Instinctive Commercial Peace Theorists? Interpreting American Views of the US–China Trade War

Abstract Existing theories of individual trade preferences do not satisfactorily explain how security concerns should affect American support for the US–China trade war that began in 2018. Although existing theories of public attitudes toward international trade—economic self-interest, sociotropism, partisanship, reciprocity, and xenophobia—all help to explain initial support for the trade war, these hypotheses do not adequately explain citizen attitudes in the context of an increasingly adversarial and securitized bilateral US–China relationship. In particular, they do not address how rising security tensions affect trade preferences. Using nationally representative original survey data (n = 1,016) and a nonrepresentative survey with an embedded experiment (n = 1,015), this article argues that securitization of the bilateral economic relationship has spurred threat perceptions and given rise to a Cold War narrative that has in turn caused a substantial share of Americans to become less concerned with the economic outcomes of trade and more concerned with trade's effect on security. These Americans demonstrate an instinctive “commercial peace” response, seeing trade liberalization as a potential deterrent to conflict. The results challenge conventional wisdom on political support for the trade war and add depth to existing theories of individual trade preferences regarding the interaction between economic, security, and psychological motivations.


Introduction
How do security concerns affect individual trade preferences? Specifically, do fears of conflict with a trade partner make individuals more or less likely to support free trade? Literature on the determinants of individual trade preferences has become increasingly sophisticated in recent years, yet most of this literature ignores foreign policy and security concerns. The few relevant existing studies point to two potential effects of heightened security concerns on trade preferences. 1 First, security concerns could generate fears that trade will lead to security externalities for the potential adversary in the form of increased military funding, which would ostensibly increase preferences for protectionist barriers. 2 Alternatively, security concerns could lead to support for trade liberalization as a means of reducing the likelihood of conflict, consistent with "commercial peace" theory.
The US-China "trade war" that began in 2018 provides an ideal testing ground for these theories given its economic scale and high degree of public salience. In 2018, the Trump administration began hiking tariff rates on select Chinese imports, eventually subjecting nearly half of all Chinese imports to tariffs. In retaliation, China subjected more than half of its American imports to tariffs. Four years later, the future of the trade war remains unclear. The trade war has been highly salient given that the United States and China constitute the world's largest trading dyad and are the world's largest two economies, and that the trade war became a key political issue during the 2016 election cycle and Trump's first years in office. This high salience helps avoid challenges with knowledge gaps as well as reliance on hypothetical questions and experiments that may not apply beyond laboratory settings.
The trade war has largely been seen as politically expedient for Trump in 2018, 3 and existing trade preference theories help explain why. Five existing theories stand out: the personal economic effects of trade (self-interest); the perceived economy-wide effects of trade (sociotropism); how "other" and undesirable a trade partner appears (xenophobia); perceptions of unfair competition and relative gains (reciprocity); and party and political leader identification ( partisanship). All five theories can help explain initial public support for the trade war. Growing attribution of manufacturing job losses to trade integration with China may have caused more Americans to attribute individual job losses and wage growth stagnation to China (self-interest) and also to blame China for broader economic stagnation in the United States (sociotropism). 4 China under Xi Jinping had become more internally repressive and externally aggressive, driving down favorable opinions of China (xenophobia). China's market reforms had largely stalled and the World Trade Organization (WTO) had become less effective at pressuring change, while China's bilateral trade surplus with the United States rose to unprecedented levels (reciprocity). And Trump politicized the trade conflict ( partisanship). Contemporaneous public opinion surveys suggest that partisanship and reciprocity frames were extremely important drivers of support: Americans perceived that trade with China was unfair, and Republicans and Trump supporters agreed that a trade war could best address this unfair competition.
Yet while these existing theories appear to have had strong explanatory power in 2018, four years later the nature of the US-China bilateral relationship has evolved, and existing theories may not explain shifts in public opinion. In particular, the relationship has been securitized and reframed as great power competition, with the two countries now widely perceived as existing on the "foothills" of a new Cold War. 5 Existing research on trade preferences does not adequately address fears of conflict, and this is particularly the case with regard to great power conflict, where individuals likely perceive the consequences of conflict as more severe. 6 This article explores how security concerns in the bilateral relationship shape individual preferences regarding the trade war, arguing that as the salience of a securitized great power competition narrative came to dominate an unfair trade narrative, security concerns related to trade came to dominate traditional explanations for individual trade preferences. Using results from two public opinion surveys, one a nationally representative survey conducted by Ipsos KnowledgePanel (n = 1,016), and the other conducted by the author with participant recruitment using Amazon Mechanical Turk (MTurk) (n = 1,015), the article tests whether these security concerns influence trade preferences, and particularly whether security concerns lead to a preference for protectionism, consistent with a realist "securities externalities" theory, or to a preference for free trade and tariff removal, consistent with liberal "commercial peace" theory.
Results indicate that individuals worried about conflict and those who perceive the United States and China as having entered into a new Cold War report declining levels of support for the tariffs, consistent with the instinctive commercial peace interpretation. More specifically, the surveys demonstrate that existing theories have strong explanatory power regarding initial support for the 2018 tariffs, but have less power explaining self-reported change in support for these tariffs between 2018 and late 2021. Instead, perceptions of a Cold War and fears of conflict with China predict increasing opposition to the tariffs. To test whether these security concerns made the potential moderating effect of trade more compelling, the article then explores support for the proposition that trade reduces the threat 3 Most trade literature looks at state-level politics or the impact of business lobbying rather than individual preferences, and the US-China trade war is no exception: For example, Liu and Woo (2018) see US policy maker concerns over growth and security as driving the trade war; Di, Luft, and Zhong (2019) argue that US federal government revenue needs drove tariff increases; and Swenson and Woo (2019) argue that the trade war was made possible by the disillusionment of "internationalist business" regarding China's economic reforms. of conflict, that is, the commercial peace. Those who believe in the commercial peace are indeed significantly more likely to have become more opposed to the trade war, and those who perceive a high threat of conflict are more likely to believe in the commercial peace. In an experimental setting, asking respondents to read a commercial peace cue unsurprisingly leads treated respondents to become less supportive of the trade war; but interestingly, those with elevated security concerns are unaffected by the cue, as they appear to have already internalized the commercial peace. These findings have important implications for conventional wisdom regarding public support for the trade war, and also add depth to existing theories of individual trade preferences in the shadow of conflict.
The article proceeds as follows: Section 2 looks at existing theories of individual trade preferences in the context of the US-China trade war, and then shows how securitization may dominate these hypotheses in the context of great power relations. Section 3 describes the data and methodology, section 4 presents results, and section 5 concludes.

Theory and hypotheses Existing trade preference theories and the US-China trade war
Despite growing interest in the subject over the past decade, individual trade preferences remain relatively underresearched. Initial research tended to focus on individual economic self-interest, using trade theory predictions to understand how individual characteristics interact with the distributional consequences of trade. 7 Recent studies concentrate more on psychological and political factors. These theories appear to have explanatory power in interpreting the domestic political logic behind the initiation of the US-China trade war, and it is therefore important to understand these theories' predictions before hypothesizing about the impact of security concerns.

Self-interest
Early research on individual support for trade policies focused on economic self-interest, in effect assuming that individuals rationally asked: What can trade do for me? These studies tended to use predictions from existing trade models regarding the distributive effects of trade to determine expected support for liberal or protectionist trade policies. Perhaps most commonly, research tested the Heckscher-Ohlin factor endowments model, in which trade preferences should be determined by skill levels, and the Ricardo-Viner specific factors model, in which trade preferences should depend on industry-level effects. 8 Yet studies setting out to see if income levels or employment sectors predicted individual preferences often found little support for these rational interpretations, particularly when tested against alternative explanations. 9 One reason for the lack of empirical support, discussed in more depth in the following text, may be low levels of trade salience and knowledge-the public may be rational, but correctly assessing trade impacts on individual bottom lines is difficult. 10

Sociotropism
Beyond effects on personal well-being, individuals may be altruistically motivated and consider the effects of trade on their fellow citizens. This complements an empirical finding from public opinion polling: Trade sentiment often follows broader national economic conditions, with individuals more supportive of trade when the economy is more robust. This effect may be largely independent of selfinterest, 11 though separating the two can be very difficult. 12 7 See, e.g., Scheve and Slaughter (2001) and Mayda and Rodrik (2005). 8 See, e.g., Nollen and Iglarsh (1990); Kahane (1996); Scheve and Slaughter (2001); Mansfield, Mutz, and Brackbill (2019); and Feigenbaum and Hall (2015). 9 See Mansfield and Mutz (2009) and Baldwin and Magee (2000). 10 On the low degree of trade policy salience, see Guisinger (2009). On the low degree of knowledge about trade outcomes, see Rho and Tomz (2017). 11 Mansfield and Mutz (2009). 12 Schaffer and Spilker (2019) find that individuals do not react to sociotropic cues but rather to information about how trade may affect their personal economic situation.

Xenophobia
Recent literature on trade preferences moves beyond economic effects and looks at psychological and group-based explanations. One important determinant of trade preferences is the degree to which a trade partner appears as an undesirable "out-group," either for racial or political reasons. According to this view, increasing ethnocentrism helps explain growing support for protectionism. 13 Ethnocentrism is negatively correlated with education levels, which may help to explain why xenophobia is at times misinterpreted as economic self-interest. 14 Although racism and ethnocentrism may drive sentiment, this category can also include more rational enmities, including historical experience of conflicts or perceptions of political undesirability. 15 Most related to the literature discussed in the following text on how security concerns shape preferences, this "out-group" literature also identifies greater support for trade with allies over adversaries, though this is not generally framed as being driven by concerns of conflict. 16

Reciprocity
From another psychological vantage point, trade preferences may also reflect reciprocity or fairness concerns. 17 Part of the explanation may be trust; 18 another part is psychological loss aversion. 19 Considerable research finds that reciprocity has been a key driver of US trade policy. 20 And there is strong evidence of reciprocity concerns in the US-China economic relationship. 21

Partisanship
Finally, partisan identification can help explain trade preferences. Partisan elites have a strong role in shaping opinion in foreign policy and economic areas in which public knowledge and expertise may be limited. 22 Information about trade and the interpretation of this information is highly conditioned by partisan framing. This helps to explain why trade is viewed much more positively when one's own party holds the presidency, even when trade policies remain constant. 23 All these theories can help explain the domestic political logic behind Trump's initiation of the US-China trade war. These theories often overlap-as discussed in the text that follows, it is often difficult to construct variables that clearly address only one-but a relatively clear story about the trade war's emergence takes shape, one that shows how existing trade preference theories all pointed toward increased protectionism in the US-China trade relationship, as summarized in the text that follows and in table 1.
• Self-interest. Prior to the initiation of the trade war, considerable research-and considerable media coverage of this research-attributed job losses and wage stagnation to trade integration with China. 24 Areas hit harder by the China trade shock were more likely to push protectionist trade policies in congress. 25 • Sociotropism. Overall, trade with China may be welfare-enhancing, but the losers from trade integration are much more prominent in the public eye than gains from price effects. He tweeted about China more than 400 times between his election and the signing of the Phase 1 trade agreement in early 2019. In a sign of how all these explanations interact, these tweets predominantly focused on how America was being cheated (reciprocity), but also appealed to nationalistic and ethnocentric narratives (xenophobia), while touting the economic benefits of the tariffs and inability to lose trade wars (self-interest and sociotropism). 28 The trade war tariffs have now been in place for four years, however, during which US politics and the US-China bilateral relationship have changed considerably, with implications for theory predictions. In terms of economic self-interest, although workers in protected sectors may feel that the trade war has personally benefited them, consumers and those in unprotected sectors or sectors that faced retaliatory tariffs should ostensibly be less supportive of the tariffs. Regarding sociotropic explanations, the costs and benefits of the tariffs have become clearer, both to academics and, presumably, to the public, yet there is no indication that these distributional and economy-wide costs have caused individuals to see tariffs as more economically harmful. In contrast, both xenophobia and reciprocity concerns should predict greater support for tariffs today. Xenophobia should have increased as China's "otherness" became more pronounced as a consequence of media focus on Xinjiang human rights atrocities and China's culpability in spreading the coronavirus. Indeed, American public favorability toward China reached its lowest level on record in 2021. 29 Similarly, although the bilateral  Newport (2018). 28 Author analysis of Trump tweets accessed using Trump Twitter Archive V2. 29 Smeltz and Kafura (2020).
deficit has shrunk somewhat, 30 China remains the largest contributor to the US trade deficit, China's market liberalization has continued to stall, and the US economy is seen as weaker than China relative to 2018, 31 all of which should spur reciprocity concerns. In terms of partisanship, continuation of the trade war tariffs by the incoming Democratic administration should facilitate greater support for the trade war among Democrats who may have greater intrinsic support for protectionist policies but opposed the trade war due to its association with Trump; in contrast, free-trade Republicans may feel liberated to oppose the tariffs with a Democratic administration in place. In sum, as the economic effects of the trade war became clearer, US-China relations deteriorated, and partisan dynamics shifted, existing theories generate testable predictions, with most implying increasing support for the trade war. These predictions are summarized in table 1.

Securitization of the US-China relationship
But the US-China bilateral relationship has also changed dramatically in the four years since the trade war began, becoming increasingly antagonistic and securitized; as documented in the following text, American perceptions of China as "unfavorable" and an "enemy" have reached unprecedented highs. Many factors contributed to this change, including both securitization of the economic relationship by the Trump administration and, beginning in spring 2020, perceptions of China's culpability in spreading a global pandemic. This article does not seek to isolate these effects, but instead briefly documents the securitization process that led Americans to see China in security-related rather than purely economic terms, demonstrating that these shifting views were not solely a consequence of China's perceived global pandemic culpability. The clearest indication of this is that American favorability toward China collapsed prior to the pandemic: Only 45 percent of Americans held unfavorable views of China when the trade war began in early 2018, a share that had remained relatively stable since 1989, but this share surged to 67 percent in February 2020, before a single documented American case of the coronavirus. 32 By explaining the creation of threat perceptions as a political process, the securitization literature helps explain the process by which the Trump administration caused Americans to see China in security-related terms. 33 Threats may be socially constructed through interactive "speech acts," in which audience understanding of a threat is shaped by policy-maker "attribution of meaning"-in other words, securitization occurs when the public audience accepts and legitimizes policy-maker identification of a new threat. 34 When trade is "tied to debates over war and peace … in the presence of strong external threats" it can become an issue of national security. 35 Securitization of the bilateral relationship by the Trump administration became apparent after the trade war began in 2018. In the lead-up to the trade war, Trump presented the bilateral relationship in terms of reciprocity and unfairness. New personnel helped usher in a shift in spring 2018. Mike Pompeo became Secretary of State and John Bolton took over as National Security Advisor, and both explicitly sought to securitize the relationship by linking economics and security and diminishing the role of Treasury and Commerce. 36 Pompeo reportedly felt that his "biggest contribution" was getting Americans to think about China's plans for global supremacy rather than merely looking at the transactional economic relationship; under Pompeo, the State Department helped to generate a 30 The shrinking deficit is in part a statistical anomaly. See Clark and Wong (2021). 31 Smeltz et al. (2021). 32 Gallup (2022). This is not to deny that the pandemic also undermined favorability to China: During 2021, the share of Americans with unfavorable views of China rose to 79 percent. 33 Buzan, Waever, and Wilde (1998). 34 As Balzacq et al. (2016) further explain: "[A]n issue is given sufficient saliency to win the assent of the audience, which enables those who are authorized to handle the issue to use whatever means they deem most appropriate. In other words, securitization combines the politics of threat design with that of threat management" (495). 35 Verdier (1994,43). 36 Interview with senior former Trump administration State Department official #1, October 1, 2021.
bipartisan "common understanding of the true nature of communist China." 37 This culminated in the November 2020 release of the State Department's Elements of the China Challenge, which in sweeping terms presented the bilateral relationship as a civilizational conflict. Although language on Chinese revisionism began as the trade war was initiated, 38 securitization of the economic relationship followed these personnel changes and attendant shifts. In particular, the administration securitized semiconductor exports in May 2019 following direct pressure from Bolton, 39 and this securitization rapidly spread to the broader technology and trade relationship. 40 Although most Americans do not read policy reports, security-related language became apparent in media coverage and Trump tweets, demonstrating the success of the securitization process. As Breuer and Johnston (2019) document, a master narrative of China as a revisionist power moved from think tank and policy worlds in 2016 and 2017 into public media by 2018, crowding out narratives that emphasized cooperation and deemphasized security. 41 Trump's China tweets themselves also demonstrate a shifting emphasis from China trade topics toward China security topics. Figure 1 documents several temporal trends that hint at the success of the securitization process. The solid gray line shows that security-related Trump tweets referencing China began to rise in early 2019 and stayed elevated through 2020. The dashed gray line shows that public interest in China trade (proxied by Google News search trends) surged at the start of the trade war in 2018 but fell beginning in 2019 as security-related Trump China tweets rose. By early 2020, the dotted black line shows a surge in Google News searches for "Cold War." In response to securitization of the relationship and the global pandemic, then, Americans became less interested in trade ties with China and began to emphasize a security-focused great power competition narrative.
Growing American interest in a new Cold War also coincided with survey data demonstrating increasing views of China as a military threat and enemy. Gallup polling shows that 45 percent of Americans in 2021 considered China to be the "greatest enemy" of the United States, up from only 11 percent when the trade war began in 2018. 42 According to the Reagan National Defense Survey, 52 percent of Americans saw China as the greatest national security threat in November 2021, up from 21 percent in 2018. 43 The same survey shows that 71 percent of Americans in 2021 were worried about war with China in the next five years. And trust in the US military to win this war has receded: In 2019, 58 percent saw US military as stronger than China; in 2021, only 46 percent did. 44

Security concerns and trade preferences
The US-China bilateral relationship thus came to be characterized by rising threat perceptions and securitized economic interactions, making the trade war an excellent test case for analyzing the effects of security and conflict concerns on individual trade preferences-the hypotheticals of trade in the shadow of conflict became more real. Existing research looking at public opinion related to trade and security concerns attempts to adjudicate between the military-boosting "security externalities" of trade and the potential pacifying role of international trade ("commercial peace"). The "security externalities" theory argues that the economic gains from trade can help adversaries divert resources to military capabilities, which should lead to support for protectionism and help explain observed tendencies to sign trade agreements with allies rather than adversaries. 45 Several studies identify this 37 Interview with senior former Trump administration State Department official #2, October 7, 2021. 38 This is seen most clearly in the National Security Strategy (December 2017) and National Defense Strategy (January 2018), which both defined China as a revisionist power. 39 This episode is described in Davis and Wei (2020). 40 For more on US technology securitization in the bilateral relationship, see Friis and Lysne (2021) and Medeiros (2019). 41 Breuer and Johnston's analysis ends in 2018, so it is unclear if the trend continued into 2019 and beyond. 42 Gallup (2022). 43 Ronald Reagan Institute (2021). 44 Kafura and Smeltz (2021). 45 Gowa and Mansfield (1993Mansfield ( , 2004. Note that signing trade agreements with allies could also be due to xenophobia concerns or simply greater opportunities for international cooperation. "intuitive neorealist" concern of security externalities and individual hostility toward trade with adversaries. 46 In an experimental setting, Carnegie and Gaikwad (2017) find that "respondents eschew trade when economic exchange is posited to increase an adversary's military capabilities." Yet other studies find that publics are sensitive to the economic costs of conflict and may believe that trade can limit the potential for conflict, consistent with the commercial peace pillar of liberal peace theory. In the 2014 Russia-Crimea conflict, media coverage highlighting the costs of conflict significantly reduced hawkish attitudes. 47 Similarly, information about trade ties between Japan and China strengthened opposition to conflict in a study of the Japanese public. 48 Broader work on democratic peace theory finds that economic variables including GDP per capita, levels of foreign direct investment, and indices of economic freedom all negatively predict individual "bellicosity," implying that propensities for interstate conflict are lowered by increasing economic prosperity and interstate linkages. 49 And although Carnegie and Gaikwad (2017), in a direct experimental comparison of the two theories, find greater support for the securities externalities theory, they also find public support for the commercial peace and that belief in these peace-inducing trade aspects can overcome aversion to trade with adversaries.
Which of these two theories should dominate in the case of the securitized US-China relationship? Two factors may make the case special: high levels of salience and great power conflict. Existing research findings are predominantly based on hypothetical survey experiments, but the high salience of the trade war and the "real" threat level observed in a nonlab setting may yield different results. Note: Security-related China tweets include all Trump tweets that reference China as well as one of the following terms: enemy, military, security, threat, defense, South China Sea, Taiwan, or communism. Both Google News trends are calculated as six-month rolling averages. All three series are converted into normalized z-scores to facilitate trend comparison. Source: Author's calculations based on Google Trends and Trump Twitter Archive. 46 Herrmann, Tetlock, and Diascro (2001). DiGiuseppe and Kleinberg (2019) find that survey respondents were much less likely to prefer agreements with military rivals. Spilker, Bernauer, and Umaña (2016) find that military alliances can lead to public support for trade agreements, though they only find support for this in one of three countries in which they run experiments. 47 Stoycheff and Nisbet (2016). 48 Tanaka, Tago, and Gleditsch (2017). 49 Jakobsen, Jakobsen, and Ekevold (2016).
Literature on individual trade preferences frequently must grapple with low levels of respondent interest and knowledge, 50 but there is reason to believe that this does not apply to the US-China trade war due to wide media coverage, Trump tweets, and economic impact. By March 2018, before most tariffs were even applied, an Economist/YouGov poll showed 75 percent of Americans had heard a little or a lot about the tariffs, 51 and the same poll in May 2019 saw this number rise to 83 percent. 52 Other polls seem to indicate that this awareness also translates into better knowledge: A 2019 Monmouth poll showed that 62 percent of Americans think US consumers will bear the cost of tariffs; the same number correctly thought that higher costs would simply be passed through. The salience of China trade issues in recent years makes direct survey responses related to China potentially more robust than hypothetical questions regarding hypothetical trade partners. 53 In addition to salience, great power conflict may lead to greater threat perceptions. According to many Americans, the United States and China are now embroiled in a new Cold War, one that many Americans fear could become a hot war. This dynamic did not exist when Trump was elected or when the trade war began. The original Cold War provides a relevant but distinct test case: relevant due to the nature of bipolarity and fears of conflict, but distinct because of the low level of trade integration between the United States and the Soviet Union. Most academic literature on bilateral US-USSR trade focuses on elite preferences rather than public opinion. 54 But in a sign of disjuncture between policy and opinion, the American public was consistently more supportive of increased trade integration with the USSR than political leadership. Starting as early as 1955, a majority of Americans supported increasing trade ties with the Soviet Union; they supported increased trade ties by a 2 to 1 ratio in the 1950s, rising to a 6 to 1 ratio in the 1970s. 55 These high degrees of support are not adequately explained by declining threat perceptions; indeed, in some ways threat perceptions were higher in the 1970s given the failure of the Vietnam War and increased perceptions of nuclear parity. 56 If similar trends hold today, then it is likely that the liberal commercial peace belief may dominate the realist securities externalities concerns with regard to China trade. There is some indicative evidence along these lines, with recent survey data demonstrating that Americans tend to think trade with China strengthens US national security: According to the 2019 Chicago Council survey, 64 percent of respondents supported this view.

Hypotheses
The preceding discussion points to several hypotheses regarding American public support for the US-China trade war. First, in the case of a highly salient great power relationship, heightened security concerns should lead to greater support for trade liberalization. Additionally, although existing trade theories are likely to explain initial levels of trade war support, they should have less explanatory power in the presence of salient conflict frames and threat perceptions: 50 On salience, see Guisinger (2009). On trade knowledge, see Bauer, Pool, and Dexter (1972) and Rho and Tomz (2017). 51 The Economist/YouGov Poll (2018). 52 The Economist/YouGov Poll (2019). 53 Much of the existing research relies on hypotheticals and the introduction of information cues, which may not translate to real-world nonexperimental settings. The two most relevant existing studies (Carnegie and Gaikwad 2017 and DiGiuseppe and Kleinberg 2019) both use conjoint studies that look at support for trade with hypothetical partners, but this approach may be ill-suited to a specific high-salience case as opposed to more hypothetical trade relationship settings. For instance, it is strange that Carnegie and Gaikwad find no significant subgroup differences in commercial peace versus security externality emphases between doves/hawks, liberals/conservatives, Republicans/Democrats, ethnocentrism, nationalism, nonisolationists/isolationists, or level of foreign interest. Differences between these subgroups are generally large in existing surveys of trade policy support, so their lack of significance may indicate responsiveness to framing that does not reflect underlying real-world views. 54 See, e.g., Levine (1974); Esno (2018). 55 Wittkopf and McCormick (1990). 56 Wittkopf and McCormick (1990) explain increased willingness to "build bridges of accommodation" in the 1970s as a consequence of American concern about "the growing capability of the Soviet Union to destroy the United States with nuclear weapons." H1a. Respondents with high reported levels of security concerns should report increasing opposition to the trade war tariffs.
H1b. Existing theories of individual trade preferences (partisanship, reciprocity, self-interest, sociotropism, and xenophobia) should explain initial levels of trade war tariff support.
H1c. Existing theories of individual trade preferences should have less explanatory power when it comes to recent changes in levels of trade war tariff support.
A second set of hypotheses relates to the channels through which security concerns should affect trade war preferences. In particular, respondents who believe in commercial peace theory should be less supportive of the trade war, all else equal, and this effect should be greater for those with elevated security concerns, who are likely to give more weight to trade's expected effects on security.
H2a. Belief in the commercial peace should be associated with lower support for the trade war tariffs.
H2b. Opposition to the trade war tariffs should be heightened for those who both believe in the commercial peace and have heightened security concerns; the effect should disappear for those who believe in the commercial peace but have low security concerns.
A final set of hypotheses looks at the role of prior information and the expected effects of experimental cues. When respondents are informed about commercial peace theory in an experimental setting, they are likely to report greater subsequent belief in commercial peace theory. However, respondents who have greater security concerns should be more likely to have already internalized this belief because they are already thinking about trade in security-related terms. These respondents should exhibit higher baseline support for commercial peace theory and should be less affected by experimental cues.
H3a. Exposure to commercial peace theory using experimental cues should increase belief in the commercial peace.
H3b. Respondents with elevated security concerns should have a greater baseline belief in the commercial peace.
H3c. Respondents with elevated security concerns should be less affected by commercial peace cues.

Empirical approach Data and variable construction
To test these hypotheses, this article draws from two surveys: a nationally representative survey conducted by Ipsos KnowledgePanel (n = 1,016), and a similarly sized survey using Amazon Mechanical Turk (n = 1,015). Ipsos KnowledgePanel (IKP) is representative of the US population, with survey respondents recruited randomly through probability-based sampling using address-based sampling methods to ensure full coverage of all households. A total of 1,969 surveys were fielded from October 22-31, 2021, with a response rate of 52 percent, yielding 1,016 completed surveys. 57 MTurk, though not representative of the population at large, can produce high-quality survey data; indeed, based on completion rates and success passing manipulation checks, which indicate high engagement, the data quality may be better than many commissioned surveys, 58 possibly because 57 Full IKP methodology can be found at Ipsos (2021). 58 Zhang and Gearhart (2020); Buhrmester, Kwang, and Gosling (2011). payment is based upon approval. 59 Many best-practice approaches to "crowdsourced sampling" have emerged in recent years to manage challenges of inattentive workers; 60 intentionally dishonest workers; 61 "bots" that mimic workers; 62 and "virtual private networks" (VPNs). 63 To address these challenges, rather than use Amazon's "Master Workers," which is more expensive and does not clearly affect data quality, 64 this study follows Agley et al. (2022) in designing quality control questions. MTurk workers had to be based in the United States and have a successful task completion rate greater than 95 percent; participants were paid $0.50 for a study that took, on average, under five minutes; and the survey included two initial questions to help screen out potential bots and non-US VPNs. 65 This article analyzes how different individual-level variables translate into initial support for the China tariffs as well as preferences regarding the removal of tariffs today. Dependent variables include self-reported initial support for the tariffs in 2018 (tariff2018), ranging from strongly opposed (1) to strongly supported (5), and self-reported change in support for tariffs at the time of the survey in late 2021 (tariff_change), ranging from much less supportive (1) to much more supportive (5). 66 Individual-level variables are constructed to correspond to existing theories regarding individual trade preferences discussed in the preceding text-self-interest, sociotropism, xenophobia, reciprocity, and partisanship-as well as theories related to security concerns. The two key security-related variables are a dummy indicating identification of a new Cold War (coldwar) and the reported level of concern that war will break out between China and the United States (warconcern). Additional independent variables correspond to theories of trade preferences, drawn from existing literature, in addition to demographic controls. The relevant questions in the survey instrument can be found in the appendix. Table A1 in the appendix reports summary statistics. 67

Methodology
The analysis incorporates three broad approaches to test the hypotheses outlined in the preceding text. First, to explore support for the initial trade war tariffs as well as self-reported change in support for the tariffs between 2018 and 2021, the article compares individual-level correlates across the two samples using ordered logit models on a group of explanatory variables. Expectations based on preceding hypotheses are that security concerns will predict growing opposition to the trade war (H1a). Additionally, all five traditional explanations for individual trade preferences should explain initial support for the trade war (H1b), but are expected to have less explanatory power related to changing views (H1c). Initial support for tariffs should increase in low income and low skill, Republican, and 59 Sheehan (2018). 60 Berinsky, Margolis, and Sances (2014). 61 Hydock (2018). 62 Buchanan and Scofield (2018). 63 Kennedy et al. (2020). 64 Loepp and Kelly (2020). 65 The first question was: "If you had an emergency, what telephone number would you dial?" Answer possibilities included 911 and three emergency numbers used in countries outside the United States. The second question was: "When you were in school, how hard did you work on your studies? In answering this question, please ignore everything else and select the final option indicating that you don't really remember." Answer options included "I worked incredibly hard"; "I worked moderately hard"; "I didn't work very hard"; and "I don't recall how hard I worked." Subjects who failed quality control tests were not paid. The informed consent statement warned participants that there would be quality control tests that, if failed, would prevent payment. The two quality control questions were placed at the beginning of the survey, ensuring that those who failed the tests did not contribute uncompensated data. 66 Overall, more than half the sample indicated that they had changed their opinion of the tariffs: Of those who changed their opinion, 58 percent became less supportive of the tariffs, while 41 percent became more supportive. 67 Looking at the summary statistics, similarities between the IKP and MTurk samples help justify the use of the nonrepresentative MTurk data. Overall, 46 percent of IKP respondents initially supported the tariffs, compared to 50 percent of the MTurk respondents. MTurk respondents appear somewhat more favorable toward China than the American public, somewhat more likely to believe that trade is good for the American economy, and somewhat less likely to believe that China's trade surplus was generated predominantly through unfair competition. The nonrepresentative MTurk survey also skews younger, more male, more white, and more educated, as in other MTurk surveys, though it has the same share of Republican respondents.
Trump-supporting individuals who perceive trade as having negative effects on the US economy, have high nationalism scores, have unfavorable views of China, and interpret China's trade surplus as based on unfair competition. Initial tariff support can be tested with both IKP and MTurk samples, though with slightly different variables; similar results help justify use of the MTurk sample. Direct assessments of changing tariff support are only available in the MTurk sample.
Second, to explore belief in commercial peace theory and how this belief affects support for tariffs, the article repeats the first analysis, including a variable corresponding to degree of belief in the assertion that trade with China reduces the risk of conflict (cp_agree), as well as interactions between this variable and the security-related variables. All else equal, respondents who report believing that trade decreases the likelihood of conflict should be less supportive of tariffs overall (H2a). However, this effect should be conditioned by respondent security concerns, indicated by a positive interaction effect between the security variables and belief in commercial peace (H2b).
Third, to explore how evolving narratives securitizing China have influenced threat perceptions and support for tariffs, an embedded experiment analyzes disparate framing effects across subgroups. Existing research shows that information cues can have a large effect on support for trade. 68 The MTurk survey randomly assigned respondents to control and "commercial peace" treatment groups that received different information cues. The commercial peace treatment group was asked to read the following cue: "Academic studies have consistently found that countries that trade more with each other are less likely to go to war. According to this 'commercial peace' theory, trade serves as an important deterrent to military conflict." The control group was not given a statement to read. Respondents were then asked about their level of support for the commercial peace as well as the conditions under which they would remove the trade war tariffs. The commercial peace cue should make respondents more likely to believe in the commercial peace and subsequently more likely to support tariff removal, all else equal (H3a). However, respondents with security concerns are more likely to have already considered and internalized a commercial peace logic; they should have a greater baseline belief in the commercial peace (H3b) and should also be less affected by the cue (H3c).

Challenges with retrospective questions
A key assumption underlying the methodological approach outlined here is that survey respondents can accurately recall both their degree of support for the trade war tariffs nearly four years prior as well as how their views of the tariffs changed in this interval. This is a demanding expectation. Although forgetting itself does not pose too large a problem given that participants can select "No opinion/do not remember," retrospective questions may lead to three recall challenges that could systematically bias results: distortions arising from projecting current attitudes backward, 69 distortions arising from seeking congruence with perceived changes in broader societal attitudes, 70 and time displacement by adopting attitudes from a point in time that does not correspond to that indicated in the survey question. Time displacement should not pose too large a challenge given the limited time horizon, 71 but distortions arising from fitting responses for consistency, either with current individual or current societal attitudes, pose a potentially greater concern.
Several factors help to mitigate-albeit not eliminate-this concern. First, given the short time horizon and the continued media coverage and campaign attention to the US-China relationship, the trade war has remained highly salient; distortions are more likely when issues have not been considered over the interval. Second, similarities between the results here and contemporaneous surveys provide some support for recall reliability: in the nationally representative IKP survey, 46 percent of respondents 68 For instance, Hiscox (2006) find that when survey respondents read an antitrade introduction linking trade to potential job losses, they were 17 percent less likely to favor increasing trade with other countries. 69 Barsky (2002); Jaspers, Lubbers, and de Graaf (2009). 70 Coughlin (1990); Himmelweit, Biberian, and Stockdale (1978); Joslyn (2003); Smith (1984). 71 The trade war tariffs had been in place for less than four years at the time of the surveys, so respondents cannot displace to a period before this. Most displacement concerns refer to a longer history. See Sudman and Bradburn (1974). Additionally, it does not particularly matter to the findings whether respondent views changed beginning in 2018, 2019, or 2020. either moderately supported or strongly supported tariffs, including 68 percent of Republicans, 28 percent of Democrats, 52 percent of men, 40 percent of women, 52 percent of white respondents, and 30 percent of nonwhite respondents, shares that are comparable to contemporaneous nationally representative public opinion surveys. 72 Third, distortion bias is likely orthogonal to the key independent variables; in particular, there is no reason to think that individuals with security concerns are more likely to distort their answers. 73 Distortions to match current societal attitudes are most likely in issue areas that have seen broad societal evolution over the intervening period, 74 and overall trade war attitudes do not appear to have significantly shifted. 75 If distortions are not systematically related to the independent variables, then they should drive results downward but not lead to systematic bias.
Finally, the key retrospective question-asking respondents directly whether their views have changed-helps to mitigate bias arising from individual consistency distortions. 76 Studies of retrospective questioning make clear that recall is improved by simplified question wording that minimizes cognitive effort, 77 and that questions with broad answer categories also help to increase recall accuracy. 78 Consequently, the question here relates to a single category with a broad categorical response, under the assumption that it may be easier to remember direction of attitude change rather than specific earlier attitudes. None of these factors fully resolve the challenge, but they do provide a degree of support for the approach.

Correlates of trade war support
Who supported the initial trade war tariffs, and how have levels of support changed? Bivariate correlations suggest that traditional trade preference theories help explain initial support for the trade war. 79 Tariff support is associated with low-income and non-college-educated respondents (self-interest); a belief that trade harms the economy (sociotropism); unfavorable attitudes toward China and high degrees of nationalism (xenophobia); a belief that unfair competition explains China's trade surplus (reciprocity); and Republicanism and Trump 2020 votes (partisanship). Regression analysis helps to better understand the interactions between these effects and their relative weights, and also to analyze whether the same variables explain reported changing levels of support. Table 2 reports coefficients from ordered logit models analyzing initial support for the 2018 tariffs (columns 1-4) and reported change in support for the tariffs (columns 5-7). Column 1 uses the IKP sample; columns 2-7 use the MTurk sample. Columns 1 and 2 include identical covariates to facilitate comparison across samples. Columns 3, 4, 6, and 7 add the two security-related variables: identification of a new Cold War (coldwar) and level of concern that war will break out between China and the United States (warconcern). Columns 4 and 7 include several additional variables that are unavailable in the IKP survey data: whether respondents believed the trade war has benefited them personally 72 See The Economist/YouGov (2018). 73 Most sociodemographic characteristics are not related to recall accuracy (Coughlin 1990), with the possible exception of higher recall by educated respondents (Joslyn 2003). 74 For example, Smith (1984) finds that ten-year prior reconstructive survey responses closely track ten-year prior marginal distributions for busing, an issue that exhibits no aggregate societal change, but not for views of Communism or racial intermarriage, which had exhibited broad societal evolution over the preceding decade. 75 Although overall societal attitudes toward the trade war have remained relatively constant, opinions may have evolved among certain subgroups. In particular, partisanship effects might predict that Republicans would broadly be more opposed to the trade war today, and therefore might be more likely to (falsely) recall initial opposition to the tariffs. However, in the IKP nationally representative data, 68 percent of Republicans recall initial trade war approval, which compares to 62 percent (The Economist/YouGov) and 74 percent (Quinnipiac) of Republicans in contemporaneous 2018 surveys, helping to alleviate concerns. 76 Such distortions would bias respondents to report zero change, which would bias downward estimated effects. 77 Krosnick (1991). 78 Beckett et al. (2001). 79 Figure A1 in the appendix breaks down tariff support by subpopulation, grouping moderate and high support and opposition to create binary categories for ease of interpretation.  Note: Columns 1-4 report coefficients from ordered logit regressions on a ranked variable indicating reported support for the 2018 trade war tariffs, with higher values corresponding to more support. Columns 5-7 report coefficients from ordered logit regressions on a ranked variable indicating reported change in support for the tariffs in 2021 since they began in 2018, with positive numbers corresponding to increasing support for the tariffs. Column 1 uses the IKP sample. Columns 2-7 use the MTurk sample. Standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1.
(tariff_goodself), whether respondents believe the trade war has benefited the US economy generally (tariff_goodeconomy), a nationalism index (nationalism), and a dummy variable indicating a vote for Trump in 2020 (trump2020). Results in table 2 largely confirm the hypotheses. Most notably, security concerns (coldwar and warconcern) have a strong negative association with change in support for the tariffs, implying that respondents with security concerns report increasing opposition to the tariffs over the past four years (H1a). Results for other individual correlates are also largely consistent with expectations. In terms of initial support, education (college) negatively predicts support; respondents who view China unfavorably (unfavorable) or who have high nationalism scores (nationalism) tend to support the tariffs; and Republicans are supportive of the tariffs, although when a variable for 2020 Trump voters (trump2020) is included, it dominates the "Republican" partisan effect, indicative of how closely these tariffs were tied to Trump. Respondents who think that China generated a trade surplus as a result of unfair trade practices (unfair) also tended to initially support the tariffs. Supporting the data quality of the MTurk survey, most coefficients are similar across samples (columns 1 and 2), with two surprising caveats: positive attitudes toward the economic effects of trade predict greater support for tariffs in the MTurk data, although they have no significance in the IKP data; and views of Chinese unfair trade practices exhibit no significant relationship in the IKP data. 80 Broadly, as hypothesized (H1c), these same factors tend to have less explanatory power when it comes to reported change in tariff support (columns 5-7). For the non-security-related variables, only two consistent effects stand out: Republicans have become more supportive of the tariffs, as expected, and individuals with unfavorable views of China, a proxy for xenophobia, have become significantly less supportive of the tariffs, contrary to predictions from existing theory. A potential explanation for this surprising finding is that individuals with high levels of unfavorability are also those who see high threat potential or Cold War emergence. This interpretation is also supported by the fact that the other variable corresponding to xenophobia (nationalism) has no significant impact, despite its explanatory power determining initial preferences. Variables corresponding to beliefs about the personal and economy-wide effects of the tariffs act in the expected direction: Those who think the tariffs benefited them personally and those who think the tariffs have benefited the US economy as a whole both report increasing support for the tariffs. 81 To facilitate comparison and interpretation, figure 2 visually presents 95 percent confidence intervals for the marginal effects of each independent variable, based on logit models run after converting the dependent variables (2018 tariff support and changing support levels) into binary form and using the same set of independent variables. The visual results highlight the earlier conclusions: Security variables and unfavorability explain declining support for tariffs, while most other variables help to explain initial support more than changing levels of support. The marginal effects imply that identification of a new Cold War leads to 15 percent less support for tariffs in late 2021, while unfavorability toward China leads to 18 percent less support.
If Americans with heightened security concerns have become more opposed to trade war tariffs, as implied by these results, then these individuals may also have different overall objectives when it comes to trade with China. In particular, the trade preferences of security-concerned individuals may not reflect economic motivations per se, but rather political goals. Table A2 in the appendix looks at how the same explanatory variables predict respondents' reported "primary goal of trade with China." As expected, coldwar and warconcern significantly predict identification of "tool for political pressure" or "containment" as the primary goals of bilateral trade with China, and are negatively correlated with identification of economic growth and job creation as primary goals. This finding lends support to the idea that security-concerned individuals do not necessarily have different underlying economic understandings of trade with China, but rather that they believe trade can fulfill noneconomic goals. 80 In these cases, bivariate relationships support the initial hypotheses, so the interpretation is that other effects are dominating unfair and postrade, though it is not clear why this is only the case in the IKP data. 81 In both cases, respondents are likely conflating broader economic trends with actual tariff effects.

Belief in commercial peace theory
An instinctive belief in commercial peace theory-that is, a belief that trade between countries lowers the risk of conflict-provides a potential explanation for the finding that individuals with security concerns report becoming more opposed to the trade war. To test this interpretation, table 3 repeats the previous analysis, including all the same independent variables (not reported to conserve space) in addition to a variable indicating degree of support for the commercial peace (cp_agree), as indicated by responses to the question: "Trade with China makes military conflict with China less likely." Columns 1-2 test whether agreement with this statement predicts initial levels of support for the trade war. Columns 3-6 test how support for commercial peace theory affects changing levels of support for tariffs. Columns 5 and 6 include interaction effects with coldwar and warconcern.
As hypothesized, respondents who express belief in commercial peace theory are significantly less likely to have supported the initial tariffs and to have become more opposed to the tariffs over time (H2a). Security concerns (coldwar and warconcern) both predict growing opposition to the tariffs even when including cp_agree (column 4). Interaction effects in columns 5 and 6 provide support for H2b: Respondents who identify a Cold War and express belief in the commercial peace have become much more opposed to tariffs, but those who identify a Cold War and do not believe in Note: All bars correspond to 95 percent confidence intervals for marginal effects from logit models. Black bars correspond to self-reported initial support for the 2018 tariffs. Gray bars correspond to self-reported change in support for the 2018 tariffs in 2021. Both outcome variables are transformed to binary form, with "1" corresponding to more support and "0" to less support. the commercial peace have not significantly changed their views (column 5). Similarly, respondents who perceive a higher risk of war with China and a belief in the commercial peace are likely to have become less supportive of tariffs, while respondents with fears of war who do not believe in the commercial peace are likely to have increased support for tariffs, perhaps reflecting a "security externalities" belief within this population (column 6). 82

Commercial peace cue treatment effects
This article hypotheses that individuals concerned about security will be more likely to have already internalized commercial peace theory. To test this, the MTurk survey randomly assigned respondents to read a commercial peace cue, as described previously. All else equal, this experimental treatment should lead to greater belief in the commercial peace and subsequently less support for the trade war tariffs (H3a). In other words, in a lab setting it should be possible to make survey respondents more supportive of the commercial peace. But in the highly salient US-China trade war case, certain respondents should be more susceptible to this information cue than others: In particular, respondents with security concerns may have already considered commercial peace theory independently, leading to a greater baseline (pretreatment) belief in the commercial peace (H3b) as well as less responsiveness to the information cue (H3c). Table 4 reports coefficients from ordered logit models analyzing the effect of the commercial peace treatment, interacted with different subgroups. To conserve space, table 4 only considers the treatment effects on the security-related variables; table A3 in the appendix repeats the analysis across the other explanatory variables. 83 As hypothesized (H3a), the commercial peace treatment has a strong Note: All columns report coefficients from ordered logit regressions. All explanatory variables from table 2 are included in each regression, but are not reported to conserve space. Standard errors in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.1.

82
Although coldwar and warconcern are related concepts and are correlated in the data (0.28 correlation coefficient), they measure different things. Cold War identification may reflect securitized messaging more directly; war concern-which is also positively correlated with unfavorable, nationalism, republican, and trump2020-may be driven in part by general hawkishness. 83 Although most subgroups in this additional analysis respond as expected to the commercial peace cue, two results differ. First, respondents who think that tariffs are good for the economy are not significantly affected by the commercial peace independent effect. Additionally, respondents who perceive that the United States and China have entered into a new Cold War are significantly more likely to agree with commercial peace theory, and they are also significantly less likely to be affected by the commercial peace cue (column 3). In slight contrast, respondents who perceive a higher likelihood of war have lower baseline belief in the commercial peace, controlling for other variables, though they are similarly unaffected by the information cue (column 4). This finding is partially in opposition to the hypotheses. It implies that certain security concerns may lead to consideration of commercial peace theory without necessarily leading to agreement with the theory; instead, types of security concerns may matter, with perceptions of great power competition leading to more liberal views and broader security fears potentially leading to more realist views (see further discussion in conclusion).
Most importantly for the analysis, these results imply that respondents with security concerns have already internalized a belief in the commercial peace. Is it reasonable to interpret this finding as a result of securitization of the relationship? The nationally representative IKP survey did not directly include analysis of the commercial peace or relevant treatments, but it does enable an analysis of whether those with more public interest and knowledge are more likely to support the removal of tariffs, providing indicative evidence. Table A4 in the appendix tests whether news awareness and knowledge of the US-China relationship leads to an emphasis on military over economic consequences of trade and support for the commercial peace. The two key variables for analysis are reported interest in public affairs and the ability to identify Xi Jinping as China's leader from a list of five Chinese names. The IKP survey did not include questions about war threat and Cold War identification, so the analysis proxies these concepts using responses that China's economic rise is a threat for military as opposed to economic reasons. Results confirm that respondents with a stated high degree of interest in public affairs and treatment, likely indicating a natural response to attitude incongruence. Second, Trump supporters are negatively affected by the cue, i.e., they are less likely to believe in the commercial peace when informed of academic support for the theory, which may indicate skepticism of expertise among this population.
respondents who can identify Xi Jinping are both significantly more likely to support tariff removal; these same variables are also associated with seeing China's economic rise in military terms and with high degrees of unfavorability toward China.

Conclusion
Conventional wisdom holds that increasing antagonism between the United States and China makes further trade decoupling likely and leads to greater support for protectionist trade measures. Existing academic literature supports the idea that xenophobia, perceptions of unfair trade, and security concerns are likely to lower support for trade. Yet the evidence presented in this article pushes back on this narrative: Americans seem to have considerable "liberal" motivations-when faced with security concerns they become instinctive commercial peace theorists. This article thus deepens our understanding individual trade preferences. The findings here have important implications for existing commercial peace research. The field of international relations has engaged in a long-running debate on whether trade is a source of peace or conflict. Liberal interdependence theorists argue that trade raises the opportunity cost of war while strengthening and incentivizing business elites to argue against war. Realists counter that interdependence increases vulnerabilities and thus the probability of war. 84 Four decades of quantitative analysis has not settled the debate, though the weight of findings appears to support the commercial peace. 85 Yet there remains very little research on the microfoundations of the commercial peace, and in particular the role that public opinion might play. Public opinion should be important, both as an explanation of the commercial peace, and also as an outcome: That is, if individuals believe that trade with potential adversaries lowers the risk of conflict, then there should be more support for trade with adversaries. Existing quantitative research findings relating peace and trade integration face an endogeneity challenge, and this article demonstrates that this challenge may be even greater than expected, especially for research finding a negative correlation between trade and peace: As tensions rise, publics may push for greater trade integration, which would bias findings toward the trade-to-conflict conclusion.
There is also room for additional research on how and when these security-based preferences affect policy making. A main channel for the commercial peace is often perceived to be the importance of elite and business lobbying, similar to research on trade policy more broadly. The lack of attention to public opinion may be due to perceptions that mass opinion does not matter for trade policy given collective action problems. 86 But public opinion does drive tariff policies in democracies, 87 and even those who think elite preferences matter more for trade (and broader foreign policy) preferences agree that public opinion matters more when issues become highly salient. 88 In the US-China trade war case, the high salience of the issue may give public opinion more influence. This article therefore has important implications for considering the future of the US-China trade relationship; with current trends pointing toward more securitization, public opinion may increasingly push toward trade "recoupling." Finally, further research is necessary to further interpret the findings and test their generalizability. Analyzing the US-China trade war has advantages in terms of salience and importance, but these advantages are shortcomings when thinking about the external validity of the article's conclusions. The analysis here does not sufficiently distinguish between general security concerns and specific concerns related to great power competition and conflict. Indeed, the security variable related to identification of a new Cold War tends to support the article's hypotheses more clearly, while results using the 84 Copeland (1996). 85 For quantitative findings supporting the commercial peace, see, e.g., Russet (1997, 1999) and Russett and Oneal (2001). 86 Gilligan (1997). 87 Kono (2008). 88 Jacobs and Page (2005). Bailey (2003) argues that public opinion on broad security issues had a strong effect on US trade policy during the Cold War. security variable measuring respondent perceptions of war likelihood are more nuanced. Whether broad security concerns outside of a salient great power conflict also lead to instinctive liberal support for trade liberalization as a way to decrease the likelihood of conflict remains untested; further research is necessary.
• Strongly supported • Moderately supported • Neutral • Moderately opposed • Strongly opposed • No opinion/do not remember Today, three years after the imposition of these tariffs on Chinese imports, have your views on them changed?
• I am now much less supportive of the tariffs • I am now somewhat less supportive of the tariffs • I have not changed my opinion • I am now somewhat more supportive of the tariffs • I am now much more supportive of the tariffs Do you think the US-China trade war has been good or bad for the American economy as a whole?
• Very bad for the American economy • Somewhat bad for the American economy • No effect on the American economy • Somewhat good for the American economy • Very good for the American economy Has the trade war has been good or bad for your personal economic conditions (employment, finances, consumption)?
• Very bad for me personally • Somewhat bad for me personally • No effect on me personally • Somewhat good for me personally • Very good for me personally The "commercial peace" treatment was randomly inserted here: Academic studies have consistently found that countries that trade more with each other are less likely to go to war. According to this "commercial peace" theory, trade serves as an important deterrent to military conflict.
Do you agree with the following statement: "Trade with China makes military conflict with China less likely." • Strongly disagree  Figure A1 breaks down tariff support by subpopulation, grouping moderate and high support and opposition to create binary categories for ease of interpretation. All five traditional explanations have some degree of explanatory power. Supporting a selfinterest interpretation, low-income and non-college-educated respondents are more likely to support imposing tariffs. Supporting a sociotropic interpretation, those who think that trade is generally good for the American economy are much more likely to oppose the tariffs. Supporting a xenophobia interpretation, those who hold broadly unfavorable views of China are much more likely to support the tariffs. Supporting a reciprocity interpretation, those who think that unfair competition explains China's trade surplus are considerably more likely to support tariffs. Finally, supporting a partisanship interpretation, Republican respondents have the largest sample gap in support of the tariffs. Table A2 looks at how individual-level variables predict respondents' reported "primary goal of trade with China," using results from the MTurk survey. As expected, coldwar and warconcern significantly predict identification of "tool for political pressure" or "containment" as the primary goals of bilateral trade with China, while they are significantly less likely to see economic growth and job creation as primary goals. Table A3 looks at the effect of the commercial peace cue across different populations, repeating the analysis in table 4 for a broader set of individual-level correlates. Table A4 tests whether news awareness and knowledge of the US-China relationship leads to an emphasis on military over economic consequences of trade and support for the commercial peace. The two key variables for analysis are reported interest in public affairs ( publicint) and the ability to identify Xi Jinping as China's leader from a list of five Chinese names (xi_knowledge). A total of 66 percent of the population reports being somewhat or very interested in politics and public affairs, and 47 percent of the population can identify Xi Jinping. Although there is a positive correlation between these two variables, more than half (52 percent) of the somewhat/very interested public interest respondents cannot successfully identify Xi Jinping. The IKP survey did not include questions about war threat and Cold War identification, so the analysis proxies these concepts using responses that China's economic rise is a threat for military as opposed to economic reasons (miloverecon). These three variables are all correlated in the data. 89 Table A4 tests whether xi_knowledge, publicint, and miloverecon predict opinions about tariff removal in expected directions. Given the high correlation between xi_knoweldge and publicint, the table repeats the baseline analysis (column 1) excluding each variable, respectively (columns 2 and 3). The table also repeats the analysis using the most similar MTurk data (columns 7-10) to show that the previous results hold when looking at the new dependent variable (conditional tariff removal), providing justification that the changed dependent variable does not drive results. To conserve space, the table only reports results for the key variables, though all regressions include the full set of explanatory variables available.

B. Additional analysis and robustness checks
The main results from table 6 (columns 1-6) confirm the hypotheses. Those with an interest in public affairs, and those who can identify Xi are significantly more likely to want to remove tariffs in case of concessions from China. Additionally, those who see China's economic threat in military rather than economic terms also support removing the tariffs in these cases. Using the Figure A1: Initial support for tariffs, by subpopulation. 89 Indeed, in regression analysis (not reported) the only significant explanatory variables for seeing China's economic rise as a military versus economic threat are the ability to identify Xi and a belief that trade is good for the US economy; controlling for these two, all other demographic and partisan characteristics are insignificantly correlated. This lends some initial support to the hypotheses.   Note: The dependent variable in each column is a binary variable taking value "1" when a respondent identifies that response as the "primary goal of the trade relationship with China," and "0" otherwise. Respondents could only choose one primary goal. "Any econ. goal" in column 7 takes value 1 if respondents selected economic growth, American jobs, or fair trade as the primary goal. All columns report marginal effects from logit regressions. Standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1. Note: All columns report coefficients from ordered logit regressions. The dependent variable in all regressions is the degree of reported agreement that trade reduces the risk of conflict. The key independent variable differs in each regression and is reported in the fourth row ("Key Ind. Var."). All regressions include the full set of explanatory variables as well as a dummy for the conflict treatment, not reported to conserve space. Standard errors in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.1. Note: All columns report marginal effects from logit regressions. All regressions include the full set of explanatory variables, excluded to conserve space. Standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1.