The left-hand sides of Figures 3 and 4 show that under a small PPS

Effect of projected power shift on decliner’s degree of containment in response to cooperative signals.

Effect of projected power shift on credibility of riser’s cooperative signals.
This absence of containment means that a hostile riser’s incentive to misrepresent increases with the PPS. By cooperating in the present, risers can completely avoid containment, thereby gaining power that will allow them to revise the international order at lower cost and higher probability of success in the future. Furthermore, the larger the PPS (below
Effects of large projected power shifts
The right-hand side of Figure 3 shows that under a large PPS
This counterintuitive result emerges because the decliner’s containment itself increases the credibility of the riser’s cooperative signals. As the PPS increases, the decliner becomes more vulnerable to revision by the riser in the future. Furthermore, as shown above, in the absence of containment the riser’s incentive to misrepresent increases with the PPS. This combination of increasing vulnerability and uncertainty means that when the PPS is beyond a certain size, the decliner is no longer willing to fully acquiesce to the riser’s cooperative signals. Instead it adopts a hedging strategy, exerting a higher degree of containment under a larger PPS.
Crucially, the decliner never escalates to full-scale preventive war with cooperative risers, no matter how large the PPS, because its hedging strategy increases the credibility of the riser’s cooperative signals. Under large power shifts, rising states anticipate that they will incur some degree of containment from the decliner even if they behave cooperatively. This reduces the incentive for hostile risers to misrepresent: their cooperation will still result in some costs of competition and a diminished power trajectory. Many hostile risers that counterfactually would have misrepresented in the absence of containment instead attempt immediate revision to their ideal points in the face of the decliner’s hedging strategy rather than continuing to endure an undesirable status quo while still incurring moderate containment.
As hostile types become less likely to behave cooperatively in anticipation of the decliner’s hedging response, risers’ cooperative signals become increasingly credible. This is because benign risers always send cooperative signals, even in anticipation of moderate containment, for two reasons. 24 First, extensive revision is inherently undesirable to benign types, whose preferences for the international order are similar to the decliner’s. But in addition, benign risers are aware that revision would prompt the decliner to escalate from moderate hedging to preventive war, whereas cooperation allows them to avoid such escalation. 25 Thus, as hostile rising states become less likely to send cooperative signals, it becomes increasingly likely that risers that do so are truly benign.
Importantly, not all hostile risers reveal themselves in response to the decliner’s hedging. Some continue to misrepresent, pooling with benign risers by attempting no revision. Thus, cooperative signals remain only partially credible, even given some degree of containment. But as hostile types become less likely to behave cooperatively in anticipation of the decliner’s hedging, cooperation becomes a more credible signal that the riser is truly benign, all else equal.
The information from the riser’s cooperative signals feeds back on the decliner’s own incentives, generating beliefs that induce it to maintain a moderate hedging strategy rather than escalating to preventive war, even under a very large PPS. As the decliner’s optimal degree of containment increases, the riser’s cooperative signals become more credible. At a certain degree of containment—the threshold k* defined above—the decliner is sufficiently optimistic in response to cooperative signals that it prefers not to escalate any further. Thus, in addition to reducing its future vulnerability to hostile risers, hedging benefits the decliner by eliciting relatively credible cooperative signals that allow it to avoid full-scale preventive war with benign risers.
Discussion
Implications for the Reassurance Debate
The dynamic reassurance game has important implications for existing theories of power shifts and interstate reassurance. As laid out above, the literature on reassurance has been sharply split on the credibility of cooperative signals and the likelihood and severity of preventive conflict in the context of shifting power. Reassurance pessimists have argued that declining states are often compelled to initiate full-scale preventive war even with cooperative rising states, because the decliner is becoming increasingly vulnerable and because risers’ cooperative signals are noncredible due to their strong incentives to misrepresent. On the other hand, reassurance optimists have often claimed that cooperative signals should remain credible even under large power shifts, such that preventive war should never occur between states with compatible goals. 26 Yet because their models have assumed a static distribution of power, optimists have heretofore lacked a deductive basis for this claim.
The dynamic reassurance game helps to reconcile the optimist–pessimist debate by formalizing reassurance under dynamic conditions. The model identifies a novel signaling mechanism in reassurance contexts that stems from a declining state’s hedging strategy. This mechanism resolves a logical contradiction in the core pessimist claim that preventive war occurs due to the lack of credibility of rising states’ cooperative signals. Specifically, anticipation of the decliner’s unconditional preventive response should mitigate a hostile riser’s incentive to misrepresent, thereby lending at least some degree of credibility to risers’ cooperative signals. Indeed, the dynamic reassurance game shows that the credibility of these signals increases with the decliner’s degree of containment, thereby reducing the decliner’s incentive to take additional preventive measures. Moreover, there is always some degree of containment short of full-scale war at which the riser’s cooperative signals are sufficiently credible to avert further escalation. Thus, even under large projected power shifts, optimists are correct that cooperative signals remain relatively credible, and full-scale war with benign risers should not occur.
Nevertheless, the dynamic reassurance game also partially supports the pessimist position. First, pessimists are correct that all else equal—that is, given a constant degree of containment—a riser’s cooperative signals become less credible as the PPS increases. Indeed, the dynamic reassurance game yields precisely this result when the decliner does not contain cooperative risers at all (i.e., when the PPS is relatively small). What pessimists miss is that all else is not equal under a large PPS: the decliner’s optimal degree of containment increases with the size of the PPS, which makes risers’ cooperative signals more credible.
Secondly, although optimists are correct that cooperative signals are always credible enough to avert full-scale preventive war in a reassurance context, they are incorrect that these signals are sufficient to avoid inefficient competition altogether under a large PPS. Indeed, the ironic result of the dynamic reassurance game is that a moderate degree of containment is actually necessary to sustain the credibility of a riser’s cooperative signals, and the degree of containment necessary to avert escalation to war increases with the size of the PPS. Thus, although the model refutes the pessimist claim that large power shifts promote full-scale preventive war between states with compatible goals, a more subtle version of this claim—that power shifts increase the degree of competition between compatible states—is borne out.
Finally, it is worth reemphasizing that the decliner’s hedging strategy does not precipitate “spiral” dynamics in which benign risers behave noncooperatively in response. 27 First, because the decliner’s preferences are already manifested in the status quo order, rising states should be well-informed of the decliner’s goals and confident that as long as they continue to refrain from revision, the decliner will not escalate from moderate hedging to full containment. Furthermore, benign risers have a built-in disincentive to revise the international order any more than is absolutely necessary because they inherently prefer an order resembling the status quo. Benign risers therefore have no incentive to respond to a moderate hedging strategy by attempting large revisions and prompting the decliner to escalate to full-scale preventive war. This result generalizes the previous finding of Andrew Kydd (1997b, 2005)—that spiraling should only occur among states with truly incompatible goals—to conditions of dynamic power.
Additional Implications
These findings of the dynamic reassurance game complement existing models of power shifts under incomplete information. As discussed above, these are models of coercive bargaining rather than reassurance—they assume that actors are completely informed of the degree of incompatibility in their goals, that is, the size of the asset at stake. A central finding of these models is that commitment problems cause preventive war under sufficiently large power shifts. This result is corroborated by the dynamic reassurance game, which shows that when the decliner is confident that the riser is hostile, it does indeed initiate preventive war. However, when the decliner is highly uncertain about the riser’s compatibility, full-scale preventive war does not occur. Thus, the findings of this article limit the scope of existing models that assume away this type of uncertainty to the context of coercive bargaining, but support their results within that context.
This article also reintroduces hedging strategies into the formal literature on power shifts, which has (usefully) focused instead on the narrower dichotomy of whether or not war occurs. In existing models of power shifts under incomplete information, the declining state dichotomously either launches preventive war or makes concessions that, if accepted, would allow the riser to gain power unimpeded for some period of time (Powell 1996, 1999; Wolford, Reiter, and Carrubba 2011; Debs and Monteiro 2014). In contrast, in the dynamic reassurance game the decliner might take preventive actions short of war that partially forestall the riser’s power gains. In doing so, the model identifies conditions under which declining states might choose various degrees of containment rather than either launching preventive war or fully acquiescing to decline.
This emphasis on hedging accords with the nonformal literature on balancing in which hedging strategies have featured prominently (e.g., Walt 1987; Glaser 1992; Layne 1993; Schweller 1994). Likewise, the literature on China’s rise has focused on how states should balance engagement and containment of China while virtually excluding serious consideration of preventive war. 28 However, whereas these literatures have characterized hedging solely as a strategy for minimizing future vulnerability, this article has shown that hedging also serves as a screening mechanism that yields valuable information about rising states’ intentions. This novel, informational benefit of hedging strategies allows the decliner to maintain some degree of cooperation with truly benign risers, while also more effectively identifying and opposing hostile risers that threaten the status quo order. Thus, in addition to reducing future vulnerability by forestalling decline, hedging strategies allow declining states to adopt more completely informed, and therefore more optimal, foreign policies toward rising states for the duration of the power shift.
Robustness Checks and Scope Limitations
Like any model, the retrenchment game omits certain aspects of the real world in the interest of tractability. This section discusses the effects of relaxing several assumptions that may appear consequential, and either demonstrates the robustness of the model’s results or delineates their scope.
Rising state can escalate to winner-take-all conflict
One obvious potential concern is that the dynamic reassurance game artificially suppresses the spiraling dynamics characteristic of the security dilemma because RS does not have an opportunity to escalate its revision following DS’s move. This is not the case, however. Because RS is completely informed, it perfectly anticipates DS’s response to its own action and can therefore preemptively escalate its degree of revision ex ante before DS’s action, just as easily as it can ex post, following DS’s move. In other words, RS’s opportunity to escalate is already “baked in” to its initial move. As such, RS would never escalate beyond its initial degree of revision even if it had the opportunity to do so, making it unnecessary to complicate the model by adding an additional move.
Allowing RS the option of escalating would reduce the decliner’s degree of hedging, all else equal. This is because the latent threat of escalation by RS in response to DS’s containment would reduce the latter’s preventive motivation. Nevertheless, this is inconsequential for the comparative statics results reported above, as the directional effects of the key variables—the size of the PPS and the decliner’s degree of hedging—are identical under either specification. Furthermore, the unrealistic assumption that RS cannot escalate creates a “hard test” for the article’s main claim that the information from a riser’s cooperative signals is sufficient to forestall preventive war, since the model artificially increases the decliner’s incentive for preventive action. Under a more realistic specification, the decliner would be even less inclined to initiate preventive war with cooperative risers.
Full containment ends game with winner-take-all conflict
Another possible objection is that the dynamic reassurance game might “stack the deck” against preventive war because full containment yields a second-round interaction instead of winner-take-all conflict. In fact, the actors’ combined first- and second-round payoffs from full containment given above are identical to what they would be if the decliner initiated winner-take-all preventive conflict instead. 29 Giving the decliner this option therefore does not affect the results at all, and indeed is already captured by the model.
Power shift endogenous to riser’s revision
The dynamic reassurance game assumes that the power shift is exogenous to the size of the riser’s revision. This is a standard assumption (Kim and Morrow 1992; Powell 1996, 1999; Wolford, Reiter, and Carrubba 2011), which also approximates reality in many important cases of long-term power shifts that stem primarily from the rising state’s internal development. 30
Nevertheless, one key finding of the model is less applicable in cases of short-term power shifts engendered largely by a change in the distribution of a disputed asset. The dynamic reassurance game can be extended to this alternate strategic context by making the power shift endogenous to the shape of the international order. Instead of an exogenous power shift, there would be a potential power shift that is realized only to the extent that the riser achieves its desired revisions (Fearon 1996).
Two fundamental results of the dynamic reassurance game are robust to this alternative specification. First, full-scale preventive war still never occurs between compatible states. Indeed, when the power shift is fully endogenous to the degree of revision, the decliner acquiesces completely in response to cooperative signals (i.e., little or no revision), instead of hedging. This is because small revisions would result in a small actual power shift, which would not significantly increase DS’s future vulnerability. Furthermore, under this kind of endogenous power shift, the credibility of the riser’s cooperative signals still increases with the size of the potential power shift.
However, the mechanism by which these results occur when the PPS is endogenous to the asset differs from the hedging mechanism identified above under an exogenous power shift. As the potential size of an endogenous shift increases, hostile risers become less willing to accept little or no revision, which would require them to forego their future power gains. This allows benign risers to send highly credible cooperative signals in which they forego both immediate revision and future power gains, which would be highly costly for hostile types.
In sum, this alternative specification reveals an important scope condition of the model. Although its findings regarding the absence of preventive war and the credibility of cooperative signals in reassurance generalize to endogenous power shifts, the hedging mechanism applies only to the extent that the power shift is exogenous. Again, however, this restriction on the scope of the hedging mechanism does not diminish the value of this finding in important cases where the power shift is largely exogenous.
Conclusion
This article has presented a formal model of reassurance under shifting power in which a declining state is uncertain of the extent to which a rising state’s preferences for the shape of the international order converge or diverge with its own. In contrast, existing models of power shifts under incomplete information have captured coercive bargaining in which the actors are completely informed of the degree to which their ideal outcomes diverge but uncertain about each other’s resolve to fight over this disputed asset.
The model has shown that in a reassurance context, benign rising states can communicate their intentions with sufficient credibility to avert full-scale preventive war even under large power shifts and high initial uncertainty. In response to cooperative signals, there is always some limited degree of containment—a hedging strategy—that the declining state prefers to full-scale war. This is because hedging itself constitutes a screening mechanism that increases the credibility of a riser’s cooperative signals. Given that the decliner will hedge, the rising state cannot completely avoid containment even if it cooperates. Hostile risers therefore have less incentive to misrepresent their intentions by sending cooperative signals, which increases the decliner’s confidence that cooperative risers are truly benign. Although the decliner’s degree of containment increases with the PPS—as it becomes more vulnerable—the resulting increase in the decliner’s confidence about the intentions of cooperative risers is sufficient to avert further escalation to preventive war, no matter how large the power shift.
This novel signaling mechanism partially supports the longstanding claims of reassurance optimists, which have heretofore lacked a strong theoretical basis. Yet contrary to optimists, the very mechanism that averts war—the decliner’s hedging—ensures that some degree of costly competition will still occur under large power shifts, even among compatible states. This article also contradicts the counterclaim of reassurance pessimists that the combination of uncertainty and commitment problems often produces full-scale preventive war between rising and declining states with truly compatible goals.
Although the contribution of this article is solely theoretical, the dynamic reassurance game has clear empirical implications that should be evaluated in subsequent work. First, in response to large power shifts and high uncertainty about the compatibility of a rising state’s goals, declining states should adopt hedging strategies rather than fully acquiescing or launching preventive war. The latter outcomes should occur only in a coercive bargaining context, when the decliner is quite confident that the riser’s intentions are benign or hostile, respectively.
Secondly, hedging should reduce incentives for rising states to misrepresent. Once policy makers in hostile rising states have recognized the decliner’s hedging strategy, we should expect them to abandon—or strongly consider abandoning—attempts at misrepresentation in favor of attempting immediate revision. 31
Finally, leaders in declining states should update their beliefs substantially in response to rising states’ cooperative behavior, provided that the decliner has employed a hedging strategy. In the absence of hedging, however, policy makers should be highly skeptical of a riser’s cooperative signals, and recognize that the riser would have strong incentives to misrepresent any hostile intentions it may hold.
This final hypothesis carries potentially important, albeit highly speculative, implications for contemporary US–China relations. The equilibria of the model appear to be manifested across different issue areas in the US–China relationship. On regional security issues in East Asia, the United States has long adopted a hedging strategy toward China (Silove 2016). This strategy has been reinforced since 2009 by the Obama Administration’s “pivot” to Asia, which has increased America’s military presence and strengthened its alliances in the region, thereby reducing the trajectory of China’s rising military capabilities (Campbell 2016). Moreover, US military hedging has been recognized by Chinese policy makers, to whom it signifies that the United States “intends to remain the global hegemon and prevent China from growing strong enough to challenge it” (Nathan and Scobell 2012, 36). In keeping with the model, since 2009 China has become increasingly “assertive” on particular regional issues—especially its territorial claims in the South China Sea (SCS)—in a marked departure from its vigorous attempts to reassure its regional neighbors of its commitment to peaceful negotiated resolution of the SCS dispute prior to the pivot (Johnston 2003; Shambaugh 2005; Kang 2007). These noncooperative actions have allowed American observers to update that China’s preferences regarding the SCS are incompatible with those of the United States (Mastro 2014).
However, on other issues, particularly in the economic realm, the United States has refrained from any sort of hedging strategy and instead maintained unilateral cooperation. As Thomas Christensen points out, “US China policy in the past few decades has been nearly the opposite of our containment policy toward the Soviets [in which] the United States made an active effort to isolate and harm the target economies through tight restrictions on trade, investment and technology transfer. Since the Chinese reform era began in 1978, no global actor has done more to assist China’s rise than the United States” (Christensen 2015, xv). Likewise, the United States has done little to restrict China’s growing influence in international institutions, encouraging its membership in the World Trade Organization, expanding its weight in the International Monetary Fund, and largely acquiescing to China’s own institutional initiatives such as One-Belt-One-Road and the Asian Infrastructural Investment Bank (Dollar 2015).
As the model predicts given this lack of economic and institutional hedging by the United States, China since the 1990s has consistently maintained a high degree of cooperation regarding these issue areas (Goldstein 2005; Christensen 2015). Alastair Iain Johnston (2013) convincingly argues that China’s “assertive turn” is almost entirely limited to the SCS and does not extend to the broader international order (p. 46). Instead, China has become heavily integrated into the US-led liberal order, and indeed, as American leadership has wavered under President Donald Trump, China has increasingly stepped into the role. China’s institutional initiatives thus far have augmented, rather than revised, the existing liberal order, and it has continued to pursue domestic economic reforms that would increase its cooperation regarding intellectual property and cybersecurity (Dollar 2015; Parameswaran 2016; Chan 2016).
The findings above imply that these cooperative signals over the course of China’s rise should have been seen as having low credibility, given the US strategy of broad accommodation. Indeed, there is strong evidence that China’s leaders have recognized these incentives, manifested most prominently in Deng Xiaoping’s famous dictum, tao guang yang hui (“hide capabilities and keep a low profile”). As Avery Goldstein (2005) has argued, China’s mid-1990s turn toward a “Bismarckian” strategy of broad-based cooperation within the status quo order was driven in large part by the latent threat of American containment. Nevertheless, even during the steepest phase of China’s rise, a large subset of the literature on China’s foreign policy interpreted its cooperative actions as credible indicators that its preferences are largely compatible with the US-led international order (e.g., Johnston 2003; Shambaugh 2005; Kang 2007).
In contrast, this article has shown that American leaders cannot both fully acquiesce to China’s rise and confidently update their beliefs in response to China’s cooperative actions. Unrestricted reciprocation of China’s cooperation over the last quarter century, although enormously beneficial to the United States, has generated strong incentives for China to conceal any revisionist ambitions it may hold and thus reduced the credibility of China’s cooperative behavior as a signal of its long-term intentions. Whether and to what degree hedging is an advisable US strategy toward China is an empirical question beyond the scope of this article. The more modest policy contribution of this article is to make US leaders aware of the previously unrecognized informational benefits of a hedging strategy, which should be taken into account when formulating policy toward a rising China.
Supplemental Material
Supplemental Material, JCR-17-0338.R1 - Hedging for Better Bets: Power Shifts, Credible Signals, and Preventive Conflict
Supplemental Material, JCR-17-0338.R1 for Hedging for Better Bets: Power Shifts, Credible Signals, and Preventive Conflict by Brandon K. Yoder in Journal of Conflict Resolution
Footnotes
Acknowledgments
For helpful comments on this paper the author wishes to thank Dale Copeland, Alexandre Debs, Eric Gartzke, Kyle Haynes, Andrew Kydd, David Lake, Kyle Lascurettes, Yon Lupu, Nuno Monteiro, John Owen, Kate Sanger, Todd Sechser, Craig Volden, Alex Weisiger, and the JCR editors and anonymous reviewers. The contents remain solely the responsibility of the author.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The completion of the manuscript was aided by funding from the Bankard Fund for Political Economy, the Malone-Gallatin Foundation, and the University of Virginia Quantitative Collaborative.
Supplemental Material
Supplemental material is available for this article online.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
