Sage Journals: Discover world-class research

Abstract

How does a great power ally's demonstration of toughness toward a common adversary influence the sense of security among its weaker ally's citizens? The literature on the effects of standing firm against adversaries has significantly evolved in recent years, but empirical evidence is scarce, especially at the mass level. By taking advantage of the unique opportunity of the 2019 North Korea–US Hanoi summit, we examine the conditions under which a powerful alliance partner's firm and uncompromising posture toward a common adversary reassures the citizens of a junior ally. Based on alliance literature, we hypothesize that a patron's hawkish policy reassures its protégé when the protégé fears abandonment more than entrapment and that the more a citizen of the protégé fears abandonment, the more reassuring the patron's hawkish policy is to the individual. Our analysis of online surveys conducted before and after the summit using a quota sample of the Japanese electorate shows that the level of threat perception is significantly lower in the post-summit sample than in the pre-summit one, especially among the respondents who fear abandonment by the US. We thus conclude that President Trump's decision not to compromise with North Korea on its nuclear programs alleviated Japanese citizens’ concerns about US alliance commitments to Japan, improving their security perception. This study is significant in showing that a patron state's policy toward a common adversary can greatly influence the perception of its ally's citizens as well as its government.

Keywords

alliance North Korea public opinion reassurance the US–Japan relationship

How does a great power ally's demonstration of toughness toward a common adversary influence the sense of security among its weaker ally's citizens? A conventional view holds that a protégé feels reassured of its patron's commitments, thus its assessments of the security environment improve when the patron takes a hard line toward their common adversary (e.g. Schelling, 1966; Snyder, 1961). In an article published in April 2020, an anonymous Japanese official argued that Japan would benefit from US President Donald Trump's second term despite the various shortcomings of the president: “For countries on the receiving end of Chinese coercion, a tougher U.S. line on China is more important than any other aspect of U.S. policy” (Y.A., 2020). A Polish official, commenting in 2020 on the country's concerns about US security commitment in Europe, remarked that “as long as Russia [was] considered as a rival by the US, some kind of support and presence in Europe [would] probably remain to preserve deterrence” (quoted in Kim and Simon, 2021).

However, recent studies have challenged the notion of reassurance through toughness by showing that a junior ally may not necessarily desire its great power ally's hardline policy toward an adversary (Henry, 2020; Krebs and Spindel, 2018). Furthermore, while the logic of reassurance by toughness is generally applied to governmental-level interactions among allies, it does not necessarily explain how the junior ally's citizens evaluate the tough attitude of its great power ally toward a common adversary. Citizens in democracies, who exert a significant effect on the security policy preferences of their leaders (Chu and Recchia, 2022; Tomz et al., 2020), may fear the risk of entanglement or entrapment into a conflict triggered by the patron's hardline policy (Kim, 2011; Snyder, 1984, 1997). Although the intuitive link between “alliance games” and “adversary games” for hawkish reassurance is widely accepted as a theoretical argument (Snyder, 1984, 1997), the academic literature has yet to find systematic empirical evidence for the link and the conditions under which a protégé welcomes its patron's hawkish stance.

Understanding how a patron state's toughness toward a common adversary may influence the security perception of citizens in its allied state is significant for three reasons. First, answering this question will provide a clue for understanding how a state can form the reputation needed for successful deterrence and alliance management (Crescenzi, 2018; Mercer, 1996; Renshon et al., 2018). Standing firm against adversaries is considered to be an important measure to reassure allies and acquire a positive reputation (Snyder, 1984, 1997). The academic literature, both qualitative and quantitative, has demonstrated the benefit of having a positive assessment by allies (Crescenzi et al., 2012; Crescenzi, 2018; Gibler, 2008; LeVeck and Narang, 2017; Miller, 2003, 2011). On the other hand, Mercer argues that “fighting to create a reputation for resolution with allies is unwise” (1996: 228), and Krebs and Spindel (2018) and Henry (2020) show that junior allies may not necessarily appreciate toughness by their stronger ally. This study will shed light on whether or under what conditions a great power ally's tough stance against a common adversary leads to a positive reputation by reassuring its weaker ally.

Second, policymakers should learn not only how the government of their allied state may view their policy toward a common adversary, but also how such a policy may be viewed by the citizens of the allied state, in particular when their ally is democratic. Even when their policy is effective in reassuring the government of their ally, the policy's effectiveness may be undermined when the ally's government is constrained by its own citizens who perceive the policy differently. In addition, understanding how ordinary citizens of a weaker ally may respond to the policy of its great power ally may contribute to the study of public diplomacy, the importance of which is increasingly recognized these days (Goldsmith et al., 2021).

Finally, our investigation provides novel empirical evidence on how citizens assess their security environment. Threat perception has been given a prominent position in theories of alliances and deterrence (Stein, 2013; Walt, 1985). Citizens, however, assess their security environment by taking into account not only the threats that they face but also the protection offered by the allies of their state. For instance, Japanese leaders perceived their security environment to be benign despite the substantial threats from the neighboring Soviet Union and China during the Cold War, because of the US’ high degree of commitment to the alliance with Japan (Dian, 2014; Soeya, 1998). Yet, to what extent Japanese citizens’ perception of their security environment is affected by the US commitment to the alliance has been little examined empirically, owing to the lack of a touchstone event that clearly makes Japanese people skeptical of the US commitment.

To address the question stated above, we examine the effect of a patron's hardline posture on the citizens of a protégé by taking advantage of a unique opportunity presented by the 2019 North Korea–US Hanoi summit. Before the summit, many Japanese had worried that US President Donald Trump might make significant concessions to North Korea that would be detrimental to Japan's national security (Gramer, 2019).¹ Trump, however, stood firm and refused to compromise, resulting in a breakdown of the summit meeting. We conducted online surveys in Japan soon before and after the summit, and analyzed how the result of the summit influenced the respondents’ security perception.

Our analysis shows that the security perception of the Japanese respondents significantly improved after the summit. More precisely, we find that the improvement was significantly driven by the respondents who had feared abandonment by the US, and that the more knowledgeable the respondents were about international affairs, the more reassured they became by the result of the summit. This was probably because these respondents understood the implications of the US stance in Hanoi for Japan's security. Given that the deadlock between the US and North Korea elevated the risk of a conflict close to Japan (for that matter, of North Korean attacks against US bases in Japan), the improvement in the Japanese respondents’ security perception was by no means guaranteed. We thus argue that the US’ firm posture toward North Korea at the Hanoi summit reassured Japanese citizens in this case because many Japanese had feared being abandoned by the patron more strongly than they had feared entrapment.²

In the next section, we review the previous literature on the notion of reassurance through toughness in the study of alliance management and extended deterrence. We then present our theoretical framework and hypotheses to be tested. Subsequently, we analyze the results of the survey and present our statistical findings. We conclude this paper by summarizing the findings and discussing this study's implications and the direction for future research.

Who sees a patron's hardline policy as reassuring?

A patron's hardline policy is welcome to its protégé when it deters adversaries and demonstrates the patron's commitment to the alliance, but it can also provoke an adversary and increase the risk of the protégé being dragged into a conflict against the adversary. Other things being equal, a protégé is more likely to welcome the patron's hawkish policy toward common adversaries when the protégé is worried about abandonment by the patron.³ In other words, a patron's “strategy of firmness, resistance, or coercion” against common adversaries “will tend to reassure an ally who doubts” the patron's loyalty (Snyder, 1984: 470). Similarly, a protégé is less likely to welcome the patron's hardline policy when the protégé is worried about entrapment into a conflict with the common adversary. Snyder (1984: 475) further notes that “if a state feels highly dependent on its ally, directly or indirectly, if it perceives the ally as less dependent, if the alliance commitment is vague, and if the ally's recent behavior suggests doubtful loyalty, the state will fear abandonment more than entrapment.”⁴

Although Snyder's argument is intuitive, it is not easy to test with many observations because fears of abandonment and entrapment are hard to measure. To our knowledge, there is almost no quantitative research that uses the variables as either an independent or a dependent variable.⁵ Scholars have, however, made arguments about the effects of standing firm in alliance politics, and the literature on this topic has significantly evolved in recent years.

Traditionally, a dominant view has been that allies find their patron state's hardline postures toward common adversaries reassuring because such postures represent demonstrations of resolve and commitment (e.g. Schelling, 1966; Snyder, 1961, 1984, 1997).⁶ Powerful alliance partners can deter adversaries and reassure junior partners by standing firm against the adversaries. The acts of standing firm include not only military interventions but also non-military hardline measures, which are costly materially or in terms of escalation risks. Such costs help the patron demonstrate its security commitments to its allies. More recently, Blankenship and Lin-Greenberg (2022) show that a patron state needs to possess sufficient capabilities to make its resolve credible and reassure allies.

In contrast, some scholars argue that allies consider their patron state's hawkish postures a threat to their security because such postures can entangle them in a conflict or deplete the resources of the patron. Through their analysis of allies’ reactions to US hardline policy in the Vietnam War, Krebs and Spindel (2018) argue that junior allies actually are inclined to oppose hawkish measures because they make it difficult for a patron to maintain alliance commitments in the short and long run. Henry (2020) finds that most US allies discouraged the US from supporting Taiwan in the First Taiwan Strait Crisis (1954–1955) because they wished to avoid the risk of a major war. Sukin (2020) argues that highly credible nuclear security guarantees provided by the US can make South Korean citizens scared of their patron using nuclear weapons in an unnecessary preventative attack and escalating a conflict. Thus, junior allies sometimes feel that their security is undermined by their patron's hardline policies.

Even for allies with direct interests at stake, the expected benefits of a patron's hardline policy must be calculated against the expected cost of a conflict that such a policy might trigger. Although alliance literature has focused mostly on a patron's fear of entrapment (Cha, 2010, 2016; Morrow, 2000: 79), junior allies have strong reasons to worry about entrapment by a patron (Kim, 2016: 68). Particularly in the case of the US–Japan alliance and North Korea, the Japanese, in theory, have stronger reasons to worry about entrapment than Americans do because North Korea is unlikely to attack Japan (most important, US bases there) unless provoked by US actions and because Japan is highly unlikely to provoke North Korea militarily.⁷

Despite the significance of these opposing views regarding this theoretical and policy debate, empirical evidence is scarce, especially at the mass level. Therefore, by taking advantage of the unique opportunity of the Hanoi summit, we examine whether or how a patron's uncompromising posture toward a common adversary affects the security perception of the people of a junior ally.

Theoretical framework and hypotheses

In order to address the question discussed in the introduction, we envision a theoretical framework that encompasses three-way interactions among a great power, its weaker ally, and their common adversary. In this framework, the weaker ally's assessment of its security environment is shaped not only by the threats from its adversary but also by the perceived effectiveness of its great power ally's security commitments. Indeed, in an asymmetric alliance where a patron state provides security to its protégé (Morrow, 1991), the credibility of the patron state's alliance commitments is crucial for the protégé's security. The patron, for its part, aims to reassure its ally of its alliance commitments because the failure to do so could prompt the protégé to pursue security through other means to the detriment of the patron's interest (Castillo and Downes, 2023). In general, the patron adopts either or both of two reassurance policies toward its protégé. One is to signal its commitments bilaterally to the protégé through such means as public statements, leaders’ visits, and troop deployments (Blankenship, 2020; McManus and Nieman, 2019). The other is to demonstrate its hardline attitude toward the common adversary; if the patron takes a compromising or soft attitude toward the adversary, the protégé may doubt the resolve of the patron to defend its interests and fear the risk of abandonment when difficulties arise (Snyder, 1984). The latter is our focus in this study, and we design a quasi-natural experiment, where the Hanoi summit outcome serves as a treatment, to examine how the patron's policy may contribute to reassuring citizens of the protégé, leading to improvements in their security perception.

Based on the framework provided above, we develop hypotheses regarding the impact of the patron's hardline policy toward a common adversary on the security assessment of the protégé's citizens. Examining the citizens’ attitudes is of great importance given not only the observational evidence that foreign policy preferences of the public and elite are congruent (Jacobs and Page, 2005; Kertzer, 2022; Page and Shapiro, 1992), but also the experimental evidence that public opinion shapes elite decisions about the use of military force (Chu and Recchia, 2022; Tomz et al., 2020).

According to Snyder's alliance security dilemma, a patron's hardline postures reduce its allies’ fear that they may be abandoned by the patron, but such postures could also increase the allies’ fear of being entrapped into unnecessary conflicts that are initiated by the patron (Snyder, 1984, 1997). This implies that a patron state's hardline policies are reassuring when allies fear abandonment more strongly than they fear entrapment; however, the same policies are troublesome when allies fear entrapment more strongly than they fear abandonment.

This leads us to the following hypothesis.

Hypothesis 1: A patron's hardline policy toward a common adversary improves the assessment of the security environment by the citizens of its protégé when the protégé fears abandonment more than entrapment by the patron.

At the same time, we assume heterogeneity of individuals’ security perceptions, and some citizens may perceive the patron's tough attitude toward an adversary as reassuring while others may perceive it as more likely to escalate tensions vis-à-vis the adversary. Furthermore, we infer that the citizens who find the patron's tough policy toward an adversary reassuring are those who were more sensitive to the risk of abandonment by the patron. Therefore, on the individual level, the following hypothesis is derived.

Hypothesis 2: The more a citizen of the protégé fears abandonment, the more the patron's hawkish policy toward a common adversary improves the assessment of security environment by the individual.

As we discuss in the next section, Japanese discourses prior to the 2019 Hanoi summit between US President Trump and North Korea's Kim Jong-un were dominated by their fear of abandonment by the US. Thus, the application of our theoretical framework to the Hanoi summit case leads us to predict that (1) Trump's uncompromising posture reassured the Japanese citizens of firm US alliance commitments, improving their assessment of the security environment surrounding Japan, and (2) the reassurance effect by President Trump's uncompromising attitude toward North Korea was felt most strongly among the respondents who were sensitive to the fear of abandonment.⁸

Japan's fear of abandonment and entrapment at the time of the Hanoi summit

On 27 and 28 February 2019, US President Trump held his second summit meeting with Kim Jong-un in Hanoi, Vietnam. Since the first summit meeting in Singapore in June 2018, the US and North Korea had continued disputing how to implement the denuclearization of North Korea. Therefore, when the announcement was made that the second summit meeting would be held, an expectation (or concern) rose that the two leaders might somehow find a compromise. The expectation did not materialize, however. President Trump judged that North Korea's offer short of complete denuclearization was unsatisfactory, and decided to walk out of the negotiation with Kim.

In the run-up to the Hanoi summit, the Japanese public was predominantly focused on the risk of abandonment by the US (in the form of a conciliatory approach to North Korea at the expense of Japan's security) rather than the risk of entrapment for several reasons. First and foremost, Japanese citizens were alarmed by the possibility that Trump might concede too much to North Korea without due regard for Japan's interests, as he had done so in the past. During 2017, the Japanese and US governments had jointly pursued the so-called “maximum pressure” approach toward Pyongyang. In 2018, however, Trump made a sudden U-turn from his previous hardline rhetoric and unilaterally decided to become the first sitting US president to meet the top North Korean leader. Even worse, while the two leaders agreed on the “complete denuclearization of the Korean Peninsula” at the Singapore summit, their joint statement was disappointingly vague and lacked specifics.⁹ The Japanese subsequently witnessed US policy softening toward North Korea, evident in Washington's offering step-by-step rewards instead of withholding “carrots” until Pyongyang's complete denuclearization (Sonoda et al., 2019).

One important way in which the US could conciliate North Korea at the expense of Japan was to strike a deal on Pyongyang's intercontinental ballistic missile (ICBM) capabilities. President Trump's focus on North Korea's ICBM capabilities was well known. Republican Senator Lindsay Graham, for example, said that Trump would not allow Kim Jong-un to have an ICBM with a nuclear weapon capability to reach the US:

If there's going to be a war to stop [Kim Jong Un], it will be over there. If thousands die, they’re going to die over there. They’re not going to die here. And … [President Trump] has told me that to my face.¹⁰

Setting aside the safety of American soldiers and citizens in East Asia, one could make an argument for prioritizing the protection of the US mainland by limiting North Korea's ICBM capabilities. If President Trump had exclusively focused on ICBM capabilities, however, Japan would have still remained under the threat of North Korea's short- and medium-range nuclear missiles. Thus, the Japanese had significant concerns about Trump's possible deal in Hanoi.

The fact that President Trump's negative attitude toward US allies and his “America First” approach had been widely known by the Japanese public did not help ease the Japanese public's concerns. Trump's anti-alliance rhetoric had attracted the attention of allies as early as in the 2016 presidential election campaign, and he continued his critical attitude toward US allies after taking office. As the 2016 Republican presidential nominee, for example, Trump said that “Japan [would] have to defend itself against North Korea” if it did not pay more to the US; he even suggested that Japan might be “better off” with its own nuclear weapons, and that he as president would consider ending the US defense commitment to Japan.¹¹ Trump's disregard for US allies was also evident in the 2018 summit with Kim, as he abruptly announced the suspension of joint military exercises with South Korea without consulting the South Korean government. Trump even remarked after the summit that he wanted to get US “soldiers out” from South Korea and to bring them back home.¹²

It should be also noted that the Japanese had formed by 2019 a more hardline view toward Pyongyang after years of North Korea's military provocations and the unresolved issue of Japanese abductees. Japan remained very critical of conciliatory approaches to North Korea, which it regarded as Japan's most immediate national security threat since the 1990s (Cabinet Office of Japan, n.d.). North Korea had test-fired missiles landing near or flying over Japan numerous times. The previous North Korean supreme leader Kim Jong-il admitted North Korea's abduction of Japanese citizens in 2002, since when the unresolved abductee problem continuously reminds the Japanese public about the nature of Pyongyang's regime. According to a survey conducted in the US, China, Russia, and Japan in 2019, Japan had the highest rate of respondents favoring an exclusive reliance on sanctions (as opposed to diplomacy) in its dealings with North Korea (17%), and 64% of the Japanese respondents answered that denuclearization should be the first priority of the international community, followed by the US (52%), China (34%), and Russia (23%).¹³ Although threats from China became conspicuous among the Japanese public after the intensification of the Sino-Japanese maritime territorial disputes in the early 2010s, North Korea's threat has been consistently identified as the most immediate threat to Japan's national security by the government.¹⁴

For the reasons described above, the Japanese public was pessimistic about the expected outcome of the Hanoi summit, and Japanese media exclusively focused on the risk of abandonment by the US. According to a poll conducted shortly before the Hanoi summit, only 16.9% of the Japanese thought that the second US–DPRK summit would lead to progress in North Korea's denuclearization (Sankei-FNN, 2019).

Moreover, as far as major Japanese newspapers are concerned, we could not identify any discourse on Japan's fear of entrapment regarding the Hanoi summit. Through the time period from 5 February 2019, when Trump announced his plan to meet Kim in Hanoi, to the dates of the meeting (27–28 February), we searched the database of two major newspapers in Japan, Yomiuri Shimbun and Asahi Shimbun, and found some articles that included such words as “concern” and “Hanoi.” All of these articles expressed concerns about the US making concessions to North Korea, but not about Japan being dragged into a conflict between the two countries—most likely because the fear of abandonment was so dominant. Asahi Shinbun newspaper, one of Japan's leading newspapers, captured the mood in Japan well, as it reported several days before the Hanoi summit:

The nightmare scenario heading into the second summit … isn’t so much “fire and fury” and millions dead. Rather, some experts fear the meeting could result in an ill-considered deal that allows North Korea to get everything it wants while giving up little, even as the mercurial leaders trumpet a blockbuster nuclear success.¹⁵

After the summit, Japanese Prime Minister Abe Shinzo remarked that “Japan completely support[ed] President Trump's decision not to make concessions easily and to continue constructive talks to urge North Korea to take concrete steps” for denuclearization.¹⁶ Abe had a real sense of relief, because the nightmare scenario for Japan would have been the US tolerating short- and medium-range missiles of North Korea, while North Korea discarded ICBMs.¹⁷ An observer of international politics in East Asia described Japan's reaction to the negotiation outcome as follows:

Fearful that President Donald Trump, in his pursuit of a big win, would not keep Japanese interests in mind, many in Tokyo expected the worst. Instead, Japanese leaders joined in relief to sing with the “no deal is better than a bad deal” chorus.¹⁸

Research design

To test our hypotheses, we compare the perception of security environment between pre- and post-summit samples of Japanese citizens. We employed the pre-post survey design, rather than the panel survey design, in which respondents are asked the same questions before and after the summit.¹⁹ Although the pre-post survey design may cause the post-treatment bias (Montgomery et al., 2018) as discussed in the later section, it is essential for our purpose to make a time interval between the two surveys as short as possible to rule out possible factors influencing the security perception of respondents other than the summit outcome. A panel survey with a short time interval may make respondents give the same answer to the threat perception question, not because of the insignificant treatment effect of the summit outcome, but because of the tendency to avoid cognitive dissonance. For example, respondents who expressed a high level of fear of US abandonment in the pre-summit survey could answer the same way in the post-summit survey even if they felt safer immediately after the summit to avoid cognitive dissonance.

The key assumptions of the experiment are as follows: first, the treatment of the negotiation breakdown was randomly assigned to respondents; second, there is no reverse causality between the negotiation breakdown and security perception of the Japanese; and third, the negotiation outcome was so unpredictable that it was not baked into the security perception before the summit.

For the first assumption, we conducted separate (non-panel) pre- and post-summit online surveys using a quota sample of Japanese citizens with a short time interval (approximately 58 hours) between each survey during which all the summit-related events took place. Quota sampling from a registered panel guarantees that the two samples share key demographic characteristics. The short time interval between the surveys ensures that respondents’ attitudes do not significantly vary across samples except for those affected by the negotiation outcome. For the second assumption, it is theoretically and empirically impossible for a change in Japanese security perception to influence the negotiation outcome in such a short period. For the third assumption, as described in the previous section, Japanese people were uncertain before the summit whether or not the US would take Japan's side by standing firm against North Korea during the denuclearization negotiations. In the previous Trump and Kim summit in June 2018, they agreed to make efforts to establish a new relationship with only vague pledges of nuclear disarmament by Pyongyang.

The pre-summit survey started at 6 pm on 24 February in Japan Standard Time (JST) and ended at 12 pm on 26 February, a few hours before the two leaders arrived in Hanoi. The post-summit survey was opened at 10 pm on 28 February, a few hours after the leaders departed from Hanoi, and it was left open until 12 pm on 3 March. The pre- and post-summit samples comprised 771 and 772 Japanese citizens, respectively. All participating citizens were aged 18 or over and were drawn from a panel of people registered with Rakuten Insight. The surveys included questions about demographic and socioeconomic attributes (age, education, gender, and income), party support, and political attitudes including security perception. For security perception, respondents were asked to answer the following questions:

Taking into consideration the current international situation, do you think there exists a risk of adversaries waging war against Japan, or no such risk exists?

Risk exists

Some risk exists

Neutral

Little risk exists

No risk exists

This question aims to measure the extent to which respondents perceive potential foreign threats to Japan in general, not specific threats from countries such as North Korea and China. Because of Japan's heavy reliance on the US for its security, this general perception of security environment should largely depend on the extent to which respondents believe in the effectiveness of US extended deterrence through the US–Japan alliance (Dian, 2014; Soeya, 1998). Since the breakdown of the Hanoi summit symbolized the continuing US commitment to Japan's security based on the US–Japan alliance, we expect the perception of security environment among the Japanese to improve after the Hanoi summit even though North Korean military threats remained unchanged.

To estimate the causal effect of the negotiation breakdown at the Hanoi Summit owing to Trump's firm stance against Kim, we use a regression discontinuity design with a categorical running variable, following Goldsmith et al. (2021). The outcome variable is ordinal: the extent to which respondent i perceives the risk of an adversary waging war against Japan around the time of the summit. Formally, $Y_{i p} = 1$ if a category $p \in {N o r i s k e x i s t s, L i t t l e r i s k e x i s t s,$ $N e u t r a l, S o m e r i s k e x i s t s, R i s k e x i s t s}$ is chosen, and $Y_{i p} = 0$ otherwise, and $\sum_{p} Y_{i p} = 1$ . The treatment variable is whether respondent i participated in the survey after the summit ( $X_{i} = 1$ ) or participated before it ( $X_{i} = 0$ ).

The running variable is the number of hours since 10 pm on the last day of the summit (the starting time of the post-summit survey) for respondent i ( $Z_{i}$ ). If $Z_{i} \geq 0$ , then $X_{i} = 1$ deterministically; otherwise, $X_{i} = 0$ . Since we focus on sufficiently short periods before and after the summit, this running variable is inherently categorical and takes only $k \times 2$ distinct levels for k hours before and after the summit. Therefore, instead of searching for the optimal bandwidth or specifying a functional form, which assumes that the running variable is continuous, we simply estimate the difference of means in the outcome variable between the treated group ( $X_{i} = 1$ ) and the control group ( $X_{i} = 0$ ) for a given value of k. Formally, we define the average treatment effect ( $π_{p k}$ ), for each category p given k hours as follows: $π_{p k} \equiv E (X_{i} = 1, k) - E (X_{i} = 0, k)$ For our analysis, we chose $k = 48$ (48 h), which covers all responses in the pre- and post-summit samples. Because it is a very short period, we can interpret any difference between the treated and control groups as being primarily due to the exposure to the news reporting the negotiation breakdown between Trump and Kim at the Hanoi Summit. For the same reason, we do not conduct in-time placebo tests, in which the treatment effects are estimated based on varying adjusted running variables, for a robustness check.²⁰

We estimate the average treatment effect defined in the equation above by running an ordered logit regression with the treatment variable. The key identification assumption is the lack of any systematic difference between the respondents in the control group and those in the treated group. We tested this assumption using the basic demographic questions included in the survey. The covariates include male dummy (1 if male, 0 if otherwise), age, household income (from 1 = less than 2 million yen to 7 = more than or equal to 12 million yen), college graduate dummy (1 if college graduate, 0 if otherwise), and party support dummies (1 if supporting respective parties, 0 if otherwise). Table 1 shows that for all covariates, we find no statistically significant difference between the treated and control groups.²¹

Table 1.

Descriptive statistics.

Covariates	Pre-summit (control)		Post-summit (treatment)
Covariates	Mean	SD	Mean	SD	p
Male	0.495	0.500	0.500	0.500	0.859
Age	48.530	15.115	48.734	14.990	0.790
Household income	0.453	0.498	0.440	0.497	0.629
College graduate	0.205	0.404	0.216	0.412	0.583
Support: Liberal Democratic Party	0.292	0.455	0.290	0.454	0.942
Support: Constitutional Democratic Party	0.082	0.274	0.076	0.266	0.701
Support: National Democratic Party	0.010	0.101	0.004	0.062	0.130
Support: Komei	0.019	0.138	0.016	0.124	0.558
Support: Japanese Communist Party	0.038	0.190	0.035	0.184	0.782
Support: Japan Innovation Party	0.040	0.197	0.038	0.190	0.788
Support: other party	0.005	0.072	0.013	0.113	0.108
N	771		772

Note: p-values are calculated using one-way ANOVA.

Results

Figure 1 shows the distribution of the security perception responses by the sample. Although the two distributions look similar, the proportion of the respondents who answered Risk exists or Some risk exists is lower by 5.7 percentage points (0.057) in the post-summit sample than in the pre-summit sample, as shown in the first row of Table 2. This difference was statistically significant.

Figure 1.

Distribution of threat perception responses of pre- and post-summit samples. Question: Taking into consideration the current international situation, do you think there exists a risk of adversaries waging war against Japan, or no such risk exists?

Table 2.

The difference in threat perception between pre- and post-summit samples.

Threat perception	Pre-summit (control)	Post-summit (treatment)	Difference	p
Proportion (“Risk exists” + “Some risk exists”)	0.541	0.484	0.057	0.027
Mean (1: “No risk exists” to 5: “Risk exists”)	3.488	3.365	0.123	0.017

In addition, the second row of Table 2 compares the means of the threat perception variable, which is coded from 1 = No risk exists to 5 = Risk exists. The means for the pre- and post-summit samples were 3.488 and 3.365, respectively, and this difference was statistically significant, indicating that the level of threat perception among the Japanese significantly declined after the summit. This is consistent with our expectation that Trump's firm posture against Kim would reassure the Japanese public and improve its sense of security.

The main results of our analysis using an ordered logit regression are presented in the first column (Model 1) of Table 3. The outcome variable, threat perception, is coded from 1 = No risk exists to 5 = Risk exists, while the treatment variable is a post-summit dummy (1 if post-summit sample, 0 if pre-summit sample). The treatment effect is statistically significant and negative, indicating that security perception is better in the post-summit sample than in the pre-summit sample.

Table 3.

Estimated coefficients from the ordered logit models

Covariates	Model 1		Model 2		Model 3		Model 4
	Estimate	Standard error	Estimate	Standard error	Estimate	Standard error	Estimate	Standard error
Post-summit	−0.229*	0.092	−0.221*	0.093	−0.087	0.211	0.538^†	0.307
Male			0.198*	0.098
Age			−0.010**	0.003
College graduate			−0.062	0.099
Household income > 8 million JPY			0.165	0.116
Support: Liberal Democratic Party			0.214	0.110
Support: Constitutional Democratic Party			0.126	0.179
Support: National Democratic Party			1.082	0.551
Support: Komei			−0.230	0.334
Support: Japanese Communist Party			−0.084	0.278
Support: Japan Innovation Party			0.255	0.245
Support: other parties			−1.064*	0.494
Awareness of international politics					0.140*	0.067
Awareness of international politics × Post-summit					−0.069	0.094
Fear of abandonment							0.196**	0.068
Fear of abandonment × Post-summit							−0.248**	0.095
Cut 1	−3.475**	0.150	−3.825**	0.231	−3.197**	0.200	−2.876**	0.255
Cut 2	−1.587**	0.082	−1.929**	0.193	−1.307**	0.157	−0.984**	0.223
Cut 3	−0.167**	0.069	−0.487**	0.186	0.115	0.152	0.440*	0.221
Cut 4	1.749**	0.087	1.462**	0.191	2.036**	0.163	2.365**	0.231
Log-likelihood	−2158.842		−2141.752		−2156.109		−2154.335
N	1543		1543		1543		1543

Note: **p < 0.01, * p < 0.05, ^†p < 0.10.

To interpret this substantively, the black dots in Figure 2 depict the changes in the predicted probabilities of the five response categories of threat perception (No risk exists, Little risk exists, Neutral, Some risk exists, and Risk exists) in percentage points. The horizontal bars indicate the 95% confidence intervals. The figure shows that the treatment of the negotiation breakdown in the summit decreases the responses of Risk exists and Some risk exists by 2.66 and 3.04 percentage points, respectively, and increases the responses of Neutral, Little risk exists, and No risk exists by 2.23, 2.73, and 0.74 percentage points, respectively. These results support Hypothesis 1.

Figure 2.

Treatment effects of summit outcomes on threat perception responses.

For a robustness check, the second column (Model 2) of Table 3 presents the coefficients and standard errors from a multivariate regression model with covariates. The results show that the treatment variable, the post-summit dummy, is statistically significant and negative, even after controlling for covariates. This reconfirms the finding that the security perception of the Japanese significantly improved after the summit.

In addition, to check whether the changes are caused by the summit outcome, we estimated the interaction effect between the treatment variable (post-summit dummy) and awareness of international politics. If the negotiation outcome of the Hanoi Summit caused an improvement in security perception among the Japanese through the appreciation of the outcome's strategic ramifications, the treatment effect of the post-summit dummy should be stronger among the respondents with a higher level of awareness of international politics. This is because informed respondents were more likely to have been exposed to the dominant discourse prior to the summit that the US may sacrifice Japan's interest (partial abandonment), and because they were also more likely to understand the implications of the summit's outcome for Japan's security.

Awareness of international politics is measured by the number of correct answers to three questions about current international affairs. The three questions asked respondents to choose the name of the supreme leader of China (“Xi Jinping”), Japan's role in its military cooperation with the US (“the provision of off-battlefield logistical support to the US”), and the name of an intergovernmental military alliance involving 29 North American and European countries (“NATO”) among the four answer choices. The mean number of correct answers to this question does not significantly differ between the pre- and post-summit samples, suggesting that the summit outcome did not affect awareness of international politics. Therefore, it is reasonable to assume that conditioning the post-summit dummy on awareness of international politics hardly causes the post-treatment bias (Montgomery et al., 2018).²²

Although neither the main nor interaction effects are statistically significant at the 5% level, the fourth column (Model 3) of Table 3 shows that the total effect is a linear function of the moderating variable (Brambor et al., 2006).²³ Figure 3 graphically shows the total effect of the treatment variable and its respective 95% confidence intervals for each value of the variable of awareness on international politics. The average treatment effect is statistically significant among the respondents who correctly answered two or three out of the three questions about international politics, but it is insignificant among those who correctly answered one or no question.

Figure 3.

Treatment effects of summit outcomes on threat perception, conditioned by awareness of international politics.

To interpret this substantively, the black dots in Figure 4 depict the changes in the predicted probabilities of the five categories of perception in percentage points among the respondents who gave correct answers to all three questions and those who gave no correct answer. The right panel of the figure shows that the treatment of the negotiation breakdown in the summit decreases the responses of Risk exists and Some risk exists by 5.37 and 5.92 percentage points, respectively, and increases the responses of Neutral, Little risk exists, and No risk exists by 4.51, 5.34, and 1.44 percentage points, respectively among the most politically informed group of respondents who answered three questions correctly. On the other hand, the left panel of the figure shows no statistically significant treatment effects on the predicted probabilities. The finding that the treatment effect increases as the level of political awareness rises is consistent with the theoretical causal mechanism we posit.

Figure 4.

Treatment effects of summit outcomes on threat perception responses, conditioned by awareness of international politics.

To test Hypothesis 2, which posits that the negotiation outcome of the Hanoi Summit caused a greater improvement in security perception among the respondents with a stronger abandonment fear, we ran an ordered logit regression that included an interaction term between the treatment variable (post-summit dummy) and fear of abandonment. Fear of abandonment is measured by asking respondents whether they think the US would or would not help Japan if China invaded the disputed Senkaku/Diaoyu Islands. This variable is coded from 1 = The US would help to 5 = The US wouldn’t help on a five-point scale, where a higher value is associated with a stronger fear of abandonment. As in the case of awareness of international politics, the mean value of the answer to this question does not significantly differ between the pre- and post-summit samples. This implies that the US’ firm posture toward North Korea did not affect the fear of US abandonment vis-à-vis China,²⁴ probably because Japanese citizens understand that the US firm posture against North Korea does not necessarily guarantee the US taking the same stance against China.²⁵ Together with the ideological roots of the citizens’ confidence in the US commitment,²⁶ this can explain why the US firm posture against North Korea did not change the level of the fear of US abandonment regarding China's aggression.

The fourth column (Model 4) of Table 3 shows that the interaction effect is statistically significant and negative, suggesting that the negative treatment effect increases as the fear of US abandonment increases. In fact, as shown in Figure 5, the total treatment effect is statistically significant and negative among all respondents, except those who more or less think that the US would help Japan if China invaded the disputed islands.

Figure 5.

Treatment effects of summit outcomes on threat perception, conditioned by fear of abandonment.

To illustrate this more substantively, Figure 6 compares the changes in the predicted probabilities of the five response categories of their perception in percentage points among the respondents who have the strongest and weakest abandonment fears. The right panel of the figure shows that the treatment of the negotiation breakdown in the summit decreases the responses of Risk exists and Some risk exists by 8.99 and 8.24 percentage points, respectively, and increases the responses of Neutral, Little risk exists, and No risk exists by 7.47, 7.74, and 2.02 percentage points, respectively among the respondents who have the strongest fear of US abandonment. On the other hand, the left panel of the figure shows no statistically significant treatment effects on the predicted probabilities among the respondents who have the weakest abandonment fear. These results support Hypothesis 2.

Figure 6.

Treatment effects of summit outcomes on threat perception responses, conditioned by fear of abandonment.

Finally, we briefly examine two possible alternative explanations for the decline in the respondents’ threat perception caused by the summit outcomes. That is, the Japanese might feel safer simply because they perceived North Korea as weaker, or because they perceived the US as stronger after the summit than before. To test these explanations, Table 4 compares the mean scores of the perceived influences of North Korea and the US on international politics on a 0–100 scale, where higher values are associated with stronger perceived influences.

Table 4.

Difference in perceived influences of North Korea and the US on international politics between pre- and post-summit samples.

Perceived influence on international politics (scaled from 0 to 100)	Pre-summit (control)	Post-summit (treatment)	Difference	p
North Korea	41.803	41.610	0.193	0.885
The US	81.795	82.472	−0.677	0.471

As shown in the table, the perceived influences of North Korea and the US do not significantly differ between the pre- and post-summit samples. This finding rules out the possibility that the decline in the post-summit threat perception among the public was brought about by their perceptions of the declining adversary's power or of the patron's increasing power. This indirectly reinforces our argument that the US firm standing against North Korea in the summit worked as a signal of US alliance commitment to Japan, thereby reassuring Japanese people of US reliability.

Conclusion

In this study, we hypothesized that the US firm posture toward North Korea reassured Japanese citizens because many Japanese feared being abandoned by the patron more strongly than they feared entrapment concerning the Hanoi summit. Consistent with our expectation derived from the literature on alliance politics, Japanese respondents perceived their security environment to be significantly better in the post-Hanoi summit sample than in the pre-summit one, especially among the respondents who feared US abandonment. Furthermore, we find that the more worried Japanese citizens were about abandonment by the US, the more reassured they became by the breakdown of the Trump-Kim summit in Hanoi. These findings provide new empirical evidence, at the mass level, that a diplomatic stance of a patron has meaningful impact on the protégé's security perception. The findings may be also useful for policymakers of the patron state when they contemplate how their behavior may influence the public opinion of the citizens as well as the leaders of the protégé.

Furthermore, we point out two features that this study differs from and thus adds values to the previous studies. First, it goes beyond the bilateral framework focused on a great power ally's signal to its ally. As Snyder (1984, 1997) emphasized in his theory and investigation of diplomatic history, “alliance games” and “adversary games” are closely linked, indicating that a dyadic analytical framework can miss many important recipients of international signaling. Our trilateral framework complements the predominantly bilateral frameworks by analyzing the impact of a patron state's policy toward a common adversary on the patron's weaker ally. Second, quantitative research on reputation of allies typically focuses on extreme events such as a failure to militarily intervene on behalf of allies or an abrogation of alliance agreements (Crescenzi et al., 2012; Crescenzi, 2018; Gibler, 2008; LeVeck and Narang, 2017). In contrast, we demonstrated the importance of what Snyder calls the “halo” of alliances, political commitment expected as a result of military commitment: “Allies expect each other's diplomatic support on issues short of the casus foederis, even though such support is not explicitly commanded in the alliance contract” (Snyder, 1997: 356). While not denying the importance of the research on explicit alliance violations, we emphasize the need to study the effect of a more subtle signal on allied relations, especially because subtle policy differences among allies occur much more frequently than dramatic ruptures of alliances. For instance, even though we observed former President Trump's numerous problematic handlings of US alliances, his behavior and rhetoric mainly violated the expectations of the US allies rather than the texts of the alliance treaties with them.

Additionally, our research, by showing that an individual leader's diplomatic stance can substantially affect the public opinion of a third-party country, may add further evidence to the recent literature that emphasizes the role of leaders in international security (e.g. Byman and Pollack, 2001; Chiozza and Goemans, 2011; Horowitz et al., 2015; Saunders, 2011). McManus (2014) shows that statements of resolve against an adversary improves the chance of prevailing in disputes, and McManus (2018) finds that leader-specific signals of support enhance extended deterrence for the recipients of the support; what we find, in contrast, is that the effects of a leader's stance toward an adversary travel over to the allied relations and play a major role in reassurance.

Future research could directly pit fear of abandonment against fear of entrapment. We have argued that a protégé's fear of abandonment needs to be stronger than its fear of entrapment for the patron's hardline policy to be reassuring for the protégé. Since fear of entrapment into a conflict with North Korea is strong in South Korea, especially among pro-engagement progressives, the negotiation breakdown of the Hanoi summit may have had a contrasting effect on the security perception of the South Korean public. One may want to conduct a cross-national survey of Japanese and South Korean public in response to a future US–North Korean negotiation. A natural extension of our research would be also to examine whether or not a tough US stance against a common adversary would cause anxiety among the Japanese public when most Japanese citizens fear entrapment. Similarly, it would be important to investigate the Japanese public's reactions to a conciliatory stance by the US toward common adversaries such as China and North Korea. It would be also interesting, although challenging, to assess how long the effect of reassurance by a patron state may last. Depending upon how long the reassurance effect last, our view of the significance of a patron state's reassuring behavior may change.

Footnotes

Acknowledgements

We would like to thank Raymond Kuo,Naoko Matsumura,and three anonymous reviewers for their helpful comments. Any remaining errors are the authors' own responsibility.

Funding

The author(s) disclosed receipt of the following financial support for the research,authorship,and/or publication of this article: This work was supported by the Japan Society for the Promotion of Science,(grant number 18K18559)

ORCID iD

Takeshi Iida

References

Blankenship

(2020) Promises under pressure: Statements of reassurance in US alliances. International Studies Quarterly 64(4): 1017–1030.

Blankenship

Lin-Greenberg

(2022) Trivial tripwires? Military capabilities and alliance reassurance. Security Studies 31(1): 92–117.

Brambor

Clark

Golder

(2006) Understanding interaction models: Improving empirical analyses. Political Analysis 14(1): 63–82.

Byman

Pollack

(2001) Let us now praise great men: Bringing the statesman back in. International Security 25(4): 107–146.

Cabinet Office of Japan (n.d.) https://survey.gov-online.go.jp/index-all.html.

Castillo

Downes

(2023) Loyalty, hedging, or exit: How weaker alliance partners respond to the rise of new threats. Journal of Strategic Studies 46(2): 227–268.

Cha

(2010) Powerplay: Origins of the U.S. Alliance system in Asia. International Security 34(3): 158–196.

Cha

(2016) Powerplay: The Origins of the American Alliance System in Asia. Princeton, NJ: Princeton University Press.

Chiozza

Goemans

(2011) Leaders and International Conflict. Cambridge: Cambridge University Press.

10.

Chu

Recchia

(2022) Does public opinion affect the preferences of foreign policy leaders? Experimental evidence from the UK parliament. Journal of Politics 84(3): 1874–1877.

11.

Crescenzi

(2018) Of Friends and Foes: Reputation and Learning in International Politics. Oxford: Oxford University Press.

12.

Crescenzi

MJC

Kathman

Kleinberg

, et al. (2012) Reliability, reputation, and alliance formation. International Studies Quarterly 56(2): 259–274.

13.

Dian

(2014) The Evolution of the US–Japan Alliance: The Eagle and the Chrysanthemum. Oxford: Chandos Publishing.

14.

Funatsu

Iida

Kafura

, et al. (2023) The US–Japan alliance in the age of crisis. Chicago Council of Global Affairs. https://globalaffairs.org/sites/default/files/2023-03/US-Japan%20Report%20PDF.pdf

15.

Gibler

(2008) The costs of reneging: Reputation and alliance formation. The Journal of Conflict Resolution 52(3): 426–454.

16.

Goldsmith

Horiuchi

Matush

(2021) Does public diplomacy sway foreign public opinion? Identifying the effect of high-level visits. American Political Science Review 115(4): 1342–1357.

17.

Gramer

(2019) Hanoi summit has Tokyo feeling left out. Foreign Policy, 26 February. https://foreignpolicy.com/2019/02/26/hanoi-summit-has-tokyo-feeling-left-out-japan-north-korea-shinzo-abe-kim-jong-un-nuclear-deal-trump-asia-security-denuclearization/.

18.

Henry

(2020) What allies want: Reconsidering loyalty, reliability, and alliance interdependence. International Security 44(4): 45–83.

19.

Horowitz

Stam

Ellis

(2015) Why Leaders Fight. Cambridge: Cambridge University Press.

20.

Izumikawa

(2010) Explaining Japanese antimilitarism: Normative and realist constraints on Japan’s security policy. International Security 35(2): 123–160.

21.

Jacobs

Page

(2005) Who influences U.S. foreign policy? American Political Science Review 99(1): 107–123.

22.

Jou

Endo

(2016) Ideological understanding and voting in Japan: A longitudinal analysis. Asian Politics & Policy 8: 456–473.

23.

Justwan

Berejikian

( (2023) Conditional assistance: Entrapment concerns and individual-level support for us alliance partners. Journal of Global Security Studies 8(3).

24.

Kafura

Nakayama

Funatsu

, et al. (2022) Strong partners: Japanese and US perceptions of America and the world. Chicago Council on Global Affairs https://globalaffairs.org/sites/default/files/2022-03/220324%20Final%20US-Japan%20Report.pdf

25.

Kertzer

(2022) Re-assessing elite–public gaps in political behavior. American Journal of Political Science 66(3): 539–553.

26.

Kim

(2011) Why alliances entangle but seldom entrap states. Security Studies 20(3): 350–377.

27.

Kim

(2016) The Supply Side of Security: A Market Theory of Military Alliances. Stanford, CA: Stanford University Press.

28.

Kim

Simón

(2021) A reputation versus prioritization trade-off: Unpacking allied perceptions of us extended deterrence in distant regions. Security Studies 30(5): 725–760.

29.

Krebs

Spindel

(2018) Divided priorities: Why and when allies differ over military intervention. Security Studies 27(4): 575–606.

30.

Laver

Benoit

(2005) Estimating party policy positions: Japan in comparative context. Japanese Journal of Political Science 6(2): 187–209.

31.

LeVeck

Narang

(2017) How international reputation matters: Revisiting alliance violations in context. International Interactions 43(5): 797–821.

32.

McManus

(2014) Fighting words: The effectiveness of statements of resolve in international conflict. Journal of Peace Research 51(6): 726–740.

33.

McManus

(2018) Making it personal: The role of leader-specific signals in extended deterrence. The Journal of Politics 80(3): 982–995.

34.

McManus

Nieman

(2019) Identifying the level of major power support signaled for protégés: A latent measure approach. Journal of Peace Research 56(3): 364–378.

35.

Mercer

(1996) Reputation and International Politics. Ithaca, NY: Cornell University Press.

36.

Miller

(2003) Hypotheses on reputation: Alliance choices and shadow of the past. Security Studies 12(3): 40–78.

37.

Miller

(2011) The Shadow of the Past: Reputation and Military Alliances before the First World War. Ithaca, NY: Cornell University Press.

38.

Montgomery

Nyhan

Torres

(2018) How conditioning on posttreatment variables can ruin your experiment and what to do about it. American Journal of Political Science 62(3): 760–775.

39.

Morrow

(1991) Alliances and asymmetry: An alternative to the capability aggregation model of alliances. American Journal of Political Science 35(4): 904–933.

40.

Morrow

(2000) Alliances: Why write them down? Annual Review of Political Science 3: 63–83.

41.

Page

Shapiro

(1992) The Rational Public. Chicago, IL: University of Chicago Press.

42.

Pardo

Kim

Desmaele

, et al. (2019) One year after the Singapore summit: What do the publics of the US, China, Japan and Russia think about the situation in the Korean Peninsula? June. KF-VUB Korea Chair. https://www.korea-chair.eu/wp-content/uploads/2019/06/Korea-Chair-Survey-Analysis_June_2019.pdf

43.

Proksch

Slapin

Thies

(2011) Party system dynamics in postwar Japan: A quantitative content analysis of electoral pledges. Electoral Studies 30: 114–124.

44.

Renshon

Allan

Huth

(2018) Leader influence and reputation formation in world politics. American Journal of Political Science 62(2): 325–339.

45.

Sankei-FNN (2019) Joint public opinion research. questions and answers. 18 February. https://www.sankei.com/politics/news/190218/plt1902180015-n2.html.

46.

Saunders

(2011) Leaders at War: How Presidents Shape Military Interventions. Ithaca, NY: Cornell University Press.

47.

Schelling

(1966) Arms and Influence. New Haven, CT: Yale University Press.

48.

Snyder

(1961) Deterrence and Defense: Toward a Theory of National Security. Princeton, NJ: Princeton University Press.

49.

Snyder

(1984) The security dilemma in alliance politics. World Politics 36(4): 461–495.

50.

Snyder

(1997) Alliance Politics. Ithaka, NY: Cornell University Press.

51.

Soeya

(1998) Japan: Normative constraints versus structural imperatives. In: Algappa

(ed) Asian Security Practice: Material and Ideational Influences. Palo Alto, CA: Stanford University Press, pp. 198–233.

52.

Sonoda

Takedan

Suzuki

(2019) Beichou, naze futatabi shunou kaidan? Kitachousen ga nerau ‘kakuhoyuukoku’ka [U.S. and North Korea, why another summit? North Korea’s attempt to become ‘nuclear power’] Asahi Shimbun, 26 February. https://digital.asahi.com/articles/ASM2R5CV4M2RUHBI022.html.

53.

Stein

(2013) Threat perception in international relations. In: Huddy

Sears

Levy

(eds) The Oxford Handbook of Political Psychology. Oxford: Oxford University Press, pp. 364–394.

54.

Sukin

(2020) Credible nuclear security commitments can backfire: Explaining domestic support for nuclear weapons acquisition in South Korea. Journal of Conflict Resolution 64(6): 1011–1042.

55.

Tomz

Weeks

Yarhi-Milo

(2020) Public opinion and decisions about military force in democracies. International Organization 74(1): 119–143.

56.

Walt

(1985) Alliance formation and the balance of world power. International Security 9(4): 3–43.

57.

Y.A. (2020) The virtues of a confrontational China strategy. The American Interest, 10 April. https://www.the-american-interest.com/2020/04/10/the-virtues-of-a-confrontational-china-strategy/.