across decision-making levels

Interest group influence on the political agenda across decision-making levels Anne Skorkjær Binderkrantz Aarhus University Anne Rasmussen Copenhagen University Leiden University Paper prepared for presentation at the Annual Meeting of The Danish Political Science Association Vejle, October 24 rd -25 th First draft. Please do not cite 1

Abstract Despite a large increase in studies on EU interest representation, the vast share of the existing literature has focused on interest group strategies rather than influence and rarely compared EU and national policy-making. To fill these gaps we analyze the patterns of perceived agenda setting influence on EU and national policy-makers. More specifically, we examine whether the explanatory potential of factors at the national, policy and interest group level vary when it comes to persuading national and EU level policy-makers to initiate new issues. The analysis is based on surveys of domestic interest groups in the United Kingdom, Denmark and the Netherlands. Our findings indicate that agenda influence on EU policy-makers is considerably lower than on national ones but finds little variation between national contexts in absolute levels of influence on national and EU decision-makers. Despite frequent criticism of the prominent role of business interests in the EU, we also do not find strong evidence of comparative advantages for business groups in dealing with EU as opposed to national policy-makers. Instead, an organization s policy portfolio has a substantial impact on the size of the gap between influence on national and EU policy-makers. Relative increases in an organization s EU focus have a positive effect on perceived agenda influence on EU policy-makers and a negative effect on its ability to influence national ones. Moreover, staff resources play a stronger role in the EU level playing field than the national one when it comes to ensuring agenda influence on an occasional basis. 2

1. Introduction European interest groups operate in a multi-level system, where they are affected by decisions made at different levels. Therefore, many domestic groups do not only operate at the national policy level but have incorporated the EU level in their political action repertoire (Beyers & Kerremans, 2012; Dür & Mateo, 2012). In fact, even when it comes to influencing EU policy, interest group lobbying in the multi-level EU system often takes place at both the national and EU level. Lobbying tends to be cumulative with most domestic groups beginning their efforts at the national level and subsequently extending their work to the European level (Beyers, 2002: 607). Even though the EU is of at least some relevance to most if not all domestic groups, existing studies demonstrate significant variation in the extent to which groups have Europeanized their lobbying efforts (Beyers & Kerremans, 2012; Dür & Mateo, 2012; Eising, 2004; Klüver, 2010). A relatively small, but cumulative literature has made significant progress in describing and explaining interest group lobbying and access to the EU. Here, it has been demonstrated that the extent to which groups target the EU level varies for example across different group types, across different policy issues and across countries (Beyers & Kerremans, 2012; Dür & Mateo, 2012; Beyers, 2002; Eising, 2004; Klüver, 2010; Dür & Mateo, 2013; Greer, Fonseca & Adolph, 2008). The issue of group access and influence in the EU is of relevance to discussions of the socalled democratic deficit in the EU policy process (Greer, Fonseca & Adolph, 2008). At the same time, studying influence on EU policy separately from influence at the national level leaves us with no good benchmark for judging whether specific influence patterns at this level are democratic. Finding that certain group types, or groups with certain policy portfolios, dominate EU agenda setting may for example not mean that the EU is any less democratic then its member states. Moreover, the multi-level character of the European political system raises the important issue of whether the same groups are influential at the domestic and the EU level. The issue of whether the 3

same actors are active at different decision-making level also echoes a classic theme in the interest group literature focusing on whether different arenas provide possibilities for different types of interests or whether the same groups are able to dominate across all political arenas (Schattschneider, 1975 [1969]; Truman, 1951). While many studies have focused solely on the factors affecting group strategies at the EU level, we therefore follow the recent study of group focus by Dür and Mateo (2013) in incorporating the domestic and EU level in a single study. We aim to bring the literature a step further by focusing on the factors affecting perceived interest group influence on the agenda of the EU compared to interest group agenda setting influence at the national level. From a democratic perspective, it is particularly interesting whether the factors affecting group influence at the national and the EU level differ. This focus allows us to address questions such as: Do resources matter more or less for influence at the EU level compared to the domestic level? Are groups who are well-integrated in national decision-making processes also more influential at the EU level? And how does a group s policy portfolio affect its level of agenda setting influence at the different levels? Interest group influence may come in many shades from behind-the doors attempts to push the political agenda of bureaucrats and politicians, over lobbying for specific amendments to regulation, to attempts at affecting the implementation of political decisions. We focus on agenda setting influence because this may be considered the least demanding type of influence, yet an important step in influencing policy making. This provides us with greater variation in the dependent variable because more groups can be expected to report success with affecting political agendas than specific decisions. We conduct a multinomial regression analysis of different degrees of perceived agenda influence at the national and EU level based on survey data with responses from almost 1,500 domestic interest groups in Denmark, the Netherlands and the United Kingdom. This allows us to 4

include factors at the group, policy and country level in a single study of perceived influence at the national and EU level. This articles contribution thus lies in: 1) the focus on perceived agenda setting influence, 2) the incorporation of the domestic and the EU level in a single study and 3) the explicit comparison of similarities and differences in what explains perceived agenda influence at these two levels. Our analysis finds similarities between the patterns of perceived agenda setting influence at the national and EU level. Agenda influence on national decision-makers is considerably higher than on European ones and this pattern is consistent no matter what country we examine. We also see that despite differences in the character of state-society structures there are hardly any differences in absolute levels of influence on EU and national decision-makers between the three countries. Moreover, we only find modest differences in perceived agenda influence between business groups and other interest groups. Overall evidence of comparative advantages for business groups in dealing with EU as opposed to national policy-makers is weak. In other respects important differences in the ability of groups to persuade national and EU level decision-makers to initiate new issues are present. An organization s policy portfolio affects the gap between perceived influence on national and EU policy-makers. The higher the relative EU policy focus of an organization the higher its perceived agenda influence on EU policy-makers and the lower its perception of national agenda influence. Conversely, formal embeddedness in national advisory bodies is a more important factor for securing agenda influence on national as opposed to EU policy-makers. Finally, staff resources are more important for exerting some degree of agenda influence on EU than national policy-makers. However, we see that whereas increases in political staff are important for influencing EU policy-makers on an occasional basis, the effect of staff increases on exerting influence in the EU playing field often is more limited. 5

We begin by discussing the role of domestic interest groups in the EU multi-level governance system and propose a set of hypotheses on the factors affecting agenda setting influence at the two levels. After a section presenting the research design, our analysis of group agenda setting influence at the national and EU level is presented and discussed. 2. Explaining group influence at different levels 2.1 Interest groups in a multi-level system European interest groups operate in a multi-level governance system, where decisions of relevance to groups are made at different levels (Eising, 2004). While domestic interest groups could traditionally stick to the national arena when pursuing their interests, nowadays most groups cannot ignore that EU level decisions affect their interest domain. With the increased decision-making powers of the EU in many policy areas, powerful incentives exist leading groups towards being active in EU. The process by which national political actors such as interest groups adapt themselves to the impact of European integration is typically termed Europeanization (Beyers & Kerremans, 2007). While the EU has been rising in importance, there is no reason to expect that the EU replaces the national level for most domestic interest groups. Some policies are still exclusive national competence and even in EU policy areas a considerable share of the policy activity takes place at the national level and/or involves national actors. Interest groups are after all tied to their national members and to the national context in which they emerged (Eising, 2004). Even empirical studies which specifically focus on EU policy-making indicate that the national level has first priority. Beyers for example concludes that: Most actors start to push their interests domestically and only after these networks are established do they extend their networks to the European level 6

(Beyers, 2002: 607). In general, groups direct the greatest share of attention to the most proximate lobbying venues (Beyers & Kerremans, 2012). The last decade has witnessed increased scholarly attention to interest group politics generally and EU lobbying in particular. However, rather than integrating the study of EU and national level interest representation, there has been a certain division of labor between scholars focusing on interest group behavior vis-à-vis national and EU policy-makers. Some studies have, however, taken a different point of departure asking to what degree national level interest groups target the EU in their political work (Beyers & Kerremans, 2012; Dür & Mateo, 2012; Beyers, 2002; Eising, 2004; Klüver, 2010; Dür & Mateo, 2013; Greer, Fonseca & Adolph, 2008; Beyers & Kerremans, 2007; Eising, 2007; Constantelos, 2007; Falkner, 2007). Here, it has been established that many domestic groups have an action repertoire including the EU level but that even in EU related areas groups may naturally also seek to affect their own government. Overall, much support for the multilevel governance perspective for understanding interest group strategies has been generated (Eising, 2004). A common finding is that the degree of Europeanization varies between groups (Beyers, 2002). The extent to which groups incorporate the EU in their political work differ with some groups sticking entirely to the national level and others developing intense lobbying strategies directed at the EU. When it comes to seeking influence on EU policy specifically, national groups are more prone to lobby national venues than EU-level venues. Beyers and Kerremans thus describe a cumulative system, where groups first tend to lobby more proximate venues before adding more distant venues (Beyers & Kerremans, 2012). The degree to which groups target the EU directly is mediated by domestic institutions, policies, culture and identities (Beyers & Kerremans, 2007). The focus in this literature is mainly on group strategies and to some extent on access to the EU. Also, much of the literature focuses on one specific policy area (Klüver, 2010; Greer, Fonseca & 7

Adolph, 2008; Beyers & Kerremans, 2007) or one specific type of group (Eising, 2004; Constantelos, 2007). As discussed below this may explain some of the partly contradictory results found. Thus, according to Beyers and Kerremans (2012: 265): the functioning of interest groups in multilevel and federal systems is, despite some recent research, a largely underexplored terrain as most empirical research remains confined to horizontal venue shopping situated at one territorial level. The most encompassing study so far has been conducted by Dür and Mateo (2012; 2013). However, there is a lack of studies which examine access and influence patterns of interest groups in national and EU level policy-making within a single analysis design. As a result, our design includes both the EU and national level in a study of perceived agenda influence of interest associations across all policy fields in several countries. 2.2 Factors affecting agenda influence The EU may be seen as an alternative for groups excluded from influence at the national level. It may also be seen a venue exploited by groups that are already successful at the national level. Beyers and Kerremans (2012) thus point to two different lines of argument concerning group presence at the EU level. One view argues that multilevel governance benefits groups that are already well integrated at the national level. Another, that EU provides a compensatory arena primarily inhabited by groups finding themselves constrained by the national context. In general, most studies conclude that EU interest representation is biased in favor of those who are already powerful in the member states (Greer, Fonseca & Adolph, 2008: 427). We formulate predictions about the factors expected to determine perceived agenda setting influence in general as well as how the effect of certain factors in influencing national and EU level policy-makers may differ. 8

Some explanatory factors may affect agenda setting influence at the EU and national level with varying strength or sometimes even in an opposite way. National as opposed to EU level agenda influence A first crucial factor to consider in explaining perceived agenda influence relates to the decisionmaking level of the actors who the groups are trying to influence. Based on the findings of comparatively low levels of attention towards the EU among domestic interest groups, we would expect them to demonstrate a lower rate of success when it comes to affecting the agenda of EU level actors than national level ones. EU policy making is more distant for the average interest group and, all things equal, the average interest group can also be expected to devote a lower share of its overall level of policy activity to EU than national matters. Many policy competences have not been transferred to the EU level. Moreover, even in the areas where competence has been transferred to the EU level, national interest groups are often not responsible for all the lobbying activities on their own but can delegate a considerable share of the work to the European-level groups. As a result, our first hypothesis is therefore: H1: Interest groups are likely to be have more influence on national than EU-level policy-makers In addition, factors at the country, policy and group level may be emphasized when it comes to specific variables affecting group influence at the national and EU level, The contextual level: country, policy focus and national embeddedness The discussion of differences in the overall structures in interest group-government relations has historically played a prominent role in the interest group literature. Adopting a macro perspective, 9

scholars have developed system level theories of how the character of competition between groups and the access of different types of groups to policy-makers varies between systems with differences in overall state-society structures. A key distinction has been between pluralist societies which provide relatively open and easy access for a broad range of different interest groups and corporatist ones with substantial entry barriers for groups and where a narrow set of groups often enjoy privileged access (Schmitter, 1974). There tends to be variation between systems with different state-society relations in how dispersed access and influence is between different types of groups. A corporatist system is well known for providing privileged access to the key groups involved in policies of crucial importance to the coordination of the economy, such as employer and employee organizations (Rokkan, 1966). In contrast a pluralist system provides more equal access to different types of groups by establishing looser bonds to a broader range of actors (Eising, 2004). However, the fact that these theories allocate different types of groups a very different role when it comes to influencing public policy does not imply that an average interest group should be more or less successful in systems with a given set of state-society structures than in another one. As an example, it is highly plausible that the advantages of some groups in corporatist systems are evened out by the disadvantages of others. As a result, we expect that H2a: Differences in state-society structures between countries do not affect the agenda influence of an average interest group on national policy-makers When it comes to looking at EU interest representation, scholars have considered whether variation in national structures of interest representation lead to differences in the access and influence enjoyed by interest groups in EU policy. The overall argument is that the national institutional 10

context channels and constrains interest group behavior and mediates the impact of European integration (Klüver, 2010: 180; Falkner, 2007). The specific nature of the country effect is, however, disputed. Some scholars focus on the general level of access at the national level and the consequences hereof at the EU level. Others focus on the degree of fit between the national mode of interest intermediation and the EU institutions (Klüver, 2010: 176). Eising belongs to the latter group of authors arguing that negotiation capacities are important for group influence at the EU level. He therefore expects groups operating in neocorporatist settings at the national level to adapt more easily to the EU system (Eising, 2004: 218). Groups in such countries can exploit bargaining skills developed in their national systems of representation at the EU level. According to such an argument, we could therefore expect a smaller gap between national and EU level agenda influence for groups in such systems than for groups in non-corporatist systems. However, Beyers and Kerremans (2012: 284) find support for a contrary line of argument with for example French organizations being more active in the EU than Dutch. They argue that this is because groups in a statist system are used to break in the system whereas groups in corporative settings are less stimulated to develop the skills to compete with other groups for access (Beyers & Kerremans, 2012: 284). Klüver supports this line of reasoning arguing that groups in statist countries are more likely to Europeanize because of their lack of access at the national level (Klüver, 2010). Reflecting the different arguments in the literature as to how country context can be expected to matter, we leave the direction of the effect open and simply hypothesize that, H2b: Differences in state-society structures between countries affect the agenda influence of interest groups on European policy-makers, 11

In addition to the national context, the policy portfolios of interest groups may affect their perceived agenda influence. Interest groups are first and foremost involved in a set of policies drawn from the types of members or causes they represent. Several authors have drawn attention to the importance of policy-specific and sector-specific characteristics, when it comes to explaining the role of EU in the political work of domestic interest groups (Greer, Fonseca & Adolph, 2008: 405; Beyers & Kerremans, 2007: 463; Falkner, 2007). Falkner has even argued that the importance of policy sector may be so significant as to reduce any generalization (Falkner, 2007: 224). Most notably, even though most policy areas are to some extent affected by EU level regulation the effect is not uniform across policy areas. One reason for many groups to stick to the national level is thus simply that they are still able to realize many of their political goals at the domestic level (Beyers & Kerremans, 2007: 477). Other groups have a policy portfolio that is much more Europeanized. According to an old pluralist argument groups mobilize in accordance to disturbances to their fundamental interests and will therefore be most active when and where there is something at stake for them (Truman, 1951). When it comes to obtaining agenda influence a group s relative EU focus is therefore likely to affect its allocation of lobby efforts between EU and national decision-makers. Given that such efforts translate into some element of lobbying success, we can therefore also expect a link between relative EU focus and perceived agenda influence. As a result, we hypothesize that, H3: Increases in the relative EU focus in the policy portfolio of an interest group has a positive effect on its perceived agenda influence on European policy-makers and a negative effect of its perceived agenda influence on national policy-makers 12

The perceived agenda influence of groups may also depend on domestic institutional factors, more specifically on the access societal interests enjoy to domestic public actors. It follows directly from some of the theories of overall difference in state-society structures discussed that interest groups are not necessarily equal. Some may enjoy privileged access to the national policy process by being represented in national advisory bodies and other types of committees. Assuming that such access is not purely ceremonial, we would also expect that the groups that have been selected for such positions enjoy a greater degree of influence on national agenda setting than other groups. Moreover, it may even be the case that such privileged positions enjoyed at the national level can affect an interest group s agenda setting influence not only on national but also EU level decisionmakers. Beyers (2002: 585-586) argues that the supposed Europeanization of domestic interest associations may be substantially stimulated or constrained by the degree to which they are embedded in domestic policy networks. The role of such integration is contested (Beyers, 2002: 591-592). On the one hand, access to domestic policy making, may be a resource that can be utilized also at the EU level. On the other, groups excluded from national policy making may seek access to the EU to a larger degree than other groups. Empirically, there is most support for the first position with positive effects of domestic access on access at the EU level (Beyers, 2002: 608). Beyers and Kerremans even conclude that:..relational variables such as access gained to domestic policymakers or the distance visà-vis governing political parties generate significant and substantial explanatory power. In comparison, organizational resources or the policy sector in which organizations are active do not significantly predict the extensiveness of multilevel venue shopping (Beyers & Kerremans, 2012: 284). 13

We therefore expect integration into national policy making to be an important predictor of influence not only when it comes to putting issues on the domestic but also on the EU agenda, even if the effect is likely to be most pronounced at the domestic level. H4: Groups integrated into domestic political systems have more influence on the national and EU agenda, but the effect is strongest at the national level. The interest group level: resources and group type Business groups have continually been singled out as more prominent in the political system than other types of groups (Schattschneider, 1975 [1969]; Schlozman, Verba & Brady, 2012), although some also question the power of business (Smith, 2000). As such we would expect business groups to play a strong role no matter whether we examine national or EU policy-makers. At the same time, different types of groups may have different rates of success when it comes to setting the EU and national agenda. Specifically, Beyers argues that European policies basically concern the deregulation of specific economic domains, and therefore that concentrated producer interests are more focused on and adapted to Europe (Beyers, 2002: 590). Business groups thus have more welldefined constituencies with concentrated gains and losses from specific pieces of EU legislation (Dür & Mateo, 2012: 972). Business associations are also more likely to possess information and expertise than other types of associations, which may be particularly useful for influencing EU policy which is often regarded as technical. The EU policy portfolio consists of a range of regulatory policies involving a high degree of complexity and relevance for especially business interests. In support of this, previous studies have found that business groups invest a greater proportion of their time on EU legislation than non-business groups, although the effect seems to interact with the possession of resources (Dür & Mateo, 2012: 983; Dür & Mateo, 2013: 27). This 14

leads us to suspect that any advantages enjoyed by business groups in agenda setting should be most pronounced at the EU level. H5: Business groups have more influence than other groups on the national and EU agenda but the effect is strongest at the EU level. The next factor of relevance is interest group resources. Resources are crucial because they enable interest groups to be continually involved in monitoring and lobbying. Several studies have found that well-endowed groups are more active and more likely to gain access at the domestic as well as the EU level. Also in the European multi-level system, several studies have found that wellendowed groups are especially able to make use of the opportunities provided by the (Beyers & Kerremans, 2012; Dür & Mateo, 2012: 983; Eising, 2004; Greer, Fonseca & Adolph, 2008: 428; Beyers & Kerremans, 2007: 462; Binderkrantz, Pedersen & Christiansen, 2012). Groups differ in their possession of resources such as finance and staff. While these differences are crucial for their level of access and influence at both the domestic and EU level, we expect such resources to matter most at the EU level. Two arguments are relevant for the expected difference in the role of resources. First, according to Klüver, the EU institutions are understaffed and in need of external expertise. The ability of groups to deliver such expertise is related to their financial and personal resources (Klüver, 2010: 179). Second, for domestic groups, being active at the national level is typically most crucial for the relation to group members. Groups will thus prioritize their resources first towards the domestic level and only be active at the European level if additional resources are available. Based on this argument we expect that: 15

H6: Group resources affect influence on the national and EU agenda but the effect is strongest at the EU level. 3. Research Design The ambition of the study is to test the importance of different factors for perceived agenda setting influence at the national level and the EU level. The study includes the United Kingdom, Denmark and the Netherlands allowing us to investigate possible country level differences. It draws on largescale surveys among domestic interest groups in the three countries. We have restricted the study to associations that is groups with members or supporters. This enables us to include a large number of groups operating in a wide range of policy areas and exhibiting different levels of resource possession and integration into national decision-making. Identifying relevant and comparable interest group populations across countries is challenging. We are interested in comparing the influence of domestic interest groups at the national level and the EU level. Therefore, the surveys were conducted among politically active groups. These groups were identified in a wide range of sources. Specifically, we identified groups in: 1) The UK: written and oral evidence to parliament, meetings with ministers, appearance in the national newspapers: the Guardian and the Telegraph, and responses to selected government consultations, 2) Denmark: letters to parliament, membership of government committees, appearance in the national newspapers: Politiken and Jyllands-Posten, and responses to consultations on bills and 3) The Netherlands: letters to and participation in meetings in the parliament and responses to government consultations. 16

We obtained response rates of respectively 15 percent (UK), 65 percent (Denmark) and 53 percent (The Netherlands) 1. An ongoing follow-up on the UK survey will hopefully lead to a somewhat higher response rate, although UK groups are known to be particularly reluctant in responding to surveys. The absolute number of respondents in the three countries are: 339 in the UK, 520 in Denmark and 636 in the Netherlands. We focus the analysis on perceived agenda setting influence. Generally, the proportion of national groups reporting to be influential in relation to the EU is rather low. Agenda setting may be considered the least restrictive type of influence and using this measure therefore provides us with more variation simply because a higher share of groups reports at least some success even at this decision-making level. More specifically, we compare survey questions where respondents are asked how frequently the EU institutions and national politicians respectively have taken up an issue suggested by an interest group. The question of how to measure interest group influence is highly debated in the literature (Dür, 2007; Leech, 2010). The advantage of focusing on perceived or attributional influence is that it allows us to rely on surveys and thus to include a large number of groups in the analysis as opposed to conducting a study of preference attainment, in which policy preferences and outcomes on specific proposals are mapped (Dür, 2007). The disadvantage is that groups may have incentives to over- or underreport their level of influence. However, studies have found a relatively high level of convergence between measures of perceived influence and other measures of group access and influence (Pedersen, 2013). Further, the potential problem is less relevant for comparisons of the factors affecting influence at different levels than for studies of the absolute level of influence 1 In Denmark, the survey was sent to a larger population of groups including groups not appearing in political arenas. In order to make the survey respondents as comparable as possible across countries these groups have been excluded in the present paper. 17

obtained by groups. There is thus no reason to expect groups to differ systematically in their evaluation of influence at the national as opposed to the EU level. Because we are interested in comparing perceived agenda setting at the national and EU level and in examining whether the explanatory factor of perceived influence vary between these two levels we use a stacked dataset. Perceived national and EU level influence is stacked in one general influence variable meaning that each respondent appears twice in our regressions. In these regressions, the unit of analysis is therefore each individual respondent-agenda measure dyad rather than the total number of survey responses. Because the background characteristics for each respondent are the same no matter whether we look at national or European level agenda influence, we cluster the standard errors by respondent to avoid artificially low standard errors (Primo, Jacobsmeier & Milyo, 2007; Moulton, 1990). Descriptives for our variables are shown in Table 1 and 2. We distinguish between Often, Occasionally and Never having agenda setting influence 2. Hence, even if the level of influence can be ordered, the Brant test indicates that the parallel regression assumption is violated, which makes ordinal logis tic regression inappropriate in the current study. We use multinomial logistic regression with the middle category as the baseline to maintain the ordinal nature of the comparison. <TABLE 1> <TABLE 2> Our first independent variable is a dichotomous measure of whether the perceived agenda influence measure relates to the national or EU level, whereas our second variable is a categorical variable for 2 Our survey question used a four point scale asking respondents to distinguish between very often, fairly often, occasionally and never having influence. However, due to the limited number of respondents reporting to have influence on EU agenda setting very often, we have collapsed the two upper categories. 18

national context. The third independent variable reports the share of policy activity of the organization in core EU policy areas. We calculated this measure based on questions where individual organizations were asked to report their degree of activity on a scale from 1 to 4 (max activity) in18 different policy areas 3. We added the overall score of activity per interest group and subsequently calculated the share of overall activity which fell with into four policy areas where the EU competences lie exclusively or to a very high extent at the EU level, i.e. a) Research, technology and communications policy, b) Industrial and consumer policy, c) Agriculture, fishery and food policy, and d) Environment and energy policy. Classifying the degree of EU competence in policy areas is of course a difficult exercise, since many general policy areas may involve an element of EU activity. However, the four selected policy areas distinguish themselves by their strong degree of EU competence. The fourth independent variable measures the national embeddedness of the group based on a survey question where groups indicate how frequently the group has been represented in public council, boards, and committees. The advantage of assessing national embeddedness based on a response question looking at actual membership of advisory bodies compared to other methods of participation with open access (e.g. national government consultations) is that the former indicates a more selective type of access than the latter (Öberg et al., 2011). The fifth independent variable is a dichotomous measure reporting whether the group in question was a business group coded based on the Interarena coding scheme for interest groups (www.interarena.eu). Our interest group survey respondents were coded into business groups and 3 The 18 policy areas were Labour market policy, Urban and housing policy, Research, technology and communications policy, Defence and security policy, Industrial and consumer policy, Refugee and immigrant policy, Religious policy, Local government and regional policy, Culture and sports policy, Agriculture, fishery and food policy, Environment and energy policy, Monetary, fiscal and tax policy, Law and order/justice policy, Social affairs and families policy, Health policy, Traffic and infrastructure policy, Education policy, and Foreign affairs. 19

remaining interest groups based on information on their websites. Finally the sixth independent variable is the number of full term staff of a given organization engaged in political work defined broaly as contact with civil servants, politicians or journalists; generating analyses or research for policy advocacy, or monitoring the political process. We start by presenting a multinominal logistic regression model with all the main effects of our independent variables followed by a model which includes interactions between the national/european agenda setting level and the different explanatory factors to test for differential effects at these two levels. Our tables present the relative risk ratios, whereas we report predicted probabilities with 95 per cent confidence intervals for all the three levels of our dependent variable in figures. 4. Analysis As expected in hypothesis 1 we see a strong effect of decision-making level in our first regression. According to model 1, the relative risk of exerting agenda setting influence never compared to occasionally is expected to increase by a factor of 8.857 when we move from national to EU policy-makers given that the other variables in the model are held constant. Conversely, we see that the relative risk of persuading policy-makers to take up an issue often as opposed to occasionally decreases by a factor of 0.239 when the targeted decision-makers are at the EU rather than the national policy level. <TABLE 3> In model 1, we find some small overall differences in agenda setting influence between countries, some of which are statistically significant. It is for example the case that the Dutch organizations are less likely than the British ones to never exert agenda setting influence as opposed to 20

occasionally whereas the Danish and British organizations do not differ in this respect, Instead, we see that the Danish organizations are somewhat less likely than the British organizations to exert agenda setting influence often as opposed to occasionally. The limited variation between countries is also apparent when we consider national and EU level policy influence separately in Figure 1. Agenda influence on EU policy-makers is consistently ranked lower no matter which country we look at. As expected in hypothesis 2a there is no significant variation in agenda influence on national policy-makers except that the British groups are slightly more likely to never have influence on national policy-makers than the Dutch ones are. All other confidence intervals overlap. Turning to influence on EU policy-makers, we also do not find evidence in favor of any of the two contradictory predictions in the literature that groups from neo-corporatist countries should be either disadvantaged or advantaged compared to groups from systems with different state-society structures. The variation in agenda influence is not statistically significant except for a few exceptions. Hence, the Dutch respondent are somewhat more likely to often and less likely to never have influence on EU policy-makers than the Danish ones. Somewhat ironically, the only significant variation we find at the country level with regard to EU agenda influence is thus between countries with similar rather than different state-society structures, which contradicts the expectation in hypothesis 2b. Despite a few smaller, statistically significant differences between countries, the overwhelming share of evidence does therefore not support that national context is an important factor in explaining variation in interest group agenda influence no matter which decision-level we examine. As a result, the gap between agenda influence on national and EU level decision-makers therefore also does not vary much between the countries with the exception of the Danish groups 21

that witness a somewhat stronger gap (especially with regard to occasional influence) between national and EU policy-makers than the other two countries. <FIGURE 1> Unsurprisingly there is no effect of the share of EU related out of total political activity on overall agenda influence in model 1. Instead, we see a strong interaction effect in the expected direction in hypothesis 3. As illustrated in figure 2, EU policy activity has a positive effect on EU level agendainfluence and a negative one on national agenda influence. The higher the share of activity in areas of EU competence, the more likely a group is to never have influence on national and the less likely it is to never have influence on EU policy-makers. Conversely, EU involvement increases having influence on EU policy-makers both occasionally and often and it decreases having influence on national policy-makers occasionally and often. We even see that for those organizations that have the highest EU involvement there is no longer a positive gap between influence on national and EU level policy-makers. Agenda influence on policy-makers at the two levels is no longer significantly different. <FIGURE 2> According to model 1, national embeddedness has the expected overall effect on agenda influence in hypothesis 4. The stronger a group is represented on national boards and committees, the lower the likelihood that it never has influence on policy-makers as opposed to having influence occasionally and the higher the likelihood that it has influence often as opposed to occasionally. Figure 3 shows that the direction of this relationship is the same no matter whether we deal with national or EU policy-makers. In other words, national embeddedness is a factor that 22

has a positive effect on influence at both levels as we hypothesized. At the same time, there is also strong evidence of an interaction effect between national embeddedness and policy level. As expected, national embeddedness exerts a stronger influence on national than EU-level policymakers. The slope in Figure 3 is considerably steeper with regard to national than EU policy when we examine the predicted probability of often having agenda influence. This means that the gap between often having influence on national and EU-level policy-makers is generally much wider for groups that are very often as opposed to never represented on national boards. This also means that whereas there is also a weak positive relationship between national embeddedness and EU agenda influence, we only find significant differences in EU agenda influence between groups located at the extreme ends of national embeddedness scale. Moreover national embeddedness does not affect the likelihood of having influence occasionally no matter whether we examine the national or EU level. <FIGURE 3> In aggregate terms, we find mixed evidence with regard to business group influence in line with the expectation in hypothesis 5. According to model 1, business groups are significantly less likely than other groups to have agenda influence never as opposed to occasionally. However, even if the co-efficient is in the expected direction there is no significant difference between business and other interest groups when it comes to comparing having agenda-influence exerted occasionally versus often. In the interaction model we also do not find differences between business and other groups in affecting the agenda of policy-makers at the two policy levels. The direction of the relationships in figure 2 point towards business influence. However, the differences between business and other groups are insignificant except for the difference in the predicted probability of never having influence on EU policy-makers, which just meet the threshold of statistical significance at the 0.05 23

level with business groups being less likely than other group types to never have influence. The figures show a weak tendency for business group to experience a narrower gap between influence on national and EU-level policy-makers as we would expect for influence exerted occasionally. However, there are only minor differences in the national versus EU agenda influence policy gap between business and remaining groups for the remaining levels of influence 4. <FIGURE 4> <FIGURE 5> Finally, we see an effect of political staff size as expected in hypothesis 6. In the overall analysis in model 1 a higher staff means a lower likelihood of never (as opposed to occasionally ) having influence and conversely a higher likelihood of often (as opposed to occasionally having influence). We also see some interesting interactions between resources and the decision-making level. When it comes to never having agenda influence, resources play a much greater role in EU as opposed to national policy. Hardly any of the organizations with the highest level of political staff have no influence at any of the two decision-making levels. In contrast, there is a large difference between small organizations when it comes to having influence on EU and national level policy-makers. Small organizations without political staff have a predicted probability of never exerting influence that is more than double as high for EU than national-level policy makers. In other words, it is clear that resources play a strong role pushing groups to exert some degree of influence not only on national but also on EU-level policy makers. As staff size increases, the share of groups that report to have influence often goes up for decision makers at both the 4 The confidence intervals for the business category are wider than the remaining interest group types not least because the number of business groups is smaller than that of other groups. 24

national and EU level. However, the win is greater for national policy makers where the relationship between staff size and exerting influence often is stronger. Instead, staff size plays a strong role for securing occasional agenda setting influence on EU policy-makers whereas there is a negative relationship between staff size and occasional agenda influence on national level policy-makers. Hence whereas resource gains play a positive role in moving from having no and to some influence on EU policy, they primarily ensure that groups get influence on EU level policymakers occasionally rather than often. 5. Conclusion The point of departure for this paper was the observation that domestic interest groups operate in a multi-level system, where important decisions are made at the EU as well as the national level. Our study draws on the existing literature demonstrating variation in the degree to which groups have incorporated the EU in their action repertoire (Beyers & Kerremans, 2012; Dür & Mateo, 2013). In contrast to most of the existing literature the different policy-making levels were included in a single analysis focusing on perceived agenda setting influence. More specifically, we examined whether the explanatory potential of factors at the national, policy and interest group level vary when it comes to persuading national and EU level policy-makers to initiate new issues. Based on surveys of domestic interest groups in the United Kingdom, Denmark and the Netherlands we were able to disconfirm some of the often-held assumptions about the nature of interest group influence. First, the variation in agenda setting influence across countries was small and the evidence therefore does not indicate that groups originating from a corporatist context should be neither disadvantaged nor advantaged compared to groups in pluralist countries. Also, despite frequent criticism of the prominent role of business interests in the EU, we did not find 25

strong evidence of comparative advantages for business groups in dealing with EU as opposed to national policy-makers. What did matter for the size of the gap between influence on national and EU policy makers was the nature of an organization s policy portfolio. Groups focusing relatively more on areas heavily affected by EU regulation were more likely to experience agenda influence on EU policymakers and less likely to affect national decision makers. Also, staff resources were as expected important for influence at both levels, but with a stronger role at the EU level when it comes to ensuring agenda influence on an occasional basis. 26

Table 1. Descriptives (categorical variables) Variable Value Label % N Dependent variable Model 1 & 2: Never 1 37,72 Agenda influence (EU Occasionally 2 38,15 Institutions & National Civil servants) Often 3 24,13 2731 Dependent variable Model 3 & 4: Never 1 37,06 Agenda influence (EU Occasionally 2 40,54 Institutions & National Politicians) Often 3 22,4 2728 National 0 50 Level EU 1 50 2990 UK 1 22,68 DK 2 34,78 Country NL 3 42,54 2990 No 0 78,95 Business Groups Yes 1 21,05 2984 Contact EU institutions Less/Never 1 52,4 At least once a year 2 30,2 At least once a month 3 11,86 At least once a week 4 5,53 2748 Never 1 15,61 Participation in national boards Occassionally 2 36,65 Fairly often 3 29,52 Very often 4 18,22 2832 Table 2. Descriptives (numerical variables) Variable min. max. N Share of EU activity 0,2337382 0,0724251 0,097561 0,470588 2438 Political staff (log staff + 1) 1,028803 0,9322081 0 5,638355 2496 27

Table 3. Multinomial logistic regression results (1) Outcome 'never (baseline 'occasional agenda setting influence) Variable Model 1 Model 2 EU Level 8.857 (0.969)*** Country (Ref. UK) DK 0.905 (0.138) NL 0.665 (0.108)** EU Level#Country (Ref. UK) 18.544 (8.642)*** 0.554 (0.136)** 0.478 (0.119)*** EU*DK 1.914 (0.525)** EU*NL 1.603 (0.454)* Business Group 0.676 (0.108)** 0.859 (0.221) EU Level#BusinessGroup 0.765 (0.221) Representation boards (Ref. Never) Occassionally 0.638 (0.126)** Fairly often 0.479 (0.098)*** Very often 0.439 (0.096)*** EU Level#Representation boards (Ref. Never) 0.603 (0.142)** 0.319 (0.095)*** 0.221 (0.091)*** EU Level*Occassionally 1.183 (0.334) EU Level*Fairly often 1.897 (0.645)* EU Level*Very often 2.791 (1.287)** Share EU activity 0.360 (0.333) 58.477 (80.531)*** EU Level#Share EU activity 0.001 (0.001)*** Political staff (log) 0.560 (0.045)*** 0.342 (0.072)*** EU Level#Political staff (log) 1.720 (0.365)** Constant 1.153 (0.333) 0.680 (0.273) 28