Interest group preferences towards trade agreements: institutional design matters

Interest groups play a key role in the political economy of preferential trade agreements (PTAs). Their support for or opposition to a planned PTA tends to be crucial in determining the fate of PTAs. But which PTAs receive support from (which) interest groups? Clearly, the design of a PTA, that is, which types of provisions are (not) included in the agreement, is essential in that respect. We argue that trade and trade-related provisions, such as those that regulate services trade or the protection of intellectual property rights, mainly increase support for PTAs among export-oriented business groups. In contrast, the inclusion of non-trade provisions, namely clauses aimed at the protection of environmental and labour standards, makes citizen groups, labour unions, and import-competing business groups more supportive of trade agreements. Relying on original data from a survey of interest groups across the globe, including a conjoint experiment, we find support for the argument that different types of interest groups value the inclusion of trade and non-trade provisions in PTAs differently. Interestingly, however, we find little difference between export-oriented and import-competing business interests. Our study speaks to research on interest groups and trade policy.


Introduction
Interest groups play a key role in trade policy-making (Schattschneider 1935;Gilligan 1997;Ehrlich 2008). Domestically, they try to influence trade policy outcomes with campaign donations (Drope and Hansen 2004) or the provision of political and technical information (Hall and Deardorff 2006). At the international level, they attend the ministerial conferences of the World Trade Organization (WTO), civil society dialogue meetings with the European Union, or ASEAN's Civil Society Conference (see e.g. Dür and De Bièvre 2007;Hanegraaff et al. 2011;Beyers and Hanegraaff 2017;Lucas 2019). Despite their strong involvement in trade policymaking, however, only little research investigates the reasons for interest groups' support or opposition towards specific trade agreements. Instead, considerable attention has been given to firms' trade preferences (Baccini and Dür 2018;Eckhardt and Poletti 2016;Kim et al. 2019;Osgood 2018). The few notable exceptions that empirically scrutinise interest groups' trade preferences tend to focus on the USA or Europe, and business associations (Dür 2008;Yildirim et al. 2021;Cezar 2022). Labour unions and citizen groups as well as interest groups from the Global South have received substantially less attention. Our article seeks to address this research gap by empirically testing prevalent, but untested, assumptions that have been made on interest groups' trade preferences. Hence, we ask: what explains interest groups' support for trade agreements?
We argue that interest groups' position towards trade agreements depends on the interaction between the specific design of trade agreements and the constituencies that interest groups represent. Depending on which issues are covered in a trade agreement, a group may be more or less supportive of that agreement. Concretely, we expect the inclusion of trade or trade-related provisions 1 , such as those that regulate services trade or the protection of intellectual property rights, to increase support for PTAs among groups that have export-oriented firms as members. Other groups, such as import-competing business associations, should be indifferent to such clauses or even oppose them, because they fear that they produce negative consequences for their members or the constituencies they represent. The inclusion of environmental provisions should enhance support for a PTA among citizen groups on the one hand and groups representing import-competing interests on the other. For the former, PTAs with environmental clauses may be a means to further their normative goals, whereas for the latter environmental clauses offer at least some protection against foreign competition. Labour clauses, finally, should boost PTA approval among labour unions, citizen groups and import-competing business groups. Similar to the argument on environmental protection, citizen groups may support PTAs with labour clauses for normative reasons. In contrast, labour unions and import-competing business associations may favour PTAs with such clauses because of their potentially protectionist effects.
To test our theoretical expectations, we implemented a survey including a conjoint experiment with more than 600 interest groups across the globe. To our knowledge, this is the first experiment carried out to better understand interest group preferences, particularly on a large and heterogeneous sample of groups from the Global North and Global South. By presenting interest groups with different types of PTAs as part of our conjoint experiment, we are able to estimate the causal effect of how much support or opposition towards PTAs changes with the inclusion of any specific trade or non-trade provision. Due to our large sample consisting of various types of organisations from across the globe, this set-up provides an excellent test for our theoretical framework. By and large, our results offer support for the theoretical expectations. We find that while business groups are more supportive of PTAs that include provisions concerning services and intellectual property rights than citizen groups and labour unions, adding environmental and labour clauses increases support for PTAs mostly among citizen groups and labour unions.
We also investigate how interest groups trade off different dimensions of the design of trade agreements. This allows us to answer questions such as: To what extent does adding non-trade clauses reduce business groups' support for a trade agreement compared to one that only includes trade provisions? And how much does adding non-trade clauses increase citizen groups' support for a trade agreement compared to one that only includes trade provisions? While the answer to the former question is 'not at all', as support even increases, the latter answer turns out to be 'a lot'. Thereby, our findings suggest that the concrete bundles of provisions included in PTAs matter substantially when trying to secure political support for them.
When disaggregating the actor types further, we find additional variation in how various types of clauses matter for PTA support. Illustratively, we find a statistically significant difference between export-oriented business groups from the Global North, on the one hand, and those from the Global South, on the other, in how IPR clauses matter for the appeal of a PTA. IPR clauses also clearly boost PTA support among business groups representing knowledge-intensive sectors, but lower PTA support among citizen groups concerned with public health and development. The size of some of these effects shows just how important the design of PTAs is for interest group support or opposition.
Overall, this study confirms some broadly held expectations concerning interest groups' trade preferences using systematic, experimental evidence. At the same time, the data allow us to shed light on issues for which no clear expectations exist in the literature. This concerns differences between interest groups from the Global North and the Global South with respect to their assessments of different provisions in PTAs; exactly how the addition of non-trade clauses affects business groups' position vis-à-vis trade agreements; labour unions' position towards environmental clauses in PTAs; and just how large the differences between business interests and nongovernmental organisations (NGOs) are in terms of their evaluation of the trade provisions in PTAs.
Going beyond its contributions to research on interest groups and trade policy, the paper adds to the broader literature on trade agreements. Studies in this research field have evaluated whether PTAs increase trade between their members (for an overview see Freund and Ornelas 2010), increase welfare (Moser and Rose 2014), or even influence other non-economic factors, such as human rights (Hafner-Burton 2005) or environmental regulation (Bastiaens and Postnikov 2017). There is also an emerging literature on public opinion towards PTAs (Hahm et al. 2019;Spilker et al. 2018Spilker et al. , 2016. So far, however, the question why PTAs contain specific trade and non-trade provisions has attracted relatively little attention (among the exceptions, see Lechner 2016;Raess et al. 2018). Our paper shows how governments can design a PTA to garner as much support from interest groups as possible. In light of our findings, more comprehensive PTAs have a higher chance of being backed by societal interests. Especially non-trade clauses can help governments to shore up support for a PTA, as there is virtually no opposition to them. By contrast, excluding such clauses substantially reduces support among non-business groups. For example, omitting labour clauses from a PTA very likely leads to strong opposition from labour unions.

PTA design and interest group preferences
PTAs are agreements that liberalise trade between two or more member countries, but without extending this liberalisation to other, non-partner countries. While all PTAs thus have in common that they liberalise trade, albeit to varying degrees, they differ strongly with respect to their design, namely which clauses they include (Dür et al. 2014). These clauses can be classified as trade or trade-related and "non-trade" clauses. Trade or trade-related clauses concern the reduction in tariffs, the liberalisation of services trade, but also the protection of foreign direct investments and intellectual property rights (IPR). Non-trade provisions, by contrast, concern issues such as labour rights or environmental regulation. Some agreements, such as the one between Japan and the EU (2018), include many trade or trade-related clauses and non-trade clauses, whereas others, such as the one between India and Bangladesh (2006), just focus on tariff cuts.
In the following, we focus on four broad and important clauses, two trade or trade-related clauses-services and IPR-and two non-trade clauses-environmental regulation and labour rights. While there exist several further clauses, these four are representative of the clauses that are often but not by default included in a modern type PTA. Concretely, according to version 2.1 of the Design of Trade Agreements dataset (2021), 26% of all PTAs concluded since 2010 include all four of these provisions, 32% include both trade provisions but neither of the non-trade clauses, whereas 7% include none of the four clauses (Dür et al. 2014). Since we are interested in how interest groups evaluate different design options, it is necessary to focus on clauses that can but do not need to be included. This is the reason why we, for example, do not include provisions liberalising trade in goods since all PTAs include such provisions. Trade-related clauses serve to increase market access, either by liberalising specific sectors or by creating an environment that allows for trade. Services clauses, in particular, liberalise an often protected sector in PTA member countries (Baccini 2019). For example, PTAs between the USA and Latin American countries led to a significant opening of telecommunications services in the latter (Castillo Mezarina 2020). Furthermore, intellectual property rights tend to ensure that innovative companies can engage in trade without fearing a violation of their patents, trademarks or copyrights. In contrast, the two non-trade clauses-environmental regulation and labour rights-serve to either alleviate concerns regarding the negative consequences of trade on the environment or workers rights (Khanna 2001;Segerson and Miceli 1998;Vogel 2009) or to protect domestic industries from (unfair) competition based on low standards of production (Bechtel et al. 2012;Lechner 2016;Mosley and Uno 2007). In many respects, our findings for these four clauses should be generalisable to other provisions potentially included in PTAs.
We argue that which of these clauses are included in a PTA is crucial for shifting preferences of different interest groups in favour or against PTAs. By interest groups, we understand formal associations for whom the exercise of influence over public policy is a major aim, but that do not run for office. Such interest groups are highly active in lobbying decision-makers with respect to PTAs. For example, a large number of business associations, including the National Association of Manufacturers, formed a coalition in support of the USA-Mexico-Canada Agreement (USMCA). In Costa Rica, trade unions strongly opposed the liberalisation of the services sector foreseen in the Central America Free Trade Agreement (CAFTA). The Seattle to Brussels Network, an international coalition of nongovernmental organisations (NGOs), even demands a stop to all trade negotiations, as in their view trade agreements harm both countries in the Global South and the environment. These are just some of the thousands of interest groups worldwide that engage in lobbying on PTAs.
The argument below combines the design of PTAs, i.e. which types of clauses are included, with five types of interest groups that can be expected to differ in their interests to explain which groups support or oppose a PTA. For this purpose, we distinguish between import-competing business groups, export-oriented business groups, business groups which are neither import-competing nor export-oriented ('Other business'), citizen groups, and labour unions. Business associations generally represent the economic interest of their members. To understand their preferences, we refer to their members' degree of import competition and export orientation. Import-competing business groups are associations that represent firms that produce goods or services that are not part of the country's comparative advantage. They hence can be expected to perceive trade liberalisation as an economic threat to their members. By contrast, export-oriented business groups are associations that (predominantly) represent firms that benefit from a liberalisation of trade relations. Since the concepts of export orientation and import competition do not apply to all business associations, we also distinguish 'other' business groups. They represent sectors with high intra-industry trade or the whole economy, or operate globally and thus face heterogeneous gains and losses from trade liberalisation.
In contrast with some of the current literature focusing mainly on individual firms, which in line with "new new" trade theory argues that we should see cleavages over trade policy within industries (see e.g. Kim 2017), our focus is on exportoriented and import-competing business associations. Methodologically, this makes sense because including firms in a survey together with business associations, labour unions, and citizen groups would have been very challenging because of the large differences across these actor types: while firms lobby for their own individual interest, business groups, by definition, represent the collective interest of their members and thus need to find a compromise between all individual interests of various firms they represent. Substantively, our focus results from the empirical observation that business associations continue to be at least as prevalent in trade policy debates as individual firms and that they do take opposing positions on trade policy. The EU's 2020 public consultation on its trade policy review illustrates this well: from the European Automobile Manufacturers' Association to WindEurope, the overwhelming majority of business actors participating in this consultation were associations. 2 Just as business associations, labour unions aim to protect the interests of their members, in essence employees in a sector or in an economy as a whole. While this defence of the economic interests of their members makes them similar to business groups, their interests do not always align with those of business groups as we will argue below. Finally, citizen groups are NGOs that do not directly defend the economic interests of their members or supporters. Instead, these groups are founded to push for values or to fight for specific rights. Examples are Greenpeace or Transparency International.
Clearly, preferences vis-à-vis the design of PTAs may differ within these five types of actors. For example, cleavages may exist between business associations from the Global South and the Global North or between public health and environmental NGOs. To keep the argument accessible, we largely elide these differences here, but return to them in a section on additional evidence below.

Trade and trade-related clauses
Trade and trade-related clauses increase market access or directly regulate a specific aspect of trade. Services clauses, in particular, liberalise an often protected sector in PTA member countries. A PTA including such clauses should thus be perceived positively by groups that represent export-oriented businesses. They will see these clauses as means to improve access to foreign markets and thus to increase the profits of their members. In contrast, interest groups that represent non-competitive, i.e. import-competing firms are likely to view services clauses negatively, as increased competition hurts the interests of their members. Moreover, most labour unions and citizen groups might perceive liberalisation of the service sector as a threat to the jobs and rights of their members or supporters and should therefore view it sceptically. They may also expect trade liberalisation to trigger so-called regulatory-chill or even race-to-the-bottom effects (Drezner 2001;Greenhill et al. 2009;Prakash and Potoski 2006;Revesz 1992). Labour unions, finally, might fear that services clauses in PTAs increase competition in the domestic market, thereby putting pressure on the wages and working conditions of domestic workers (Lechner 2016). The expectation, therefore, is for business groups representing import-competing interests, citizen groups and labour unions to prefer PTAs that do not include services.
The constellation of interests should be very similar for IPR clauses. In highly developed countries, IPRs anyhow are strongly protected, so for imports to these markets, IPR provisions in PTAs play a subordinate role. Among business interests from the Global North, therefore, opposition to IPR provisions should be virtually absent. By contrast, export-oriented business groups from these countries should welcome such clauses as they reduce competitors' ability to copy technology or compete based on imitation. In turn, this facilitates both exports and foreign direct investments by export-oriented companies. In less developed countries, importcompeting business groups should oppose IPR clauses. For most citizen groups, IPR clauses should play a subordinate role in their evaluation of PTAs. Some of them oppose IPR clauses, for example because they may make access to medicines more difficult in developing countries, but many others-such as environmental NGOsshould be largely indifferent because IPRs matter little for their most pressing concerns. The same applies to labour unions. We derive the following two hypotheses from this discussion:

H1
The inclusion of provisions liberalising services has the largest positive effect on PTA support among export-oriented business groups.

H2
The inclusion of provisions protecting intellectual property rights has the largest positive effect on PTA support among export-oriented business groups.

Non-trade clauses
In terms of non-trade clauses included in PTAs, we focus on environmental and labour provisions. The inclusion of environmental clauses should mainly affect the PTA preferences of citizen groups. These groups are concerned about PTAs potentially undermining specific regulations. They may fear that in the face of trade liberalisation, governments are tempted to cut back regulation that could decrease firms' competitiveness. Such regulatory race to the bottom or regulatory chill concerns also pertain to environmental standards. For these groups, the inclusion of non-trade clauses regulating environmental issues in a PTA could therefore be seen as a means to ensure (the maintenance of) high protection levels. 3 An early illustration of this mechanism at work is the North American Free Trade Agreement (NAFTA). Especially in the USA, environmental groups protested heavily against the agreement fearing both a race to the bottom with regard to environmental standards between the three member countries-the USA, Canada, and Mexico-and a significant increase in the import of goods produced under low environmental standards especially from Mexico. However, former US president Bill Clinton was able to secure ratification of the agreement after promising to add side agreements on labour rights and environmental protection (Steinberg 1997;Johnson and André 1996).
A second constituency that can be expected to view PTAs more positively when they include environmental clauses are import-competing business groups. On one hand, they can hope that trade disputes over non-trade clauses lead to the imposition of trade barriers (in form of retaliatory duties) if the standards established in these clauses are not met by at least one member country. This then would provide welcome protection to import-competing firms. On the other hand, import-competing business groups in the Global North can hope that higher environmental standards in Southern partner countries increase production costs there. Their own production costs should not be affected since typically the prevailing environmental standards in most industrialised countries are already rather strict (and typically stricter compared to the PTA clauses). The inclusion of non-trade clauses can thus provide an additional shield of protection to cushion import-competing groups in the Global North from at least some of the increased competition arising from firms in partner countries with lower standards and thus lower production costs (Lechner 2016). Export-oriented business groups, by contrast, should be either indifferent or opposed to the inclusion of environmental clauses in PTAs. They incur economic losses if these clauses lead to trade disputes (mainly Northern export-oriented business groups) or if these clauses increase production costs (Southern export-oriented business groups) (Spilker 2013).
In terms of labour clauses, the key constituency benefiting from their inclusion in PTAs are labour unions. On the one hand, pushing for labour rights is a key raison d'être for labour unions. On the other hand, labour unions (especially in the Global North) may also view labour clauses as a way to reduce foreign competition. If less developed countries are forced to increase labour standards, this may lead to higher production costs there, and hence may make imports from these countries more expensive. In turn, this may reduce pressure on firms in the Global North to reduce labour costs. Also citizen groups, on average, can be expected to prefer PTAs that include labour clauses. Especially development NGOs and human rights NGOs should link their support for a PTA to the presence of labour clauses. For them, PTAs can be an instrument to get recalcitrant states to sign up to labour rights. Finally, for the reasons mentioned in the discussion about environmental clauses, also import-competing business groups should increase their support for a PTA when it includes labour clauses while export-oriented business groups should look on labour clauses in the same way as they view environmental clauses in PTAs. Summarising this discussion, we again arrive at two hypotheses: H3 The inclusion of environmental clauses has the largest positive effect on PTA support among citizen groups and import-competing business groups.

H4
The inclusion of labour clauses has the largest positive effect on PTA support among labour unions, citizen groups, and import-competing business groups.

Sample
To test these theoretical propositions, we conducted a survey with interest groups across the globe on their trade policy views. For this purpose, we created an interest group respondent pool that is as encompassing as possible of the universe of interest groups engaged in trade-policy lobbying worldwide from countries with at least a minimum level of freedom of association. In particular, we sampled interest groups by inferring trade policy interest if a group (a) attended a WTO ministerial meeting, (b) selected trade policy as a field of interest in the EU's Transparency Register, (c) participated in one of the EU's Civil Society Dialogues with third countries, or (d) appears as one of the organisers of the ASEAN Civil Society Conference. Evidently, our sampling frame has some shortcomings, especially for groups from outside Europe. Illustratively, groups from the Global South that attend a WTO ministerial meeting may have a more positive view of trade and trade agreements than the "average" interest group from the Global South. Currently, however, no better sampling frames for a worldwide survey on interest groups' trade policy views are available.
Relying on this approach, we identified 2841 business associations, labour unions, and citizen groups. We invited them by email to participate in a survey on their trade policy views. To keep the costs of participating as low as possible, we kept the online questionnaire short. After two reminders, we received 609 complete responses, resulting in a response rate of 21.4%. This response rate is very good given the difficult nature of contacting groups from across the globe and the heterogeneous nature of the groups in our sample. Furthermore, some of the groups that we contacted may simply not have been operational anymore. Different tests suggest that the non-responses do not introduce a systematic bias in the sample (see Table A3 in online Appendix). In other words, our sample is representative for the 2841 groups in our sampling frame.
The resulting sample includes 382 business associations (62.7%), 176 citizen groups (28.9%), and 51 labour unions (8.4%). Of these groups, 74.1% operate in the Global North. 4 The respondents were mostly secretary generals (39,3%), policy officers (23,9%) or presidents (16,8%) of the groups contacted and explicitly responded to the survey as representatives of their group. As most respondents hold prominent positions in their respective group, we are confident that their answers indeed represent the group's position. To our knowledge, while larger surveys have been carried out of European interest groups, our study is one of most comprehensive attempts to survey interest groups across the world. For more descriptive statistics on the sample, see the corresponding section in the online Appendix.

Choice experiment
To measure the outcome of interest, support for a trade agreement with a specific design, we utilise a survey embedded conjoint experiment (Hainmueller et al. 2014). The general idea is to capture interest groups' underlying preferences by asking them to rate different, randomly designed trade agreements. In this specific choice experiment design, respondents compare two proposed trade agreements side by side. For each agreement, we randomly vary whether it includes (Yes) or excludes (No) two trade clauses (service liberalisation and intellectual property rights protection) and two non-trade clauses (labour rights protection and environmental protection). 5 With this approach, we implicitly keep constant other aspects on which agreements can vary, such as the extent to which they liberalise trade in goods. Survey respondents then choose the design they prefer more (choice) 6 and rate each design individually on a scale from 1 (not favourably at all) to 10 (very favourably). In our empirical analysis, we utilise the rating outcome because it allows us to better understand the popularity of different options, irrespective of the choice of one agreement over the other. We provide the findings for the choice outcome as a robustness check. Table 1 illustrates our design.
Following this empirical design, the unit of analysis is one trade agreement proposal assessed by one respondent. In the aforementioned setup, respondents see two proposals at a time, two times. This results in 2436 observations (609 respondents × 2 designs × 2 rounds). As we are mostly interested in groups' reaction to specific design features, we analyse the rating outcome conditional on the inclusion of different institutional design features and assignment to groups. Thereby, we follow  Hainmueller et al. (2014) and Leeper, Hobolt and Tilley (2020: 12) and estimate the causal effects of the inclusion or exclusion of each dimension using a regression model and cluster-robust standard errors which account for multiple observations per respondent (see Eq. 1). In particular, we estimate where i represents the proposal, and j represents the respondent. We thus regress a proposal's rating ( rating i ) on all four conjoint dimensions ( ser i , ipr i , env i , lab i ). This set of predictors is interacted with the assigned actor type, for example, exportoriented business or labour union ( actor j ). Since we only have two options for each dimension (inclusion and exclusion), we avoid an arbitrary picking of a reference category (as discussed in Leeper et al. 2020). Additionally, we are not interested in the absolute rating for a PTA but in the marginal effect the inclusion of a dimension has on the rating. For this, it is irrelevant whether the effect comes from the backlash following the non-inclusion of a clause or from the support for including a clause. Utilising a choice experiment has several advantages vis-à-vis simply asking respondents whether they perceive certain design features to be important for them (Hainmueller et al. 2014). For one, because of the random assignment of different design features, we can estimate the average causal effect in the overall support for a specific agreement induced by including or excluding a specific provision. Additionally, choice experiments help mitigate social desirability concerns and ad-hoc rationalisation (Wallander 2009), as it is substantially harder for respondents to trade off several dimensions simultaneously and ponder potentially conflicting socially desirable responses.

Operationalisation of the predictor
Based on our theoretical argument, we anticipate variation in different types of actors' reactions to certain design features. Specifically, we distinguish (1) exportoriented business groups and (2) import-competing business groups, (3) other business groups, (4) citizen groups, and (5) labour unions. To assign respondents to these actor types, we first manually coded groups as business groups, citizen groups, or labour unions. We define business groups as associations that have firms or other business groups as members. Their purpose is to represent the economic interests of businesses. The category of business groups also includes professional associations (such as lawyers' or medics' associations), as theoretically it is not clear why we should expect professional associations to react differently to PTA design features than business associations narrowly defined. Moreover, our sample only contains few professional associations, meaning that any estimates for them would be highly uncertain. Just as business associations, labour unions aim to protect the (economic) interest of their members. However, they have employees as members and advocate for workers' interests. Citizen groups are all groups that do not directly protect the (1) rating i = 0 + 1 ser i + 2 ipr i + 3 env i + 4 lab i + 5 actor j + 6 ser i * actor j + 7 ipr i * actor j + 8 env i * actor j + + 9 lab i * actor j + i economic interests of their members. Rather, their core motivation is to defend and advocate for specific norms or values.
In order to disentangle export-oriented, import-competing, and other business groups, we assigned each business group the International Standard Industrial Classification (ISIC) code that best describes the industry it is active in. The level of detail of the coding ranges from ISIC classes, e.g. ISIC rev. 4 class 0112 'rice', to sections, e.g. ISIC rev.4 section A 'Agriculture, forestry and fishing'. We then used HS6 trade data from Comtrade for trade in goods as well as BaTiS data for trade in services and aggregated exports and imports to the respective ISIC levels using official conversion tables. To capture whether an industry is import-competing or export-oriented in the group's home country, we divided exports by total trade. 7 This variable is naturally scaled between 0 and 1. A value of 0.5 indicates that a sector exports as much as it imports, whereas values above 0.5 indicate more exports than imports and values below 0.5 indicate more imports than exports. We categorise a business association as export oriented if it represents a specific sector which exports substantially more than it imports ( eo ci > 0.55 ). Import-competing groups represent specific sectors with large imports relative to exports ( eo ci < 0.45 ). Business associations that operate globally (such as the International Chamber of Commerce), represent the whole economy (e.g. domestic chambers of commerce) or operate in sectors that are fairly balanced in terms of imports and exports ( eo ci between 0.45 and 0.55) are considered 'Other business groups'.

Results
We start the empirical analysis by providing descriptive evidence before presenting the experimental results. Figure 1 shows the distribution of rating scores (on the x-axis) over all groups and disaggregated by group type. In panel a) of Fig. 1, the rating is mildly skewed to the left and the lowest rating is the most frequently selected response. This means that a considerable number of interest groups in our sample oppose PTAs independent of their exact design. The second most popular category is the middle of the scale, whereas the highest rating categories are rarely While this descriptive evidence provides interesting insights, it does not allow to causally attribute the support (or lack thereof) of specific groups to design features of PTAs. In the remainder of the section, we will thus scrutinise how the 7 In form of an equation: eo ci = X ci X ci +M ci , where eo ci is the export orientation of the group's home country c in industry i, X are exports, and M are imports. theoretically derived group types react to the inclusion of certain design features in PTAs. We will start by discussing trade related clauses before moving on to the non-trade issues. All results are shown in Fig. 2. The x-axis shows the conditional average marginal component effect (that is the change in rating when the dimension shown on the right is included).

The effect of trade clauses
Starting with service provisions (top panel in Fig. 2), we expect according to H1 that export-oriented business associations should be the core advocates for the inclusion of such clauses. Other groups should be either indifferent or critical of service liberalisation. Figure 2 generally confirms this expectation. Export-oriented business associations indeed welcome the inclusion of service provisions, as shown by the second estimate in the figure. In contrast, citizen groups are indifferent and labour unions oppose such measures. Interestingly, however, import-competing business associations, and other business groups also become more supportive of PTAs when this design feature is present. These findings provide some but not unambiguous support for H1. Moving on to IPR clauses (second panel in Fig. 2), following H2 we again expect export-oriented business groups to favour PTAs that regulate IPRs. The findings are, however, only partly in line with this expectation since all business groups, regardless of whether they face import competition or are oriented towards exports, favour PTAs with IPR clauses. Both labour unions and citizen groups are significantly less supportive towards these provisions compared to business groups.
To summarise, our findings for trade-related clauses suggest that business associations (regardless of their trade competitiveness) support the inclusion of the two trade-related clauses in PTAs, namely service liberalisation and IPR protection. At the same time, citizen groups are indifferent and labour unions are indifferent regarding IPR provisions and opposed to the inclusion of service provisions in PTAs. Overall, the findings thus are partly consistent with H1 and H2.

The effect of non-trade clauses
We now move to the effects of non-trade clauses-environmental protection and the protection of labour rights. We argued above that environmental provisions should increase support for a PTA particularly among import-competing business groups  Table A4 in online Appendix provides the underlying regression results for the visualisation in Fig. 2 and citizen groups. The third panel of Fig. 2 provides the findings that relate to this argument. All actor types support the inclusion of environmental provisions. Even import-competing business groups, for which we estimate the smallest positive effect, increase their support for an average PTA by roughly one point on the 10-point scale when it covers provisions on environmental protection. This is a substantial effect. That the estimated effects are even larger for export-oriented business groups and labour unions is not completely consistent with H3. In line with H3, citizen groups are the most supportive actor type. They increase their support for a PTA with environmental provisions by almost two points. Turning to labour clauses, we anticipated that citizen groups, labour unions and import-competing business groups increase their support for a PTA when it includes labour clauses. These are indeed the groups for which we find the largest positive effects. However, while the estimate for import-competing business groups is substantial in size, it is not statistically significantly different from export-oriented and other business. Citizen groups are more supportive than the reference category, 'other business groups'. Finally, as expected, labour unions are substantially more supportive than every other actor type. They increase their support for a PTA with labour clauses by more than 2.6 points on a 10-point scale, a massive effect. Thus, these findings are by and large in line with H4.
Overall, these findings for non-trade clauses support H3 and H4, with the exception that again we do not observe much difference between import-competing and export-oriented business groups, especially regarding environmental provisions. What comes out most clearly from the results is that neither environmental nor labour clauses in PTAs meet any real opposition among interest groups. In fact, nearly all groups favour PTAs including these non-trade provisions over PTAs that do not include them.
Several reasons could explain this finding. First, business groups might believe that these non-trade provisions are merely cheap talk and are thus not worth the paper they are written on. While we cannot rule out this explanation, we think it is rather unlikely since at least for environmental clauses in PTAs several studies have shown that they do lead to improvement in domestic environmental protection levels (Brandi et al. 2020;Lechner and Spilker 2021). Similarly, but less pessimistically, business groups, especially export-oriented groups from the Global North, might reckon that if they already comply with such standards in their home country, they prefer levelling the playing field such that also firms in the Global South need to comply with these standards. As we show below, however, only for labour clauses the evidence goes in the direction expected by this explanation (see Fig. 4). For environmental clauses, if at all, it seems that it is business groups from the Global South that are more supportive. In addition, our results might be a product of social desirability, that is, business actors simply pretending to support environmental and labour clauses, when in fact they work against them behind the curtain. Again, we do not think this is likely the case as conjoint experiments are exactly the experimental tool designed to avoid such social desirability biases (Hainmueller et al. 2014). Finally, our results might reflect either genuine support for the environment and labour rights or a belief that modern types of PTAs cannot overcome domestic opposition if they do not include non-trade provisions (Spilker et al. 2018;Cezar 2022), both of which are explanations that seem plausible given all of our results as discussed also below.

Trading off trade and non-trade clauses
So far, we have focused on the marginal effects of adding or subtracting individual provisions for interest groups' average support of PTAs. Here we go beyond this analysis by treating trade and non-trade provisions as bundles. This allows us to probe deeper and see how adding or subtracting trade or non-trade provisions affects different interest groups' PTA evaluations and also which kind of PTA (e.g. very narrow or very broad) would receive most support from each interest group type. Additionally, marginal effects mask how much interest groups support these bundles. Potentially, including a certain provision may increase support but the entire PTA may still fail to get an acceptable rating and thus may not garner enough political support to be feasible. To do so, we re-ran the analysis interacting various design dimensions with each other (see Model 3R in Table A4) and then calculating predicted ratings for four scenarios: an agreement with neither trade nor non-trade provisions (this would be an agreement just cutting tariffs); an agreement merely with  Fig. 3 Actors, trade-offs, and PTA support. Note: Ranges represent 90% and 95% confidence intervals, calculated based on cluster-robust standard errors. Model 3R in Table A4 in online Appendix provides the underlying regression results for the visualisation in Fig. 3. Specifically, the model is used to predict the support level by group type for four scenarios: The 'Bare-bones' scenario includes neither trade nor non-trade provisions. The 'Trade' and 'Non-Trade' scenario includes both trade and non-trade provisions, respectively. The 'Comprehensive' scenario includes all provisions. The dotted line represents the scale's middle point of 5.5 trade provisions; an agreement with non-trade provisions; and an agreement with trade and non-trade provisions. Figure 3 shows the results. Contrary to what might be expected (but in line with the findings above), for export-oriented business interests a full PTA is actually the most attractive option, even more attractive than a PTA with only trade provisions included. The same applies to import-competing and other business groups. Citizen groups and labour unions are most positive about PTAs that include non-trade provisions, but adding trade provisions does not reduce their PTA approval (whereas they consider a PTA just with trade provisions as clearly inferior option). This evidence thus indicates that for a government aiming to maximise interest group support for a PTA, signing a comprehensive agreement is the best strategy. If we only consider agreements acceptable that receive a rating above the scale's middle point (the dotted line in the figure), the most comprehensive agreement is the only bundle that garners support among all groups.

Additional evidence
The five actor types that we have focused on so far are relatively broad. Within each type, preferences towards trade agreements may still be heterogeneous. This may explain why we found only partial support for our hypotheses. In this section, we split these actor types based on theoretical considerations about which groups should be most supportive towards specific PTA provisions. The following results come with considerable uncertainty as for some of the sub-types that we distinguish, we only have a relatively small number of respondents. Still, this evidence nicely complements the findings we reported before.
We start by splitting export-oriented business groups according to whether they are from the Global North or the Global South. Groups' geographical origin might matter as companies from the Global South generally tend to produce less intellectual property. Similarly, the service sector tends to be of lesser importance in developing countries than in industrialised countries (Kolcava et al. 2019). In terms of non-trade provisions, countries from the Global South often compete in the global economy based on cheap labour and most of them also have less stringent environmental regulations which hands them an advantage. Mainly export-oriented business groups from the Global North, hence, can expect gains from service and IPR clauses in PTAs. Moreover, they anyhow face relatively strict labour and environmental legislation in their domestic markets, so they should not be too much concerned about the inclusion of labour and environmental clauses in PTAs. Taken together, support for PTAs including any of the four design features should be larger among Northern export-oriented business groups.
For Fig. 4, we hence recoded export-oriented business groups into Northern and Southern based on the location of their headquarters, with Northern groups being those from the OECD or the EU. Whereas we observe a substantial difference between the two sets of groups when it comes to services provisions, the coefficient is not statistically significant ( b = −1.08 , p = 0.18 ; see Model 4R in A5).
Additionally, the direction of the effect runs against our expectation. While this is surprising, there are at least two plausible explanations. The Global South likely has a comparative advantage in certain service industries. Additionally, some export-oriented firms in the Global South likely benefit from service liberalisation that improves their access to services from the Global North. The differences are substantial in terms of IPR provisions. For Southern export-oriented business groups, the inclusion of IPR provisions increases support by approximately 0.22 points. In contrast, support for an average PTA increases by more than one and a half points (1.64) for Northern export-oriented business groups when IPR protection is included. The difference between the two sets of groups is significant at the 10% level (see IPR × Northern export-oriented business Model 4R in A5). With regard to environmental provisions, both groups are approximately equally fond of PTAs including such provisions (the difference between the two estimates is close to zero and not statistically significant, see the Environ. × Northern export-oriented business estimate in Model 4R in Table A5). Finally, and most interestingly, the groups substantially differ in their attitudes towards PTAs that incorporate labour rights. While Southern export-oriented business groups view these provisions as potentially harmful (as indicated by the negative, albeit non-significant estimate shown in Fig. 4), Northern groups clearly support the inclusion of labour provisions. The difference in average marginal component effects is 1.19 on a ten-point scale. The interaction term is not statistically significant ( p = 0.15 ; see Table A5 in online Appendix). Overall, these findings again are partly but not fully in line with our expectations.
Next to the distinction between Northern and Southern export-oriented business groups, heterogeneity within our five large actor types may also result from other divisions, some of which we already alluded to in the theory section. Starting with services clauses, they may be viewed differently by business groups representing  Table A5 in online Appendix provides the underlying regression results for the visualisation in Fig. 4 tradeable service industries. The distinction between tradeable services groups and the rest of the business actors arises because groups representing manufacturing industries and also some services sectors, such as public services or the health sector, are only indirectly affected by international competition in services. In fact, in the first panel of Fig. 5, we find that groups representing firms in tradeable service industries are supportive of service liberalisations while the rest of the business actors are largely indifferent. That is, the former increase their support by 1.45 points on a ten-point scale while the latter do not update their views of PTAs conditional on the inclusion of service provisions. The difference between the two sets of groups is significant at the five-percent confidence level (see Model 5R Table A5).
Regarding IPR clauses, we compare business groups from knowledge-intensive industries to the rest of the business actors and citizen groups that can be expected to be particularly concerned with health policy to other citizen groups. The argument here is the following: business groups that represent knowledge-intensive sectors should be particularly supportive of PTAs that include IPR provisions in order to protect their intellectual property. We coded this variable based on the OECD's categories of highmedium high technologies manufacture and ICT manufactures, and our own coding of service sectors concerned with copyrights (e.g. audiovisual or software producers). In contrast, citizen groups focusing on health policy should be strongly opposed to IPR provisions because they fear that such provisions make access to medicines more difficult (Dür and De Bièvre 2007). The findings back these expectations. Business groups in knowledge-intensive industries are substantially more supportive of PTAs including  Table A5 and Model 7R-8R in Table A6 in online Appendix provides the underlying regression results for the visualisation in Fig. 5 IPR provisions than all other business groups (by almost 1 point). This difference is statistically significant at conventional levels (see Model 6R in Table A5). However, also other business groups support the inclusion of IPR provisions, albeit to a lesser extent (2.05 points on a ten-point scale among knowledge intensive business groups vs 1.10 points among other business groups). Finally, for citizen groups, we also observe a pattern in line with our expectation. Citizen groups focusing on health policy lower their PTA support upon the inclusion of IPR provisions by approximately 1.35 points. In contrast, other citizen groups generally are indifferent towards the inclusion of IPR provisions. Their rating is on average slightly more positive if IPR provisions are included (estimated at 0.17 points). Next, while we anticipate that all citizen groups are supportive of environmental protection, clearly the desire for such clauses should be strongest among environmental citizen groups. The third panel in Fig. 5 confirms this notion. The data suggest that other citizen groups increase their support for a PTA by 1.59 points on the 10-point scale, whereas environmental groups show an increase in their PTA support by 3.38 points when environmental clauses are included. In other words, a PTA without such provisions will find it hard to garner support among these groups. This difference is also statistically significant, as indicated by the interaction term in Model 7R in Table A6 in online Appendix. Finally, we separate citizen groups that should be particularly concerned with labour rights (human rights, development, and social welfare NGOs) from all other citizen groups. The former, however, are only slightly more supportive of PTAs including labour clauses. To be specific, their support for a PTA is, on average, 1.53 point higher when labour clauses are included, whereas this value is 1.14 for all other citizen groups. As this effect is not large, the interaction term fails to reach conventional levels of significance (see Fig. 5 and Model 8R in Table A6). Overall, however, we found support for four of five expectations related to these more specific actor types.

Robustness checks
In order to assess the robustness of our findings, we replicated all findings presented above using the choice outcome, that is, the question whether respondents prefer agreement A or B. Tables A7, A8, and A9 in online Appendix fully support the findings presented above. If anything, the results are even more pronounced for the choice outcome than for the ratings. Moreover, we re-ran the analysis with a dummy variable as outcome that indicates whether a rating is > 5 or not. Online Appendix Tables A10, A11, and A12 show that also with this dependent variable, the results are very similar to those presented before. Finally, Section A.4.3 in online Appendix replicates the main findings using the approach outlined by Leeper et al. (2020).

Conclusion
What explains interest group attitudes towards trade agreements? We have put forward an argument that stresses the design of trade agreements and actor types. Trade and trade-related provisions should enhance support for PTAs among export-oriented business interests. We also expected environmental clauses to lead to greater PTA backing among citizen groups and import-competing business groups. Finally, labour clauses should mainly push up support for a PTA among labour unions, citizen groups, and import-competing business groups. An unprecedented survey with more than 600 interest groups from across the globe that included a conjoint experiment has allowed us to test the argument. We found some but not unambiguous support for these theoretical expectations. Contrary to what we expected, for example, we detected hardly any differences in PTA attitudes between export-oriented and import-competing business groups. When further disaggregating our actor types, differences between them became more pronounced. Illustratively, we find a subset of citizen groups that can be expected to care about access to medicine to clearly lower their support for PTAs that include intellectual property clauses. At the same time, environmental NGOs show much greater support for PTAs that include environmental clauses.
These findings add to our understanding of interest group preferences towards PTAs in several ways. First, while some differences between business interests on the one hand and citizen groups and labour unions on the other were to be expected based on anecdotal evidence, our data show just how large this effect is for some of the clauses that we deal with here. For example, if a government wants to avoid opposition from labour unions to a PTA, adding labour clauses is absolutely imperative. Second, for business interests more seems to be strictly better than less. That comprehensive PTAs that include non-trade provisions enjoy greater support from business interests than PTAs only with trade provisions is a particularly remarkable finding. In fact, a government maximising interest group support should find it relatively costless to add non-trade clauses to trade agreements. This finding may explain why we have seen the increasing scope of PTAs over the past two decades that previous research has uncovered (Dür et al. 2014). Third, for citizen groups and labour unions to support trade agreements, the inclusion of non-trade issues is a necessary condition, but moving from an agreement with only non-trade clauses to a comprehensive one hardly changes their PTA evaluation.
Fourth, the near absence of differences between export-oriented, import-competing, and other business groups is a novel finding. Interest group preferences seem to reflect more the kind of members a group has (firms or not) than the level of international competitiveness of these members. This suggests that economic interest is not the only factor driving business groups' attitudes towards trade agreements. Fifth, we find considerable support for services and environmental provisions even among export-oriented business groups from the Global South. This finding is intriguing, as existing research would lead one to expect greater scepticism at least towards environmental clauses in developing countries. Mostly, such clauses have been seen as part of an agenda pushed by the Global North on the Global South (for example, because some countries strongly opposed the inclusion of such non-trade clauses in World Trade Organisation negotiations). The most plausible interpretation for this finding is that these actors may see PTAs as a way to pressure their governments into implementing stricter environmental and labour standards. For them, PTAs may facilitate the kind of boomerang effect that describes how local actors leverage a foreign power to achieve their domestic aims (Keck and Sikkink 1998).
Finally, our findings also shed light on the political feasibility of PTAs with specific designs. By understanding PTAs as bundles of provisions, we can scrutinise how groups trade off different combinations of trade and non-trade provisions. Our findings suggest that comprehensive agreements have the highest chance of being politically feasible, as on average they garner positive ratings among all interest group types. To the extent that policymakers are interested in maximising interest group support for a PTA (rather than, for example, maximise its effectiveness or trade-liberalising impact), they hence should negotiate comprehensive PTAs that include both trade and non-trade clauses.
Future research could advance on this study in several respects. For one, we need surveys that include a larger number of interest groups from the Global South. Studies may even just focus on groups from developing countries to increase our knowledge of these groups' views of PTAs. Second, we put forward a conjoint design with four dimensions and two levels per dimension. While insightful, these rough categories mask substantial variation in the actual design of PTAs (cf. Dür et al. 2014). Additionally, we kept the partner countries with which an agreement is signed constant. For citizens' attitudes towards trade agreements, however, we know that partner country perceptions matter (Spilker et al. 2016). It is thus plausible that this might matter for interest groups, too. This seems particularly likely for business groups, as their members' competitiveness may vary strongly by partner country. For them, the choice of partner determines whether they need to fear import competition or can benefit from access to interesting and profitable markets. Finally, interview data may allow to elicit interest groups' underlying rationale when evaluating PTAs. Such an approach may, for example, allow to get a better understanding of why import-competing and export-oriented business associations behave similarly. By providing systematic and experimental data on interest groups' attitudes towards trade agreements, this paper thus has made a first step towards a potentially very fruitful research agenda.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.