Folk Theories of Trade: Geopolitics, Sociotropic Concerns, and Consumer Safety*

Transcription

1 Folk Theories of Trade: Geopolitics, Sociotropic Concerns, and Consumer Safety* Frederick R. Chen Jon C.W. Pevehouse Ryan M. Powers November 15, 2017 Abstract Early work on mass trade preferences emphasized the role of material self-interest in shaping free trade attitudes. In recent years, however, more attention has focused on how community- and national-level expectations about the economic and geopolitical effects of international commerce shape support for trade openness. What are missing from these accounts, however, are explorations of the underlying political and economic logics employed by the public as they form preferences over trade policy. In this paper, we argue that the public relies on a number of folk theories of trade in order to form trade policy preferences and to draw inferences about how trade with particular types of states will affect the economic and geopolitical fortunes of the country as a whole. In some cases, these logics appear consistent with the theories of international politics and economics, while in other cases they decidedly do not. We explore these logics using a survey experiment in which we ask respondents about their support for trade with states that have different economic and political features. We vary economic size, wage levels, military size, regime type, and alliance status, and we show that these features have systematic effects on trade preferences. Further, we demonstrate that expectations about trade s effect on a number of important economic and political outcomes are strongly conditional on the political and economic profiles of potential trading partners. Our findings contribute to ongoing efforts to better understand the sources of mass trade preferences and the conditions under which the public can be induced to support trade openness. *We thank Stephen Chaudoin, Nikhar Gaikwad, Jonathan Renshon, Jessica Weeks, and the participants in the University of Wisconsin Madison Experimental Politics Workshop for helpful comments and suggestions. Additionally, we thank the approximately 2,500 mturkers who participated in our study. Ph.D. Student in Political Science at the University of Wisconsin Madison. frederick.chen@wisc.edu. Professor of Political Science at the University of Wisconsin Madison. jcpevehouse@wisc.edu. Post-doctoral Associate with the Leitner Program in International and Comparative Political Economy at the MacMillan Center for International and Area Studies at Yale University. ryan.m.powers@yale.edu. Web: 1

2 In the past two decades, a rich literature has emerged seeking to understand the origins of individual trade preferences. Much of the recent work in this area demonstrates that concerns about how the country as a whole will be affected by trade openness are key determinants of support for trade openness. In some accounts, concerns about the effect of trade on the national economy are primary (Mansfield and Mutz 2009). In some cases, education plays an important explanatory role (Hainmueller and Hiscox 2006), while in others, community- or identity-based interests predominate (Ahlquist, Clayton, and Levi 2014; Guisinger 2017). A recent study also suggests an addition to sociotropic or community-based concerns: that of national security (Carnegie and Gaikwad 2017). Recent contributions notwithstanding, however, there is a dearth of research on the influence of geopolitics on trade policy attitudes. We find this oversight puzzling for two reasons. First and with respect to the scholarly literature on this topic, the voluminous literature exploring the international political motivations for and implications of international trade provides a number of readily-available empirical implications for individual-level preferences. A large empirical literature suggests that international commerce follows the flag because allies can bolster each other economically and militarily. Moreover, trade is affected by domestic political dynamics that arise deferentially in democracies and autocracies. Second and with respect to the how trade politics are manifest in the real world, there has been renewed interest in using the geopolitical implications of trade openness to sell new trade agreements to the public over the last decade. For example, within the recent past, both Barack Obama and Mike Pence made the national security implications of the Trans-Pacific Partnership (TPP) central to their attempts to generate support for that agreement among the public. In particular, they singled out China as a threat to U.S. influence in Asia and framed TPP as a key instrument through which the U.S. could reassert its economic and political dominance, or at least forestall China s influence, in the region. Both the efforts of Obama and Pence have, of course, been overtaken by a series of unfortunate events, but these examples do suggest that political leaders with a diverse set of political and economic constituencies expect the geopolitical implications of international trade with particular kinds of states to be salient and convincing to the public. Of course, these geopolitical pro-trade arguments fly in the face of the more nationalistic even mercantilist rhetoric concerning trade policy seen today, especially in the United States (where much of the data on trade attitudes has been compiled). President Trump s continuous attempts to blame China or South Korea or Mexico for U.S. job losses has attempted to score political points with a growing U.S.-versus-them rhetoric. Indeed, recent work has found growing signs of these othering attitudes. Guisinger (2017), for 2

3 example, shows how racial attitudes may mediate sociotropic concerns. Mutz and Kim (2017) show that community-based concerns about trade implications could come at the cost of concerns for the international implications of trade restrictions. These protectionist arguments usually present a country (and workers) as the victim of a globalist trade agenda. In this paper, we provide a novel set of tests about the role of sociotropic and geopolitical concerns in shaping trade policy preferences using survey experiments. We argue that a number of folk theories of trade motivate the public to favor or oppose trade with particular kinds of states because of the geopolitical and economic implications of trade with those states. In contrast to much of the recent work on this topic, we avoid framing trade as having any particular implications for the United States. Instead, we directly ask American respondents about their level of support for encouraging trade with countries of varying economic and geopolitical profiles. We follow up with questions about how trade with the country in question will affect the economic and geopolitical fortunes of both the United States and the trading partner as well as respondents personal financial situation. Using causal mediation analysis, we estimate the relative explanatory power of particular mechanisms. We find that, consistent with recent work (Carnegie and Gaikwad 2017; Rho and Tomz 2017; Mansfield and Mutz 2009, 2013), trade preferences are driven by concerns about how trade openness affects the country as a whole. Importantly, however, we show that these expectations are systematically affected by the geopolitical and economic profile of potential trading partners. While our respondents nearly always expected to benefit the U.S. economy, this is especially the case when the trading partners have large economies, high per capita incomes, democratic institutions, or an alliance relationship with the U.S. Similarly, international trade is expected to benefit U.S. national security when it occurs with allies, democracies, and states with large militaries. We find little evidence to suggest that the public expects the pacifying effects of trade to be conditional on the economic or political features of a given trading partner. Instead, at least among our respondents, trade is broadly seen as having peace promoting effects regardless of the political or economic profile of the trading partner. We proceed as follows. The first two sections briefly review existing literature on trade policy preferences and the relationship between geopolitics and trade, respectively. The third section draws out a number of individual-level implications of both geopolitical and economic concerns for trade policy preferences, followed by our specification of hypotheses and proposed mechanisms. The fourth section lays out the details of the survey experimental design. The fifth section presents the empirical results. The last section discusses 3

4 the findings and concludes. 1 Past Work on Mass Trade Preferences There is now a vast literature on the individual determinants of trade policy preferences. 1 Perhaps the biggest debate in this literature is the relative importance of self-interest vs. other-regarding concerns. Early work on this topic focused on the the role of economic self-interest in shaping support for trade openness among the public. The logic here is similar to that of pocketbook theories of voting behavior (e.g., Fiorina 1978): Individuals are assumed to be rational and self-interested and, as such, favor public policies that will benefit them economically while eschewing policies that they expect to be economically costly. This intuition is generally paired with one or more distributional models of trade policy. Most commonly, scholars adopt either the factor endowments or specific factors model. In the former framework, where factors can move costlessly across sectors, a more open trade policy changes relative product prices, which in turn, changes relative factor prices by the Stopler-Samuelson theorem. Therefore, in the context of mass preferences for trade in the developed world, analysts posit that skilled labor (a relatively abundant factor) is more likely to favor more open economic policies, while unskilled labor (a relatively scarce factor) is not. In contrast to the factor-endowments framework, the specific-factors dynamics assume that factors of production are immobile between all or some industries in the domestic economy. At the individual level, adopting specific-factors model entails predicting that individuals employed in import-competing industries will oppose trade liberalization while employees in export industries will either be indifferent over trade policy choices or in favor of more open economic policies. Although Mayda and Rodrik (2005) provide evidence to support both the factor endowments and specific factors models, other research on individual-level trade policy preferences finds significant support for the factor-endowments framework but limited evidence of the dynamics as the specific-factors model predicts. Scheve and Slaughter (2001) find that, at least in the U.S. case, factor type dominates industry of employment in explaining individual trade policy preferences. 2 They interpret these findings as evidence that the cleavages predicted by the specific-factors model are not as salient as those predicted by the factor-endowments model 1 For a recent and comprehensive review of trade preferences literature, see Kuo and Naoi Scheve and Slaughter (2001) use education and income to measure individual skill. Nonetheless, Hainmueller and Hiscox (2006) argue that the effects of education on trade preferences are not directly related to the distributional concerns linked to job skills. Instead, the exposure to economic ideas and information in college classrooms plays an essential role in shaping individuals opinions on trade and globalization. 4

5 and other non-material determinants of trade preferences. 3 Several other studies also find additional support for the factor-endowments dynamics (e.g., Ardanaz, Murillo, and Pinto 2013; O Rourke and Sinnott 2001). New-New Trade Theory (NNTT) has revolutionized the trade literature in recent years due to the theory s ability to explain a number of previously puzzling empirical patterns in disaggregated trade data. The relevant insight of NNTT for models of individual-level trade preferences is that firms differ in their productivity within a given industry and these differences in productivity levels determine whether the firm can enter the export market and/or compete effectively against imported products. To the extent that producer concerns drive trade preferences, NNTT-based models of trade policy preferences expect employees of higher productivity firms to favor trade openness while employees of lower productivity firms should oppose it. In one of the first tests of the implications of NNTT-based models on public opinion, however, Rho and Tomz (2015) find no relationship between trade policy preferences and the perceived productivity or profitability of an individual s employer. 4 Hays, Ehrlich, and Peinhardt (2005) suggest that government programs that are designed to protect individuals and to offer side compensation for the losses from trade liberalization mitigate opposition to free trade. This study indirectly concurs with the economic self-interests explanations for trade policy preferences. On the other hand, Baker (2005) investigates public attitudes toward economic globalization from the perspective of consumer tastes. Since trade liberalization has disproportionate effects on the prices of exportable goods and imported products, heavy consumers of exportables tend to be more protectionist visa-vis heavy consumers of imports and imports-competing goods. Similarly, Naoi and Kume (2015) show that individuals primed to consider their interests of consumers are more likely to support trade openness than those primed to consider the interests of workers. Mansfield and Mutz (2009) argue that individual trade policy preferences are guided less by material selfinterests than by the perceptions of how a trade policy affects the U.S. economy at large. Partly because of economic ignorance (Rho and Tomz 2017), it is difficult for the general public to connect between government trade policies and personal economic well-being. Instead, people formulate their trade preferences based on collective, national-level outcomes. While framed differently, this is also an implication of work showing that education plays a key role in moderating protectionist trade policy preferences (Hainmueller and Hiscox 3 Scheve and Slaughter (2001) also find that, in addition to current factor incomes driving preferences as in standard models, asset values have an independent effect on public support for trade protectionism. Homeowners in regions with a high concentration of activities in comparative-disadvantage sectors generally oppose freer trade policies. 4 The results of Rho and Tomz (2015) also do not support the factor-endowments model or the specific-factor model either. 5

6 2006). Fordham and Kleinberg (2012) question the causal interpretation of this study, but other work does show that exposing respondents to the sociotropic implications of trade openness in an experimental context has similar effects to those observed by Mansfield and Mutz (Rho and Tomz 2017; Spilker, Bernauer, and Umaña 2016). We should note here that a large number of other factors may also play a considerable role in shaping public trade policy preferences. Factors that have been examined so far include issue framing (e.g., Ardanaz, Murillo, and Pinto 2013; Hiscox 2006), gender (e.g., Mansfield, Mutz, and Silver 2015; Guisinger 2017, 2015), ideology (e.g., Herrmann, Tetlock, and Diascro 2001; Rathbun 2014), geographic orientation (e.g., Kaltenthaler, Gelleny, and Ceccoli 2004), risk orientation (Ehrlich and Maestas 2010), and level of cognitive mobilization (e.g., Kaltenthaler, Gelleny, and Ceccoli 2004). In broad terms, however, we view these explanations as largely mediating or moderating the more fundamental explanations that we review above. 2 The Geopolitics of Trade While the factors that scholars have examined so far may well shape people s attitudes toward trade liberalization/protectionism, these explanations mostly ignore a large literature on trade policy that is concerned with the political-military context in which trade takes place. 5 There are strong reasons to believe that the international politics of trade will have important implications for individual-level trade policy preferences. As we explain below, geostrategic factors affect the patterns of trade in highly visible ways. At the same time, however, trade flows have important implications for (inter)national security. A voluminous literature argues that geopolitics has an important effect on trade between states. First, conflict onset and diplomatic relations between states drive commerce considerably. Pollins (1989) suggests that trade follows the flag, that is, trade patterns are correlated with the friendliness of political relations between states. Traders consider not only the price, quality, and quantity of goods and services but also the origins of these products and the political relationship between the importing and exporting nations. Since firms are sensitive to the disruption of trade as well as the possibility of hurting friends or aiding foes, the level of trade between two states is associated with their diplomatic relationship (cf. Keshk, Pollins, and Reuveny 2004). Some research demonstrates that wars and military conflicts suppress and reduce trade flows (e.g., Anderson and Carter 2001; Li and Sacko 2002), although Barbieri and Levy (1999) find that non-major 5 For a recent exception, see Carnegie and Gaikwad

7 power wars generally do not have a significant influence on trading relationships. Second, states take into account potential trading partners alliance status. States can utilize the gains from trade to develop and enhance their potential military power. Realists argue that political leaders are less likely to be driven by the possibility of absolute gains from trade than by concerns with relative gains. In consequence, trade creates security externalities. Following this logic, Gowa and Mansfield (1993) argue that tariff games between allies are systematically different from those played between (potential) adversaries because trade with an ally produces positive security externality while commerce with an adversary generates negative security externality. Therefore, open trade is more likely between allies than between adversaries (cf. Gowa 1994; Mansfield and Bronson 1997). Furthermore, Mansfield and colleagues suggest that the effect of alliance on trade is even more manifest in a bipolar international system because of less credible exit threat and clearer responsibilities for alliance stability, that the interaction between alliances and preferential trade agreements is fundamental to explaining commerce patterns, and that allies that include a major power trade considerably more than their counterparts. In addition, Morrow (1998) finds that trade flows are greater if states share more similar interests. Nonetheless, interests similarity here is conceptualized and operationalized based on states alliance portfolios. Third, regime type plays an important role in determining trade patterns between states. Limited government in democracies ensure that they are more likely to resist pressure to protect home industries and domestic firms from outside competition and that trading partners interests will be free from arbitrary treatment, which in turn, encourages trade flows (Morrow 1998). In a study of U.S. foreign trade, Dixon and Moon (1993) suggest that trade liberalization is more likely between states with similar political regimes. They find that the United States exports more to democracies than to other types of regime. Since the focus of this research is the US, this study at least indicates that a democratic state tends to trade more with another democratic regime. Concurring with Dixon and Moon (1993), Mansfield, Milner, and Rosendorff (2000) suggest that democratic pairs generally share much more open trade relations than mixed pairs (i.e., between a democracy and an autocracy). By emphasizing the role of protectionist legislatures in democracies, they argue that democratic dyads negotiate lower trade barriers than do nondemocratic dyads. 6 An equally large body of literature debates whether increasing trade influences interstate hostilities and 6 The conclusion Mansfield, Milner, and Rosendorff (2000) draw is that democratic pairs trade more than a pair consisting of a democracy and an autocracy, but not that democratic dyads universally trade more than other pairs (i.e., mixed pairs and authoritarian pairs). They further demonstrate that the level of trade liberalization between autocracies is contingent upon the preferences of the policy-makers as well as upon the differences between their political regimes. Thus, whether a democratic pair has more extensive trade than an autocratic pair is largely indeterminate. 7

8 conflicts. Existing work on trade-conflict relationship can be mainly categorized into three groups: (1) the liberal argument that trade (unconditionally) fosters peace; (2) the proposition that trade increases conflict; and (3) some realists belief that trade is not correlated with conflict. The logic of the trade-promotes-peace argument is straightforward: states emphasize their material interests and they try to avoid opportunity costs that are caused by international conflicts. International trade offers valuable gains, and thus, is beneficial for states. However, international conflict is risky and costly. When a conflict between states occur, combatants often raise trade tariffs or set unfavorable quotas against their adversaries. Thus, commercial ties provide a material incentive for states to avoid conflicts; increasing opportunity costs creates a sufficient deterrent to conflict when trade ties are extensive (Polechek 1980). Additionally, economic interdependence conveys resolve for states without resorting to military conflict. The credible signals associated with interdependence help obviate the need for costly military contests (Gartzke, Li, and Boehmer 2001). Given the effect of conflict on trade, scholars suggest that the pacifying effect of interdependence still holds even when accounting for the reciprocal effects of trade and conflict (e.g., Heger, Oneal, and Russett 2010). Furthermore, Mansfield and Pevehouse (2000) argue that preferential trading arrangements (PTAs) reduce interstate conflict among their members. However, some scholars claim that increasing trade is a potent source of stress and conflict between states. Since contact is a prerequisite for disputes, close interdependence raises the prospect of at least occasional conflict (Waltz 1979, 138). In the meantime, some realists maintain that trade is of low importance for policy-makers; factors with regard to high politics outweigh economic ones in explaining the decline in the use of force (e.g., Buzan 1984; Gilpin 1987). Other research demonstrates that trade interdependence may have a mixed or indeterminate outcome with regard to international conflict (e.g., Morrow 1999). Other scholars notice that the consequences of trade are not universally beneficial or detrimental, but can have countervailing effects. For instance, while it is probable that symmetric trade facilitates peace, asymmetric peace may stimulate conflict. Asymmetric trade dependence may serve as a new source of power for the less dependent state to exert political pressures to the more dependent party (Keohane and Nye 1977), which in turn, raises the tensions between trading partners (Gasiorowski 1986). In addition, Copeland (2015) argues that whether economic interdependence fosters peace is contingent upon states expectations of the future trade and investment environment; economic interdependence is able to contribute to peace only if a dependent state has positive expectations. Pevehouse (2004) makes an effort to reconcile the debate by reconsidering the nature of competing ar- 8

9 guments of interdependence theory and the measurement of international conflict. He suggests that trade interdependence may simultaneously increase the likelihood of international conflict and keep these conflicts from being endemic between trade partners. Moreover, Gartzke and Li (2003) show that alternative variable operationalization can explain why existing empirical research has contradictory findings. Trade dependence decreases the possibility of conflict onset, whereas trade share, which is negatively correlated with trade openness, reversely increases the likelihood of international conflict. Of course, these arguments about geopolitics and trade all function at the state or firm level. The mechanisms of these arguments rarely rely directly on popular attitudes. Yet, the same is true of the Heckscher- Ohlin/Stopler-Samuelson model, the Ricardo-Viner model, and the New-New Trade Theory. They are macroeconomic theories from which scholars have deduced individual-level preferences about trade policy, then subjected these predictions to empirical tests (Hiscox 2001). It is possible, however, that geopolitical theories of trade are too far removed from the every-day experiences of individual voters whether Great Britain is strengthened by its trade with the U.S. is possibly not as observable to the average voter as the economic implications stemming from opening or shrinking trade with Great Britain. In the next section, we suggest how intuitive understandings of the geopolitics likely shape mass attitudes towards trade. We also articulate how similar, intuitive understandings of economics and consumer safety also shape these attitudes. 3 Folk Theories of Trade: Communities at Home and Abroad Folk theories of trade are intuitions about the effects of trade held by the mass public and informed by political narratives about the economic, social, and political effects of trade. It is less important that individuals have personally experienced good or bad impacts from growing (or shrinking) trade. Rather, we argue that individuals hold logical intuitions about how trade influences their own life, the lives of those around them, the nation as a whole, as well as international trading partners. 7 These intuitions lead the public to favor trade with some states but not others, based on their understandings of how trade influences various economic and political factors. Three highly-salient narratives about trade relate to its effect on U.S. geostrategic interests, on the national economy, and on the safety of consumer goods. 7 In some ways this parallels Guisinger s discussion of the role of self, community, and nation in considering the effects of trade openness and forming trade policy preferences. See Guisinger (2017). 9

10 Geopolitical narratives about trade emphasize the ability of trade to increase national security by strengthening the U.S. economy and that of our allies. This was especially salient in the cold war, but also made a resurgence in the TPP debate, as we noted above. At the same time, other aspects of this narrative focus on the need for robust industry at home in the event of war. Recent calls in the U.S. of national security tariffs on steel and/or aluminum imports are one example of this. These narratives can be used to motivate both liberalization and protection, depending on the identity of the trading partner. We posit that three factors, as discussed above, are likely to play an important role in folk narratives about the geopolitical factors and trade. Hypothesis 1 (military size): All else equal, the public will be more likely to support trade with a small military than a large military. Hypothesis 2 (regime type): All else equal, the public will be more likely to support trade with a democracy than a non-democracy. Hypothesis 3 (alliance status): All else equal, the public will be more likely to support trade with an ally than a non-ally. Theories of geopolitics and trade expect that the treatment effects we observe here will be mediated through trade s effect on the level of expected conflict between trading partners and, potentially, the generation of other negative externalities like economic or military growth in states that might be strategically worrisome to the public (non-democracies, non-allies, or states with large militaries). Our guess is that folk theory versions of these arguments will not be mediated by expectations of conflict. Economic narratives about trade emphasize the adjustment costs and unfair competition from abroad. Outsourcing and off shoring are highly salient in the minds of the public (Mansfield and Mutz 2013). This leads to expectations that trade with low wage states will be bad for the economy. Such narratives are especially salient in the wake of negative macroeconomic shocks (Powers 2017; Mansfield, Mutz, and Brackbill 2016). They also help drive nostalgia about legacy industries. On the flip side, these narratives can be used to motivate support for trade by emphasizing the need for the U.S. to expand access to export markets, especially when the trading partner has a large economy. Generally, the economic profile of the state dominates in these narratives. 10

11 Given the dominance of these economic narratives in current trade debates, plus the robustness of sociotropic theories of trade policy preference formation, we expect support for trade openness to depend crucially on the economic features of a potential trading partner being conducive to economic growth, but to also result in a minimal dislocation of American workers. As such, we predict that: Hypothesis 4 (economy size): All else equal, the public will be more likely to support trade with a large economy than a small economy. Hypothesis 5 (income level): All else equal, the public will be more likely to support trade with a highincome than a low-income economy. Additionally, we expect that the treatment effects observed in each of these conditions will be mediated through concerns about how trade with such partners will affect the U.S. economy. Consumer safety narratives emphasize the regulatory environment of the trading partner. Here the political and economic profile matters. For example, in 2007, China recalled millions of consumer goods sold in the U.S. and elsewhere around the world as a result of consumer safety concerns including. The recalls covered toys, pet food, and cosmetics and garnered significant media coverage and hearings in the U.S. Congress (Lipton and Barboza 2007). A Pew survey conducted in 2008 found that 81 percent of Americans had heard about the recalls of Chinese-made goods (Pew Research Center 2008). Thus, these concerns are highly salient in the minds of the public and linked to particular kinds of countries. Importantly, we do not test for a direct treatment effect of consumer safety narratives. It would be shocking to find anything but a strong, negative influence on trade preferences of product safety concerns. Rather, we attempt to discern whether concerns over product safety among the public vary across the political and economic features of potential trading partners. Several previous studies have tested whether geopolitical factors such as alliance and regime type determine public support for trade or for selecting potential partners of preferential trade agreements (e.g., Carnegie and Gaikwad 2017; Herrmann, Tetlock, and Diascro 2001). 8 This research have yielded understandings of how non-economic factors drive public policy preferences over trade or PTA. Nonetheless, we argue that they suffer from two drawbacks. First, both Carnegie and Gaikwad (2017) and Herrmann, Tetlock, and Diascro (2001) investigate the impact of geopolitics separately, overlooking other potential determinants (e.g., social, economic, and consumer safety factors) that may simultaneously shape popular attitudes toward 8 The central research question of Herrmann, Tetlock, and Diascro (2001) is how ideas guide Americans thinking about trade. 11

12 trade along with geopolitical aspects. Consequently, even if geopolitics does affect people s trade policy preferences as these scholars suggest, we know little about the relative importance of geopolitics vis-a-vis traditional economic considerations. Second, although both studies demonstrate that alliance and/or regime type matter, the underlying causal mechanisms behind the findings are not entirely clear. Therefore, to provide a more comprehensive understanding of the relationship between geopolitics and mass trade policy preferences, we are interested in isolating any effects of the geopolitical process as they compete with or compliment more traditional economic motivations for trade liberalization/protectionism as well as in exploring the causal mechanisms of why geopolitical factors may make a difference. The survey experiment we implement seeks to tap those traditional studies to investigate whether the factors they identify indeed shape individual trade policy attitudes and why this is the case. One final note concerning existing studies of trade preferences. In general, this research program has relied on framing experiments in which respondents are told directly about the implications of trade openness and then asked about their trade policy preferences. When the consequences are sufficiently dire, support for trade predictably declines. While this kind of work provides rich information about the particular appeals that may be salient to the public in attempting to generate support for trade openness, it is of less utility when seeking to understand the underlying structure of mass beliefs about the economic and political implications of trade openness. On the one hand, these kinds of experiments are useful and important since they tend to reflect the way in which public s learn about the implications of trade and other policy options in the real world. Political leaders, for example, frame policies as either beneficial or harmful to the country as a whole or the interests of particularly important constituencies in order to build support for their preferred policies. On the other hand, however, framing experiments, in general and particularly in the context of low-information issues like trade policy, provide respondents with information about the implications of particular policy choices that they may not be exposed to in real world policy debates. To the extent that this information affects the preferences of individuals it may be because they do not have strong or politically-relevant opinions on the issue at all or because of social desirability bias. 4 Research Design We use survey experiments and causal mediation analysis to investigate the relative importance of geopolitical concerns in the trade policy attitudes. 12

13 Economy size Introduction Income level Military size Regime type Trade Preferences DV Effects of trade Mechanisms Demographics Alliance status Treatments Figure 1: Experimental design Experimental design The manipulation The experimental design can be seen in Figure 1. We began the experiment by providing all respondents with a common introduction. The introduction read, There is much concern these days about international trade. We are going to describe a situation related to international trade policy that the U.S. could face in the future. For scientific validity, the situation is general and is not about a specific country in the news today. Some parts of the description may strike you as important, while other parts may seem unimportant. Please read the details carefully. After describing the situation, we will ask your opinion on a policy option. To verify that you have paid close attention, we may ask you to recall some details of the situation later in the survey. We then told respondents about a hypothetical scenario in which the U.S. government was about to set trade policy for a particular country. The description of the country randomly varied on five theoreticallyrelevant dimensions: 1) regime type, 2) military size, 3) alliance status, 4) size of the economy, and 5) per capita income. The text of the hypothetical scenario read: The U.S. government is deciding between two trade policies that will either encourage or discourage trade between the U.S. and another country. By encouraging trade, we mean that the policies would make it easier for U.S. goods to be sold in another country and for that country s goods to be sold in the U.S. By discouraging trade, we mean that the policies would make it harder for U.S. goods to be sold in another country and for that country s goods to be sold in the U.S. The country at which these policies would be targeted has the following features: The country s economy is among the [smallest/largest] in the world. The country s workers earn incomes that are among the [lowest/highest] in the world. 13

14 The country is [a democracy/not a democracy], which means it [holds/does not hold] free and fair elections on a regular basis. The country s military is among the [smallest/largest] in the world. The country [is/is not] an ally of the U.S. At first blush, the scenario with which we have presented respondents may not look too much like the way the U.S. generally makes trade policy. For example, U.S. commitments under the WTO s most-favored nation rule do not generally permit trade policies that single out a particular country for better or worse treatment. We think, however, that this scenario actually very much like the key calculation that policy makers must make when setting U.S. trade policy for three reasons. First, this scenario is reminiscent of the kind of policy choices that are faced by the U.S. International Trade Commission (ITC) in deciding whether to approve new and whether or not to extend existing temporary trade barriers. These barriers, which include antidumping and countervailing duties are often targeted at a small set of products imported from a given country or set of countries, but they can have broader effects on the level of trade between the U.S. and the targeted countries. The literature on this kind of administered protection shows further that these decisions are not purely technocratic, but are instead strongly influenced by the interests of members of Congress who choose to testify at ITC hearings, the interests of members of Congress on the ITC s oversight committee, and the broader level of support for trade openness among the public. Second, despite the U.S. s standing international trade policy commitments via the WTO, the Trump administration and President Trump, in particular, have proposed trade policy measures explicitly designed to encourage or discourage trade with particular countries. For example, Trump has proposed both an import tax of 45 percent on all goods imported from China (Haberman 2016) and called for a preferential trade agreement between the United States and the United Kingdom. Third and finally, we view the proposed U.S.-U.K. agreement as a special case of the more normal process of deciding whether or not to invite a particular country to join a broader preferential trade agreement. That is to say that because of the trade diverting effects of preferential trade agreements, including or excluding a given country in a particular PTA is, in effect, deciding whether U.S. trade policy will encourage or discourage trade with that country. The most prominent and recent example of this dynamic was the exclusion of China from the Trans-Pacific Partnership. U.S. policy makers very much viewed the TPP as a key mechanism through which the U.S. would encourage trade with TPP member states at the expense of trade with China. Still, to get a sense of how the public s trade opinion on average differs from their opinion on trade with the 14

15 particular hypothetical countries that we described we also assigned some respondents to a baseline condition in which we asked if the respondent was told that the U.S. was deciding between two trade policies that will either encourage or discourage trade between the U.S. and other countries around the world. Throughout each of the the manipulations, we avoided linking any of the features of the trading partner to any expected outcome. For example, we did not say or imply that the size of the trading partner s economy or its alliance status would be good or bad for the economy or national security of the U.S. We thus manipulated the profile of the trading partner, but not the implications of trading with such a country that already exist in the minds of the respondent. Measuring support for trade Following the manipulation, we asked respondents if the U.S. should adopt new trade policies to either encourage, discourage, or neither encourage nor discourage trade with country. We followed up this question by asking those respondents that responded encourage or discourage if wanted U.S. policy to encourage/discourage trade a lot or a little. We asked those that selected neither if they lean towards encouraging or lean towards discouraging trade. The resulting responses allow us to build a four-point scale of support for trade which ranges from Discourage trade a lot to Encourage trade a lot. Immediately following this question, we asked respondents to explain why they answered the way that they did in a few words. Potential mediators Above we reviewed the literature on individual-level trade preferences and the geopolitics of trade. We outlined a number of causal mechanisms draw from past theoretical and empirical work linking geopolitical concerns to trade policy preferences. We presently operationalize those mechanisms. Because trade is a dyadic process and has political and economic implications for both the importing country and the exporting country, we asked respondents about a variety of effects that trade might have on both the U.S. and the potential trading partner. This allows us to get a sense of whether the treatment effects are driven by the effect that respondents expect trade to have on the U.S. or on the other country. In particular, we asked about encouraging trade with the potential trading partners would affect: the national security of each country the national economy of each country 15

16 100% 80% % support trade 60% 40% Baseline democracy non democracy Regime type ally non ally Alliance status among largest among smallest Miltiary size Figure 2: Effect of political treatments on support for trade. Left most panel shows support for trade in the baseline condition in which respondents were asked only about support for trade in general. The middle and right two panels shows support for trade in each of the treatment conditions. the likelihood of military conflict between the U.S. and the other country the personal finances of the respondent the safety of consumer goods in the U.S. Potential for confounding resulting from treatments Some combinations of features might evoke particular countries in the mind of respondents. For example, while not perfectly accurate on all dimensions, the following description might evoke China or Russia in the minds of respondents: an economy that is among the largest in the world, per captia income among the lowest in the world, is not a democracy, has a large military, is not an ally of the U.S. As Dafoe and colleagues argue that this kind of priming can led to confounding because individuals might favor or oppose trade with China or Russia for reasons other than the stimulus to which they were exposed. To counter this threat to inference, we asked respondents if the treatment caused the them to think about any country in particular. The results are unchanged. 16

17 5 Results Effect of treatments on support for trade We present the results for the political treatments in Figure 2. The left-most panel of the figure displays support for trade in our baseline condition that asked respondents about encouraging or discouraging trade with other countries in general. Support for trade is high among the sample, with just over 82 percent of respondents calling for the U.S. government to encourage trade with other countries a lot or a little. Importantly, as soon as respondents are given any specifics about the identity of the trading partner, support for trade drops substantially. The most dramatic treatment effects obtained in the alliance and regime type treatments. Respondents were much more supportive of trade with a country if they were told it was an ally of the U.S. About 68 percent of those in the ally condition said they wanted to encourage trade a lot or a little, while just over 50 percent of those in the not an ally condition said the same. On our six point scale, those in the ally condition were about.71 units more supportive of encouraging trade than those in the not an ally condition (t = 9.96, p <.000). Similar effects can be seen in the regime type treatment. Just about 70 percent of those in the democracy condition wanted to encourage trade a lot or a little, while about 53 percent of respondents in the not a democracy condition reported the same. On our six point scale, those in the democracy condition were about.80 units more supportive of encouraging trade than those in the not a democracy condition (t = 11.28, p <.000). For the military size treatment, the effects were substantively small (right-most panel of Figure 2). About 57 percent of respondents in the military among the largest treatment were wanted to encourage trade a lot or a little while about 60 percent of those in the military among the smallest treatment group said the same. On our six point scale, this amounted to a difference of.17 units (t = 2.39, p =.017). Given arguments about the potential security externalities generated by trade with other countries we might expect military size treatment to moderate the effect of regime type or alliance status. We find that this is the case for regime type, but not for alliance status. Among both democracies and allies, the size of the trading partner s military does not affect support for trade, as Figure 3 shows. However, having a large military drives down support for trade with non-democracies, relative to non-democracies with small militaries. This is not the case, however, for non-allies. 17

18 100% 80% % support trade 60% 40% democracynot democracy Regime type Large Military democracy not democracy Regime type Small military ally non ally Alliance status Large military ally non ally Alliance Status Small military Figure 3: Effect of alliance and regime type conditional on military size. Left two panels show that effect of democracy depends on military size, but this is not the case for alliance status. We present the results for the economic treatments in Figure 4. Support for trade does not appear to be conditional the size of the trading partner s economy. Of those in the economy among the largest condition, about 58 percent of respondents wanted to encourage trade a lot or a little, while about 60 percent of respondents in the economy among the smallest condition said the same. On our six point scale, the difference was.46 units (t = 1.17, p =.240). The effect of the wage levels treatment was rather large. Of those in the high income treatment, about 67 percent wanted to encourage trade a lot or a little, while in the low income treatment the same figure was only 52 percent. On our six-point scale, the resulting difference was.51 units (t = 7.11, p <.000). This suggests that the public is broadly concerned about the potential competition from low wage workers abroad. Further, these concerns depend on the size of the economy in question, as Figure 5 shows. Among large economies, the low wage treatment generated a difference of about 19 percentage points, while in the small economy treatment, the effect was only about 10 percentage points. Importantly, however, this effect is driven not by reduced support for trade with large economies that pay low wages, but by modest increases in support for trade with high-wage, large-economies. While difference of means tests tell us a good deal about the direction and magnitude of the treatment effects, the estimated effects may be confounded or imprecise since we have ignored the other treatments in each of the comparisons. It is worth considering if and how the results might change if we control for the effects of the other treatment conditions and/or a variety of pre-treatment demographic characteristics. We do this with a series of linear regression models that are presented in Table 1. The estimated effects are largely consistent with our initial analysis, but are unsurprisingly more precisely estimated. Indeed, we see that the 18

19 100% 80% % support trade 60% 40% Baseline among largest among smallest Economy size among the highest among the lowest Income levels Figure 4: Effect of economic treatments on support for trade. Left most panel shows support for trade in the baseline condition in which respondents were asked only about support for trade in general. The middle and right two panels shows support for trade in each of the treatment conditions. 100% 80% % support trade 60% 40% among highest among lowest Income levels Large economy among highest among lowest Income levels Small economy Figure 5: Effect of wage levels by economy size. Wage level effects are moderated by economy size. Notably, this effect is driven by increased support for trade in large economies with high wages, rather than by reduced support for trade with large economies with low wages. 19

20 level of income, size of the economy, regime type, and alliance status all moved support for trade openness. In contrast, military size did not seem to have any independent effect on support for trade; the estimated effect size was much smaller relative to the other treatments and in comparison with the standard errors. Recall that we asked respondents if the treatments reminded them of any particular country. About 28 percent of respondents responded in the affirmative. Of those respondents, about 42 percent said they thought of China, about 10 percent said they thought of North Korea, and about 8 percent said they thought of Russia. As model 3 in Table 1 shows, controlling for this potential source of confounding has almost no effect on the magnitude of the estimated treatment effects. The pattern of results suggests that the public has rather nuanced views about which kinds of states the U.S. should encourage trade relations. The public appears to condition their support for trade for trade on the political and economic profiles of trading partners in ways consistent with folk theories of trade. Causal mechanisms The results above demonstrate that both the economic and political profiles of potential trading partners have important effects on support for trade among the public, but they tell us little about the particular consequences of trade that motivate respondents to either favor or oppose trade with particular types of countries. To investigate this question, we reminded respondents about the features of the treatment country and asked them to report how they thought encouraging trade between the U.S. and the country in question would affect a variety of economic and political outcomes. Geopolitical concerns We asked respondents about the effect of trade on U.S. national security. We plot the results in Figure 6. In general, respondents are not optimistic about trade s ability to beneift U.S. national security. Still, a number of important patterns emerge across treatments. Respondents expect trade with democracies and allies to be more beneficial to U.S. national security than trade with non-democracies and non-allies, suggesting that the public does view trade as having important security implications. Above, support for trade with countries with large militaries was not conditional on alliance status. In the case of respondent expectations about trade s effect on national security, however, the effect of military size is conditional on alliance status (Figure 7). For allied trading partners, about 43 percent of respondents said that trade with countries with 20