Trade Rage: Audience Costs and International Trade

Abstract

Politicians frequently issue public threats to manipulate tariffs but only sometimes follow through. This behavior theoretically ought to generate audience costs. We therefore test the validity of audience costs in trade war settings through a vignette-based survey experiment. The vignettes describe a hypothetical situation involving the U.S. and a second country (China, Canada, or unspecified) with whom the U.S. has a trade deficit. The president (Democrat, Republican, or unspecified) either maintains the status quo, threatens to impose tariffs and backs down, or threatens to impose tariffs and follows through. Our findings highlight differences between security and trade conflict when it comes to audience costs and presidential approval. While Americans sanction the president for issuing a threat to raise tariffs, they generally support backing down. Regression modeling and text analysis of a free response question from our surveys suggest this is because consumers are wary of paying the costs of tariffs.

Keywords

audience costs trade tariffs public opinion

On January 17, 2018, President Trump issued his first threat to sanction China over the alleged theft of intellectual property. Trump also lamented the U.S.-China trade deficit and expressed a desire to insulate American workers from Chinese imports. Five days later, the President followed through on his threat, imposing tariffs on all imported washing machines and solar panels, both of which the United States buys in large volumes from China. On September 7, 2018, the President threatened further tariffs on $267 billion of Chinese imports. Two weeks afterward, a 10 percent tariff on $200 billion worth of Chinese imports went into effect. The Trump administration concomitantly threatened to increase that tariff rate to 25 percent by January 1, 2019. But on December 1, the President announced a delay in the pending tariff increase until at least early March. By February 25th, he had extended the March deadline on an open-ended basis.

The president’s persistent use of public economic threats, and the apparent variation in the extent to which he has followed through or backed down on these threats, pose an important puzzle for international relations scholars broadly and the conventional theory of audience costs more specifically. On the one hand, a raft of theoretical and empirical evidence suggests that leaders pay approval costs for issuing public threats and subsequently backing down.¹ Scholars explain this effect as one comprising public intolerance of both belligerence and inconsistency, as different segments of the population disapprove of threats in general versus the perceived weakness that comes with backing away.² On the other hand, it is unclear whether audience costs ought to manifest outside of the security-focused cases on which the theory has mostly been tested or in instances where reputational considerations may be less salient to the public than policy outcomes.³ Perhaps the perceived stakes are higher in military conflict and crises than in economic affairs, for instance, driving greater concern about threatening military action or maintaining consistency between words and deeds.⁴ But this literature has yet to assess whether leaders should face similar approval costs to those found in military conflict when their threats and subsequent actions are confined to tariff policy.⁵

We therefore address the following question: Does the logic of audience costs extend to tariff policy? There are at least two compelling reasons why this might be the case. First, and as suggested above, leaders commonly issue public threats on economic matters, and especially trade.⁶ Second, these public threats engage concerns about the leader’s reputation both internationally and domestically. The conventional mechanism behind audience costs is that citizens want their leader to be perceived as tough and willing to follow through on their statements—hence the stress on consistency between word and deed. Consistency in turn conveys the credibility of the leader’s signal to international audiences. If economic affairs more broadly and tariff policy specifically are no different than traditional audience costs settings, the general parameters of the theory ought to apply. In fact, scholars have shown that threats to impose more generic economic sanctions do generate the predicted audience costs effects.⁷

However, if there is something distinct about tariff policy in particular, then audience costs may be a poor fit for explaining public attitudes towards leaders who threaten and/or impose tariffs. We argue that audience costs are unlikely to manifest in the context of presidential threats and actions on tariff policy. This is because tariff policy, unlike military intervention or more generic economic sanctions, carries well-defined economic consequences for domestic publics in the issuing country, and citizens on average tend to lose from higher tariffs. Under these conditions, we believe that citizens prioritize economic self-interest considerations over reputational concerns. Moreover, we argue that this is true regardless of which theoretical framework for economic self-interest is applied. Indeed, irrespective of whether skill-based accounts like Heckscher-Ohlin (H-O),⁸ industry-based theories like Ricardo-Viner (R-V),⁹ or consumer cost theories¹⁰ are applied, the theoretical prediction is identical: Americans should on average show an aversion to both threats to impose tariffs and subsequent tariff hikes. In other words, citizens should oppose presidential tariff threats, but support the president backing down from these threats.

One crucial caveat here is that citizens must be able to connect shifts in tariff policy to their economic wellbeing. Extant research shows this is often not the case.¹¹ However, because of the context in which we field our original survey experiments (namely the United States in January and April 2019), we anticipate that respondents should have been able to link tariffs with welfare. During this period, tariff policy was particularly salient, as the U.S. and China had been engaged in a costly trade war for just over one year, generating substantial public debate and economic reverberations. Indeed, Figure 1 shows that our experiments were deployed at moments of particular salience for trade-related issues. As a result of the high-information environment in which we conduct our research, we conclude that respondents can likely connect changes in tariff policy to their economic wellbeing. We therefore apply the well-established audience costs logic to the context of an international trade dispute. However, we hypothesize that citizens should on average be unsupportive of both tariff threats and subsequent actions to increase tariffs, as they bear in mind the economic costs of such policies.

Figure 1.

Salience of tariffs. This plot captures Google search trends for the word “tariffs” over the period 2016-2019.

In our experiments, we vary whether or not a hypothetical president threatens to impose tariffs and subsequently follows through. We also vary the party of the president (Democrat or Republican) and the target country (Canada or China). We find strong evidence to support our argument—Americans revise their opinions of the president downward when he issues threats to impose tariffs and when he follows through on said threats. In other words, Americans broadly disapprove of the president issuing trade threats in the first place, yet they are not likely to sanction him if he fails to implement the threatened tariffs. On the contrary, there may be a consistency cost to the president for going ahead with the tariffs. This suggests that a basic level of tariff-driven economic self-interest transcends the logic of audience costs.

To probe the drivers of this result, we examine several explanations derived from existing literature on public opinion towards trade. These include general costs to consumers and individual-specific levels of skill and industry of employment. We do not find evidence that low-skill workers or those employed in import-competing industries approve more strongly of tariff aggression than others. Therefore, we suggest that the relatively indiscriminate consumer costs associated with tariff hikes are behind our results.

We bolster these conclusions by analyzing qualitative responses to an open-ended survey question which asked participants about the factors they considered in assessing the president’s handling of the situation. Even though President Trump provided the motivation for our experiment, the qualitative responses suggest that our findings are not merely the result of respondents projecting his image onto our hypothetical president; only ten out of 1,415 respondents (0.007%) mention Trump at all. Rather, respondents express concern that tariffs are dangerous for both their personal economic wellbeing and that of the U.S. economy overall. Though respondents do sometimes worry about the president looking weak for backing down, consumer cost concerns appear to overwhelm these considerations.

The rest of the paper proceeds as follows. First, we develop our theory in the context of the literatures on audience costs and public opinion of trade. Second, we present our data and methods. Third, we discuss our findings and offer concluding remarks.

Audience Costs & Tariff Policy

Why might audience costs carry over from security to the economic realm? On the one hand, to the extent that the public cares about consistency between a leader’s words and deeds, there is reason to believe that the standard audience costs logic ought to apply. On the other hand, scholars of international political economy have offered strong evidence favoring economic self-interest as the microfoundation of public opinion toward matters related to trade. These factors indicate that respondents may care more about how trade policy affects economic outcomes than any inconsistency from leaders. However, this research has only recently begun testing whether there are electoral or approval costs associated with free trade or protectionist policies.¹² We therefore draw on insights from the audience costs and public opinion of trade literatures to explain the bases of support and opposition for trade threats. We argue that economic welfare considerations are likely to overwhelm concerns about inconsistency between threats and actions with respect to tariff policy.

Since Fearon’s (1994) seminal article, numerous scholars have explored the conditions under which citizens punish leaders who back down from publicly issued threats. While Tomz’s (2007) classic “repel invader” experiment provided the first empirical validation of the audience costs phenomenon, several follow-on studies have shown that the effect is not necessarily automatic or exclusive to democracies.¹³ For instance, the incidence of audience costs may hinge on how the leader justifies his decision to back down,¹⁴ whether the leader “backs into” a foreign policy crisis by promising not to intervene then taking action anyway,¹⁵ or whether the leader threatens action before “backing up” toward a less hawkish policy.¹⁶ This points to the importance of specifying the conditions under which we expect audience costs to emerge.

Recent research has helpfully disaggregated audience costs into two components: belligerence and inconsistency.¹⁷ This means that audience costs do not purely constitute a words-deeds mismatch; citizens differ on whether they care more about the president backing down from the threat or having issued it in the first place. Some (doves and liberals in particular, who typically display greater warmth toward foreign countries and international organizations) are generally wary of threats in the first place, generating belligerence costs. However, once a threat has been made, other citizens (hawks and conservatives especially, who tend to view foreign countries and international organizations less favorably) prefer that their leaders follow through on said threat.¹⁸ Otherwise, leaders look weak, sacrificing reputation in the public eye and incurring inconsistency costs for deviating from their promised course of action. As is now standard in the literature, we test for this double-barreled treatment effect in our experimental design.

Scholars have previously applied the audience costs logic to the economic realm, though these studies have concentrated on the politics of economic sanctions. This body of work suggests that commitment strategies affect the duration and effectiveness of sanctions,¹⁹ while backing down from threatened sanctions may carry public opinion costs.²⁰ Most relevant to our own research, Thomson (2016) employs a survey experiment to assess the domestic political consequences of threatening economic sanctions and subsequently reneging. She finds that such inconsistency does generate audience costs, but only in situations where the relevant crisis is deemed to threaten national security—which again suggests that context matters for identifying these effects.

Our research extends and refines this literature by looking specifically at the domestic politics of presidential threats to impose tariffs. The closest existing test comes from Chaudoin (2014), who examines whether the president suffers greater approval costs for pursuing trade policies that are inconsistent with existing trade agreements. He finds that audience costs only manifest for those respondents that lack strong preferences about trade policy. Otherwise, trade policy preferences mute any concerns about inconsistency.

We make similar predictions to Chaudoin, honing in on the intersection between trade policy preferences derived from economic self-interest and reputational concerns associated with inconsistency between a president’s words and deeds. However, our research design seeks to more closely replicate the types of vignette-based survey experiments that have been employed in the security domain by simulating the experience of consuming news about trade policy. Critically, Chaudoin’s experimental design does not include a public threat to impose tariffs on which the president can either follow through or back down. Yet the president’s issuance of a public threat is a crucial aspect of any audience costs experiment since it involves an explicit and observable commitment.²¹ Without a public threat and subsequent decision to renege or follow through, Chaudoin is testing for the effect of inconsistency between the status quo of an agreement (which may or may not have been signed by the sitting president) and protectionism as opposed to inconsistency between a president’s own words and deeds. We therefore build on his approach by examining whether promising to alter tariff policy and subsequently backing down produces similar public opinion shifts.

We identify two compelling reasons that audience costs might apply to tariff policy, derived from the aforementioned literatures. First, as our motivating example illustrates, leaders often make public threats on economic matters, and particularly on tariffs. Threats must be issued publicly for audience effects to manifest. Second, as traditional audience costs studies and the sanctions literature described above suggest, citizens should be concerned about consistency between words and deeds when it comes to economic policy, as backing down might make the leader look weak or threaten his domestic and/or international reputation. To the extent that leaders want to appear resolute and credible before both types of audiences and across issue areas, the logic of audience costs predicts that they should not issue threats on which they are unprepared to follow through.

However, there are also reasons to doubt that audience costs will manifest in this area. For one, citizens and foreign leaders often privilege policy outcomes over inconsistency concerns.²² For instance, President Barack Obama famously issued but did not back up his “red line” threat over Syria’s use of chemical weapons in 2013.²³ This ought to have greatly damaged his international reputation. Yet inconsistency here did not prevent Obama from subsequently negotiating two landmark international agreements on climate change mitigation and Iran’s nuclear program. Similarly, British Prime Minister Boris Johnson repeatedly threatened to force a no-deal Brexit before reaching a transition agreement with the European Union in 2019.²⁴ These examples suggest that leaders not only see value in threats as a negotiating or deterrence tactic, but also that backing down did not impede their ability to forge new commitments. This is because leaders recognize two things: first, that others’ perceptions of the leader’s resolve and credibility are not perfectly transferable across cases;²⁵ and second, that citizens often care about results over process.

We suggest that citizens should express greater concern for outcomes over inconsistency especially when it comes to economic rather than security issues. This is because the impact of government policy hits far closer to home. Indeed, in the case of tariffs, we contend that economic self-interest is likely to overshadow any concern about inconsistency between words and deeds. For starters, tariffs are different from broader economic sanctions because the former tend to be applied against countries with whom a state already has extensive trade ties, while the latter are typically levied on countries not deemed to be critical trade partners. This means that tariffs are more likely than sanctions to affect economic outcomes in the issuing state. Several recent studies have reinforced the fact that consumers and companies bear the brunt of tariffs,²⁶ which ought to engage their economic self-interest, and thus, their concern over policy outcomes.²⁷ To the extent that tariffs affect a broad range of goods and reduce consumers’ purchasing power, and to the degree that consumers are aware of this reality, we anticipate that self-interest will outmatch concerns stemming from an audience costs logic.

Additionally, the public opinion of trade literature offers abundant evidence that economic self-interest shapes perceptions of trade policy, highlighting the importance of both skill level and industry of employment as predictors of trade opinion.²⁸ Traditional economic models such as Heckscher-Ohlin (H-O, and its corollary, Stolper-Saumuelson) and Ricardo-Viner (R-V) stress the importance of factor mobility both within and across industries in determining citizens’ trade preferences.²⁹ This literature typically operationalizes skill level with measures of education and income. Those possessing the scarce factor or who work in import-competing industries should oppose free trade and be more supportive of protectionist policies such as tariffs.³⁰

Importantly, regardless of which of the aforementioned models of trade opinion are active in the contemporary American context, we would anticipate identical results when it comes to audience costs. Under H-O, the abundant factor in the U.S. is high-skilled labor, and this population should oppose tariff aggression; because there are more high- than low-skilled workers, this should outweigh any support from the low-skill contingent. Similarly, per R-V, only import-competing workers should approve of tariffs, and these workers are only a minority of total workers in the economy. Last, given that tariffs are typically paid by all consumers in the U.S., we might expect all respondents to anticipate or have experienced some loss of buying power and oppose tariffs as a result. Therefore, we would anticipate resistance to both tariff threats and actions in the current American context under any of these theoretical frameworks so long as respondents can connect their economic situations to trade policy. And to the extent that tariffs are relatively indiscriminate in inflicting economic costs, they may lead public opinion to converge on a less confrontational trade policy stance instead of yielding heterogeneous effects by hawkishness.³¹ In other words, our logic affirms belligerence costs arguments from the audience costs literature, but rejects inconsistency costs where tariffs are concerned.

Rho and Tomz (2017), however, show that Americans are generally unaware of how trade policy ought to affect their personal wellbeing. But after receiving a prime with information about the economic effects of trade policy, respondents can intuit which policies would improve their own economic situation. A related literature shows that when Americans can clearly identify trade as the source of their economic misfortune, they become more protectionist.³² This work highlights the importance of information—citizens need cues to link economic outcomes to trade policy.

We believe that respondents can do so because of when and where we field our surveys, namely the United States in January and April 2019. During this period, tariff policy was a particularly salient topic, as the trade war between the U.S. and China was entering its second year. Moreover, the costs of aggressive tariff policies were starting to affect Americans’ pocketbooks. For example, Trump’s first round of tariffs increased the price of washing machines by 20 percent.³³ Meanwhile, farmers were hit so hard by retaliatory Chinese tariffs on American agricultural goods that Trump initiated a $16 billion farm bailout program just a few months after our surveys were fielded.³⁴ Given the high salience and broad economic impact of tariffs during this period, we contend that respondents should be able to discern the effects of tariff policy on their economic wellbeing. As a result, Americans generally ought to punish the president for issuing threats to impose tariffs, and they should not punish him for failing to implement threatened trade barriers. This is because tariffs hurt the majority of Americans consumers, who otherwise on net tend to benefit from freer trade, and so our main hypothesis is as follows:

Hypothesis 1: Citizens should on average disapprove of the president issuing a tariff threat (belligerence cost) but approve of the president backing down from that threat (absence of inconsistency cost).

Had we tested for audience costs in a different time period or local context where tariff policy was less salient, we might expect different results. This is because in the absence of both economic shocks resulting from substantial changes to trade policy and widespread public debate on the topic, it might be more difficult for citizens to determine how tariff policy influences their economic fortunes. This could lead fewer respondents to have strong and defined policy preferences, making them more susceptible to audience effects in line with Chaudoin (2014). Conversely, under different circumstances, citizens might have fallen back on partisan or ideological priors for cues about how to assess the tariffs’ impact.³⁵ As such, we do not argue that economic self-interest explanations should always dominate other mechanisms when it comes to public opinion over tariff policies. Rather, scholars should think hard about the context in which they are looking for audience versus economic versus social explanations of public opinion trends.

Data & Methods

Using Amazon Mechanical Turk, we collected 700 survey responses in January 2019 and 715 additional responses in April 2019, which we pool into a combined sample of 1,415 observations.³⁶ Fielding two surveys provides confidence in the robustness of our results, and in the analyses presented below, we include fixed effects for each survey wave. Our experiment is a 3 × 3 × 3 factorial design in which respondents receive a randomly selected vignette from a total of twenty-four hypothetical scenarios that we modeled verbatim on Associated Press (AP) news releases about trade statistics.³⁷ This design is consistent with the setup of previous audience costs experiments, though we pivot from the “repel invader” scenario to one involving a trade deficit between the United States and another country. We replicate AP vignettes to make the content appear as legitimate as possible, leveraging the AP’s neutral language and just-the-facts style of reporting to simulate the experience of consuming news media.³⁸ To avoid feeding respondents unbalanced information about the potential distributional consequences of trade policy, we employ phrasing taken directly from the AP to offer competing views about whether trade deficits are a sign of economic strength or weakness.³⁹

Each vignette follows a standard pattern.⁴⁰ The first paragraph references a Bureau of Economic Analysis report on the trade deficit with the country in question during a previous month. The second paragraph mentions that top U.S. officials are meeting with their counterparts in the target country (Canada, China, or unspecified) for talks aimed at reducing the trade deficit. The third and fourth paragraphs are identical in all vignettes and discuss different views on trade deficits among politicians and economists. The final paragraph contains the president’s party (Republican, Democrat, or unspecified) and indicates whether the president has threatened to impose tariffs or declared that U.S. trade policy would not change. The number of respondents that received each treatment can be found in Table 1.⁴¹ The full set of vignettes can be found in Online Appendix 6, and a sample vignette that includes a Republican president threatening China appears in Figure 2. Additionally, the pure control vignette (unspecified target country and party of president; no threat issued) can be found in Figure 3. We acknowledge challenges in constructing a proper control condition given the prevalence of trade conflict at the time our survey was administered. Nonetheless, we take care to emphasize the hypothetical nature of our vignettes.

Table 1.

Size of Treatment Groups. For a More Detailed Breakdown, See Online Appendix Table A1.

	No Threat; No Action	Threat; No Action	Threat; Action	Sum
Democratic President	185	103	101	389
Republican President	178	103	97	378
China	173	98	99	370
Canada	188	105	95	388
Subtotal	724	409	392

Figure 2.

Sample treatment vignette. This particular scenario involves a Republican president threatening to impose tariffs on China over its trade deficit with the United States.

Figure 3.

Sample control vignette. This scenario discusses no specific country or political party and presents only aggregate deficit statistics. No threat is issued by the president.

Respondents in the “threat” treatment group are then taken to another screen where they are informed about the president’s subsequent actions (either following through or backing down). If the president pursued tariff action in line with the threatened behavior, they are told: “The president HAS followed through on implementing the threatened tariffs.” If the president backed down, they are told: “The president HAS NOT followed through on implementing the threatened tariffs.”⁴²

At this point, we ask respondents whether they approve of the president’s behavior. Our main dependent variable (approval) comes from the following survey question:

On a scale from 1 to 10, with 1 representing “strong disapproval” and 10 representing “strong approval,” do you approve or disapprove of the way the U.S. president handled this situation?

Beyond the treatment conditions, we collect data on a host of demographic covariates—balance tests and descriptive statistics for which appear in Online Appendix Tables A2 to A4. Covariates include income, gender, age, minority status, education, industry of employment, and employment status.⁴³ Consistent with the broader literature, we include both income and education as proxies for skill. We also gather information on respondents’ political preferences (their party identification and whether they are co-partisan with the president in question) and general foreign policy leanings. Party ID could be Democrat, Republican, or independent, which serves as the reference group in subsequent models. Respondents are coded as co-partisan with the president if they share the party of the president from the vignette. Meanwhile, we measure foreign policy dispositions by asking how much respondents trust other nations.⁴⁴ Last, we include a manipulation check to screen for whether respondents understood and accepted the treatment. Specifically, we ask respondents to report whether or not the president issued a tariff threat in the scenario they read. Those who failed the manipulation check were removed from the sample in our primary analyses, leaving a total of 1,124 observations.⁴⁵ As an additional screening measure to eliminate concerns about bots, we manually review and remove responses featuring content typical of machine respondents, such as text pulled from web searches separated by ellipses.⁴⁶ The complete survey questionnaire can be found in Online Appendix 6.

From the descriptive statistics, it is evident that the samples are not representative of the U.S. population. As with most MTurk respondent pools, education levels and income are relatively high and respondents are mainly white. Additionally, there are more liberals and Democrats in the sample than conservatives and Republicans. While the standard MTurk caveats apply given the non-representativeness of the subject pool, a number of recent articles suggest that MTurk samples are adequate for social science research. In fact, they tend to produce substantively similar results more quickly and cheaply relative to nationally representative surveys,⁴⁷ and there is compelling evidence that liberals and conservatives on MTurk exhibit similar psychological divisions to liberals and conservatives in the mass public.⁴⁸ As a result, we have no reason to doubt the external validity of our findings, but future research could test our theory on a more representative sample.

Results

We find that the president’s tariff threats and actions affect his approval in a manner consistent with our hypothesis. To start, we estimate bootstrapped average treatment effects.⁴⁹ Additionally, the non-bootstrapped mean approval in each treatment group can be found in Table 2.⁵⁰ We find evidence of significant belligerence costs, as the leader’s bootstrapped approval is 1.2 points lower on our ten-point approval scale after issuing a trade threat (p = 0.00). This represents a 12 percentage point cost to the president. The substantively large, negative effect of belligerence is striking and suggests that tariff threats are broadly unpopular among our respondents.

Table 2.

Mean Approval by Treatment Group.

Treatment	Mean Approval
No Threat; No Action	5.87
Threat; No Action	5.01
Threat; Action	4.67
Democratic President	5.47
Republican President	5.17
Unspecified Party	5.34
China	5.11
Canada	5.62
Unspecified Country	5.23

Meanwhile, on inconsistency, we find little evidence of approval costs from backing down after making trade threats. In fact, we find evidence of consistency costs, as the average bootstrapped effect of inconsistency on approval is statistically significant and positive (0.34 with p = 0.045). The bootstrapped effect of backing down corresponds to a 3.4 percentage point approval bump for the president. These results support our hypothesis, as respondents are generally disapproving of both trade threats and subsequent tariff action. However, on net, belligerence costs appear to outweigh consistency costs by roughly eight percentage points, lending some support for one portion of the audience costs perspective.

Notably, we find similar results for an alternate outcome measure in which we ask respondents to assess whether the president’s handling of the situation helped or hurt the reputation of the U.S. in the world.⁵¹ The bootstrapped treatment effect of belligerence is negative and highly statistically significant (−1.2 with p = 0.00), while the effect of inconsistency is positive and statistically insignificant (0.21 with p = 0.121). The consistency of our results across two outcome measures suggests that we have identified a stable effect of trade aggression on public attitudes, and our findings point to concerns about outcomes over process.

Next, to account for individual-level covariates, we model approval costs in a regression framework. The results can be found in Figure 4. Our findings are again consistent with the idea that respondents punish the president for belligerency on tariff threats, but there is no evidence of an associated inconsistency cost. Respondents remain supportive of backing down from threats to impose tariffs. The effect of issuing a trade threat on approval is −0.849 on a 10-point scale and statistically significant at conventional levels, while the effect of following through with tariff action is −0.388 (though this coefficient is only significant at the 0.1 level).⁵²

Figure 4.

Full model results. Model type is OLS. Wave fixed effects are included. Confidence intervals are plotted at the 0.9 level. Table results can be found in Online Appendix Table A6.

By contrast, our secondary treatment attributes—the party of the president and the identity of the target state—mostly do not affect the way that respondents think about tariffs. Changing the president’s party, for instance, does not seem to matter.⁵³ We find counterintuitive results when it comes to the identity of the target state of tariff hikes, where we might expect our subjects to prefer sanctioning adversaries rather than allies. Respondents are significantly more likely to approve of tariff threats and increases when they are directed at Canada, an ally, than they are to approve of those directed at China, an adversary. We speculate that respondents may be worried about Chinese retaliation, which is borne out in our qualitative findings below.

Among the covariates, party identification influences approval in a manner consistent with today’s partisan environment. Republicans strongly approve of tariffs threats and actions, while Democrats weakly disapprove. While this is out of step with the traditional positions taken by each party, wherein Democrats have historically been more anti-trade than Republicans, we find no such evidence. The current political climate offers one potential explanation. Recent polling data on trade indicates that Republicans have on average become more protectionist of late, though they are split based on whether or not they support Trump, with anti-Trump Republicans still backing free trade on par with the party’s longtime stance.⁵⁴ By contrast, the trust variable does not enter the model as statistically significant. This suggests that tariff disapproval is widespread regardless of whether respondents are open to cooperation with other countries.

In line with Brutger and Kertzer (2018), we include both income and education as proxies for skill level. Income has a negative and statistically significant relationship with presidential approval, potentially suggesting that more skilled (and thus better compensated) workers are less supportive of trade threats and tariff actions. Given that the U.S. is capital abundant and tariffs tend to insulate industries that intensively utilize the scarce factor (labor), this matches the H-O mechanism for trade opinion posited by other scholars. However, education fails to achieve significance in either model.⁵⁵ Additionally, unemployed workers are no more supportive of tariff threats and actions. We further unpack the viability of H-O, R-V, and consumer cost models of trade opinion on our sample in the next subsection.

We run a host of sensitivity analyses to bolster confidence in our main results. First, we replicate our primary analysis for the full sample of respondents including those that failed the manipulation check; this does not change the results (see Online Appendix Table A9). Second, we perform randomization inference to test the sharp null hypothesis of no effect, which indicates how unusual our results are compared to a large number of possible randomizations. This is a useful tool for surveys like ours that have relatively small samples.⁵⁶ The results of randomization inference with 1,000 simulations appear in Online Appendix Figures A3 and A4. In the threat treatment, (Online Appendix Figure A3), our ATEs fall at the extreme left tail of the distributions regardless of whether we include covariates (p = 0.00). For the action treatment (Online Appendix Figure A4), the results are not quite as strong, but still fall well into the left tail of the distribution (p = 0.037 without covariates and p = 0.049 with covariates). All in all, our core findings stand up to a more rigorous estimation procedure. With respect to belligerence, respondents clearly punish the president for issuing tariff threats. When it comes to inconsistency, we can decisively conclude that backing down from a threat to impose tariffs does not lead Americans to sanction the president; if anything, respondents reward the president for failing to follow through with the erection of tariff barriers.

Exploring Economic Explanations: Regression Analysis

While the initial results point to the importance of belligerence rather than inconsistency in approval of tariff threats, we now turn to the competing economic explanations laid out in our theory—H-O, R-V, and consumer costs. First, we add interaction terms between measures of skill (income and education) and our primary treatments (threat and action) in order to more closely test for the presence of the H-O mechanism.⁵⁷ When we visualize these interaction results in Figures 5 and 6, the plots collectively suggest that disapproval of tariff threats and actions is fairly uniform across different levels of education and income. In other words, tariffs are unpopular among most types of workers in our sample.

Figure 5.

Interaction plots (education). These figures show the effect of trade threats and actions on presidential approval conditional on the education levels of the respondent. (a) Education and threat. (b) Education and action.

Figure 6.

Interaction plots (income). These figures show the effect of trade threats and actions on presidential approval conditional on the income levels of the respondent. (a) Income and action. (b) Income and threat.

Next, in Figure 7, we add industry-level dummy variables to our main regression to check for R-V effects. Existing work suggests that R-V is potentially a more appropriate theoretical framework than H-O in the American context given that there is low factor mobility in the U.S.⁵⁸ This framework posits that opinion toward trade policy should be closely tied to the tariffs’ industry-specific effects. We focus in particular on low-skill and import-competing industries such as agriculture, mining, construction, and manufacturing (both durable and non-durable) where approval of tariffs would theoretically be strongest according to this model. Our analysis provides scant evidence of such a phenomenon, as none of the coefficients register statistical significance.⁵⁹

Figure 7.

Industry test results. Model type is OLS. Wave fixed effects are included. Confidence intervals are plotted at the 0.9 level. Table results can be found in Online Appendix Table A7.

Finally, for robustness, we test for the non-linear effect of education to probe possible socialization/sociotropic explanations for our findings. Some work suggests that moving from high school to college education has a unique effect on trade preferences, whether because respondents learn that trade increases aggregate welfare⁶⁰ or because the college-educated have a more cosmopolitan worldview.⁶¹ Therefore, we swap out the linear education term from our main models for a binary equal to 1 if a respondent has obtained a college degree and 0 otherwise. Results are nearly identical to our main models, as shown in Online Appendix Table A10—income remains negative and statistically significant, while neither the unemployed nor the college education terms achieve statistical significance. These findings cast some doubt on the validity of sociotropic and socialization perspectives for our sample.

These results are consistent with a story about economic self-interest generally, but only match one of the more specific existing explanations for public opinion toward trade policy (consumer costs). The H-O model expects heterogeneity in support for tariffs by skill level; there is at best limited evidence for this here. The R-V framework anticipates heterogeneity in support for tariffs by industry of employment, yet we do not observe such effects. The main takeaway is that tariffs are unpopular at most levels of education and income and in many industries—suggesting that citizens may be aware of the welfare-diminishing effects of tariffs, specifically those related to the increased costs borne by consumers under higher tariffs. We examine this possibility more systematically in the next subsection.

Exploring Economic Explanations: Text Analysis

To further disentangle how economic self-interest versus audience cost considerations factored into the way that respondents formed their trade views in our samples, we use evidence from a qualitative question that we ask at the tail end of both surveys. The question is as follows:

What considerations did you factor in when assessing the president’s handling of the trade-related situations?

We pool respondents’ written answers to the question across samples and conduct descriptive text analysis to test for systematic variation in responses across the main treatment conditions.⁶² After preprocessing the raw text, we estimate a series of structural topic models (STM), using the stm package in R to both tune the number of topics and estimate the parameters.⁶³ STM uses unsupervised learning to model the underlying text as semantically interpretable clusters of words called topics. The method allows us to combine the unsupervised discovery of latent semantic structure with our respondents’ treatment information to better understand how our subjects formed their expressed preferences. The left panel of Figure 8 captures the change in topical prevalence as we move from the no threat; no action to the threat; no action condition. The right panel indicates the same statistic as we move from the threat; no action to the threat; action condition. Online Appendix Table A13 contains representative responses that illustrate how we assign labels and substantive interpretations to the topics.

Figure 8.

Changes in topic prevalence, pooled partisan and country treatments.

The results in the left panel show that moving from no threat; no action to threat; no action heightens respondents’ sensitivity to empty threats, diminishes their concerns about the trade deficit, and leads them to cite the benefits of trade. Respondents’ dislike of threats is quite palpable in statements such as “Making threats to other nations is already questionable…” and “Slapping tariffs on other countries is the worst possible way to handle this situation.” Meanwhile, the right panel suggests that moving from threat; no action to threat; action leads concerns about empty threats to lose salience, while worries about the costs of tariffs for consumers and other countries’ potential retaliation become prominent. Respondents are worried not just about their ability “to buy foreign goods the way we used to,” but also the prospect of a spiraling economic conflict: “Trade wars are no good, the Chinese will charge companies more for their goods and that will be passed down to the consumer, no body wins.”; they fear “the impact on consumers in the US and on the manufacturing plants and jobs lost due to the trade war that could arise.”

These topline results not only point to the prominence of belligerence costs and distributive consequences, but also suggest that a portion of our respondents are thinking about the costs of the president issuing empty threats, in line with audience costs’ canonical inconsistency mechanisms. It could be the case that those most sensitive to audience costs concerns are also those with weak trade policy preferences, as Chaudoin (2014) suggests. However, we lack the data needed to test this explanation explicitly since we do not have information on our subjects’ general feelings about trade, as we wanted to avoid priming effects in our survey design. Alternatively, respondents may simply have weighed inconsistency concerns against economic self-interest and ultimately determined that the latter was more important.

To explore the robustness of these findings across our sample, we break the responses down further according to the partisan (Democrat or Republican president) and country (Canada or China) treatment arms. If economic concerns are truly driving the effects that we have observed thus far, then we would expect them to appear irrespective of partisan or country treatments. Similarly, if the statistical results are again masking underlying inconsistency concerns, our approach should pick this up as well. Our models and corresponding topic labels within the country and partisan treatment arms appear in Online Appendix Figures A5 and A6, while example responses for the depicted topics within each treatment can be found in Online Appendix Tables A14 and A15.

Restricting the underlying data to these subsamples largely confirms that economic self-interest, as proxied by the focus on costs to consumers, substantially drives our findings. In the top left panel of Online Appendix Figure A6, when the president is a Democrat, those in the threat; no action condition are unwilling to tolerate the threatened policies “because the taxes trickle down to the consumer who is barely making ends meet as it is.” These respondents believe that tariffs are economically consequential policies that should not “be treated lightly or played with back-and-forth like a hot potato.” Relative concern for trade deficits again drops away in response to the president issuing a threat. Further, we find that the same topics change in prevalence when the treatment moves from threat; no action to threat; action in the top right panel, with costs to consumers remaining prominent. Respondents understand that when the president enacts tariffs, “it harms us because the other countries will in turn put tariffs on the goods we export to them.” They become less concerned with trade deficits and instead focus on the potential harm that tariffs pose. Notably, even in our relatively liberal MTurk samples, respondents are willing to sanction a Democratic president for threatening and imposing tariffs as a result of the perceived economic costs. Also of note is that concerns about empty threats do not appear salient among those in the Democratic president treatment arm.

The bottom panels of Online Appendix Figure A6, in which the president is a Republican, are most plausibly where we might expect some Trump effect, even though we do not refer to the president by name in our vignettes. This is especially true given the emphasis that President Trump has placed on tariffs, frequently touting trade barriers against China and the European Union as signature policy successes. Given the overrepresentation of liberals in the MTurk subject pool, we might have anticipated an anti-Trump effect in the written responses. Yet the changes in topical prevalence across treatments do not lend themselves to this interpretation. Respondents in the threat; no action condition still care more about costs to consumers and empty threats, and less about action to fix the trade deficit, relative to those who receive the no threat; no action condition. One respondent who is employed in the soybean industry exhibited clear concern for the tariffs’ distributive consequences, attesting that being a farmer “has shown me a first hand account this past year of how terrible these tariffs can be for Americans.” In the threat; action condition, only diminishing concern for empty threats registers as a significant change in topical prevalence. Thus, as in Figure 8, some respondents remain concerned about the president looking weak for making false threats.

At the same time, we find little explicit evidence to suggest that respondents projected President Trump onto our hypothetical Republican leader. In fact, just ten of our 1,415 respondents mentioned the president by name in their open-ended answers. These findings are consistent with recent work on the role of elite and informational cuing in the trade domain, which shows that misinformation and threatening rhetoric do not move public opinion on this issue.⁶⁴ In fact, citizens are broadly pro-trade even when presented with anti-trade misinformation.⁶⁵

An examination of the target country treatments in Online Appendix Figure A5 similarly reveals two primary sets of concerns—one about taking tariff action against an ally and the other about the cost to consumers of trade conflict with China. In the threat; no action treatment for Canada, respondents are again attuned to the costs of tariffs to consumers and concerned about empty threats. These participants are worried about “whether tariffs are generally a good idea…whether this tariff would do more harm than good” and think that the president should “be more measured and say what they will do instead of just throwing words around.” For the threat; action treatment, respondents instead focus on the U.S.’s global image, worrying about “The image we give to the rest of the world. The need to stake actual stands after you threaten to do so.” and whether imposing tariffs “came off as too aggressive…arbitrarily slapping tariffs on something does little to help.” On the whole, these results again suggest that while respondents may be attentive to inconsistency costs, they assign greater weight to the consequences of harsh policies against an ally and the potential impact on trade relationships.

When it comes to China, however, the results reinforce the notion that American consumers primarily make economic arguments in opposition to tariffs. In the threat; no action condition, respondents fret about tariffs as “destructive,” “just another name for taxes,” and making “a certain group of purchases…more expensive.” They are also cognizant of how empty threats may hurt the U.S.—“The President threatened to do something then went back on his word, making the United States look weaker”; “The incoherence of this process does not look good.” For the threat; action condition, respondents continue to factor in “consumer demand for imported items,” how tariffs would make these goods “less enticing,” and how they would represent a “rising tax on consumers.” The persistence of the costs to consumers topic under both the threat and action conditions suggests that the tariffs’ distributive consequences are the main factor behind our findings.

We conclude that respondents’ overriding concerns about the positive aspects of trade and the tariffs’ negative impact on consumers, both of which are consistent with economic self-interest, overwhelm any audience costs considerations. We contend that this result manifests because of the information environment in which we field our study. Both of our surveys were administered at the height of U.S.-China trade tensions. Americans became more cognizant of the costs of tariffs as they came to bear on consumer prices.⁶⁶ Moreover, popular debate among politicians and in the media stirred more general interest in tariffs, as we show in Figure 1. Therefore, we believe that the high salience of trade policy during these periods allowed our respondents to ably connect tariffs to economic self-interest. We acknowledge that consumers are often unable to do so,⁶⁷ which might leave them without strong opinions on trade and therefore more susceptible to audience costs dynamics.⁶⁸ Therefore, we encourage scholars to contextualize public opinion research broadly and work on trade more specifically by discussing the relevant information environment.

Conclusion

In this paper, we apply the logic of audience costs to the realm of trade wars broadly and tariff policy more specifically. While we suggest that audience costs might apply in this domain because leaders commonly issue public threats to impose tariffs and because citizens ought to care about the reputational costs of inconsistency between words and deeds, we find that the distributional consequences of tariff policy overwhelm these concerns. Consumers push back on threats to impose tariffs and any subsequent tariff hikes because they recognize the economic consequences of protectionism. Regression analysis, structural topic modeling, and a battery of additional tests offer support for this argument.

We believe that this logic ought to apply to any situation where tariffs are salient enough that citizens can correctly deduce the economic welfare ramifications of changes to tariff policy. In information environments like the U.S. in early 2019, where tariffs were a major topic of public debate and where fluctuations to tariff levels have recently driven economic dislocations, we believe this to be a safe assumption. However, as existing research indicates, trade is not always salient nor can citizens always make the correct conclusions about the relationship between trade policy and welfare.⁶⁹ In low information environments, citizens might be more sensitive to inconsistency concerns, as their priors about which trade policy is best might be weaker.⁷⁰ Therefore, we suggest that whether audience costs concerns or economic factors win out is likely context-dependent.

This research contributes to several important strands of literature. First, we extend and revise the body of work that has applied the audience costs framework to economic domains such as sanctions⁷¹ and trade agreements⁷² by looking at its applicability to a new area—tariff policy. Second, our research engages debates about the extent to which citizens are aware of the economic consequences of trade policy, addressing the conditions under which they might be able to deduce the ramifications of policy shifts on economic welfare.⁷³

Our findings point to several potential avenues for future research. First, we suggest that scholars ought to continue testing whether audience costs dynamics operate in economic settings and/or under different configurations of threats and actions. For example, future work might examine whether respondents sanction the president for promising to pursue trade openness and subsequently backing down, how the state receiving the tariffs views the other leader’s trade aggression,⁷⁴ or when concerns about retaliation become most salient. Another option is to look at Americans’ perceptions of formal economic conflict in forums like the WTO and ICSID.⁷⁵ Additionally, future work should examine whether our results carry over to settings beyond the United States. We suggest that our core theoretical insights—namely that respondents disapprove of trade aggression in line with economic self-interest—should generalize when economic conflict is salient enough for respondents to properly deduce the implications of leaders’ economic policies. This includes both democratic and autocratic settings; indeed, the broader audience costs literature offers some evidence that autocratic leaders incur public opinion costs when they back down from public threats.⁷⁶

Last, our findings carry potential electoral and policy implications. Our survey suggests that only very small subsets of the population, including the least skilled workers, are willing to sanction the president for failing to erect threatened trade barriers. These findings support those from Snyder and Borghard (2011), who argue that the public cares more about policy substance than consistency, in conjunction with recent polling showing that more Americans are pro-trade than at any point in recent memory.⁷⁷ Given the volume of American politics literature pointing to the importance of opinion leadership and elite rhetoric, and the fact that both parties have shifted against free trade in the U.S., this is a surprising result.⁷⁸ Both major parties have moved in an anti-trade direction, yet our respondents are not following their lead. Perhaps then there is more room for pro-trade elements in American politics despite the recent elite convergence in the direction of protectionism—especially as Americans internalize the costs of higher tariffs.

Supplemental Material

Supplemental Material, sj-pdf-1-jcr-10.1177_0022002721994085 - Trade Rage: Audience Costs and International Trade

Supplemental Material, sj-pdf-1-jcr-10.1177_0022002721994085 for Trade Rage: Audience Costs and International Trade by Don Casler and Richard Clark in Journal of Conflict Resolution

Supplemental Material

Supplemental Material, sj-zip-1-jcr-10.1177_0022002721994085 - Trade Rage: Audience Costs and International Trade

Supplemental Material, sj-zip-1-jcr-10.1177_0022002721994085 for Trade Rage: Audience Costs and International Trade by Don Casler and Richard Clark in Journal of Conflict Resolution

Footnotes

Acknowledgments

We thank Allison Carnegie, Nikhar Gaikwad, Josh Kertzer, Erik Lin-Greenberg, Keren Yarhi-Milo, two anonymous reviewers, participants at the 2019 Annual Meetings of the International Studies Association (ISA) and the American Political Science Association (APSA), and members of the Columbia University IPE Working Group for their thoughtful comments. All remaining errors are our own.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research is supported by grants from the Columbia University Experimental Laboratory for Social Science (CELSS) and was approved under Columbia University IRB protocol AAAR9356.

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Achen

Christopher H.

Bartels

Larry M.

. 2016. Democracy for Realists: Why Elections Do Not Produce Responsive Government. Princeton, NJ: Princeton University Press.

Alt

James E.

Gilligan

Michael

. 1994. “The Political Economy of Trading States: Factor Specificity, Collective Action Problems and Domestic Political Institutions.” The Journal of Political Philosophy 2 (2): 163–92.

Amiti

Mary

Redding

Stephen J.

Weinstein

David E.

. 2019. “The Impact of the 2018 Tariffs on Prices and Welfare.” Journal of Economic Perspectives 33 (4): 187–210.

Autor

David

Dorn

David Allan

Hanson

Gordon

Majlesi

Kaveh

. 2017. “Importing Political Polarization? The Electoral Consequences of Rising Trade Exposure.” Working Paper, National Bureau of Economic Research. Accessed February 5, 2021. https://www.nber.org/papers/w22637.

Berinsky

Adam J.

Huber

Gregory A.

Lenz

Gabriel S.

. 2012. “Evaluating Online Labor Markets for Experimental Research: Amazon.com’s Mechanical Turk.” Political Analysis 20 (3): 351–68.

Blonigen

Bruce

. 2011. “Revisiting the Evidence on Trade Policy Preferences.” Review of International Economics 85 (1): 129–35.

Bowers

Jake

Panagopoulos

Costas

. 2011. “Fisher’s Randomization Mode of Statistical Inference: Then and Now.” Unpublished manuscript. Accessed February 5, 2021. http://www.jakebowers.org/PAPERS/BowPan-Fisher.pdf.

Brewer

Paul R.

2004. “Public Trust in (or Cynicism About) Other Nations across Time.” Political Behavior 26 (4): 317–41.

Brutger

Ryan

Kertzer

Joshua D.

. 2018. “A Dispositional Theory of Reputation Costs.” International Organization 72 (3): 693–724.

10.

Burns

Sarah

Stravers

Andrew

. 2020. “Obama, Congress, and Audience Costs: Shifting the Blame on the Red Line.” Political Science Quarterly 135 (1): 67–101.

11.

Cavallo

Alberto

Gopinath

Gita

Neiman

Brent

Tang

Jenny

. 2019. “Tariff Passthrough at the Border and at the Store: Evidence from US Trade Policy.” NBER Working Paper No. 26396. Chicago, IL: Becker Friedman Institute International Economics Initiative.

12.

Chaudoin

Stephen

. 2014. “Promises or Policies? An Experimental Analysis of International Agreements and Audience Reactions.” International Organization 68 (1): 235–56.

13.

Clifford

Scott

Jewell

Ryan M.

Waggoner

Philip D.

2015. “Are Samples Drawn from Mechanical Turk Valid for Research on Political Ideology?” Research and Politics 2 (4): 1–9.

14.

Crisman-Cox

Casey

Gibilisco

Michael

. 2019. “Estimating Signaling Games in International Relations: Problems and Solutions.” Political Science Research and Methods. doi:10.1017/psrm.2019.58.

15.

Dorussen

Han

Jongryn

. 2001. “Ending Economic Sanctions: Audience Costs and Rent-seeking as Commitment Strategies.” Journal of Conflict Resolution 45 (4): 395–426.

16.

Fearon

James D.

1994. “Domestic Political Audiences and the Escalation of International Disputes.” American Political Science Review 88 (3): 577–92. https://www-jstor-org.web.bisu.edu.cn/stable/2944796.

17.

Flaaen

Aaron

Pierce

Justin

. 2019. “Disentangling the Effects of the 2018-2019 Tariffs on a Globally Connected U.S. Manufacturing Sector.” Working paper. Accessed February 2, 2021. https://www.federalreserve.gov/econres/feds/files/2019086pap.pdf.

18.

Flynn

D. J.

Horiuchi

Yusaku

Zhang

Dong

. 2011. “Misinformation, Economic Threat, and Public Support for International Trade.” Unpublished manuscript. Accessed February 5, 2021. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3264674.

19.

Gerber

Alan S.

Green

Donald P.

. 2012. Field Experiments: Design, Analysis, and Interpretation. New York: Norton.

20.

Grieco

Joseph M.

1988. “Anarchy and the Limits of Cooperation: A Realist Critique of the Newest Liberal Institutionalism.” International Organization 42 (3): 485–507.

21.

Gries

Peter

. 2014. The Politics of American Foreign Policy: How Ideology Divides Liberals and Conservatives Over Foreign Affairs. Stanford, CA: Stanford University Press.

22.

Guisinger

Alexandra

. 2009. “Determining Trade Policy: Do Voters Hold Politicians Accountable?” International Organization 63 (3): 533–57.

23.

Guisinger

Alexandra

. 2017. American Opinion on Trade: Preferences without Politics. Oxford, UK: Oxford University Press.

24.

Guisinger

Alexandra

Saunders

Elizabeth N.

. 2017. “Mapping the Boundaries of Elite Cues: How Elites Shape Mass Opinion across International Issues.” International Studies Quarterly 61 (2): 425–41.

25.

Hainmueller

Jens

Hiscox

Michael J.

. 2006. “Learning to Love Globalization: Education and Individual Attitudes Toward International Trade.” International Organization 60 (1): 469–98.

26.

Hart

Robert A

Jr . 2000. “Democracy and the Successful Use of Economic Sanctions.” Political Research Quarterly 53 (2): 267–84.

27.

Harvey

Frank P.

Mitton

John

. 2016. Fighting for Credibility: US Reputation and International Politics. Toronto, Canada: University of Toronto Press.

28.

Herrmann

Richard K.

Tetlock

Philip E.

Diascro

Matthew N.

. 2001. “How Americans Think about Trade: Reconciling Conflicts among Money, Power, and Principles.” International Studies Quarterly 45 (2): 191–218.

29.

Hiscox

Michael J.

2002. “Commerce, Coalitions, and Factor Mobility: Evidence from Congressional Vote on Trade Legislation.” American Political Science Review 96 (3): 593–608.

30.

Huff

Conor

Tingley

Dustin

. 2015. ““Who Are These People?” Evaluating the Demographic Characteristics and Political Preferences of MTurk Survey Respondents.” Research and Politics 2 (3): 1–12.

31.

Ioffe

Julia

. 2016. “How Russia Saw the ‘Red Line’ Crisis.” Accessed February 5, 2021. https://www.theatlantic.com/international/archive/2016/03/russia-syria-red-line-obama-doctrine-goldberg/473319/.

32.

Jakel

Ina

Smolka

Marcel

. 2013. “Individual Attitudes Towards Trade: Stolper-Samuelson Revisited.” Open Economies Review 24 (4): 731–61.

33.

Jakel

Ina

Smolka

Marcel

. 2017. “Trade Policy Preferences and Factor Abundance.” Journal of International Economics 106 (C): 1–19.

34.

Kees

Jeremy

Berry

Christopher

Burton

Scot

Sheehan

Kim

. 2017. “An Analysis of Data Quality: Professional Panels, Student Subject Pools, and Amazon’s Mechanical Turk.” International Organization 85 (1): 141–55.

35.

Kertzer

Joshua D.

Brutger

Ryan

. 2016. “Decomposing Audience Costs: Bringing the Audience Back into Audience Cost Theory.” American Journal of Political Science 60 (1): 234–49.

36.

Krustev

Valentin L.

Morgan

Clifton T.

. 2011. “Ending Economic Coercion: Domestic Politics and International Bargaining.” Conflict Management and Peace Science 28 (4): 351–76.

37.

Lenz

Gabriel S.

2012. Follow the Leader? How Voters Respond to Politicians’ Policies and Performance. Chicago studies in American Politics Chicago. London, UK: The University of Chicago Press.

38.

Levendusky

Matthew S.

Horowitz

Michael C.

. 2012. “When Backing Down Is the Right Decision: Partisanship, New Information, and Audience Costs.” The Journal of Politics 74 (2): 323–38.

39.

Levy

Jack S.

McKoy

Michael K.

Poast

Paul

Wallace

Geoffrey P. R.

. 2015. “Backing Out or Backing In? Commitment and Consistency in Audience Costs Theory.” American Journal of Political Science 59 (4): 988–1001.

40.

Lin-Greenberg

Erik

. 2019. “Backing Up, Not Backing Down: Mitigating Audience Costs through Policy Substitution.” Journal of Peace Research 56 (3): 559–74.

41.

Mansfield

Edward D.

Mutz

Diana C.

. 2009. “Support for Free Trade: Self-interest, Sociotropic Politics, and Out-group Anxiety.” International Organization 63 (3): 425–57.

42.

Margalit

Yotam

. 2011. “Costly Jobs: Trade-related Layoffs, Government Compensation, and Voting in U.S. Elections.” American Political Science Review 105 (1): 166–88.

43.

Mason

Winter

Suri

Siddharth

. 2012. “Conducting Behavioral Research on Amazon’s Mechanical Turk.” Behavior Research Methods 44 (1): 1–23.

44.

Mayda

Anna Maria

Rodrik

Dani

. 2005. “Why Are Some People (and Countries) More Protectionist than Others?” European Economic Review 49 (6): 1393–430.

45.

O’Rourke

Kevin H.

Sinnott

Richard

. 2001. “The Determinants of Individual Trade Policy Preferences: International Survey Evidence.” Brrokings Trade Forum. Accessed February 5, 2021. https://www-jstor-org-443.web.bisu.edu.cn/stable/25063160.

46.

Paolacci

Gabriele

Chandler

Jesse

Ipeirotis

Panagiotis G.

. 2010. “Running Experiments on Amazon Mechanical Turk.” Judgment and Decision Making 5 (5): 411–19.

47.

Plouffe

Michael

Kuo

Jason

. 2018. “Do Electoral Campaigns Influence Public Support for Trade? Evidence from the 2016 US Presidential Election.” Unpublished manuscript. Accessed February 5, 2021. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3278778.

48.

Potter

Philip B. K.

Baum

Matthew A.

. 2014. “Looking for Audience Costs in All the Wrong Places: Electoral Institutions, Media Access, and Democratic Constraint.” The Journal of Politics 76 (1): 167–81.

49.

Rho

Sungmin

Tomz

Michael

. 2017. “Why Don’t Trade Preferences Reflect Economic Self-interest?” International Organization 71 (S1): S85–108.

50.

Roberts

Margaret E.

Stewart

Brandon M.

Tingley

Dustin

Lucas

Christopher

Leder-Luis

Jetson

Kushner

Shana

Albertson

Gadarian Bethany

Rand

David G.

. 2014. “Structural Topic Models for Open-ended Survey Responses.” American Journal of Political Science 58 (4): 1064–82.

51.

Scheve

Kenneth F.

Slaughter

Matthew J.

. 2001. “What Determines Individual Trade-policy Preferences?” Journal of International Economics 54 (2): 267–92.

52.

Snyder

Jack

Borghard

Erica D.

. 2011. “The Cost of Empty Threats: A Penny, Not a Pound.” American Political Science Review 105 (3): 437–56.

53.

Thomson

Catarina P.

2016. “Public Support for Economic and Military Coercion and Audience Costs.” The British Journal of Politics and International Relations 18 (2): 407–21.

54.

Tingley

Dustin

. 2017. “Rising Power on the Mind.” International Organization 71 (S1): S165–88.

55.

Tomz

Michael

. 2007. “Domestic Audience Costs in International Relations: An Experimental Approach.” International Organization 61 (4): 821–40.

56.

Weeks

Jessica L.

2008. “Autocratic Audience Costs: Regime Type and Signaling Resolve.” International Organization 62 (1): 35–64.

57.

Weiss

Jessica Chen

. 2013. “Authoritarian Signaling, Mass Audiences, and Nationalist Protest in China.” International Organization 67 (1): 1–35.

58.

Whang

Taehee

McLean

Elena V.

Kuberski

Douglas W.

. 2013. “Coercion, Information, and the Success of Sanction Threats.” The British Journal of Politics and International Relations 57 (1): 65–81.

59.

Zaller

John

. 1992. The Nature and Origins of Mass Opinion. Cambridge, UK: Cambridge University Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.09 MB

0.00 MB

2.48 MB