Public Support for TTIP in EU Countries: What Determines Trade Policy Preferences in a Salient Case?

Transcription

1 Public Support for TTIP in EU Countries: What Determines Trade Policy Preferences in a Salient Case? Paper prepared for presentation at the ECPR Graduate Student Conference, University of Tartu, July 2016 Nils D. Steiner Johannes Gutenberg-University Mainz Department of Political Science Mainz Germany steiner@politik.uni-mainz.de Web: Abstract: Research on attitudes towards international economic integration usually measures support for economic openness via survey questions on abstract preferences in contexts where policies on international economic integration are, arguably, of low salience in the public mind. The ongoing controversy over the Transatlantic Trade and Investment Partnership (TTIP) represents an opportunity to re-examine the correlates of individual-level preferences over international economic integration on a specific real-world case of relatively high salience. In this paper, I draw on three recent rounds of Eurobarometer surveys conducted in the 28 EU member states and previous approaches to explaining preferences over international economic integration to explore the correlates of preferences towards a free trade and investment agreement between the EU and the USA. I find that, at the individual level, attitudes towards TTIP are most strongly related to general attitudes towards economic globalization and the European Union as well as satisfaction with the functioning of (national) democracy. At the aggregate level, support for TTIP is unrelated to a number of a priori likely correlates such as general levels of economic openness and economic ties to the US. Suggestive evidence is obtained that the prevalence of post-materialist values in a society might to some extent account for cross-country variation in opposition to TTIP. Overall, the findings add to growing skepticism that attitudes towards economic openness can primarily be explained by economic self-interest and point to politically richer explanations that reserve a role not only for socio-cultural attitudes towards international openness, but also for the broader political processes surrounding the dispute over a particular trade policy. Acknowledgements: Previous versions of this work were presented at a research seminar at the Department of Political Science and at the Mainz Workshop on Foreign Direct Investment and Multinational Corporations, both at the Johannes Gutenberg University Mainz. I thank the participants for their helpful and constructive criticisms and suggestions.

2 1. Introduction Since July 2013 the EU and the USA are negotiating over a free trade and investment agreement named Transatlantic Trade and Investment Partnership (TTIP). With tariffs between the USA and the EU already down to a low level, the main focus of TTIP lies in the reduction of non-tariff barriers to international trade and investment (see Council of the European Union 2013). The broad reduction of market access restrictions and regulatory hurdles to international trade and investment amounts to a particular ambitious, deep form of international economic integration. Even the sheer sizes of the involved economies, together amounting to almost half of worldwide GDP, render TTIP a preferential free-trade agreement of extraordinary high scope. 1 As compared to the usual low public salience of international trade policies in the EU member countries, TTIP has gained a remarkable amount of attention in the media and public spheres of at least some of the EU member states. This public attention has brought stiff controversies and fierce opposition with regard to the negotiation process itself and over the substance of the potential future treaty. An international alliance named Stop TTIP collected over 3.25 million signatures against TTIP (and CETA) in an unofficial European Citizen s initiative from 7 October 2014 to 6 October 2015 (Stop TTIP 2015) and a large demonstration against TTIP with about 150,000 participants took place in the German capital on 10 October 2015 (FAZ 2015), for example. According to the most recent available Eurobarometer data from November 2015 (see below), a free trade and investment agreement between the EU and USA is still supported by a clear majority of Europeans, but opposition is on the rise and opponents outnumber TTIP supporters (already) in four of the EU member states. On the European side, the ratification of a TTIP will ultimately require approval by the European Parliament, the European Council and, in all likelihood, ratification within the member states in accordance with their national regulations, i.e. in most cases approval by the national parliaments. In some countries (e.g. the Netherlands), even a public referendum is not unlikely to follow after a TTIP has passed the national parliament. As the stakes involved in a potential future transatlantic free trade agreement are high and public opinion will matter in some way for the final outcome of the negotiations, it is important to know what explains individual attitudes towards TTIP. Yet despite the high relevance of this issue no previous study systematically studies the correlates of attitudes toward TTIP, to the best of my knowledge. 2 The present contribution represents an initial attempt towards addressing this research gap. Support for TTIP is, however, not only of interest in itself, but forms a very interesting case for the academic study of preferences towards international economic integration more generally. This is for a 1 For a general treatment of preferential free trade agreements, see Dür and Elsig (2015). 2 A partial exception is a working paper by Andreas Dür (2015) who fielded an original survey on TTIP in France, Germany and the United Kingdom in early Yet, Dür is not interested in the correlates of support for TTIP per se, but rather uses the case to study the role of interest group cues and substantial frames via survey experiments. He finds that interest group cues do not matter for support for TTIP, while substantial frames pointing to positive or negative effects of the trade agreement have an effect that is, moreover, moderated by self-reported prior information levels on TTIP and general attitude strength on the question of free trade vs. protectionism. 1

3 number of related reasons. First, there is by now a sizable literature on preferences towards international economic integration and specifically towards international trade, yet these studies usually measure preferences over economic integration through abstract questions in contexts where trade policy is, arguably, of rather low salience in the public mind (for evidence on this low salience with regard to the US, see Guisinger 2009). It is questionable whether such survey expressions are anchored in welldeveloped opinions on the issue (on this general non-attitude problem, see Converse 1964 and Zaller 1992) and, therefore, whether they matter for preferences in specific real-world controversies over trade policies. Against this background, it seems wise to re-examine the role of different approaches to the explanation of preferences toward international economic openness in a real-world case of relatively high salience. In this paper, I focus on three such approaches: (1.) the economic self-interest argument, (2.) socio-cultural attitudes towards openness and internationalization, and (3.) cues from and attitudes towards political elites. Do these approaches help us to explain support for TTIP as well? Are attitudes towards TTIP shaped by general preferences towards economic openness in the first place? A second related issue is that the usual focus on general preferences towards free trade leaves open the question of free trade with whom (see also Kuo and Naoi 2015: 102). In this regard, studying a specific preferential free-trade agreement is interesting, inter alia, in that it allows us to test theories of economic self-interest in a more focused way by drawing on specific predictions on which individuals (or countries) are likely to gain or lose from this particular deal. Such predictions provide us with additional empirical leverage for assessing the validity of theories of economic self-interest. A third advantage is that studying preferences for specific salient trade policies allows us to consider the impact of the broader political process surrounding it. It might be the case, for example, that individuals oppose a specific trade policy out of their distrust of those political elites who initiated the proposal. Again, such potential insights are lost if we only measure and study general preferences towards international trade. 3 In short, studying public preferences over specific salient and controversial trade policies means studying attitudes towards trade policy when and in a form in which they are most likely to matter for concrete political processes and outcomes. This study examines the correlates of support for TTIP via an observational design. I draw on three recent rounds of Eurobarometer surveys conducted in the 28 EU member states in November 2014 (EB 82.3), May 2015 (EB 83.3) and November 2015 (EB 84.3). These surveys contain a binary measure of approval 3 Similar arguments are raised in two studies on Costa Rica s referendum on the Central American-Dominican Republic Free Trade Agreement (Hicks et al. 2014; Urbatsch 2013). I build on these more recent attempts to study trade policy preferences in a more focused and disaggregated way. In this context, we should also mention Rho and Tomz (2015) approach of studying preferences towards protectionism for specific industries. Another related and clever idea is to study preferences towards hypothetical preferential trade agreement via conjoint experiments (Spilker et al. forthcoming; Strezhnev 2013). All these recent studies, and some more, confront the problem identified by Kuo and Naoi (2015: 102) in their recent state-of-the-art review on individual attitudes toward trade policy: [W]hile the menu of trade policy options that governments can adopt is complex in reality, the majority of survey instruments used have asked for respondents opinions about placing new limits on imports or increasing trade [ ] While these questions might allow scholars to solicit citizens gut-based, general reactions to trade, these survey instruments often diverge from the day-to-day context in which citizens think about trade policy. 2

4 of a free trade and investment agreement between the EU and the USA and a number of interesting measures that tap into attitudes likely to affect support for TTIP (even though not all of these measures are perfect for the present purpose). I combine these surveys with data on economic aggregates. The resulting dataset is analyzed by means of multilevel regression models. The analysis mostly focusses on the individual level factors correlated with support for TTIP in line with the main theoretical interest of this paper and in light of the limitations to studying country effects via multilevel models when the number of countries is modest, especially in non-linear models (e.g. Bryan/Jenkins 2016; Stegmueller 2013). I cautiously introduce some country-level predictors in supplementary analyses of a more exploratory nature. I find that, at the individual level, attitudes towards TTIP are most strongly related to general attitudes towards economic globalization and the European Union as well as satisfaction with the functioning of (national) democracy. At the aggregate level, support for TTIP is unrelated to a number of a priori likely correlates such as general levels of economic openness and economic ties to the US. Suggestive evidence is obtained that the prevalence of post-materialist values in a society might to some extent account for the amount of opposition towards TTIP causing a negative association between GDP per capita and approval of TTIP. Overall, the findings add to growing skepticism that attitudes towards economic openness can primarily be explained by economic self-interest and point to politically richer explanations that reserve a role not only for socio-cultural attitudes towards international openness, but also for the broader political processes surrounding the dispute over a particular trade policy. The remainder of this paper is organized as follows. Section two reviews the three main approaches to explaining preferences towards international economic integration: economic self-interest, socio-cultural attitudes towards internationalization and cues from and attitudes towards political elites. Section three applies these general approaches to the particular circumstances of TTIP in order to derive expectations on the correlates of support for TTIP on the individual and aggregate level. In doing so, I add considerations that are specific to the case of TTIP and cannot be meaningfully subsumed under the umbrella of one of the three main approaches. Section four describes the data and discusses estimation issues. Section five presents the results. A final section concludes. 2. Explaining preferences over international economic integration: Three main approaches Previous research on individual preferences towards international economic integration has, mostly focusing specifically on international trade policy, identified three major sets of explanatory factors: (1.) determinants related to economic self-interest, (2.) socio-cultural attitudes towards openness and international integration and (3.) cues from and attitudes towards political elites. 4 In this section, I briefly 4 This categorization is roughly in line with Hicks et al. s (2014: 106) three-fold classification into a bottom-up approach that focusses on economic self-interest, another bottom-up approach dealing with non-economic attitudes and a top-down approach suggesting that voters are often uncertain about their preferences or can be swayed by 3

5 review each of these three approaches which I will then apply to the particular circumstances of TTIP in section three in order to derive a set of hypotheses to be tested in the empirical part. One obvious first source for divergent preferences towards international economic integration is economic self-interest: Those who are likely to economically gain from economic openness should be in favor of more integration, those who are likely to economically lose should oppose further integration. A rich body of literature in mass international political economy draws on this basic argument. Typically focusing on preferences towards trade policy, these studies employ some theory of international trade in order to derive predictions on who is likely to materially gain from further trade integration, and should be in favor of free trade, and who is likely to lose, and should hold protectionist attitudes. The underlying models used to develop expectations on the distributive consequences of increased openness in this literature are mostly of the traditional Ricardo-Viner or Heckscher-Ohlin type. In the Ricardo-Viner world (i.e. the specific factors models), winners and losers are defined along sectoral lines with workers or capital-owners in comparatively advantaged industries, i.e. those industries producing at lower relative opportunity costs as compared to the production abroad, gaining from liberalization. From the Ricardo-Viner perspective, we would expect those who work in industries that are comparatively advantaged to be in favor of free trade; while those working in comparatively disadvantaged industries should tend to favor protection. In the Heckscher-Ohlin world (i.e. the factor endowment model), where factors of production are mobile between industries, the distributive cleavages divide different factors of production instead of sectors. According to the Stolper-Samuelson theorem that is based on this model, capital and/or skilled-labor profits from international economic integration in relatively capital and/or skill abundant countries (see: Rogowski 1989: chapter 1). In global comparison, in the developed countries high-skilled labor and capital are the relatively abundant factors, while low-skilled labor is relatively scarce. As high-skilled workers are the ones objectively expected to gain from an increase in economic openness in the developed countries, they should be more supportive of increased economic openness in general according to the logic of economic self-interest. The seminal contribution by Scheve and Slaughter (2001) tests the model implications of the sectoral and the factor endowment model against each other using data from the US. Scheve and Slaughter find that those with higher skill levels are less protectionist as predicted by the Heckscher-Ohlin model, while international competitiveness of the sector in which individuals work is not consequential in their model. A number of other studies confirm the basic pattern reported by Scheve and Slaughter that, in relatively rich developed countries, preferences are divided across factoral lines with higher educated individuals tending towards free trade as predicted by the Heckscher-Ohlin model (Hays et al. 2005; Mayda and Rodrik 2005; O Rourke and Sinnott 2001; Scheve and Slaughter 2006). Some of these studies also point to political elites who, because of their public position, resources, and information, have the capacity to influence public opinion. 4

6 the simultaneous effects of sector of employment consistent with the Ricardo-Viner model (e.g. Hays et al. 2005; Mayda and Rodrik 2005). In addition to the Heckscher-Ohlin and Ricardo-Viner perspective, a more recent contribution to this literature (Ardanaz et al. 2013) has drawn on the new new trade theory (see the seminal contribution: Melitz 2003) to rationalize a skill level-gap in support for economic openness that holds even in relatively skill-scarce countries. The underlying logic of this heterogeneous firms-model is that, from an intraindustry perspective, it will be highly productive firms that draw heavily on input from high-skilled labor that profit from increased export opportunities. This, in turn, drives up the demand and reward for skilled labor. The economic-self-interest argument has come under attack by researchers who argue that the differences in attitudes toward trade between individuals with higher and lower individual attainments are actually caused by differences in exposure to the idea of free trade being beneficial (Hainmueller and Hiscox 2006) and/or socio-cultural attitudes towards international openness (Mansfield and Mutz 2009) (see again below). This critique has, in turn, stimulated rejoinders defending the notion of material self-interest shaping trade policy preferences (Fordham and Kleinberg 2012). The question whether and how economic self-interest matters for trade policy preferences appears not definitely settled yet, despite rising skepticism (see also some further recent contributions: Hicks et al. 2014; Rho and Tomz 2015; Urbatsch 2013). Apart from economic self-interest effects along distributive divides on the individual level, so called sociotropic gains or losses that concern the nation as a whole could be of relevance. Mansfield and Mutz (2009) show that perceived effects of trade on the US economy are strongly correlated with support for free trade. While the idea that expected welfare gains or losses on the national level matter for trade policy preferences is plausible, it is unclear whether such a correlation reflects a causal effect of perceived sociotropic effects on trade preferences: individual survey responses on the perceived socio-tropic economic consequences of trade are likely to be endogenous to an individual s general stance towards free trade (Fordham and Kleinberg 2012). The role of expected aggregate welfare gains should thus be reconsidered through more objective measures of aggregate welfare effects. A potential second source of divergent preferences over economic openness are socio-cultural attitudes towards openness and international integration. Margalit (2012), for example, argues that individuals view economic integration as being part of a broader openness package. He shows that opposition towards economic globalization does not only reflect economic concerns, but also social and cultural ones, like fear of foreign influences and culture or concern over Americanization specifically, that individuals attach to economic globalization. This view is actually in line with findings from earlier studies which focused on the economic self-interest argument, but found that socio-cultural attitudes towards internationalism like patriotism and ethnic chauvinism (O Rourke and Sinnott 2001), nationalism (Hays et al. 2005) and cosmopolitanism (Mayda and Rodrik 2005) matter as well. The main idea of this line of thought is that individuals do not neatly distinguish between economic and socio-cultural dimensions of 5

7 openness and internationalization. This implies, in turn, that a strong attachment to the national community and an opposition to the erosion of national borders in the social, cultural and political realm can affect preferences towards internationalization on the economic dimension. 5 Moreover, if higher educated individuals differ from lower-skilled individuals systematically on socio-cultural attitudes towards openness and international integration with regard to such factors as ethnocentrism, nationalism and isolationism, then a skill-gap in trade policy preferences might arise without being caused by economic selfinterest (Mansfield and Mutz 2009). A third set of explanations of individual s preferences over international economic integration which I subsume under the umbrella cues from and attitudes towards political elites has only recently gained traction (see: Hicks et al. 2014; Jensen and Shin 2014; Naoi and Urata 2013; Urbatsch 2013). Under this rubric I summarize somewhat heterogeneous theoretical arguments that start from the common idea that trade policy preferences are not formed in a political vacuum and emphasize influences of the broader political processes surrounding the dispute over a particular trade policy. More specifically, trade policy preferences are thought to be powerfully influenced by elite cues. On the level of individual attitudes, orientations towards political elites and general political orientations that influence an individual s susceptibility to different elite messages and framing strategies are thus likely to be influential. Moreover, citizen s general ideological orientations might give them heuristic reasons, i.e. act as information short-cuts, to support or oppose specific trade policy proposals. A plethora of studies on party identification, for example, has shown that individuals often adopt the issue positions advocated by their party (e.g. Achen and Bartels 2006). It is more than likely that similar effects of elite cues exist for a rather complex and abstract political issue as trade policy preferences; at least in circumstances were elites send clear messages. This conjecture is borne out by recent empirical evidence. In a study on Costa Rica s referendum on the Central American-Dominican Republic Free Trade Agreement (CAFTA-DR), for example, Hicks et al. (2014) find strong partisan divides. In their verdict, it was in the end the political cueing and framing done by the leading party, the PLN (Hicks et al. 2014: 115) that helped to tilt the balance in favor of the approval of CAFTA-DR in the referendum (see also Urbatsch 2013). Other studies have also documented robust effects of party attachment on citizens attitudes towards rather complex questions of international economic policy like preferences towards international bail-outs in the Eurozone crisis (Bechtel et al. 2014) or voting behavior in Iceland s second Icesafe referendum on sovereign debt resettlement (Curtis et al. 2014). Going beyond party attachments, research on voting behavior in EU referenda has identified satisfaction with the national government as an important predictor (e.g. Hobolt 2015). In addition to attitudes towards political elites, like party attachments or orientations towards the national government, general ideological orientations like placements on the left-right scale might also help citizens to form preferences on such specific and relatively complex issues (Bechtel et al. 2014: 839; Urbatsch 2013: 202). 5 Note that this is not to say that this bundling is necessarily irrational or wrong. It is quite plausibly that increased economic openness leads to increased exposure to foreign cultural influences. 6

8 3. The correlates of support for TTIP: Theoretical expectations In this section, I discuss expectations on the determinants of support for TTIP on the individual and aggregate level. Drawing on the distinction from the previous section, I consider expectations that might follow from an economic self-interest perspective (3.1), socio-cultural attitudes towards openness and international integration (3.2) and cues from and attitudes towards political elites (3.3). I complement these perspectives with a discussion of specific additional factors potentially relevant for TTIP (3.4) that cannot be classified as belonging to one of these three groups of explanations. 3.1 The economic self-interest perspective In the context of TTIP, preference divides related to material self-interest might enter at the individual level and the country level. I consider both perspectives in turn. Following the literature on trade policy preferences reviewed above, distributive divides on the individual level can potentially take place along sectoral lines or divide different factors of production. In case of TTIP and the EU countries, predictions on which individuals should gain (and which lose) are complicated by the fact that comparative advantages of sectors as compared to the US and factor endowments potentially differ between EU member states. Therefore, the individual divides might differ in a complex way across countries. Here, I focus on the general effects of potential divides along factoral lines 6 that are not obvious either. The typical prediction (and finding, see above) in studies on general preferences towards international economic openness is that, in the developed countries, higher-skilled workers tend to prefer more openness. Recall that this is argued to be the case, because in global comparison, in the developed countries high-skilled workers are the relatively abundant factors; and the relatively abundant factors are those predicted to gain from economic integration. Therefore, the reverse pattern should hold in developing countries, at least from the logic of the Heckscher-Ohlin model. Interestingly, for the relatively developed countries of the European Union the effects of skill-level with regard to specific trade agreements should also differ depending on the factor endowment of the partner countries. This nuance is lost in studies that examine general support for free trade, although such variation seems highly useful to test additional implications on how economic self-interest should shape preferences towards economic integration from the factor endowment perspective. With regard to the distributional consequences of a potential trade deal between the US and the EU, Felbermayr et al. (2014: 40f.) argue that the structure of factor endowments across the Atlantic is not too different. This leaves little scope for Stolper-Samuelson type effects. So, there are reasons to believe that distributional consequences from a TTIP should be limited, too. From the Heckscher-Ohlin perspective, we would therefore expect no meaningful differences in support for TTIP across individuals with different 6 Note right away that the survey data used below contain no sufficiently detailed information on sector of employment such that a detailed discussion on sectoral divides would proof rather moot in the context of this study in any case. 7

9 educational attainments and skill-levels. The new new trade theory would, however, expect relatively larger gains for higher skilled-individuals. Available economic estimates on the potential impacts of TTIP in the EU across different skill-levels point to largely similar gains across skill-levels. In Eggert et al. s (2015) estimates the income gains are even marginally higher for lower skilled workers; while they are marginally lower for skilled workers according to Francois et al. (2013: 71). While there is uncertainty as to the relative gains across workers with different skill-levels, there is at least no clear expectation of relatively more gains for high-skilled labor. From the self-interest perspective, no significant differences across people with different educational attainments and skill-levels are, therefore, to be expected in the case of TTIP. Interestingly, this contrasts clearly with theoretical expectations and empirical findings on general preferences towards trade (in relatively developed countries). With regard to the country-level it should first be noted that major studies on the economic impact of a transatlantic free trade and investment agreement point to meaningful overall aggregate welfare gains (Eggert et al. 2015; Felbermayr et al. 2015; Fontagné et al. 2013; Francois et al. 2013). Partly a natural consequence of the fact that the particular content of a potential future TTIP is as yet unclear, but also reflecting different methodological approaches to estimating the effects (for short discussions, see: Felbermayr and Larch 2013: 16f; Rodrik 2015), the particular estimates of potential long-run gains vary widely: While Francois et al. (2013) estimate a potential increase in real GDP per capita of 0.48% for the EU, Felbermayr et al. (2015) predict a long-run increase of 3.9%. Despite giving rise to overall welfare gains, the particular magnitude of these gains will likely differ across countries. The economic self-interest argument would hold that across-country variation in support for TTIP is associated with expected economic welfare gains at the country level. One could in principle, i.e. if such measures were available, capture this by measuring citizens subjective expectations on overall gains or losses for their country. Yet such self-assessment would introduce strong endogeneity concerns (see above). The approach taken in this contribution is to explore whether cross-country differences in support for TTIP are related to a number of factors that proxy for whether publics might expect gains from TTIP. One might expect that support is higher in countries with higher degrees of economic openness to begin with as these tend to be more dependent on and accustomed to international economic openness. In a similar vein, economic ties to the US specifically could matter. Publics in countries that have profited from FDI from the US in the past might expect to profit in the future as well, for example. Another approach is to draw on existing measuring of expected gains at the country-level from economic studies on the impact of TTIP (Aichele et al. 2014; Felbermayr et al. 2015). Taking the economic self-interest argument seriously, the empirical analysis below explores how support for TTIP at the country-level is related to such variables. 3.2 Socio-cultural attitudes towards openness and international integration With regard to socio-cultural attitudes towards openness and international integration in general, there is little reason to suggest any specific patterns originating from the particular features of TTIP. In 8

10 accordance with the previous studies discussed above on preferences toward economic openness in general, we would thus assume that those who are more strongly attached to their nation and who object to socio-cultural and political aspects of internationalization are also more likely to oppose TTIP. 3.3 Cues from and attitudes towards political elites Given the high levels of politicization of TTIP at least in some of the EU member states, there are ample possibilities for elite cues and, consequently, attitudes towards political elites to influence attitudes toward TTIP. I focus here on four explanatory factors on the individual level that seem of high relevance and are largely similarly applicable across all European countries. The first factor are general attitudes towards the EU. On the European side, TTIP is initiated and negotiated by EU institutions. Moreover, especially the European Commission is actively advocating TTIP, emphasizing its advantages and thereby trying to persuade the European public (see e.g. European Commission 2015a). If there is any single elite-level entity being perceived as being unequivocally in support of TTIP and similarly so across all member states it is the EU. Against this background, attitudes towards the EU should serve as a powerful force shaping individual s stances towards TTIP. Those who generally oppose the EU, should also be more likely to oppose TTIP given that it is an EU initiative that is clearly advocated by central EU actors. In contrast, those with more positive orientations towards the EU should from the outset be more sympathetic to the EU initiated TTIP and also more likely to buy into the framing strategies and persuasion efforts by mainstream EU elites. Second, a similar argument can be put forward with regard to national governments, although the connection is less clear than for the EU. The mandate for the TTIP negotiations from June 2013 stems from the European Council that authorized the Commission to open the negotiations in the first place. At least at this initial stage, most, if not all, national governments should be classified as being generally supportive of TTIP, but over time some have taken a more ambiguous public position (also consider that a number of governments have changed in the meantime). At least in cases where national governments publicly take a decidedly positive position towards TTIP, those who trust the national government should be more likely to support TTIP, according to the elite-cueing idea (also see: Sattler/Urpelainen 2012: 1112f.). On average, one might expect a positive relation between trust in the national government and support for TTIP. This general expectation is tested below. Yet it might hide interesting heterogeneity between countries. While data on the position of individual governments has not been assembled yet, I consider heterogeneity in the association between trust in the national government and support for TTIP below in exploratory analyses that examine the individual correlates of support for TTIP on a country specific basis. The third factor are general ideologies in the form of left-right orientations. These could matter in two related ways: For one thing and in line with the reasoning above, individuals might follow the messages of 9

11 their preferred parties which will differ according to the left-right position of parties. 7 Alternatively, individuals might themselves derive their positions from their general ideological orientations. These mechanisms are hard to disentangle empirically, but give similarly rise to the expectation of an inverted U- shaped relation between left-right-orientations and preferences over TTIP: Opposition to TTIP might be most pronounced on both extremes of the scale, but for different reasons. While parties and individuals on the left are likely concerned about the economic and environmental consequences of TTIP; those on the extreme right might oppose further integration out of opposition towards internationalization. In France, for example, the Front National has by now taken a clear position against TTIP (EurActiv 2015) alongside opposition from the Greens, from left-wing parts within the Socialist Party and left-wing civil society groups such as Attac. Similar dividing lines are apparent from the voting records on a motion for resolution dealing with the TTIP negotiations adopted by the European Parliament on July 8 th, 2015 (Votewatch EU 2015). One might expect a similar inverted U-shaped pattern on the individual level. 8 Finally, attitudes towards political elites and the political system more broadly might matter at a yet more fundamental level. To the extent that citizens do not strongly differentiate between what they perceive as political elites, a rejection of what are perceived to be elite-initiated policy proposals might be fueled by a general disaffection with the functioning of the political system. Although this specific reasoning seems (to my knowledge) largely neglected in previous research on the topic, it could be the case that opposition to TTIP acts as a welcome channel for the expression of latent feelings of political disaffection and antisystem attitudes. Anecdotal observations on the reasoning Dutch citizens gave for their rejections of the EU-Ukraine association agreement in the recent referendum, whose outcome is hard to make sense of on substantial grounds only, for example, supports the view that despite anti-eu orientations, broader dissatisfaction with the political system played a motivating role (The Guardian 2016). Accordingly, we might expect that, in addition to attitudes towards specific political elites, broader (dis-)satisfaction with the functioning of the political system might drive attitudes towards TTIP to some extent. 3.4 Specific additional factors So far the discussion has concentrated on structural explanations for support and opposition towards TTIP that draw on previous literature on citizen s attitudes towards trade openness. This approach allows us to re-examine existing theories on preferences toward economic integration on a salient case. Yet, as touched upon in the introduction, the planned TTIP is a preferential trade agreement of extraordinary scope and depth. The specifics of this agreement and the public controversy surrounding it give rise to a 7 Note that there is no measure in the Eurobarometer survey used below on preferences for specific parties. Leftright orientations is used as a reasonable proxy for partisan divides in this context. 8 As with attitudes towards the national government, there are good reasons to expect heterogeneity across countries in the effect of left-right orientations on support for TTIP. Country specific patterns are likely to result from the fact that party endorsements for TTIP differ across countries with, e.g., opposition to TTIP on the right being present in some party systems, but not in others. 10

12 set of specific additional considerations that are discussed in this section. TTIP proponents and opponents tend to highlight different aspects of the agreement, framing the discussion over TTIP in different ways. TTIP proponents tend to emphasize the potential economic benefits of a TTIP. TTIP opponents (more visibly in some European countries than in others) have focused on specific substantial concerns like the controversial investor-state dispute settlement (ISDS) procedure or fears over weakened consumer or environmental protection as well as on procedural criticisms relating to the allegedly low transparency of the negotiation process. Given this specific discussion over TTIP which differ from the usually largely exclusively economically framed discussions over trade agreements (to the extent that there are strong public controversies at all), it is debatable to what extent preferences over TTIP are related to general preferences for economic openness at all. Previous research suggests that they closely are (Dür 2015). Yet, it is important to reevaluate this finding in the present study. No matter whether the different viewpoints economic benefits on the one hand, concerns over consumer and environmental protection as well as national sovereignty on the other in the discussion over TTIP reflect real trade-offs or merely different framing strategies, they are likely to affect citizen s orientations towards TTIP. In the only existing study on attitudes towards TTIP (I know of), Dür (2015) shows via survey experiments that support for TTIP tends to be higher in the context of positive economic frames and lower if embedded in negative frames relating to ISDS. Against this background and with regard to citizens preexisting views, we might expect those who prioritize economic issues to be more supportive of TTIP. In contrast, those who prioritize consumer protection, concerns over the sovereignty of national democracy and protection of the environment, should be more likely to oppose TTIP. Such issue priorities might reflect underlying differences in basic value priorities and/or be a product of considerations that affect issue attention in the short term. Inglehart s (1977) classic distinction between materialist and post-materialist values seems like a good approximation to conceptualize differences in basic value priorities that should matter in this regard. Below, I consider both: issue priorities (for the economy and the environment, specifically) and post-materialist values as a measure of underlying basic value priorities. In addition, previous research on preferences towards economic integration with specific partner countries (Jensen/Lindstädt 2013; Spilker et al. forthcoming; Strezhnev 2013) shows that characteristics of the partner countries matter for preference formation as well. With regard to TTIP, opposition might be related to generally negative opinions of the US. Anti-Americans might be more skeptical with regard to the US, because they are critical of the US government s intentions during the negotiations, fearing an agreement that is disadvantageous for the EU; or due to simpler affective mechanisms. 11

13 4. Data and estimation This section introduces the data used in the empirical analysis and describes the empirical methodology. I first give a brief overview on the public opinion survey(s) used below and discuss the individual-level measures that are assembled from these survey data (section 4.1). Second (section 4.2), I describe the macro-level measures that are added to the individual-level data. Finally, I discuss the statistical models used to analyze the correlates of support for TTIP (section 4.3). 4.1 Individual-level measures For the empirical analysis, I utilize the data from the Standard Eurobarometer (EB) surveys 82.3 (November 2014), 83.3 (May 2015) and 84.3 (November 2015) (European Commission 2015b, 2015c, 2016) on the 28 EU member countries. These three surveys are pooled for two reasons: they contain homogenous measures on (almost) all constructs of interest and their combination maximizes precision in the empirical estimations. Each of the surveys used a multistage random sampling procedure within the respective countries. The universe consists of the population of the respective countries of age 15 or older excluding all residents who do not hold a nationality of one of the 28 EU member states. Interviews were carried out face-to-face. Overall, 83,340 completed interviews are part of the pooled dataset with sample sizes ranging from about 1,500 cases for Cyprus, Luxembourg and Malta and usually about 3,000 cases for the rest (with equal shares for the three EB waves). 9 The dependent variable, support for a free trade and investment agreement between the EU and the USA, is part of a broader battery on a number of EU policies or policy proposals. The exact question wording is as follows: What is your opinion on each of the following statements? Please tell me for each statement, whether you are for it or against it. [ ] A free trade and investment agreement between the EU and the USA. Answers are coded into the categories for (unweighted overall percentage: 58.9%), against (26.1%) and don t know (15.0%). I code this into a binary measure of support for TTIP with don t knows being coded as missing values and excluded from the analysis. Two further issues are worth discussing: First, the question does not ask verbatim about support for TTIP. It is obvious that the survey designers intended to measure support for TTIP, but opted for a more general wording in order to avoid non-responses of those not familiar with the term. It is likely that a clear majority of respondents were familiar with the ongoing negotiating of a TTIP and understood this question as asking about their 9 Note that the Eurobarometer contains separate regional samples in the case of Germany (West and East) and the UK (Great Britain and Northern Ireland) with an oversampling of the smaller regions. I combine these regional samples at the country level for the purposes of my analysis in order to add macro-level covariates that are measured at the country level (see below). Note further that the regional differences in support for TTIP (pooled over time) are modest in these cases: The (unweighted) fraction of those in favor of TTIP is 36.4% in Eastern Germany (of 1,356 non-missing observations) and 40.3% in Western Germany (n=2,571), and 75.1% in Great Britain (n=2,544) and 79.2% in Northern Ireland (n=678). 12

14 orientation towards this particular proposal. Others without any knowledge of TTIP presumably just expressed their general attitude towards such as a measure. While it could matter whether someone belongs to the first or the second group, the survey contains no item on familiarity with the TTIP negotiations one could use to take this systematically into account. Second, as said, the question is part of a battery on EU policies and policy proposals such that respondents were interviewed on their stances towards the economic and monetary union, a common foreign policy, further EU enlargement, a common defense and security policy, a common European policy on migration and a common energy policy as well (in randomized order). This entails the possibility that a general orientation to such EU policies spilled over to the TTIP question due to the questionnaire design to some extent. As to the independent variables related to the economic self-interest argument, I measure educational attainment through the age at which respondents finished their full-time education (or current age for those still in education). To remove any potential distorting influence of extreme outliers, I set all values below 13 (the 5 th percentile) to 13 and all values above 27 (the 95 th percentile) to 27. Additionally, the resulting measure was rescaled to have a range of one to ease comparison of coefficients. I also make use of an item on the respondent s current occupation distinguishing between rather higher-skilled and rather lower-skilled occupations and different inactive categories (see Table A 1 in the appendix for details). 10 Concerning socio-cultural attitudes towards openness and international integration, the survey unfortunately contains no direct measures of relevant latent constructs such as patriotism, nationalism or cosmopolitism. There are two proxies available: exclusive national identity (though only for EB 83.3 and 84.3) and trust in the United Nations. Exclusive national identity is based on a question that asks individuals whether they see themselves as exclusively French, German, etc. or as European, or both. This information is coded into a dummy variable differentiating those who only identify with their nation from all others. This indicator gives us information on an individual s attachment to her nation state, but could only be used in additional analyses excluding EB Trust in the United Nations, as an aspect of internationalization on the political dimension, is measured through a binary indicator on whether respondents tend to trust (1) or tend not to trust (0) the United Nations. As to cues from and attitudes towards political elites, I draw on an item recording the respondent s image of the EU thereby capturing a general orientation towards the EU as a whole. The exact question wording reads as follows: In general, does the EU conjure[s] up for you a very positive, fairly positive, neutral, fairly negative or very negative image (recoded to range from zero for very negative to one for very positive ). Trust in the national government comes from the same item battery as trust in the UN and is, accordingly, measured through a binary indicator on whether respondents tend to trust (1) or tend not to trust (0) the national government. In addition, I draw on self-reported positions on the left-right scale (1-10) as a measure of the general ideological orientation. To ease interpretation, this information is recoded 10 Unfortunately, the survey does not contain information on the industry of employment which would make it possible to test for sectoral divides. Information on income is also unavailable. 13

15 in two dummy variables separating those on the left (positions 1 to 3) and those in the center (positions 4 to 7) from those on the right (positions 8 to 10). 11 General satisfaction with the way democracy works in the respondent s country is measured on a four point scale ( not at all satisfied, not very satisfied, fairly satisfied and very satisfied ) that has been recoded to range from zero to one and introduced as a general measure of political (dis-)satisfaction. The specific additional factors that could be measured on the individual level 12 are operationalized in the following way: Issue priorities, or problem attentions, are measured with a question on the two most important issues facing the respective country at the moment. Respondents could choose maximally two items from the following categories: crime, economic situation, rising prices/inflation/cost of living, taxation, unemployment, terrorism, housing, government debt, immigration, health and social security, the education system, pensions and the environment, climate and energy issues. Problem attention directed at the environment is measured via a dummy variable marking respondents that picked the last category. Attention for the economic situation is considered to be high (=1) if respondents chose two of the three issues that deal with aspects of the economic situation, i.e. economic situation, rising prices/inflation/cost of living and unemployment, and moderate (=0.5) if they picked (only) one of these. For general attitudes towards economic openness, we can draw on agreement with the statement globalization is an opportunity for economic growth on a four-point scale. Again, this variable has been recoded to range from zero to one to allow for a rough comparison of coefficient sizes. In addition to these theoretically motivated variables, the statistical models control for gender (see e.g. Mansfield et al. 2015) and age in life years (divided by 10). 4.2 Macro-level data Beyond country-level differences that arise from individual level correlates, i.e. so called composition effects, differences in support for TTIP might also be caused by macro-level phenomena. Specifically, I introduce a number of measures of economic aggregates to test for the idea that support for TTIP might correlate with general levels of economic openness and specific estimates of economic gains from TTIP. In addition to that, I use aggregate information on individual attitudes that I would prefer to have on the individual level, but that are not included in the surveys. As to economic aggregates, I add the index of economic globalization from the KOF globalization index data (Dreher 2006) as a summary measure of a country s level of economic openness, and FDI instocks and outstocks (UNCTAD 2016a) as well as flows of international trade (World Bank 2016) as more 11 Additional analyses with a linear and squared term for left-right position included led to similar results as those presented below. 12 I return to measures for post-materialism and anti-americanism below for which only aggregate-level measures are available. 14

16 disaggregated measures of economic integration. From the World Development Indicators, I also take information on GDP per capita (in PPP) as well as GDP as a measure of a country s economic size. 13 Economic ties specifically to the US are measured via data on bilateral trade volumes (International Monetary Fund 2016) and inward and outward FDI stocks (UNCTAD 2016b). Note that information for 2015 is not yet available for any of these measures. In all cases, I made use of the most recent data which I added as country-level information (rather than on the country-time level). 14 This shouldn t be problematic, however, given that I am interested in differences across countries rather than time and as the patterns of cross-country differences in these measures are quite stable relative to short-run changes in these series. Country-level estimates on the effects of TTIP on real income are available from Aichele et al. (2014) and Felbermayr et al. (2015). 15 In addition to these economic data, I consider aggregate measures of attitudes that could not be measured via the EB survey data. As discussed in section 3, general attitudes towards the US and post-materialist values might affect individual attitudes towards TTIP. To allow for an initial exploration of these factors, I add external information from previous EBs: The share of anti-americans in European countries in 2006 as reported in Lawson and Hudson (2015); 16 and the number of post-materialists in 2008 as listed in European Commission (2008). 17 These variables need to be used cautiously as some shifts might have occurred in the meantime (more recent data are not available) and because these are aggregate measures of constructs that we want to ideally measure on the level of individuals. 4.3 Estimation The dataset resulting from this combination of EB survey data with macro-level measures is of a hierarchical multilevel structure. Overall, the dataset contains 83,340 individuals (level 1) which are nested in 84 country-survey/time combinations (level 2) which are themselves nested in 28 countries (level 3). 18 The proper analysis of such data requires multilevel models that contain random intercepts for the higher 13 Note that smaller countries tend to more dependent on and used to international economic openness. 14 The data from UNCTAD (2016a), the World Bank (2016) and the IMF (2016) are for In the case of Malta, some values for 2014 are not yet available from the WDI. In these cases, I took the most recent estimates available in the WDI (for 2013 and, for, trade openness, for 2011). The most recent KOF data refer to The data on bilateral FDI stocks (UNCTAD 2016b) are generally for 2012 though contained some missing values for this year (especially in cases of very low FDI volumes). In these cases, I took the most recent available information instead. 15 I make use of their preferred estimates without spill-over effects, i.e. the deep TTIP estimate from Aichele et al. (2014) and the benchmark estimate from Felbermayr et al. (2015). 16 According to Lawson and Hudson s (2015) measure an anti-american is someone who holds a negative view on the United States role with regard to world peace, the fight against terrorism, the fight against world poverty, protecting the world environment and the growth of the world economy. 17 This measure of post-materialism draws on the classic scale proposed by Inglehart (1977). 18 Alternatively, the country-survey/time combinations could be modeled as nested in a cross-classified structure of countries and the three points in time/survey waves. In the present case, this setting would make the estimation much more difficult (considering the very small number of survey waves) without any need. Note that the analysis is not interested in explaining the rather modest changes over time (see below), but mainly in the associations between individual level predictors and support for TTIP and, on a subordinate level, in the differences across countries (that are largely stable over time). 15

17 level units in the individual-level model together with the individual-level predictors (see Gelman/Hill 2007 for an introduction to multilevel modeling). The analysis below proceeds in two steps: In the first, I concentrate on the individual level correlates of support for TTIP. In this step, I estimate a simplified multilevel random-intercept regression model with two levels only: individuals are nested in 84 country-survey/time combinations that each receive a random-intercept. Given the binary response variable, I estimate binary logistic multilevel models as carried out through Stata s melogit command. The simplified structure on the higher levels doesn t make any recognizable difference for results on the individual level part of the model (see below), but leads to a more time-efficient estimation that is, moreover, less susceptible to problems of model convergence. In the second step, I explore the association between country-level measures and support for TTIP. Here, it is more important to take the nested structure on the higher levels properly into account. (Note again that all macro-level information are measured on the country level, not the countrytime level.) Accordingly, I estimate multilevel logistic regression model with random-intercepts at the country and country-time level. In order to test for an association of macro-level measures, the intercepts are modeled as a function of macro-level predictors (and a random component with variance to be estimated from the data). Observations with missing values are listwise excluded. 5. Results The results are discussed in three parts. I begin with a brief descriptive look on the dependent variable (section 5.1). As mentioned above, the regression analyses first concentrates on the individual level (section 5.2) before turning to aggregate-level correlates of support for TTIP (section 5.3) Descriptive findings: Support for TTIP in European comparison In a first step, consider how attitudes toward TTIP vary across countries and over time. Figure 1 below shows that support for TTIP varies strongly across the EU. Overall, a clear majority is (as of November 2015) in favor of a free trade and investment agreement between the EU and the USA. The supporters outnumber the opponents in 24 of the 28 EU member states with the four exceptions being Austria, Germany, Luxembourg and Slovenia. At one extreme, only 22% of Austrians in November 2015 are in favor of TTIP, while 70% are opposed. At the other extreme, in May % of Lithuanians and Maltese approve of a TTIP with a minuscule opposition of about 7% in both countries. The share of don t know responses is well below a quarter in most cases (the mean across country-time observations is 14.5%) and exhibits no trend over time. Between November 2014 and 2015 (country mean) support has modestly decreased from 62% to 56%. Some countries show very stable figures (e.g. Lithuania, Italy, Spain and the UK) and others pronounced declines in support (e.g. Austria, Germany, the Netherlands and Slovenia). Overall, variation between countries is, however, much more pronounced than variation within countries over time. 16

18 Figure 1: Support for TTIP across EU member states over time Note: Post-stratification weighted. 17

19 5.2. Individual-level correlates of support for TTIP Figure 2 below plots average marginal effects (alongside their confidence intervals) on the probability of supporting TTIP from random-intercept models with only the individual-level predictors included. The plotted estimates show how the probability of approving TTIP shifts with a one-unit increase on average over the observed data. These plots are based on models 1 and 2 which are shown in detail in Table A 2 in the appendix. Model 1 on which the upper panel is based does not include attitudes towards economic globalization in general; these are included in model 2 from which the lower panel of Figure 2 is constructed. General attitudes towards globalization are excluded in a first step, as these are conceptually close to the response of interest and likely endogenous to some of the individual-level predictors: e.g., considerations of economic self-interests might shape attitudes on whether globalization is economically beneficial. Recall that all variables (except for age) are scaled to vary between zero and one to allow for some comparison of strength of associations between the predictors. The empirical analysis supports most of the theoretical expectations. As to skill-level divides on the individual level, there are overall no consistent effects in either direction: There are zero discernable differences between those in rather low-skilled and rather high-skilled occupations. As regards education, we observe a negative coefficient showing that those with more years of education tend to be less likely to approve of TTIP. These results are interesting as they run counter to the usual result on skill-level divides in general trade policy preferences in the developed countries that high-skilled workers are less protectionist. This pattern clearly does not hold when it comes to the specific preferences over TTIP. Turning to socio-cultural attitudes towards openness and internationalization, we find the proxy variable trust in the UN to be positively related with support for TTIP. Separate analysis of EB 83.3 and 84.3 (see Figure A 2 in the appendix) shows that having an exclusive national identity tends to be negatively associated with support for TTIP as well (though this is clearer for EB 83.3 than EB 84.3). As to attitudes towards political elites, positive images of the EU are strongly associated with support for TTIP. Trust in the national government is positively, but only weakly related to support for TTIP. With regard to leftright positions, there is no evidence supporting an inverted U-shaped association: Opposition to TTIP is clearly most pronounced among those who position themselves on the left, but those in the center are somewhat less likely to approve of TTIP than those on the right, on average. 19 Satisfaction with the working of democracy shows a strikingly strong association with attitudes towards TTIP. There is also support for the expectation that those who think of an environmental problem as being most pressing on the country s agenda are, on average, less supportive, while those who think of issues relating to the country s economic situation are more likely to be in favor of TTIP. 19 One might think that this pattern reflects post-treatment bias, i.e. that it only holds in a model including covariates (e.g. EU image, trust in the UN) that might lead to higher opposition on the right. This is not the case, however. In a model that includes only demographics and left-right orientations, we obtain the same pattern. 18

20 Figure 2: Average marginal effects on the probability of supporting TTIP from multilevel models Note: Detailed estimation results for the statistical models on which these estimates are based are shown in Table A 2 in the appendix. Both models are two-level (individuals nested in survey-country combinations) random-intercept models with only individual level predictors included. The upper panel is based on model 1 from Table A 2. The lower panel is based on model 2 that differs from model 1 in additionally including attitudes towards globalization being an opportunity for economic growth. The lower panel of Figure 2 additionally includes attitudes towards potential economic benefits from globalization in general and shows that these views are strongly related to support for TTIP potentially shifting the probability of supporting TTIP by about 30 percentage points, on average. While some of the other effects of the other variables are reduced in size, the basic patterns remain unaffected by the inclusion of this predictor. Overall, support for TTIP is most strongly related to general attitudes on the economic benefits of globalization, followed by orientations towards the EU, satisfaction with democracy, trust in the UN and a position on the left of the left-right scale. 19

21 The pooled models presented in Figure 2 may hide interesting variation across countries. One could suspect that (some of) the determinants of support for TTIP are shaped by contextual factors at the country level, e.g. differences in patterns of elite-level support for TTIP (see the arguments above with regard to trust in the national government and left-right positions) or the frames that dominate the national discussions over TTIP. In order to explore such heterogeneity, I estimated a series of simple logistic regression models by country (including dummy variables for the survey waves) thereby relaxing the assumption of common slopes. Analogous to Figure 2, I present average marginal effects from two sets of models: without (Figure 3) and with (Figure A 2 in the appendix) the inclusion of general attitudes towards globalization. With regard to the strongest correlates of TTIP identified above, we observe a high degree of homogeneity across countries: Attitudes towards the EU and globalization are virtually always significantly positively related to support for TTIP. Likewise, trust in the UN and satisfaction with democracy exhibit statistically significant positive effects on support for TTIP in most of the cases, albeit with a few more exceptions. With regard to the main correlates of support for TTIP, individual variation in support for TTIP thus follows a common pattern across countries. There is more heterogeneity across countries where we would expect it: with regard to trust in the national government and left-right positions. Trust in the national government is estimated to have statistically significant (p<0.05) positive effects in support for TTIP in (only) five countries (Austria, Belgium, Estonia, Germany, Greece; with respect to the models presented in Figure 3), though never a (statistically significant) negative effect. Heterogeneity in the effect of left-right orientations is even more pronounced. There are countries with a strong divide similar to the one found of the pooled model, i.e. those on the left being least supportive of TTIP and those on the right being most supportive, on average (e.g. Finland, Germany, Netherlands, Spain). In other countries, there is no discernible divide along left-right positions whatsoever (e.g. Lithuania, Hungary, Portugal). In still other cases, those on the left are less supportive than those in the center and right, but there s no difference between the latter two (e.g. Belgium, Denmark, France). Finally, some countries support my initial expectation of an inverted U-shaped pattern with those in the center of the left-right scale being most likely to support TTIP (most clearly Malta, and in tendency also and Italy and the UK). 20

22 Figure 3: Average marginal effects on the probability of supporting TTIP from logit models estimated separately by country: General attitudes toward economic globalization not included Note: Pseudo-R² is based on McKelvey/Zavoina. 21

23 The effects of education and skill-level are also heterogeneous, but with no obvious pattern (e.g. a skillgap varying according to a country s income level). Heterogeneity is also observed with regard to problem prioritization of environmental issues and economic issues. Also note that these two variables are seldom significant in the expected direction in the individual country models. One likely explanation is that much of the (interesting) variation in these variables is between countries rather than within. The country mean of problem attention to economic issues is as high as 0.73 for Cyprus and as low as 0.17 for Germany and even 0.15 in the case of Malta. Problem attention to environmental issues is almost non-existent in a few countries: At one extreme, in Greece only about 0.5% of respondents chose environmental issues as being most important for the country, for example. 20 At the other extreme, about 22% of Danes name environmental issues. In this context, it is noteworthy, that we tend to find a statistically significant negative effect of issue attention to the environment in richer and older EU member states where prioritization of environmental issues is relatively widespread to begin with (i.e. in Belgium, Denmark, Finland, France, Sweden) Macro-level correlates of support for TTIP So far we have looked at individual-level covariates of support for TTIP. In section 5.3 we have seen that support for TTIP varies strongly across EU members. The individual-level variables considered above explain part of this across-country variation, statistically speaking. In this section, I consider macro-level variables that might contribute to the explanation of country-level differences. Given the relatively small country-level n, i.e. 28, and the limitations to studying country effects via multilevel models with limited numbers of level-two observations, especially in non-linear models (e.g. Bryan/Jenkins 2016; Stegmueller 2013), I proceed cautiously introducing predictors individually or in small batches rather than trying to estimate one model with more than ten predictors and 28 level-two observations. Against this background, this analysis is of an exploratory nature and will reject some explanations of cross-country variation in support for TTIP and suggest that others might be more helpful. Table 1 shows results from overall ten different random-intercept models with three levels (as explained above). The first thing to note is that the findings on the individual-level predictors of support for TTIP are similar to those from the simpler model discussed above. The following discussion will concentrate on the macro-level predictors. 20 Due to this limited variation problem attention to environmental issues had to be dropped (because of perfect prediction) from one estimation entirely (Cyprus) and has huge confidence intervals in others (e.g. Latvia and Lithuania). 22

24 Table 1: Multilevel binary logistic regression models for support for a free-trade and investment agreement between the EU and the USA country level predictors (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Individual level Occupation (ref: low-skill) High-skill occupation (0.034) (0.034) (0.034) (0.034) (0.036) (0.035) (0.035) (0.035) (0.036) (0.035) Student (0.034) (0.034) (0.034) (0.034) (0.036) (0.035) (0.035) (0.035) (0.036) (0.035) Retired (0.036) (0.036) (0.036) (0.036) (0.038) (0.038) (0.038) (0.038) (0.039) (0.038) Homemaker (0.057) (0.057) (0.057) (0.057) (0.062) (0.061) (0.061) (0.061) (0.062) (0.061) Unemployed (0.041) (0.041) (0.041) (0.041) (0.043) (0.043) (0.043) (0.043) (0.045) (0.043) Education *** *** *** *** *** *** *** *** *** *** (0.044) (0.044) (0.044) (0.044) (0.047) (0.047) (0.047) (0.047) (0.048) (0.047) Trust in UN 0.47 *** 0.47 *** 0.47 *** 0.47 *** 0.41 *** 0.41 *** 0.41 *** 0.41 *** 0.41 *** 0.41 *** (0.024) (0.024) (0.024) (0.024) (0.026) (0.026) (0.026) (0.026) (0.026) (0.026) EU image 1.59 *** 1.59 *** 1.59 *** 1.59 *** 1.15 *** 1.13 *** 1.13 *** 1.12 *** 1.12 *** 1.13 *** (0.052) (0.052) (0.052) (0.051) (0.056) (0.055) (0.055) (0.055) (0.057) (0.055) Trust in national government 0.10 *** 0.10 *** 0.10 *** 0.10 *** * * * * * * (0.026) (0.026) (0.026) (0.026) (0.028) (0.027) (0.027) (0.027) (0.028) (0.027) Left *** *** *** *** *** *** *** *** *** *** (0.036) (0.036) (0.036) (0.036) (0.038) (0.038) (0.038) (0.038) (0.039) (0.038) Center *** *** *** *** *** *** *** *** *** *** (0.031) (0.031) (0.031) (0.031) (0.033) (0.033) (0.033) (0.033) (0.034) (0.033) Satisfaction with national democracy 0.69 *** 0.69 *** 0.69 *** 0.69 *** 0.51 *** 0.50 *** 0.50 *** 0.50 *** 0.52 *** 0.50 *** (0.045) (0.045) (0.045) (0.045) (0.049) (0.048) (0.048) (0.048) (0.050) (0.048) Most important problem: economic 0.17 *** 0.17 *** 0.17 *** 0.17 *** 0.18 *** 0.19 *** 0.19 *** 0.19 *** 0.19 *** 0.19 *** (0.034) (0.034) (0.034) (0.034) (0.037) (0.036) (0.036) (0.036) (0.037) (0.036) Most important problem: environment *** *** *** *** *** *** *** *** *** *** (0.044) (0.044) (0.044) (0.044) (0.046) (0.046) (0.046) (0.046) (0.046) (0.046) Globalization 1.73 *** 1.74 *** 1.74 *** 1.75 *** 1.76 *** 1.75 *** (0.044) (0.044) (0.044) (0.044) (0.045) (0.044) Age/ *** *** *** *** * * * * * (0.0097) (0.0097) (0.0097) (0.0097) (0.010) (0.010) (0.010) (0.010) (0.011) (0.010) Female * * * * * (0.022) (0.022) (0.022) (0.022) (0.023) (0.023) (0.023) (0.023) (0.023) (0.023) 23

25 Intercept *** *** 10.2 ** * (1.22) (0.54) (0.30) (2.23) (0.37) (0.53) (0.24) (3.17) (0.28) (0.22) Country level KOF economic globalization (0.015) FDI stock inward (log) 0.12 (0.13) FDI stock outward (log) (0.081) GDP (log) (0.085) Trade with US (% of GDP, log) (0.31) FDI in and from US, stocks (% of GDP, log) 0.58 (0.89) Estimated gains from TTIP 0.48 *** Felbermayr et al. (2015) (0.14) Estimated gains from TTIP Aichele et al. (2014) (0.079) GDP per capita (log) *** (0.30) Percent of post-materialists * * (Lawson/Hudson 2015) (0.023) (0.023) Percent of anti-americans (EB 69, 2006) (0.0082) Random effect variances Intercept (country level) 0.42 *** 0.41 *** 0.40 *** 0.39 *** 0.36 *** 0.28 *** 0.42 *** 0.29 *** 0.37 *** 0.36 *** (0.12) (0.12) (0.12) (0.11) (0.11) (0.083) (0.12) (0.084) (0.11) (0.10) Intercept (countryxtime level) *** *** *** *** *** *** *** *** *** *** (0.014) (0.014) (0.014) (0.014) (0.016) (0.016) (0.015) (0.015) (0.015) (0.016) Model statistics Residual intraclass corr. (ICC) Bayesian inf. crit. (BIC) Country-time-combinations (level 3) Countries (level 2) Individuals (level 1) Note: Results from random-intercept models; standard errors in parentheses; model 5 excludes Luxembourg as it constitutes an extreme outlier with regard to FDI in and from the US; model 9 omits Croatia due to missing data on the prevalence of anti-americanism; + p < 0.10, * p < 0.05, ** p < 0.01, *** p <

26 The first eight models test various ideas on how support for TTIP might be related to economic aggregates. First, one might think that support is higher in countries with higher degrees of economic openness to begin with as publics in these countries are more accustomed to international economic openness. 21 This is not the case. Support is not higher in the more economically globalized countries (model 1), in those that have attracted a larger share of FDI (model 2) or own more FDI abroad (model 3). Nor is support systematically higher in countries with smaller economies (model 4). In this regard, economically open Luxembourg, which is one of the four countries in which TTIP opponents outnumber supporters (see above), is not an outlier from a general association. Second, it could be the case that it is not general economic openness that matters, but rather economic ties to the US specifically. This is not the case either: publics in countries with stronger economic ties to the US, as operationalized by bilateral trade volumes and FDI stocks, are not generally more supportive of TTIP (model 5). Perhaps existing economic ties are not decisive, but rather the expected gains from TTIP. We indeed find a statistically significant positive association between the quasi-objective measure of gains as estimated by Felbermayr et al. (2015) (model 6), but the coefficient is essentially zero when using the alternative estimates from Aichele et al. (2014) instead (model 7). These inconsistent findings make it hard to believe that support is driven by quasi-objectively expected gains and illustrate fundamental problems with this argument given that even experts are unsure about and do not agree on the pattern of relative gains across countries. 22 Model 8 probes how support for TTIP is related to GDP per capita and shows that support for TTIP tends to decrease with rising levels of GDP per capita. This is an intriguing pattern. Yet, it is unclear which mechanism is behind this association. One possibility is that it reflects differences in values across countries that are related to income levels. This idea is taken up by model 9 which includes the aggregate measures of anti-american and post-materialist attitudes. According to the model findings, the percent of post-materialists in a society is negatively related to cross-country variation in support for TTIP, while the share of anti-americans is not. 23 The final model 10 shows that the negative effect of post-materialism holds up once we exclude the measure for prevalence of anti-americanism. This suggestive findings on the effect of the prevalence of post-materialist values complement some related findings from the individual-level analysis above: those who name issues related to the economic situation as the country s most pressing ones are more likely to support TTIP; those who chose the environment are less likely to support TTIP; issue attention to the environment is consequential especially in richer and older EU member states where prioritization of environmental issues is relatively widespread in the first place. Potentially then, differences in materialist vs. post-materialist values might to some extent account for the 21 Note that this reasoning suggest that a country s economic openness might affect support for TTIP via citizens general attitudes towards economic globalization. In order to properly test this idea, the first four models do not include attitudes towards economic globalization. 22 The correlation between the two measures is Again, given the measurement problems this nil finding should not be considered as definitely rejecting the idea that orientations towards the US are influential. 25

27 negative association between per capita GDP and support for TTIP observed across EU member states. 24 In contrast, it is difficult to account for cross-country differences in support for TTIP with economic aggregates related to a logic of (expected) economic gains and losses. 6. Conclusion This paper has sought to re-examine the determinants of preferences towards international economic integration on an unusually salient and controversial case: the planned Transatlantic Trade and Investment Partnership. In addition to being of interest in itself, analyzing attitudes towards TTIP is of more general theoretical relevance as it allows us to re-examine the determinants of preferences towards international economic openness within a salient real-world context. Such a setting gives us the chance to study specific and, potentially, more consolidated preferences towards international economic openness. Moreover, a more detailed look at attitudes towards specific and salient controversial trade policy proposals opens the eye for ways in which the surrounding political processes might shape attitudes on specific trade policies. In this regard, studying preferences for real-world trade policy proposals is a useful complement to research that analyzes survey data on general preferences towards international economic openness. Analyzing three recent rounds of Eurobarometer surveys conducted in the 28 EU member, this study has identified a number of robust correlates of support for TTIP. At the individual level, attitudes towards TTIP are most strongly related to general attitudes towards economic globalization and the European Union as well as satisfaction with the functioning of (national) democracy. It seems particularly noteworthy that support for TTIP correlates strongly with different measures of attitudes towards political elites. Insofar, the findings add to growing skepticism that attitudes towards economic openness can primarily be explained by economic self-interest and point to politically richer explanations that reserve a role not only for socio-cultural attitudes towards international openness, but also for the broader political processes surrounding the dispute over a particular trade policy. At the aggregate level, support for TTIP is unrelated to a number of a priori likely correlates such as general levels of economic openness and economic ties to the US. Suggestive evidence is obtained that the prevalence of post-materialist values in a society might to some extent account for cross-country variation in opposition to TTIP. This latter finding complements findings on the individual-level effects of problem priorities according to which those who name issues related to the economic situation as the country s most pressing ones are more likely to support TTIP and those who chose the environment are less likely to support TTIP. Interestingly, we observe no skill-level divide with regard to preferences over TTIP. This contrasts with earlier findings on preferences towards economic openness in general. Yet, it makes sense from the perspective of economic self-interest. While it is difficult to ultimately know whether this reflects the 24 The country-level correlation between the number of post-materialists and GDP per capita is Postmaterialism correlates with (weighted country mean) problem attention for economic issues with r=-0.56 and with r=0.66 with problem attention for environmental issues. 26

28 absence of dividing lines in individual self-interest or the absence of economic self-interest considerations in the case of TTIP, it illustrates the value of studying the economic self-interest via specific real-world trade policy proposals. It might seem reasonable to expect that there would be strong skill-level divides in the case of salient free trade agreements with other parts of the world that clearly differ in factor endowment levels. China and the EU are currently negotiating over a future investment agreement, for example, and it would be interesting to contrast attitudes towards such an agreement with attitudes towards TTIP, especially once the EU-China deal has increased in political salience. Studying such cases when they appear on the political agenda will enhance our understanding of what drives individual preferences towards international economic openness in specific empirical cases going beyond what we already know from work on general preferences towards international economic integration. The present study has a number of noteworthy limitations. For one thing, it has focused on structural explanations for support for TTIP and insofar largely ignored agency. The campaigns of interest groups and their success surely additionally affects support for TTIP and might lead to differences in support for TTIP across countries that appear idiosyncratic from the structural perspective adopted here; though the cross-national differences in issue priorities investigated in this study might affect the success of such campaigns. Moreover, the usual caveats attached to an observational study when it comes to establishing causality apply. Insofar, it would be prudent to add (survey-)experimental evidence to some of the findings presented here. Finally, future work could follow-up on the analysis of effect heterogeneity across countries that was of a merely exploratory nature in this study, but revealed interesting variation. References Achen, Christopher H., and Larry M. Bartels (2006): It feels like we re thinking: The rationalizing voter and electoral democracy. Working paper presented at the Annual Meeting of the American Political Science Association, Philadelphia. Aichele, Rahel, Gabriel Felbermayr, and Inga Heiland (2014): Going Deep: The Trade and Welfare Effects of TTIP. CESIFO Working Paper No Ardanaz, Martin, M. Victoria Murillo, and Pablo M. Pinto (2013): Sensitivity to Issue Framing on Trade Policy Preferences: Evidence from a Survey Experiment. International Organization, 67(2): Bechtel, Michael M., Jens Hainmueller, and Yotam Margalit (2014): Preferences for international redistribution: The divide over the Eurozone bailouts. American Journal of Political Science, 58(4): Bryan, Mark L., and Stephen P. Jenkins (2016): Multilevel modelling of country effects: a cautionary tale. European Sociological Review 32(1): Burgoon, Brian A., and Michael J. Hiscox (2008): The gender divide over international trade: Why do men and women have different views about openness to the world economy? Working Paper, Harvard University, URL: Council of the European Union (2013): Directives for the negotiation on a comprehensive trade and investment agreement, called the Transatlantic Trade and Investment Partisanship, between the European Union and the United States of America as adopted at the Foreign Affairs Council (Trade) on 14 June Converse, Philip E. (1964): The nature of belief systems in mass publics. In: David Apter (ed.): Ideology and Discontent. New York, NY: Free Press of Glencoe. Curtis, K. Amber, Joseph Jupille, and David Leblang (2014): Iceland on the Rocks: The Mass Political Economy of Sovereign Debt Resettlement. International Organization 68(3):

29 Dreher, Axel (2006): Does Globalization Affect Growth? Evidence from a new Index of Globalization. Applied Economics 38(10): Dür, Andreas (2015): Interest group influence on public opinion: A survey experiment on the Transatlantic Trade and Investment Partnership. Working Paper, University of Salzburg, URL: Dür, Andreas, Manfred Elsig (2015): Trade Cooperation: The Purpose, Design and Effects of Preferential Trade Agreements. Cambridge, UK: Cambridge University Press. Egger, Peter, Joseph Francois, Miriam Manchin, and Douglas Nelson (2015). Non-tariff barriers, integration and the transatlantic economy. Economic Policy 30(83), EurActiv (2015): National Front joins anti-ttip lobby. URL: European Commission (2008): Eurobarometer 69 Values of Europeans. URL: European Commission (2015a): The Transatlantic Trade Investment Partnership: Making trade work for you. URL: European Commission (2015b): Eurobarometer 82.3, November TNS opinion, Brussels [Producer]; GESIS, Cologne [Publisher]: ZA5932, data set version 2.0.1, doi: / European Commission (2015c): Eurobarometer 83.3, May TNS opinion, Brussels [Producer]; GESIS, Cologne [Publisher]: ZA5998, data set version 1.0.0, doi: / European Commission (2016): Eurobarometer 84.3, November TNS opinion, Brussels [Producer]; GESIS, Cologne [Publisher]: ZA6643, data set version 2.0.0, doi: / FAZ Frankfurter Allgemeine Zeitung (2015): Demonstranten protestieren gegen TTIP. URL: demonstranten-protestieren-gegen-ttip html. Felbermayr, Gabriel J., and Mario Larch (2013): Transatlantic Free Trade: Questions and Answers from the Vantage Point of Trade Theory. CESIFO Forum 2013(4): 3-17 Felbermayr, Gabriel J., Benedikt Heid, Mario Larch and Erdal Yalcin (2015): Macroeconomic potentials of transatlantic free trade: A high resolution perspective for Europe and the world. Economic Policy 30(3): Fontagné, Lionel, Julien Gourdon, and Sébastien Jean (2013): Transatlantic trade: Whither partnership, which economic consequences? CEPII Policy Brief No. 1. Fordham, Benjamin O., and Katja B. Kleinberg (2012): How can economic interests influence support for free trade? International Organization 66(2): Francois, Joseph, Miriam Manchin, Hanna Norberg, Olga Pindyuk and Patrick Tomberger (2013): Reducing Transatlantic Barriers to Trade and Investment: An Economic Assessment. Centre for Economic Policy Research, London. URL: Gelman, Andrew, and Jennifer Hill (2007): Data analysis using regression and hierarchical/multilevel models. Cambridge, UK: Cambridge University Press. Guisinger, Alexandra (2009): Determining trade policy: Do voters hold politicians accountable? International Organization 63(3): Hainmueller, Jens and Michael J. Hiscox (2006): Learning to Love Globalization: Education and Individual Attitudes Toward International Trade. International Organization 60(2): Hays, Jude C., Sean D. Ehrlich, and Clint Peinhardt (2005): Government Spending and Public Support for Trade in the OECD: An Empirical Test of the Embedded Liberalism Thesis. International Organization 59(2): Hicks, Raymond, Helen V. Milner and Dustin Tingley (2014): Trade Policy, Economic Interests, and Party Politics in a Developing Country: The Political Economy of CAFTA-DR. International Studies Quarterly 58(1): Hobolt, Sara Binzer (2009): Europe in question: Referendums on European integration. Oxford, UK: Oxford University Press. Inglehart, Ronald (1977): The Silent Revolution: Changing Values and Political Styles Among Western Publics. Princeton, NJ: Princeton University Press. International Monetary Fund (2016): Direction of Trade Statistics (DOTS). URL: Jensen, Nathan M. and René Lindstädt (2013): Globalization with Whom: Context-Dependent Foreign Direct Investment Preferences. Working Paper. 28

30 Jensen, Nathan M. and Mi Jeong Shin (2014): Globalization and Domestic Trade Policy Preferences: Foreign Frames and Mass Support for Agriculture Subsidies. International Interactions 40(3): Lawson, Colin W., and John Hudson (2015): Who Is Anti-American in the European Union? SAGE Open 5(2): Kuo, Jason and Megumi Nao (2015): Individual Attitudes. In: Martin, Lisa L. (ed.): The Oxford Handbook of the Political Economy of International Trade. Oxford, UK: Oxford University Press Mansfield, Edward D. and Diana C. Mutz (2009): Support for Free Trade: Self-Interest, Sociotropic Politics, and Out-Group Anxiety. International Organization 63(3): Mansfield, Edward D., Diana C. Mutz, and Laura R. Silver (2015): Men, Women, Trade, and Free Markets. International Studies Quarterly 59: Margalit, Yotam (2012): Lost in globalization: International economic integration and the sources of popular discontent. International Studies Quarterly 56(3): Mayda, Anna Maria and Dani Rodrik (2005): Why Are Some People (and Countries) More Protectionist than Others? European Economic Review 49(6): Melitz, Marc J. (2003): The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica 71(6): Naoi, Megumi and Shujiro Urata (2013): Free Trade Agreements and Domestic Politics: The Case of the Trans-Pacific Partnership Agreement. Asian Economic Policy Review 8(2): O Rourke, Kevin H. and Richard Sinnott (2001): The Determinants of Individual Trade Policy Preferences: International Survey Evidence. Brookings Trade Forum 2001: Rho, Sungmin, and Michael Tomz (2015): Industry, Self-Interest, and Individual Preferences over Trade Policy. Working Paper, Stanford University, URL: Rodrik, Dani (2015): The War of Trade Models, URL: Rogowski, Ronald (1989): Commerce and Coalitions: How Trade Affects Domestic Political Alignments. Princeton, NJ: Princeton University Press. Sattler, Thomas, and Johannes Urpelainen (2012): Explaining public support for international integration: how do national conditions and treaty characteristics interact with individual beliefs? Journal of Politics 74(4): Scheve, Kenneth and Matthew J. Slaughter (2001): What Determines Individual Trade-Policy Preferences? Journal of International Economics 54(2): Scheve, Kenneth and Matthew J. Slaughter. (2006): Public Opinion, International Economic Integration, and the Welfare State. In: Pranap Bardhan, Samuel Bowles, and Michael Wallerstein (eds.): Globalization and Egalitarian Redistribution. New York, NY: Russell Sage Foundation. Spilker, Gabriele, Thomas Bernauer and Victor Umaña (forthcoming): Selecting Partner Countries for Preferential Trade Agreements: Experimental Evidence From Costa Rica, Nicaragua, and Vietnam. International Studies Quarterly. Stegmueller, Daniel (2013): How many countries for multilevel modeling? A comparison of frequentist and Bayesian approaches. American Journal of Political Science 57(3): Stop TTIP (2015): The ECI result in numbers. URL: Strezhnev, Anton (2013): The Effect of Trading Partner Democracy on Public Attitudes Toward Preferential Trade Agreements. Working Paper, Harvard University, URL: The Guardian (2016): 'Politicians have lost our trust'. URL: UNCTAD (United Nations Conference on Trade and Development) (2016a): Foreign direct investment: Inward and outward flows and stock, annual, URL: UNCTAD (United Nations Conference on Trade and Development) (2016b): Bilateral FDI Statistics, URL: Urbatsch, Robert (2013): A Referendum on Trade Theory: Voting on Free Trade in Costa Rica. International Organization 67(1):

31 Votewatch EU (2015): Negotiations for the Transatlantic Trade and Investment Partnerhip (TTIP): Vote details. URL: World Bank (2016): World Development Indicators, URL: Zaller, John R. (1992). The nature and origins of mass opinion. Cambridge, MA: Cambridge University Press. 30

32 Appendix Table A 1: Recoding of current occupation Categories in Eurobarometer NON-ACTIVE Responsible for ordinary shopping and looking after the home, or without any current occupation, not working Student Unemployed or temporarily not working Retired or unable to work through illness SELF EMPLOYED Farmer Fisherman Professional (lawyer, medical practitioner, accountant, architect, etc.) Owner of a shop, craftsmen, other self-employed person Business proprietors, owner (full or partner) of a company EMPLOYED Employed professional (employed doctor, lawyer, accountant, architect) General management, director or top management (managing directors, director general, other director) Middle management, other management (department head, junior manager, teacher, technician) Employed position, working mainly at a desk Employed position, not at a desk but travelling (salesmen, driver, etc.) Employed position, not at a desk, but in a service job (hospital, restaurant, police, fireman, etc.) Supervisor Skilled manual worker Other (unskilled) manual worker, servant Category in occupation variable Homemaker Student Unemployed Retired Low-skill occupation Low-skill occupation High-skill occupation Low-skill occupation High-skill occupation High-skill occupation High-skill occupation High-skill occupation Low-skill occupation Low-skill occupation Low-skill occupation Low-skill occupation Low-skill occupation Low-skill occupation 31

33 Table A 2: Multilevel binary logistic regression models for support for a free-trade and investment agreement between the EU and the USA individual level predictors only (1) (2) Individual level Occupation (ref: low-skill) High-skill occupation (0.034) (0.035) Student -0.12* -0.21*** (0.052) (0.054) Retired (0.036) (0.038) Homemaker (0.057) (0.061) Unemployed (0.041) (0.043) Education -0.22*** -0.22*** (0.044) (0.047) Trust in UN 0.47*** 0.41*** (0.024) (0.026) EU image 1.60*** 1.13*** (0.051) (0.055) Trust in national government 0.10*** 0.067* (0.026) (0.027) Left -0.57*** -0.53*** (0.036) (0.038) Center -0.19*** -0.18*** (0.031) (0.033) Satisfaction with national democracy 0.69*** 0.50*** (0.045) (0.048) Most important problem: economic 0.17*** 0.19*** (0.034) (0.036) Most important problem: environment -0.35*** -0.35*** (0.044) (0.046) Globalization 1.74*** (0.044) Age/ *** * (0.0097) (0.010) Female * (0.022) (0.023) Intercept *** (0.099) (0.10) Random effect variances Intercept 0.48*** 0.50*** (0.077) (0.079) Model statistics Residual intraclass corr. (ICC) Bayesian inf. crit. (BIC) CountryXsurvey wave (level 2) Individuals (level 1) Note: Results from random-intercept models; standard errors in parentheses; + p < 0.10, * p < 0.05, ** p < 0.01, *** p <

34 Figure A 1: Average marginal effects on the probability of supporting TTIP from logit models estimated separately by country: General attitudes toward economic globalization included 33