Whose Side are You on? Explaining the Extent to Which National Interest Groups Support States in Global Politics

This article builds on the growing literature on lobbying in a multilevel context. Whereas previous studies have sought to explain why national groups start lobbying at the transnational level, this articles examines the policy positions of interest groups once they are active at this level. The main argument is that group type is a key factor for explaining the level of interest group support for states, yet the effect of group type is conditional on the institutional and issue contexts. The data for analysis are derived from 198 interviews conducted with representatives from businesses and NGOs at the Ministerial Conference of the World Trade Organization and the UN Climate Summit negotiations in 2011 and 2012. The results confirm that business groups support states much more often than NGOs do. Yet, the amount of support varies substantially across a country's level of development and democracy, and the salience of issues.


Introduction
Opening transnational decision-making to interest group input has been both welcomed and scrutinized. Proponents have argued that increased interest group participation could enhance decision-making legitimacy by providing additional opportunities for marginalized groups to be heard (Fairbrass and Jordan, 2001;Keck and Sikkink, 1998;Mazey, 1998;Pollack, 1997). Later, more critical voices highlighted that interest group mobilization is essentially a biased process in which privileged groups tend to profit more. The fear was that, as in domestic politics, business groups would eventually come to dominate global institutions (Beyers, 2002;Mateo, 2012, 2014;Klüver, 2013;Tallberg et al., 2013).
Any normative evaluation of interest group roles in transnational politics, however, requires an empirical assessment of the extent to which business groups and NGOs become involved at this level, and why. There are two ways to approach this question. The first, and currently most common, approach is to start at the national level and see whether business groups or NGOs are more dominant in domestic politics and how this translates in a multilevel strategy. These studies have shown that, overall, business groups are more closely aligned to policy-makers in a domestic context than NGOs are, and are also better able shift their focus to the transnational level (Beyers, 2002;Mateo, 2012, 2014;Klüver, 2011;Schneider et al., 2007;Tallberg et al., 2013).
The second approach, and the focus of this article, is to start at the transnational level and see whether business groups or NGOs are closer aligned to state positions at this level. One would expect business groups and state positions to be more closely aligned at the transnational level, given their more privileged and influential position in a domestic context (Beyers, 2002;Mateo, 2012, 2014;Klüver, 2011;Tallberg et al., 2013). Yet, no study has analyzed whether this is actually true, and whether the preferential position of business groups in domestic politics also leads to more business than NGO support for states at the transnational level. The first aim of the article is therefore to explore the extent to which national business groups and NGOs support state positions during global political negotiations. The argument that is posed is that business groups indeed more often support states than NGOs across borders, just as we see in most domestic political systems, but that there is still much variation across particular countries and issues.
This article's second contribution is to explain the variation observed. Hereby the argument is that support for states across business groups and NGOs is conditional on the salience of issues, and a country's level of development and democracy. Recent studies show that business groups and NGOs make strategic decisions differently in relation to similar contextual influences (Binderkrantz and Rasmussen, 2015;De Bruycker, 2016;Dür et al., 2015;Dür and Mateo, 2014;Hanegraaff et al., 2016;Weiler and Brändli, 2015). With this in mind, I argue that business groups are more critical of their country's positions at global venues when an issue is more salient or when the groups stem from a less developed country. NGOs are more critical of their own governments when they originate from countries that are more democratic.
Original data stem from extensive interviews with 198 national interest group representatives from all over the globe (54 countries) active at two key global venues, the World Trade Organizations and the United Nations climate summits. Interviews were conducted between 2011 and 2013 at the key negotiation fora of these institutions.
Respondents were asked to indicate the extent to which they supported their government's positions at the conferences. These answers were linked to various key attributes of the organization, the issues they indicated to work on, and their country of origin. The analysis of the resulting data provides support for the hypotheses. In what follows, I first discuss the broader embedding of the article in the multilevel lobbying literature. I then explain my research design, and analyze the data. I end with some concluding remarks and provide an avenue for future research.

I. National Interest Groups in Transnational Political Systems
A key concern in debates in the political science literature relates to the opportunities global political systems provide to business groups and NGOs (Beyers and Kerremans, 2007;Dür and Mateo, 2014;Tallberg et al., 2013;Zürn, 2014). One hope associated with the rise of interest group activity in transnational politics is that marginalized groups have more opportunities to counter frequent business capture within states. Building on an early observation by Schattschneider (1960) that the losers in a conflict have an interest in shifting to a friendlier venue or in activating a friendlier audience, many scholars have argued that NGOs could bypass governments with which they disagree and become active beyond domestic borders to galvanize support among likeminded political actors (see also Dalton et al., 2003;Tarrow, 2005). This way they would compensate domestic losses via transnational means. However, this reasoning assumes that business groups would not actively pursue their interests abroad. Therefore, later scholars hypothesized that domestic winners would also have an incentive to develop multilevel strategies and would likely do so to a larger extent (Beyers and Kerremans, 2007;Dür and Mateo, 2014). For the same reasons that business groups are more dominant at home, they would also be more powerful transnationally and this way persist on domestic success at the transnational level. Ultimately, whether the international level serves more as a continuation of domestic politics or as a correction to it, is an empirical question. For this reason, scholars have studied the routes and underlying conditions of national groups within transnational political systems. For instance, studies have analyzed whether groups with an insider position in domestic political structuressuch as resourceful groups, groups with more access domestically, or those with strong networksalso become more active in transnational politics (Beyers and Kerremans, 2012;Dalton et al., 2003;Dür and Mateo, 2012;Kriesi et al., 2007). However, given the substantial normative concerns underlying this debate, one key focus has always been to analyze the multilevel strategies of business groups and NGOs to see whether they develop varying multilevel trajectories (Beyers and Kerremans, 2012;Mateo, 2012 Klüver, 2011;Kriesi et al., 2007).
Overall, this literature has concluded that national business groups more often lobby outside domestic borders and gain more access at this level. For instance, Dür and Mateo (2012) show that business groups and NGOs, controlled for domestic access, vary considerably in their access to a series of EU political institutions across Germany, Ireland and Spain. Similarly, Kriesi et al. (2007) find in a number of European countries that business is more likely to develop favourable positions at home and in the EU across seven European countries, France, Germany, the United Kingdom, the Netherlands, Spain and Italy. However, the effect of group type is not entirely undisputed; some scholars find no or little difference. For instance, studying interest groups in the trade policy field in Belgium, France, the Netherlands and Germany, Beyers and Kerremans (2012) find that organization type has only limited effect on the probability that groups compensate for domestic losses. Rather, both business groups and NGOs with a more favourable position at home are more likely to develop robust multilevel networks in Europe.
The literature to date has done well to identify which groups have a more dominant position in national political settings, and how this affects their propensity to engage in multilevel lobbying. Yet, what is less clear is how this translates in the support states get from business groups and NGOs at the transnational level. Can we simply assume that (most) business groups support states in global politics, while NGOs more often challenge their government? Or, is there a more complex story to tell? At this point we simply do not know, while answering these questions would provide an important addition the literature. First, and most importantly, the overall amount of support and conflict with which policy-makers are generally confronted, substantially affects the position they take in policy debates (see Baumgartner et al., 2009;Mahoney and Baumgartner, 2014). Hence, a closer view of the structure of conflict among interest groups at transnational political venues helps us understand why countries take certain policy positions and ultimately should help us understand why certain decisions are made at the transnational level, and what the role is of national groups in a multilevel setting.
Second, more broadly, the findings also contribute to the debate on whether the international level should mainly be seen as a continuation of national politics, in line with the persistence perspective on multilevel lobbying, or as a correction to it, following the logic of the compensation perspective (see Beyers, 2002;Beyers and Kerremans, 2007;Keck and Sikkink 1998;Kriesi et al., 2007). True, the data I use cannot directly link the full range of the multilevel strategies national interest groups employ to the positions they take abroad vis-à-vis their states, as I do not know which groups have not reached the international level. Yet, if we find similar patterns of support and opposition at the transnational level with what we know from domestic politics, this is a further indication that the international domain mainly serves as a continuation of groups' domestic influence by other, transnational, means (Dür and Mateo, 2014). In contrast, if we find that NGOs are closer aligned to state positions in transnational politics, this is a clear suggestion that interest group politics functions quite differently at the global than at the national level. An additional benefit of approaching this debate by looking at national groups already active at the transnational level is that it allows inclusion of a wide array of countries in the analysis. This allows examination of how country level factors affect the nature of interest group politics at a transnational level, something that is far less feasible when starting in the countries themselves.

II. Explaining Interest Group Support for States in Transnational Politics
This goal of this section is to construct expectations for why business groups or NGOs are more likely to agree with their government's positions abroad and why this degree of agreement might vary across countries of origin and issues. I begin with the main effect of group type and the amount of support the two groups have for their governments outside domestic borders. Given former findings, business groups seem more likely to support their government at the transnational level, whereas NGOs would be more inclined to oppose their government's views when active abroad. That is, overall, business groups have a more dominant position at home (Beyers, 2002;Dür and Mateo, 2012;Klüver, 2011;Schneider et al., 2007); hence, their interests should be more greatly reflected in the positions countries take at global venues (and vice versa for NGOs). Obviously, the fact that business groups are more dominant at home should not necessarily lead them to support their government more than NGOs do. Business groups may only become active at these venues once they have lost a domestic battle, despite their overall dominance. However, as a starting point, I assume that business groups are more influential in shaping domestic policy positions; therefore, they are more likely to support their government when the government is active abroad. NGOs, on the other hand should be less successful in shaping state positions, hence they should be more inclined to oppose their state abroad.
H1 : Business groups are more likely than NGOs to support their government when they are active outside domestic borders.
This objective of this article is not only to determine whether business groups and NGOs support states to a varying degree but also whether this support is conditional on the context in which they operate. I see at least two relevant conditional factors. Both have been mentioned as important factors driving interest groups' motivation to lobby in a multilevel setting (see Binderkrantz and Rasmussen, 2015;Dür and Mateo, 2016;Beyers and Hanegraaff, 2017), and I apply them to explain why business groups and NGOs support states at a transnational level.
The first contextual factor to consider is the national-institutional context from which business groups stem. More specifically, I focus on variations in the developmental and democratic levels in countries. Regarding the former, I expect variation in the position of business actors but not for NGOs across the level of development in countries. Many NGOs from developing countries are financed by developed countries. This fact implies that decisions to support or oppose government positions are in most cases likely not made in the country of operation but by decision-makers in NGOs located in developed countries (Bob, 2005;Dalton et al., 2003). Consequently, an NGO's position in a developing country reflects the NGO's position in the developed country. This fact is certainly in line with studies showing that southern NGOs strongly adjust to northern NGOs' strategic considerations because they do not want to jeopardize the funds they receive (Bob, 2005;Dalton et al., 2003;Thomas and Hrebenar, 2008). This fact means that the stakes of northern NGOs will often be reflected in the stakes of southern NGOs, and the level of support by NGOs across states with different levels of wealth will vary in rather limited ways.
For business groups, this situation is different. Business groups from developing countries do not depend on business groups in developed countries and should be able to formulate a position independently. As a result, we should find greater variation in support for states across countries based on the developmental level. More precisely, I expect business groups in developed countries to be less critical of their governments than business groups in developing countries because business groups from the latter countries have more reasons to be critical. In most cases, transnational negotiations are dominated by developed countries at the expense of developing countries. To illustrate, the current deadlock in the WTO's Doha Development Agenda is highly criticized by the developing world because it benefits developed countries while limiting the opportunities for economic growth within the developing world. However, despite this ongoing criticism, no progress has been made, nor is there any realistic outlook for an agreement in the near future. From this perspective, the business community from developed nations seems more content with the way in which their governments defend their interests globally than the business community in developing nations. Hence, I expect business groups from developing countries to be more critical of their governments than business groups from developed nations.
For the level of democracy, I expect variation among NGOs in the positions they take toward states. In non-democratic states, NGOs are not free to position themselves toward their governments. In most cases, they simply must follow their government's position, especially regarding the public good issues NGOs tend to defend, such as the environment, labour standards, or human rights (Bob, 2005;Dalton et al., 2003;Thomas and Hrebenar, 2008). This situation is different in democratic countries in which NGOs are free and can criticize their government. Hence, NGOs in democratic states are more likely than NGOs in non-democratic states to be critical of their government. In addition, various studies in the US and the EU have shown that increased NGO dependency on government funding leads them to be less critical of their government (Sanchez-Salgado, 2014). In democratic states, public funding is not the only source of NGO income; they are extensively funded by private donations as well. This situation is far less common in non-democratic states. In these countries, foreign financial aid to local NGOs is often forbidden or hindered, meaning that NGOs in these countries are more often financially dependent on the state (Bob, 2005;Thomas and Hrebenar, 2008). A consequence is that they are less critical of their key donor.
For business groups, the situation is different. First, they defend private goods that are less controversial for state leaders in non-democratic countries because others can still be excluded from these lobbying gains. As such, business groups can have an opposing opinion in non-democratic states with less risk than NGOs, making it more likely they will do so (Thomas and Hrebenar, 2008). In addition, business groups in non-democratic states are less dependent financially on their governments because they are financed by firms, often with an international network and cash flow (Fuchs, 2007). As a result, business groups, in contrast to NGOs, can afford to be more critical without worrying about the associated financial repercussions. Combined, these factors mean that business groups have fewer reasons to vary in the extent to which they support their state officials based on how democratic these states are. Rather, their support will be affected by other factors (such as the developmental level). For NGOs, we should expect differences directly related to their country's democratic nature.
Hence, the following two hypotheses will be tested: H2 : Business groups from developing countries are less likely than business groups from developed countries to support their government, while for NGOs there is no variation in position across countries with varying levels of development.
H3 : NGOs from non-democratic countries are more likely than NGOs from democratic countries to support their government, while for business groups there is no variation in position across countries with varying levels of democracy.
The third contextual factor I explore relates to issue context, namely, an issue's salience. Salience here refers to the amount of attention an issue gains from a country's media. The question is whether we can expect a difference between business groups and NGOs in the extent to which they support their states at transnational conferences given varying salience levels. The answer should be yes based on the types of interests both groups defend. As stated previously, business groups defend private interests, whereas NGOs defend public interests. On issues that garner limited media attention, all things being equal, business groups can lobby for private gains without overly considering public consequences. For instance, if discussions at climate change negotiations generate limited media attention, business groups can push for private gains because they do not have to defend their position to the public. However, once the negotiations gain traction in the media, business groups must link their private gains to public benefits to avoid criticism from the public and the media. As a result, on issues with low salience, the business community can fully commit to their government's positions, whereas on issues with high salience, they should move closer to NGO positions and be somewhat more critical of their government Mateo, 2012, 2014). In contrast, NGOs already defend public goods. This fact means that NGOs do not have to change their position when more citizens pay greater attention. Although they might invest more in lobbying, particularly in outside lobbying, once salience increases (Dür and Mateo, 2016), I do not expect this investment to have a significant effect on the position they communicate vis-à-vis policymakers. In other words, salience should affect the position taken by business groups, in that they become more critical of their government, but it should not affect the postion of NGOs. These factors lead to the following hypothesis: H4 : Business groups are more likely to support their government on issues that have low salience, while for NGOs there is no variation in position across issues with varying levels of salience.

III. Research Design
To test the hypotheses, this article relies on interviews conducted with 198 lobbyists who were active during transnational negotiations. More specifically, the data were collected during three major transnational negotiation rounds: first, in the field of global trade, the 2012 Ministerial Conference of the WTO in Geneva (MC hereafter); second, in the field of global climate change, the 2011 and 2012 Conferences of the Parties of the UN Framework Convention on Climate Change (COPs hereafter) in Durban (South Africa, 2011) and Doha (Qatar, 2012). Both negotiation rounds have major implications for domestic politics. Therefore, their decisions to employ multilevel strategies can be assumed to be related to strategic considerations rather than to a lack of knowledge about the venues or a lack of interest in them. The two conferences also differ on other dimensions. First, they vary in the attention they garner among interest groups and the broader public. The climate conference attracted much more attention from interest groups. Between 1997 and 2017, more than 10,000 organizations were active at climate negotiations (Hanegraaff 2015). This number is substantial. For instance, the EU transparency register lists approximately 11,000 organizations. In contrast, the WTO attracts more limited attention. Between 1995 and 2017, approximately 2,500 organizations were active (Hanegraaff et al., 2011). Moreover, the two conferences vary in the outcomes of the negotiations. The climate negotiations have resulted in a substantial agreement (Paris 2015), whereas the WTO has been in deadlock on most issues since 2001. This difference increases the generalizability of the findings. Still, because both conferences revolve around major negotiation rounds, one should be careful in externalizing these findings to cases in which negotiation outcomes are less inclusive and perhaps less consequential for countries.
At these conferences, a small team of three to four research assistants randomly asked lobbyists to participate in an interview of 30 minutes. During the interviews, respondents were asked to mention a specific issue they were working on and the strategies they used to influence policy-makers regarding this issue. At these three conferences, 348 lobbyists (from both national and global organizations) were interviewed; 198 of these interviews were included in this article. Some of the interviews could not be used because they were with representatives from global organizations (n = 71); others were not representatives of business groups or NGOs (N = 51) but were instead, for instance, representatives of research organizations. Finally, for some interviews, the dependent or independent variables could not be measured or key information was lacking. For instance, some attendees (n = 28) interviewed indicated that they did not lobby in favour of or against a specific issue. 1 The dependent variable 2 is the level of interest group agreement (see also Figure 1 in the Empirical Analysis section), ranging from full support to full opposition, with government delegates' positions during transnational negotiations. This variable is measured with one key interview question: 'On this issue [respondents were asked to indicate one issue being discussed during the negotiations that they were primarily interested in], to what extent do you agree with the position of your government regarding this issue on a scale of 1-5? 1 means completely agree; 5 means completely disagree'.Several independent variables are included. Hypothesis 1 addresses the main effect of group type. 'Business groups' combines all actors that defend business interests in any form, such as representatives from firms or business associations. In contrast, NGOs defend interests that are not directly related to their members and supporters' professions. These interests include, for example, environmental questions, human rights and social developmental organizations (Dür and Mateo, 2012). The study includes 71 representatives from business groups and 127 from NGOs. To test Hypotheses 2-4, an interaction variable is included between group type and other variables. To test Hypothesis 2, group type interacts with a dummy indicating whether an interest group stems from a developed or a developing country. Data regarding the distinction are retrieved from the World Bank, whose categorization is a combined construction of gross national income (GNI) per capita, the human asset index (HAI), and the economic vulnerability index (EVI). 3 For Hypothesis 3, group type interacts with a country's democratic nature, which is derived from the Polity IV index. Variation across this variable is skewed, with almost half of the countries scoring an 8 or higher on this index. Given that the Polity IV index considers countries that score an 8 or higher to be democratic countries, I compare democratic states 1 For a further description of the data collection, see online Appendix S1 and S2. 2 See online Appendix S3 for an overview of the variables used in the article. 3 See http://www.un.org/en/development/desa/policy/cdp/ldc_info.shtml.

Marcel Hanegraaff
(score of 8 or higher) with non-democratic countries (score of 7 or lower on this index; see also Hanegraaff et al., 2015). For Hypothesis 4, interaction between salience and group type, I rely on a question to respondents to indicate whether the issue gained much (0) or limited (1) attention in their country's media. Note that the results are thus based on self-reported information about the topic's perceived salience. Admittedly, actual salience does not necessarily correspond with how actors subjectively judge a particular issue's salience. Still, subjective measures are justified because advocates do not respond to an objective reality; their activities are mediated by their subjective appreciation of the context in which they operate. Moreover, given that lobbyists are specialists on these issues, respondents are unlikely to report much media attention if the media coverage was scarce, and vice versa. Therefore, a high correlation is likely to exist between actual and perceived salience.
Finally, several control variables are included. First, the article controls for interest group resource endowment. The lobbyists were asked about the number of staff their organization employs for advocacy and public affairs. Due to the skewed distribution, this variable is log-transformed. Second, the potential access points in domestic politics are controlled. If interest groups have options to lobby opposition actors, they could be more inclined to seek support domestically and invest less in transnational advocacy. For this reason, a measure is included for the level of opposition party support. Participants were asked to indicate the extent to which they received support from the opposition on the issue of concern on a scale of 1 (completely agree) to 5 (completely disagree). Third, I control for the number of groups active on an issue, because on issues with only limited interest group activity, policy-makers and interest groups are likely to be closer aligned. Leech (2001, p. 1207) suggest that 'in the cases where few interest groups are involved, we can be almost certain that few representatives of labor, citizens, or the nonprofit sector will be heard'. To make sure such a mechanism does not affect the results, I control for the number of groups working on similar issues. Finally, I control for resource competition, because groups that face strong competition to acquire resources, including from states, could be inclined to be less critical of their government (see Sanchez-Salgado, 2014). For this, I rely on a survey question, which asks respondents to what extent (high, medium or low) groups perceive that they face competition over resources. Again, this is a perceived measure, but, as argued earlier, I expect groups to respond mostly to what they perceive as true, rather than what is true.

IV. Empirical Analysis
First, some descriptive results are displayed. Figure 1 plots the overall variation in support and contestation across the entire sample. The results show great variation. In fact, the respondents in the sample show a normal distribution in the degree of support with government positions. Of the respondents, 34 percent completely or mostly agreed with their government delegates' positions. In contrast, 38.5 percent mostly or completely disagreed with their government delegates.
Finally, 27.5 per cent showed a moderate degree of agreement with their delegates. These findings illustrate nicely that there is a vibrant interest community active at these conferences to which both supporters and opponents of state politics find their way. From a democratic perspective, having a multitude of views mobilized in political processes tends to be a good thing (see Binderkrantz et al., 2017). However, this does depend on how these positions are distributed among types of group, as this might highlight certain biases or inequalities that might still exist in the system. For this, we turn to the explanatory part of the article.
The analysis presented in this article as well as the predicted probabilities figures are based on OLS regressions. 4 Additionally, several countries have repeated measures, which implies that I cannot assume entirely independent residuals. For this reason, I produce robust standard errors with a correction term based on the observed raw residuals (54 clusters on countries). The independent variables included in the model are those described previously with the addition of an interaction variable that combines the developmental levels and group type. The results are presented in Table 1, model 1. 4 I interpret the ordinal scores on a scale from 1 to 5 and apply a linear regression analysis. This can only be done if the distribution is normally distributed, which it is (see Long and Freese, 2006). Yet, to make sure this doesn't affect the results, I also ran ordered logistic regression analyses. The results however were almost identical, with no differences in the independent variables related to the direction of the coefficients or the levels of significance (see online Appendix S3). Because OLS regressions are easier to interpret for readers, especially for the predicted probabilities, I opted to include these in the main text, and the results of the ordered logistic regressions in online Appendix S4.    198 198 198 198 Note: linear regression; coefficients with standard errors in parentheses. Standard errors are adjusted for within-nation clustering (54). Significance levels: * p < 0.1; ** p < 0.05; *** p < 0.01.
In models 1 to 3, the dependent variable is the level of agreement with the government's position on an issue of concern mentioned during the negotiations. To recap, a high score indicates an interest group's opposing attitude toward its state's position during transnational negotiations, whereas a low score indicates a supportive attitude to their state's position during the negotiations. First, model 1 tests the organizational type variable. The results demonstrate, as expected, that organizational type is a strong predictor of support for state positions during transnational conferences. Business groups support states more often than NGOs do. This nicely complements earlier studies highlighting how business groups in many, albeit Western democratic, countries are more inclined to engage in transnational activity as a result of their beneficial position in domestic politics (Beyers, 2002;Beyers andKerremans, 2007, 2012;Kriesi et al., 2007;). These results portray that at the aggregate global level, business groups also support states more frequently than NGOs do. This certainly does not mean that all business groups support states or that all NGOs oppose their government at the global level. Yet it does show that, just as in domestic politics, states are more aligned to the interests of business groups, than those of the NGOs. This is an important finding as it suggests that interest group politics at the international level much resembles national political power structures among interest groups (Dür and Mateo, 2014).
I now turn to the conditional effects. First, I focus on the effects of the institutional structure. To this end, an interaction variable between group type and the level of development is added in model 3. In Figure 2a these effects are even more clearly portrayed. The figure shows that in developing countries, NGOs and business groups almost equally support policy-makers at international political venues. However, in developed countries, we see much more support from business actors (ŷ business drops from 3.14 in developing countries to 2.52 in developed countries), whereas the level of NGO support remains equal across developmental level (ŷ NGOs remains almost constant: 3.27 in developing countries to 3.26 in developed countries). The results support Hypothesis 2, that business actors support states more when active in wealthy countries, but for NGOs, the developmental level does not affect the propensity to support or oppose the state in global politics. More substantially, this finding supports the view that NGOs from developed countries are highly dependent on their donors, often NGOs, in developed countries (Bob, 2005;Dalton et al., 2003). These results highlight how this dependency might also affect the positions NGOs from developing countries take in global politics. For business groups such dependencies hardly exist, hence they are more likely to form their own positions at the conferences. On a positive note this also means that NGOs are less divided in their positions across countries with varying levels of development. This should make it easier for NGOs to form a homogenous lobby coalition across countries, hereby increasing the collective pressure they can apply to states. For business groups this seems less viable, given that their positions vary much more across countries with varying levels of development.
For the level of democracy, the opposite effect was hypothesized. In Table 1, model 2, the interaction effect is tested. The results indicate that democracy has a different effect for both group types. To visualize this effect, the predicted probabilities are displayed in Figure 2b. The results show that NGOs in non-democratic states are more critical of their governments than are NGOs in democratic states. In non-democratic countries, the support is slightly lower than 3 (on a scale of 1-5), while in democratic countries NGOs score 3.5 indicating that they are more critical of their government. Moreover, this difference is statistically significant. For business groups, this difference is neither statistically significant, nor substantially, different. This result means, in line with Hypothesis 3, that varying level of democracy has a strong effect on NGOs' positions at the conferences, whereas for business groups no differences are observed. In particular the finding that NGOs in nondemocratic countries are the least critical of their government is telling. It highlights the problematic position these groups are in if they wish to criticize their government or, even worse, that NGOs which have been critical of their government in the past have simply not survived. The findings on development and democracy combined suggest that business groups are more driven by material incentives, whereas NGOs are more driven by non-material, or normative, incentives, and this motivation leads them to be more receptive to contextual forces, either private (such as the developmental level) or public (such as a country's democratic nature). Finally, in model 3 (Table 1), the conditional effect of salience is tested. Again, the results show a different effect on the positions of business groups and NGOs. Figure 2c shows the predicted probabilities for both groups' positions at varying salience levels.
The results clearly show that NGOs are unaffected by an issue's salience (scores ranging from 3.3 to 3.2), whereas business groups vary according to an issue's salience. For low salience issues, business groups have a predicted score of 2.5, whereas on high salience issues this is 2.9. This means that for high salience issues, the support level of business groups for their government is lower and statistically even similar to the level of support of NGOs for their government. This result indicates that Hypothesis 4 is accepted. Importantly, these findings are in line with studies highlighting that business groups prefer to work in the shadows (Baumgartner and Leech, 2001;Culpepper, 2010;Dür and Mateo, 2014). That is, these results seem to suggest that governments are less inclined to align with business groups once the media is paying attention to an issue, while on issues with limited media attention, governments align much closer to the business community. Given that most studies indicate that international politics is becoming increasingly politicized, this might be a positive finding. That is, it might indicate that governments will increasingly become receptive to the opinions of the public in transnational politics, if this trend continues.
The control variables also yield interesting results. First, the results show that the resource endowment of interest groups is positively associated with their level of support. These findings suggest that, similar to the situation within most domestic political structures, states are closer aligned to well-endowed interest groups at transnational conferences. Another interesting finding is that increased domestic embeddedness leads to more agreement with governments at the conferences. While earlier research has already shown that groups that are more embedded in domestic political systems are more likely to mobilize transnationally (Beyers, 2002;Beyers andKerremans, 2007, 2014); this result indicates that this likely leads to a situation in which states are supported at global conferences by more embedded groups, and challenged by groups which domestically act more in the political periphery. Combined, the fact that states' positions are closer aligned to business, more resourceful and more embedded organizations provides an interesting view on how the opening up of transnational conferences for interest groups has worked out. At the very least, it questions the potential of international venues to compensate for any skewed power constellations in domestic politics. Rather, it seems that the inclusion of non-state actors in transnational politics has reinforced traditional power structures among states. These results warrant further focus on the support that different sorts of groups provide to states, and what this means for decision-making and power constellations at this level.
Finally, in contrast to our expectation, both competition and issue density have no effect on the level of agreement with governments. For competition, this might be an optimistic finding, as it suggests that organizations do not adjust their position once they are threatened by a lack of funding. Yet, it could also be that more specific data are needed here, for instance what type of funding is gained (state funding or from members), or where it comes from (their own state or through foreign aid). Finally, whether or not more interest groups are active on an issue, at least at the conferences, does not seem to affect the average level of support for states. This somewhat contradicts Baumgartner and Leech's (2001) assertion that for low interest group density issues, organizations should find it easier to convince policy-makers to align with their position. Yet, we need better measures to analyze whether the positions of states officials have indeed changed due to the increase (or decrease) of interest groups activity. This falls beyond the scope of this article, but, again, would be an interesting avenue for future research.

Conclusion
This article builds on the growing literature on lobbying in a multilevel context. Whereas previous studies have sought to explain why national groups start lobbying at the transnational level, this article examines the positions of interest groups once they are active at this level. My main argument is that group type is a key factor for explaining the level of interest group support for states, yet the effect of group type is conditional on institutional and issue contexts. The results confirm this assertion. Overall, business groups support states much more often than NGOs do. Yet, business groups support for states varies substantially across a country's developmental level and an issue's salience (whereas for NGOs it does not), and NGOs' support for states varies strongly across a country's level of democracy (whereas for business groups it does not).
These findings have several consequences for the literature, of which three especially warrant our attention. First, the findings build on former studies that have indicated that the group of actors with a more privileged position in domestic politics, to which business groups belong, are also more engaged in transnational politics. This article adds to this finding the idea that this fact has consequences for how states are supported by interest groups in global politics. Business groups more often defend their positions abroad, whereas NGOs more often challenge their governments' positions. This finding is an important addition because we know from former research that individual lobbying behaviour is ineffective even when these actors are powerful; instead, interest groups' true power resides in the strength of their entire side (see Baumgartner et al., 2009;Dür et al., 2015;Klüver 2013;Mahoney and Baumgartner, 2014). This article's findings show that the most powerful actors in global politics, business groups, are mostly on the governments' side. This fact severely limits the opportunities for weaker actors, such as NGOs, to compensate effectively for domestic losses. As such, these findings are directly relevant to the compensation versus persistence debate, providing yet another indication that opening the transnational domain to interest group input has not led to greater equality in global politics. Rather, instead of being an alternative to domestic action for a group with limited resources or limited access, transnational action is primarily a continuation of groups' domestic influence by other, transnational, means (Dür and Mateo, 2014).
Second, by adding conditional effects to explain interest group positions in transnational politics, this article builds on the recent literature that shows that business groups and NGO strategic choices are often affected differently by contextual influences (Binderkrantz and Rasmussen, 2015;De Bruycker, 2016;Dür et al., 2015;Dür and Mateo, 2014 ). The logic underpinning these findings is that NGOs and business groups defend different interests and have conflicting goals leading contextual forces to have a different effect, at least to a different degree, on the strategic choices these groups make. To avoid analyses that are underspecified, future studies should more frequently include the conditional effects when explaining interest group strategies and, ultimately, how these strategies affect these groups' influence in the political arena (see Dür et al., 2014;Dür and Mateo, 2016).
Third, most studies focusing on multilevel lobbying by business and NGOs focus on the European Union. Less commonly studied is the role of both actors in world politics in a single research design (but see Tallberg et al., 2018). This is surprising because in most cases, business groups and NGOs fight on opposite sides of a conflict (Dür et al., 2014). This means that any evaluation of either one's role and impact is linked to the other's activities and success. Therefore, more studies comparing the group types would help us better understand the role and impact of NGOs and business groups in world politics (Tallberg et al., 2018).