Can China Back Down? Crisis De-escalation in the Shadow of Popular Opposition

Many pundits and analysts, both inside and outside China, claim that public sentiment is an important driver in China's coercive diplomacy in the East and South China Seas. They argue that China's leadership faces potential domestic instability arising from ethnic unrest, socioeconomic inequality, environmental degradation, and a slowing economy. The regime therefore has an interest in pursuing tough external policies, as variants of diversionary conflict theory might suggest. Moreover, owing to socialization into the “Century of Humiliation” narrative (namely, that China was bullied by stronger imperialist powers from 1840 to 1949), the Chinese public reacts angrily to any hints of government concessions on territory and sovereignty issues. Public opinion, therefore, can create the worst of all worlds for Chinese leaders' ability to control the escalation of conflicts. It explains coercive external behavior that could give rise to militarized crises, but it also limits the regime's options to de-escalate once militarized crises have broken out.

We are not yet convinced that the potential for domestic political unrest can fully explain China's maritime behavior over the last decade. The top leadership exercises considerable control over official media messaging. It sits on a massive coercive apparatus, consisting of internal security services, a paramilitary, and the military. Moreover, it does not face elections. Thus, other factors may be at work as well—among them, a reaction to activities of other contenders in the East and South China Seas that Beijing perceives as challenging its definition of the status quo distribution of territory; an increasingly intense security dilemma with the United States; organizational interests being pushed by China's military; and sensitivity to both elite and mass opinion.¹ The alternative explanations for China's coercive diplomacy have generally not been rigorously tested, and the case for a diversionary conflict argument has yet to be made.²

We are not arguing, however, that domestic popular opinion does not constrain the intra-crisis behavior of China's leaders in particular conflicts with particular adversaries.³ Relations with Japan appear to be particularly susceptible to popular emotions, given how the destructive history of Japanese imperialism in China is emphasized in education and media in China.⁴ Were a militarized crisis to break out with Japan over competing claims of ownership of the Diaoyu (in Chinese)/Senkaku (in Japanese) Islands, many analysts suggest that public opinion would limit the regime's options, reducing China's ability to signal intra-crisis restraint. Some senior Chinese officials also worry that public opinion may reduce China's maneuverability in a crisis.⁵ From a crisis management perspective, then, it is in the interests of the regime to have flexibility of action.⁶ Assuming that Chinese officials' preference against having their hands tied by public opinion in a crisis is more or less genuine, the question for the regime is how to influence public opinion so as to reduce the costs of backing down or making concessions if needed after issuing threats, but without providing incentives to the challenger to push for even more concessions.⁷ More generally, all states in the East Asia region, as well as the United States, have an interest in expanding the political space for Chinese leaders to make concessions. Even advocates of imposing costs on China to curb its coercive diplomacy over maritime disputes should hope that China's leaders pay domestic costs sufficiently low to allow them to back down in a crisis.

Based on an original national-level survey experiment conducted in China in 2015, we ask what kinds of strategies—behavioral and rhetorical—might help minimize the public opinion costs of restraint in a crisis over disputed territory. In the experiment, we developed a real-world scenario in which Japan engages in building structures on disputed islands, and the Chinese leader threatens to respond with force. We then use randomized treatments to gauge changes in support for Chinese leaders under a range of different policies. We find that there are some strategies and reasons that do reduce the public opinion costs from backing down—for example, the threat of economic sanctions instead of the use of force; China's invocation of its peaceful identity; the existence of an offer of mediation by the United Nations; and acknowledgment of the economic costs of war. In our experiment, the one reason for backing down that seemed to have the potential to increase, rather than reduce, public opinion costs was making concessions in the face of U.S. military threats. Our respondents did not react positively to a leader who backed down in the face of these U.S. cost-imposition threats.

These findings about the effects of using different strategies to reduce the public costs of backing down in a crisis have important implications for crisis bargaining, particularly when the purpose is not to prevail outright over an adversary in a crisis but to manage its de-escalation,⁸ or when the actors worry that above a certain threshold of violence neither side will want to back down.⁹ The standard logic of the audience cost argument suggests that whereas signaling resolve may be easier given audience costs, signaling a genuine desire to de-escalate may be harder.¹⁰ Moreover, it becomes harder as audience costs increase with greater levels of commitment to the conflict.¹¹ Thus to de-escalate, leaders have to reduce the domestic public opinion costs of bluffing or making concessions. Evidence that different policy actions can reduce these costs by providing the political space needed for concessions suggests that audience costs and the threat of punishment for bluffing may not be as constraining as leaders think.

In sum, our study suggests that Chinese leaders may have more agency in the face of public opinion during a crisis than they themselves may believe. Put differently, our findings suggest that if the Chinese leadership were willing to de-escalate in the early stages of a crisis with Japan over the Diaoyu/Senkaku Islands, there are a number of strategies that would give them the flexibility that they apparently prefer. Our findings suggest that other countries therefore have a general interest in supporting certain de-escalation strategies available to the Chinese leadership. Other countries may have, for example, an interest in encouraging United Nations offers of mediation. They may have an interest in tactfully pointing out the economic costs to China should a crisis with Japan escalate. They may also have an interest in moderating their opposition to a Chinese decision to threaten economic sanctions against Japan, insofar as such a threat allows China's leaders more room to compromise in an immediate crisis. They do not necessarily have an interest, however, in threatening to impose greater military costs on China. Of course, all but the last of these interests are stronger the more tightly the Chinese leadership is constrained by public sentiments, as many pundits and analysts seem to believe.

The article proceeds in four stages. We first discuss the literature on audience costs and the question of how much agency leaders may have to manipulate these costs. We then describe how we designed and implemented the survey experiment in China. Next, we analyze the effects of the various treatment conditions on baseline public opinion costs imposed on a leader who backs down. We conclude with a discussion of the implications of our findings for the theoretical possibilities of leader agency in the face of audience costs. We also offer comments on the tensions between some of these findings and the notion of “cost imposition” by the United States on China for its handling of maritime disputes.

Domestic audience costs are the losses in domestic approval suffered by a national leader for backing down after making a threat in an international crisis. According to the standard audience cost argument, the relevant publics are concerned about inconsistencies between their leader's words and deeds, on the one hand, and the implications of these inconsistencies for the country's reputation or honor, on the other. For political leaders, lower levels of approval could redound in the form of losses in influence or difficulties in pushing a particular political agenda.¹² If audience costs increase during a crisis, leaders have less incentive to back down from their threat; hence, as the crisis proceeds, more information about their resolve is revealed. For political leaders, the upside of audience costs is the enhanced credibility of signals. The downside is being pressured by domestic audiences to take escalatory actions.¹³

Over the years, however, critics have raised a number of tough questions about the audience cost argument. Among them are three major critiques that together suggest a lack of realism and thus limited generalizability.¹⁴ First, the survey and experimental evidence for audience costs rests almost exclusively on evidence from American respondents.¹⁵ Thus the generalizability to other countries and political systems is unclear, particularly as regards authoritarian societies where regimes have tried explicitly to socialize their populations into uncritical support for nationalist and sovereignty-centric foreign policies. Second, these respondents are asked to react to underspecified hypothetical scenarios that may or may not matter when they judge the performance of real leaders.¹⁶ Usually the scenario involves an unnamed country invading an unnamed neighbor.

In our view, however, the most fundamental critique is that hypothetical leaders in these experiments have little or no agency and do not act strategically to control audience costs.¹⁷ Most of the audience cost literature assumes that leaders do not have the option of strategically using public rhetoric or nonmilitarized policies to reduce criticisms of their bluffing. Yet it would seem obvious that, in a crisis, rational leaders might want to preserve their flexibility, including the option of backing down with minimal domestic costs to their legitimacy as leaders. In the course of an emerging crisis, leaders may acquire new information about the military, political, economic, or personal costs of following through with their initial public threats. They may want to have the flexibility to negotiate more advantageous outcomes once they have signaled their willingness to use force against an adversary. Yet the audience cost literature generally does not test for the ways in which leaders might use specific policies or framing devices to reduce the domestic costs of concessions or moderation.

To be sure, there are exceptions in the audience cost literature in this regard. Matthew Levendusky and Michael Horowitz show that by merely claiming that s/he has new information, a leader can reduce disapproval rates for backing down.¹⁸ In their survey experiment, Graeme Davies and Robert Johns offer an explicit justification regarding the human costs to the military.¹⁹ Together these two pioneering studies show the potential limits of audience costs in the face of leaders' framing of the reasons for backing down. Still, they test the effects of only a relatively limited range of cost-based reasons. In reality, leaders are likely to have a number of specific reasons to justify bluffing, some of which may work better than others to reduce public disapproval. Much remains unknown about the conditions or strategies that can moderate the public political costs of backing down. In particular, analysts do not know if mediation by international institutions, intervention by an external great power, or calculations arising from economic interdependence can change how the public reacts to their leader's retreat from a crisis. In the literature on conflict resolution, the role of international institutions,²⁰ third-party intervention,²¹ and economic interdependence²² are three of the most prominent factors that are also frequently emphasized in the press and by policymakers. All three, however, remain untested in the context of audience costs and crisis de-escalation.

Nor do the two studies mentioned above present audiences with realistic disputes to see which of a range of plausible justifications might have more real-world effects. Typically, audience cost experiments use vague scenarios of U.S. intervention in an unnamed foreign dispute. It is unclear how seriously respondents take these scenarios as reflective of a real-world judgment they might make about their leaders' choices. Will respondents be more skeptical about their leader's reasons for backing down if the dispute is a highly salient one, “ripped from the headlines” so to speak? Or will respondents be more supportive of their leaders, knowing how dangerous the real-world crisis might become? Or are there traits of respondents that predict variation in the face of a real-world dispute? Those ranking high on measures of nationalism or hawkishness, for example, might be tougher on leaders in a scenario pertaining to territorial disputes than those who rank lower on these variables.

What factors and framing devices can leaders use to reduce the public political costs imposed on them if they bluff? To explore this and other questions, we fielded an original survey in the second quarter of 2015 that covers all provinces and capital municipalities in mainland China.²³ We developed a scenario based on the existing Sino-Japanese dispute over ownership of the Diaoyu/Senkaku Islands. As far as we are aware, this is the most realistic scenario heretofore used in academic studies of audience costs. In the actual dispute, Tokyo claims sovereignty and exercises administrative control over the islands. Since 2012, when the Japanese government purchased some of the islands from their private owners, China, which also claims sovereignty, has challenged this control by regularly sending government vessels inside the 12-nautical-mile territorial waters limit.

In our scenario, Japan begins building structures on the disputed islands. From the Chinese perspective, Japan's action is highly provocative because it represents an enhancement of Japan's “actual control” (shiji kongzhi) over the islands. We examine support for Chinese leaders under two settings. The first is a control setting, where the leader issues a threat to use force that is ultimately not carried out (a bluff). The second is a treatment setting, where the leader engages in a bluff accompanied by one in a range of specific and plausible economic, political, and strategic actions or conditions.²⁴ For reasons we explain below, we examine how levels of approval are affected when bluffing is accompanied by a UN mediation offer; when the United States issues deterrent threats; when the leader invokes Chinese people's peaceful identity; when the leader indicates the costs of conflict to China's economic development; and when the leader deploys economic sanctions as an alternative to the use of military force.

We believe that our crisis setting is not an unrealistic scenario. Some Japanese officials have raised the possibility of building government or military facilities, or both, on the Diaoyu/Senkaku Islands, in part to preempt any such effort by China.²⁵ Officials on both sides worry that the other will move first to radically enhance the symbols of administrative control over the islands. Most observers believe that any such action could trigger an armed conflict, which could even compel the United States to use force to defend Japanese forces (Washington has clearly signaled that the United States is obligated under its alliance treaty with Japan to assist in the defense of the islands if they are attacked). It is the plausibility of this scenario that we believe adds to the authenticity and construct validity of our measurements of respondents' reactions.²⁶

The advantage of using a real-world case in which Chinese public opinion has historically been hostile toward Japan is that this should be a hard case for showing the positive effect on approval ratings of different strategies and rationales for backing down.²⁷ If these strategies and reasons do increase approval rates in the face of strong anti-Japanese sentiment, leaders may have a fair amount of agency in reducing the domestic constraints on their foreign policy options.

In our experiment, the respondents were asked to read about a possible international crisis in the future. To keep our vignette as clean as possible, our experiment used a short vignette that focused solely on the objective facts of the crisis. Respondents were presented with the following crisis scenario: “China and Japan have a long-standing dispute over the sovereignty of a piece of territory. In a recent turn of events, Japan started to install structures on the disputed territory.” The dependent variable measures public approval over how the leader handled the crisis.

We randomly assigned our respondents to a control condition or one of six treatments under a between-subjects design.²⁸ By design, the control and all six treatments have the same crisis scenario—the only difference is in the way the Chinese leader responded to the crisis. The treatments cover a plausible range of options that we might observe in the real world, and we suggest different ways in which these conditions might affect the costs of backing down.

The control condition was a statement indicating that, after threatening military force, the leader backed down (i.e., bluffed). Respondents randomized into the control condition read the following scenario: “The Chinese leader said that if the installation continued, China would take military action. Japan continued to install structures on the disputed territory. In the end, the Chinese leader decided not to take military action against Japan.”²⁹ Meanwhile, respondents randomized into one of the six treatment groups read a similar scenario as in the control condition, except that their scenario included an additional piece of information that encapsulated a particular treatment.

The first treatment involved China's leader backing down after a UN offer to mediate the crisis. We included this scenario for two reasons. First, empirical evidence suggests that third-party mediation may be more successful than bilateral agreements in prolonging agreements over territory.³⁰ Some scholars have suggested, however, that this success reflects the application of multilateral agreements to easy cases, where the issue at stake is relatively minor. Instead, most states prefer bilateral agreements because the parties—particularly the one with the power or stakes advantage—may be better able to control the outcome. States, therefore, tend to select into bilateral agreements and avoid third-party mediation.³¹ In view of this literature, we are interested in whether Chinese respondents are willing to allow third-party mediation for a high-profile territorial dispute such as that over the Diaoyu/Senkaku Islands. Second, precisely because of perceptions of high stakes and the emotional commitments embodied in territorial disputes, third-party mediation may offer “political cover” to leaders who face domestic opposition to making concessions, thus reducing the constraints of domestic opinion.³² In recent maritime disputes, the Chinese government has explicitly indicated a preference for bilateral negotiations, and has opposed-third party involvement, including arbitration from international legal tribunals. The public may be aware of this and indeed may endorse it. Backing down in the face of third-party mediation, therefore, may not reduce the costs from bluffing. On the other hand, the Chinese government and the public treat the United Nations as the most legitimate intergovernmental institution that should be at the center of global order. Thus, offers of UN mediation may be seen by some members of the Chinese public as a viable alternative to bilateral conflict. To the extent that the possibility of UN mediation reduces the costs of backing down over a baseline level, public preferences may be flexible. In our study, respondents randomly assigned to this treatment read that the UN secretary-general had called for peace and had offered mediation. They then read that the Chinese leader decided not to take military action against Japan. We test the following hypothesis: Compared to the control condition, respondents' approval for leaders will increase if backing down is accompanied by a UN offer to mediate the dispute.

The second treatment had China's leader backing down in the face of U.S. deterrent threats. We chose this treatment because, given U.S. treaty obligations, it reflects a highly likely U.S. response to a Chinese threat to use force. A straightforward rationalist prediction might be that the domestic costs of backing down should drop in the face of deterrent threats from a stronger actor. That is, audiences should judge that a leader who backs down is doing so because s/he faces higher costs of conflict. This would be a prudent response to deterrent threats from more powerful states. One might therefore expect public approval to increase in the face of such threats. It is also plausible, however, that backing down in the face of U.S. threats could trigger an emotional public response to being bullied by a stronger player, and thus lead to a decrease in public support for a leader. In our study, respondents assigned to this treatment first read that the United States had threatened to intervene militarily if Japan were attacked, and then read that the Chinese leader decided not to use military force against Japan. We test the following mainstream rationalist hypothesis: Compared to the control condition, respondents' approval for leaders will increase if backing down is done in the context of U.S. deterrent threats.

The third treatment involved invoking a claim about China's alleged peaceful identity as a reason for not using force. Role theory in social psychology suggests that when people identify themselves as members of a social in-group (e.g., an ethnonational group), they will tend to act in ways consistent with the constitutive norms of that identity.³³ The motivation may be normative; it may have to do with the desire for social-liking through group conformity; or it may be a psychological discomfort from being accused of hypocrisy, among other mechanisms.³⁴ Role identities vary depending on appropriate social circumstances. In a conflict with the outside world, a role identity likely to resonate with the Chinese population is the claim that the Chinese people are inherently peaceful. Although this exceptionalist claim is empirically unsubstantiated, it is clear that ordinary Chinese people, on average, strongly believe that they as a people are inherently peaceful, certainly much more so than other peoples. For instance, in a survey of Beijing residents conducted in 2015, respondents were asked to assess the essential nature of the Chinese people as a people on a 7-point semantic differential scale anchored by peaceful at one end and warlike at the other. Fifty-four percent of those surveyed put the Chinese people at 1 (peaceful) on the scale, and 80 percent put the Chinese at 1 or 2. In contrast, 59 percent of respondents put the Japanese people as people at 7 (warlike), and 94 percent of respondents put the Japanese at 6 or 7 on the scale. This rhetorical tool is regularly invoked by Chinese leaders when claiming that China's rise does not pose a threat to anyone. Premier Li Keqiang has referred to the ancient Confucian phrase that “peace/harmony is the most valued” as the essence of traditional Chinese culture.³⁵ Party General Secretary Xi Jinping has noted that the Chinese people are by nature “peace-loving,” and that there is “no gene for invasion in the Chinese people's blood.”³⁶

Role theory suggests that if leaders invoke the peaceful identity of the Chinese people as a reason for not following through with threats of force, the costs of backing down may decline, as respondents support actions that are perceived to be consistent with their “peaceful identity.” Survey respondents randomized into this treatment read that the Chinese leader had decided not to take military action against Japan, and had declared that the Chinese people were a peaceful people who would try to resolve the conflict without force. We test the following hypothesis: Compared to the control condition, respondents' approval for leaders will increase if backing down is accompanied by leaders invoking the Chinese people's “peaceful identity.”

In our fourth treatment, the leader of China invoked the economic costs of military conflict as a reason for not following through with the military threat. Our argument here is that given the importance of economic growth to the material well-being of ordinary citizens and thus for the legitimacy of Communist Party rule, one would expect Chinese leaders and publics alike to worry about the negative impact of military conflict on China's economic development. Japan is an important source of technology and a critical market for Chinese exports. A conflict with Japan might plausibly have severe economic costs, especially if the United States came to Japan's defense. Invoking these costs may reduce the political costs for Chinese leaders of bluffing in a crisis. Respondents in this treatment read that the Chinese leader had decided not to take military action on the grounds that a military conflict would derail China's economic development. This treatment tests the following hypothesis: Compared to the control condition, respondents' approval for leaders will increase if backing down is accompanied by leaders invoking the costs to China's economic development of a conflict with Japan.

The fifth treatment had the Chinese leader using economic sanctions instead of military force in reaction to Japan's actions. We chose this treatment because economic sanctions have become a tool of first resort to punish another state for provocative political-military behavior short of major military attacks. Although their effectiveness in eliciting desired responses in the target is open to debate, sanctions can reduce the pressures to escalate militarily in response to provocations while still imposing at least some concrete material costs on the target. China has typically not employed economic sanctions at nearly the same frequency as the United States or Europe, often labeling sanctions as interference in internal affairs. That said, Beijing has used economic pressure against the Philippines over territorial disputes in the South China Sea and against South Korea over the deployment of missile defenses. It has joined UN sanctions against Iran and North Korea, and in 2010 it threatened to impose sanctions on U.S. companies involved in arms sales to Taiwan. Some analysts claim that China may have used export cut-offs in reaction to a dispute with Japan in 2010. The imposition of economic sanctions may reduce the costs of backing down, as such action lowers the risk of major war while still signaling a coercive response. In our study, respondents randomized to this treatment read that the Chinese leader decided not to take military action against Japan, but implemented economic sanctions and boycotts on Japan. With this treatment, we test the following hypothesis: Compared to the control condition, respondents' approval for leaders will increase if backing down is accompanied by leaders imposing economic sanctions on Japan.

The sixth treatment involved the Chinese leader issuing a vaguer initial threat in order to retain maneuverability. While decisionmakers sometimes believe that their bargaining power is enhanced by claiming (perhaps even demonstrating) that their options are limited by domestic opinion, they nonetheless often prefer to have freedom of action in crises. Issuing specific ultimatums can be risky because, if rejected, these tend to shine a brighter light on whether the leader escalates or capitulates. Vaguer threats can blur this distinction, making it harder for audiences to observe bluffing and thereby reducing the costs of backing down. A specific ultimatum can also force the other side to respond more aggressively in order to demonstrate its resolve in the face of the threat. For this reason, a major principle of crisis management is to avoid issuing ultimatums. History suggests that leaders frequently frame their threats with some ambiguity.³⁷ Robert Trager and Lynn Vavreck, in particular, have argued that the more explicit the threat is, the greater the cost of backing down.³⁸ In our experiment, respondents in this treatment did not read about an explicit threat to Japan to remove the structures on the islands. Rather they read that China's leader had warned Japan that it “must be held fully responsible for the consequences arising therefrom” if Japan did not remove these structures.³⁹ We therefore test the hypothesis that compared to the control condition, respondents' approval for leaders will increase if backing down is preceded by ambiguous threats to impose costs on Japan.

We used a two-stage elicitation procedure to construct the dependent variable. Respondents were first asked whether they “approve, disapprove, or neither approve nor disapprove of the way the Chinese leader handled the situation.” The first question gives us the raw percentage of respondents who approved (disapproved) of the leader's action. In the second question, respondents who answered “approve” (“disapprove”) in the first question were asked how strongly they approve (disapprove). Those who answered “neither approve nor disapprove” were asked if they leaned toward approving or disapproving, or if they leaned neither way. Answers to the two questions generate an approval rating on a 7-point scale from 0 (strong disapproval) to 6 (strong approval). Our results report both the approval rating and the raw percentage of respondents who approved of the leader's action.

The treatment effect is measured based on the difference in public approval in the control condition compared to a specific treatment condition. As all respondents were randomly assigned, the control and treatment groups would be statistically identical across all observed and unobserved characteristics. Because the control and treatment conditions were the same except for the additional sentence that embedded the treatment, any systematic difference in the costs of backing down can be attributed to the treatment variable.

Figure 1 shows that four of the six treatments significantly reduced the public opinion costs a Chinese leader would face for backing down after threatening to prevent Japan from building structures on the Diaoyu/Senkaku Islands. In one treatment (U.S. deterrent threat), the public costs rose rather than decreased, though the significance of that change depends on certain conditions, which we discuss below. In general, though, it appears that U.S. deterrent threats do not create political space for a Chinese leader, in the view of the public. Below we discuss the treatment effects in detail.

The average approval for the leader increased by 11 percentage points under the UN mediation treatment when compared against the control condition, a statistically significant difference (p < 0.001).⁴⁰ More precisely, the average approval rating (7-point scale from 0 to 6) was higher at 3.70 in the treatment group compared to 3.24 in the control group (p < 0.001). As a robustness check, we used ordered logit models to estimate the relationship between public approval and the treatment dummy. The treatment variable, UN_mediation, is coded 1 if a respondent was randomly assigned to treatment 1 (UN mediation) and 0 if assigned to the control group.⁴¹ Table A1 in appendix 2 shows that the ordered logit analysis yields a positive and statistically significant UN_mediation variable with or without control variables.⁴²

Taken together, the findings suggest that the Chinese public is amenable to UN offers of mediation in a Sino-Japanese conflict scenario; that is, UN mediation may reduce the cost of backing down for Chinese leaders, giving them more flexibility in their actions concerning the Diaoyu/Senkaku Islands.⁴³

Evidence that offers of third-party mediation can reduce public disapproval of concessions has important implications for territorial dispute resolution. Most states eschew third-party mediation in resolving territorial disputes in favor of bilateral dispute resolution, even though the former may be more effective in reducing conflict over the long term than the latter.⁴⁴ Sometimes this is because the stronger side believes that its capabilities can compel an advantageous bilateral distribution of gains. But if opposition to third-party mediation stems in part from perceived domestic political costs, particularly for more equally matched dyads, then evidence that offers of third-party mediation (in this case, through the UN) can improve levels of support may give leaders more political space to make concessions. In this regard, our findings provide micro-level evidence for the argument that third-party mediation can provide domestic political cover for leaders seeking to de-escalate disputes.⁴⁵

Our findings also have interesting implications for the claim that durable third-party mediation of territorial disputes occurs only in less important or more easily resolved cases. Put differently, as we noted earlier, some scholars have argued that selection effects explain why third-party mediation of territorial disputes is more effective than, say, bilateral negotiations. These scholars claim that states are more willing to let third parties play a mediating role in low-stakes disputes than in high-stakes disputes. The Sino-Japanese contention over the Diaoyu/Senkaku Islands, however, includes strong opposition to third-party mediation (Japan publicly refuses to acknowledge the existence of a dispute; China refuses to take the case to the International Court of Justice), as well as military and paramilitary preparations by both sides for fighting over the islands. This is clearly not an “easy” case with low stakes.

In our second treatment, approval for the leader fell by 3 percentage points, though the difference is statistically insignificant. More precisely, the average approval rating was lower, at 3.04, in the treatment group compared to 3.24 in the control group, with statistical significance at the marginal threshold (p = 0.13). Ordered logit analysis suggests a negative relationship between public approval and the US_threat treatment, with p-values in the range of 0.10 to 0.30 depending on the model specification (table A2 in appendix 2).

This result suggests that respondents may not always be rationally calculating the costs of a potential conflict over the Diaoyu/Senkaku Islands. Both Chinese officials and citizens widely acknowledge that the United States is militarily more powerful than China. Yet respondents remain generally unforgiving of a leader who backs down in the face of U.S. threats. They are not willing to increase support for concession as leaders rationally calculate the costs of U.S. intervention, contrary to what the standard rationalist inference might suggest.

Approval for the leader increased by 9 percentage points under the peaceful identity treatment (p < 0.01). The average approval rating is significantly higher in treatment 3 (3.60) compared to the control group (p < 0.01). Table A3 in appendix 2 shows ordered logit estimates across different combinations of control variables that confirm a positive and significant relationship between public approval and the peaceful identity treatment.

This result suggests that Chinese respondents are sensitive to the cueing of the constitutive norms of their role identities. We are not able to ascertain whether a concern about social status, social-liking, or normative considerations are at work. However, the invocation of Chinese identity as uniquelypeaceful—an invocation often used by Chinese leaders in their foreign policy statements—appears to lead respondents to reduce the costs on a leader who is bluffing but who nonetheless claims to act consistently with this identity.

Approval for the leader increased by 8 percentage points when respondents read a statement from the leader about the economic development costs to China of a conflict with Japan (p = 0.01). The average approval rating increased under this treatment (3.60) compared to the control condition (p < 0.01). Ordered logit analysis shows a positive relationship between public approval and the economic costs treatment (table A4 in appendix 2).

The outcome of this treatment suggests that Chinese respondents will sometimes make material cost-benefit calculations about the wisdom of backing down in a crisis. In this instance, the costs to China's economic development—arguably the most important material benefit the Communist Party leadership has provided ordinary Chinese people—can serve as a persuasive reason for backing down in a crisis with Japan over territory.

Public approval for the Chinese leader jumps 21 percentage points if the leader backs down but imposes economic sanctions on Japan (p < 0.001). The average approval rating saw a significant surge to 4.11 in this treatment compared to 3.24 in the control condition (p < 0.001). Table A5 in appendix 2 displays the ordered logit estimates across different model specifications. The results reaffirm the positive and significant effect on public approval driven by the sanctions treatment.

Respondents appear to believe that punishing Japan economically is a much-preferred alternative to backing down. Many respondents favored sanctions over all other reasons and responses accompanying a decision to back down. In short, if a Chinese leader chose economic sanctions instead, this would best minimize the costs of backing down compared to the bluff base condition.⁴⁶

Approval for the Chinese leader is essentially unchanged under the ambiguous-threat treatment: approval for the leader was 40 percent compared to 39 percent in the control condition (p = 0.86). The average approval rating was 3.19 in the treatment group compared to 3.24 in the control group (p = 0.72). Table A6 in appendix 2 displays the ordered logit estimates.

This result suggests that the historically employed rhetoric used to threaten the use of force is not read by the public as a credible signal that China's leaders are considering military coercion. Apparently, Chinese respondents are not especially familiar with this history, even though, as foreign specialists of China's use of force point out, this particular rhetorical formula clearly connotes “a potential response by force.”⁴⁷ There is thus an intriguing asymmetry in how foreign experts view the seriousness of this threat and how the Chinese public views the same threat and imposes the political costs thereof.

In an international crisis, such as the one we posit, it is possible that other political and foreign policy attitudes will mediate the degree of approval for a leader who backs down under various conditions. If so, it is in the interests of leaders to mobilize or suppress these attitudes in order to further reduce popular pressure when thinking about backing down in a crisis. In particular, as important research on audience costs by Joshua Kertzer and Ryan Brutger suggests, nationalism and hawkishness are likely to be the most relevant sets of attitudes.⁴⁸

Nationalism is a complex, multidimensional set of sentiments. It includes basic favorable attitudes toward the national in-group, but it also includes a more nativist element that is strongly attached to country and territory and that denigrates outsiders.⁴⁹ In general, nationalism is associated with worldviews that are sensitive to relative gains over out-groups.⁵⁰ As a result, nativist nationalists can be hard-liners in foreign policy, preferring toughness in their country's military or economic interaction with outsiders.⁵¹ As a general expectation, then, strong nationalist sentiments in the population should reduce the leader's ability to control the costs of backing down.

For a measure of nativism, we asked respondents their level of agreement with the following statement: “People should support their own country even if what it does is wrong.”⁵² We then coded responses using a dummy variable, with 1 for those who strongly agreed or somewhat agreed with the statement and 0 for those who somewhat disagreed or strongly disagreed.

Overall, as one might expect, nativists were not especially convinced by reasons for backing down absent any punishment of Japan (see figure 2). When it came to the UN offer of mediation, there was no significant improvement for the nativists in their levels of approval for the leadership over the baseline bluff condition. In contrast, for the non-nativists there was a strong and statistically significant rise in support for the leader.⁵³ Nativists were also indifferent to the invocation of Chinese people's peaceful role identity; their support for the leader increased slightly but not significantly. In contrast, non-nativists' support for the leadership surged with this treatment.⁵⁴ Both nativist and non-nativist support for the Chinese leader, however, jumped significantly when the leader used economic sanctions to punish Japan.⁵⁵

However, regarding the threat of U.S. intervention, nativists were willing to impose even higher costs on the leadership for backing down in the face of U.S. threats than were non-nativists. Nativists' approval for the leader drops significantly. For non-nativists, there is virtually no difference in approval levels between the control and treatment.⁵⁶ This result suggests that, to the extent that Chinese leaders are sensitive to more nativist elements in public opinion, some types of deterrent threats—”cost imposition” strategies—by the United States may in fact be counterproductive. In short, nativists were significantly more supportive when the leader adopted the next-strongest response to the use of force against Japan (i.e., imposing economic sanctions). Yet they were also willing to punish the leadership more harshly for backing down in the face of U.S. military pressure. Non-nativists, on the other hand, were more responsive to UN mediation and the invocation of Chinese role identity.

As for hawkishness, we constructed a dummy variable using answers to the following question posing a guns versus butter trade-off: “As a developing country, the government of China has to allocate its budget to different areas given limited resources. Some people think that for building a better social welfare system, the government should reduce its spending on national defense. Do you agree with this view?” Those who strongly agreed or somewhat agreed with the statement we labeled “doves”; those who somewhat disagreed or strongly disagreed we labeled “hawks” (see figure 3).

Both doves' and hawks' approval ratings for leaders who back down in the face of UN mediation offers are higher than in the control condition.⁵⁷ Interestingly, doves and hawks were also more or less equally convinced by appeals to Chinese role identity as a peaceful people to increase their support for a leader who backs down.⁵⁸

Doves' support increased significantly when the economic costs of military force were invoked by the Chinese leader, whereas hawks were not convinced by this argument.⁵⁹ Meanwhile, approval rates from both doves and hawks increased significantly when the leader chose economic sanctions.⁶⁰ In the case of U.S. deterrent threats, doves' approval rates did not change compared to the control group. Hawks discounted the costs of greater conflict with the United States, however, and they punished the Chinese leaders for concessions in the face of U.S. threats.⁶¹

In short, the distribution of nativist and hawkish opinion can influence the effectiveness of the regime's strategy in reducing the public costs of concessions in a crisis over territorial disputes with Japan. Not surprisingly, non-nativists and doves are more susceptible to most of the reasons and justifications we tested, and their mobilization will provide leaders with even greater political space in a crisis. Nonetheless, it remains unclear how large these constituencies are in the country as a whole. In our sample, about 41 percent were non-nativists. In a 2008 national survey administered by Peking University, about 51 percent of that sample were non-nativist.⁶² As for doves, we could not compare our sample to the 2008 survey, as the latter did not ask a similar question about guns versus butter trade-offs. Instead, the best we could do was to compare our survey with the 2015 Beijing Area Study survey of Beijing residents.⁶³ In our study, doves constituted about 33 percent of the sample compared to about 17 percent in the Beijing survey. Despite these problems of comparison, we are confident that these pools of moderate opinion, though probably not constituting a majority, are not tiny either. The question is whether the regime is aware of these sizable pools of more moderate opinion, and whether it is interested in mobilizing them to preserve flexibility in a crisis.

Our study relates generally to agency and options for governments constrained by public pressures in a crisis. Although the public costs of backing down enhance bargaining leverage, they can also entrap leaders looking for ways to prevent further escalation in a crisis. Our findings expand on previous work by specifying a wide range of realistic strategies available to governments—democracies and non-democracies alike—to de-escalate in the face of public pressure to remain tough. Some of these strategies appeal to nonrational thought processes in the public—role identities, for example. Others appeal to apparently rational calculations—economic costs, for instance. In cases where audience costs may exist but appear to be relatively ineffective in constraining leaders, scholars might look to see whether leaders have employed some of these strategies. We also show that the costs imposed on leaders for backing down or bluffing in a crisis are influenced by some of the population's basic beliefs about nationalism and basic preferences regarding military spending. This finding suggests that the success of backing down or exercising restraint in a crisis can also rest on the distribution of nativist nationalists and hawks, as well as on the ability of the regime to mobilize non-nativists and doves.

Our study also relates specifically to the Chinese government's options and agency. Our research question assumes that leaders, for whatever reason, decide that the military option is too dangerous, or not desirable, or—at the very least—that they want the flexibility to consider the option of de-escalation. Once they make that decision, the issue then becomes how to reduce the domestic costs to the lowest possible level. We have shown that there may be a range of realistic strategies and reasons for backing down that may indeed reduce the costs imposed on Chinese leaders by their public.⁶⁴ The Chinese public may be susceptible to arguments about the economic costs to using force. The Chinese public, on average, may also be more supportive of restraint if this is seen as normatively consistent with their perceived role identities as a peaceful people. They apparently also see benefits from third-party mediation by the UN. They may, in particular, also be willing to accept economic sanctions as a substitute for the use of force in a territorial conflict. Different groups in society may be more or less susceptible to different appeals. In principle, this means that the Chinese leadership could combine different, but logically congruent, strategies to maximize its flexibility. Thus, for example, invoking China's peaceful identity, the economic costs of military conflict, and a threat of economic sanctions might lead to higher levels of support than invoking any single option.

Our findings suggest that Chinese leaders have certain diplomatic and practical options for giving themselves more domestic space to pursue concessions in high-stake crises. Our findings also suggest, however, that controlling emotional nativist tendencies, and mobilizing more moderate “butter over guns” or dovish constituencies, may be critical if the regime wants to minimize the public costs of making concessions in a crisis. Mobilizing opinions in a well-calibrated way is obviously not easy to do in an emerging crisis over territorial issues. In the tensions created by nationalist rhetoric and military moves, mobilizing moderate public opinion can be particularly difficult. The regime would likely have to communicate quietly with the United States that it would use rhetoric and actions that give it more room for self-restraint and concessions—such as some combination of symbolic economic sanctions and requests for UN mediation—in return for the United States minimizing its public threats of intervention.

This latter reassurance from the United States appears to be important for Chinese leaders' agency. As our findings suggest, the Chinese public—particularly those segments of the public with nativist and hawkish beliefs—might increase rather than lower the costs for leaders should the latter back down in the face of U.S. military threats, especially highly public ones. This likelihood creates a tension in the arguments some pundits make in Washington for a tougher U.S. deterrence strategy against China in maritime disputes in East Asia. Given our findings, one cannot both argue—as some in Washington do—that popular nationalism constrains the Chinese leadership and that certain cost-imposition strategies (e.g., military deterrent threats) will constrain China's coercive diplomacy. If general deterrence has failed and China and Japan face a crisis over the Diaoyu/Senkaku Islands, imposing costs for immediate deterrence purposes runs into this contradiction. If public opinion does indeed constrain Chinese leaders, then high-profile immediate deterrent threats are likely to reduce the leadership's flexibility and, in turn, reduce the effectiveness of these deterrent threats.

The authors thank Zenobia Chan for her excellent research assistance; Deborah Larson, the anonymous reviewers, and the panel participants at the 2016 annual meeting of the American Political Science Association for their useful feedback; and Hong Kong University and the Research Grants Council of Hong Kong for their financial support.