Is the EU different? Comparing the diversity of national and EU-level systems of interest organisations

Abstract

The European Union interest group population is often characterised as being biased towards business and detached from its constituency base. Many scholars attribute this to institutional factors unique to the EU. Yet, assessing whether or not the EU is indeed unique in this regard requires a comparative research design. We compare the EU interest group population with those in four member states: France, Great Britain, Germany and the Netherlands. We differentiate system, policy domain and organisational factors and examine their effects on interest group diversity. Our results show that the EU interest system is not more biased towards the representation of business interests than the other systems. Moreover, EU interest organisations are not more detached from their constituents than those in the studied countries. Everywhere, business interest associations seem to be better capable of representing their members’ interests than civil society groups. These findings suggest that the EU is less of a sui generis system than commonly assumed and imply the need for more fine-grained analyses of interest group diversity.

Keywords:

• Interest organisations
• interest group populations
• business bias
• European Union
• lobbying

Is the European Union interest group population different from other interest group populations? According to most experts the answer is yes, especially on two dimensions. On the one end, there is widespread concern that the EU is biased towards business interest organisations because their proportion is larger in the EU than in domestic interest group populations, leading observers to conclude that the EU system is more receptive to business interests (e.g. Berkhout et al., forthcoming).

On the other end, interest groups in Europe are also thought to be more detached from constituents than equivalent groups in most domestic systems (Kohler-Koch 2013: 16). This relates to the organisational tension identified by Schmitter and Streeck (1999: 21) between the logic of membership and the logic of influence which means that interest associations continually have to manage demands from both members and policy-makers. In particular regarding the EU, the concern is that interest groups are more focused towards policy-makers at the expense of membership demands. This is problematic from a democratic perspective as interest groups might be more willing to listen to the arguments of policy-makers than represent the interests of their members.

To date, however, we do not know whether these concerns are valid as there have been no direct comparisons of the interest group population in the EU with other political systems (but see Mahoney 2008). To put it differently, while we know that there are more business groups active in Europe than other types of interest organisations, we do not know whether this relates to the openness of the EU system, reflects an indication of bias, or simply is because economic issues are more discussed at this level. Likewise, while we know that groups at the European level are rather detached from their membership base, we do not know whether this is substantially different from membership involvement in domestic political systems. At this point we cannot meaningfully assess such questions because comparable data of interest group populations across political systems are lacking. To overcome this problem, we aim to provide a first comparison of different interest group populations in a single research design. We collected interest group population data in four countries (France, Great Britain, Germany and the Netherlands) and in the EU based on similar sampling strategies and coding schemes. This design provides us with the best opportunity to analyse the uniqueness of the EU compared to domestic interest group populations across several dimensions.

In our analysis we focus on two dimensions of diversity in interest group populations. First, we analyse whether there is a business bias in the EU interest group population. Several studies have suggested the existence of an increasing mobilisation bias toward business groups at supranational governance levels compared to national levels (Baumgartner et al. 2009; Beyers and Kerremans 2007; Hanegraaff 2015). A particular, sui generis EU version of this argument can be found in earlier studies on European integration (Eising 2004). To examine a potential EU business bias, we compare the proportion of business groups at the EU system level with those in the four countries included in this study. More importantly, we also assess differences in business mobilisation at the level of policy domains. This is important because normative concerns about business bias are related to business influence on (specific) public policies. To do this in a meaningful manner, we compare the share of business groups active in similar domains such as in agricultural, environmental, or energy policy. If we observe a higher proportion of business interest groups within these domains at the EU level compared to the domestic level, this is indeed an indication that the institutional characteristics of the EU produce a more severe bias. Yet if there is no significant difference between the EU and the four countries this suggests that characteristics of the policy domains attract business interests rather than the specific EU institutional environment.

Second, we study variation in the representational capacity of groups active across different interest group populations. Representational capacity refers to the ability of interest groups to operate as mediating organisations by aggregating societal interests and translating these interests to policy-makers. In other words, we examine how interest groups in the EU and in the four countries balance policy interests with membership involvement and hence are capable of operating as transmission belts. This issue has attracted much attention as it taps directly into the debate on whether or not the EU interest groups increasingly become part of the so-called ‘Brussels bubble’, thereby potentially ignoring the transmission belt function interest groups should play in democratic systems (e.g. Kohler-Koch 2013; Streeck and Schmitter 1991). While most current studies emphasise that EU groups are substantially detached from their constituency (Kohler-Koch 2013: 16; Schmitter and Streeck 1999: 21, 54), some argue otherwise. The fact that European associations ‘outsource’ membership-oriented activities to national associations or affiliates means that these organisations have developed relatively advanced internal structures for membership engagement (De Bruycker et al. 2016). As a result, the representative capacity, or the ability to translate membership preferences to policy-makers, could in fact be more advanced at the EU level. Such strengths can, on the one end, reduce the bias of the interest group system by empowering non-business groups, but, on the other end, it can also enhance a business bias in case business interest associations appear more qualified to ‘transmit’ the concerns of members to policy-makers. We therefore also examine whether the representational capacity of different types of interest groups varies across the EU and at the domestic level.

In the following, we set out our theoretical framework. We discuss our two analytical concepts of diversity – business bias and bias in representational capacity – and how we expect them to vary across the EU and national governance levels. We subsequently present our research design, after which we test our hypotheses. We end with some concluding remarks and an avenue for future research.

Organisational diversity across interest group systems

The diversity of interest group systems has been a traditional concern in the interest group literature (Baumgartner and Leech 1998; Beyers et al. 2008). A biased pattern of interest representation is potentially problematic as it may imply that some interests have a magnitude of influence that is disproportional to their interests in society (Lowery and Gray 2004; Lowery et al. 2015). This is a valid concern, and especially for the EU interest group population, as empirical studies repeatedly show a substantial skewness in the pattern of interest representation, such as a dominance of business interests in supranational systems such as the EU (Rasmussen and Carroll 2013; Schlozman et al. 2015).

The question, however, is whether the observed skewness is indeed an indication that the EU interest population is biased. To answer this question, we take up three challenges: first, we follow Lowery and Gray’s (2004: 21) advice that the assessment of diversity or bias ‘should entail comparisons across time, jurisdictions or venues’. While ‘bias’ may be impossible to identify in absence of a proper benchmark of ‘unbiasedness’ (Lowery et al. 2015), comparisons allow us to identify relative bias. Second, there is a theoretically unresolved puzzle regarding the relationship between the level of government and the likely bias of the group system. That is, the collective action problems of diffuse interests seem to be more easily resolved at lower levels through the provision of expressive and solidary selective benefits (Wilson 1974: 30–55). This should make it easier for social movements and civil society groups to maintain themselves locally rather than regionally, nationally, or internationally. At the same time, the expansion of the scope of conflict from the lower to higher level potentially favours exactly those groups that are typically excluded from narrowly defined local conflicts, and in that way increase diversity (Schattschneider 1960), a phenomenon or mechanism similar to what Keck and Sikkink (1998) label the ‘boomerang pattern’. This theoretical conundrum is only infrequently studied empirically: an important omission in the population literature is a lack of studies that pay attention to multi-level dynamics in interest group populations by comparing geographical levels within communities (but note such attention in the United States case, e.g. Baumgartner et al. 2009). Third, there is an EU-specific discussion related to theories of European integration about the relative tendency of several types of interests to mobilise at the European level. These theoretical debates, in broad strokes, relate to incentives on the part of European institutions to seek the backing of especially major business interests (Beyers and Kerremans 2007; Eising 2007).

As said, however, suspicions of bias in the EU interest group population remain mostly tentative. We lack comparative studies to empirically assess whether the EU interest group population is indeed different from those in other political systems. To overcome these problems we rely on a comparative design to analyse two dimensions of interest group diversity: business bias and representational capacity. Both have important normative consequences for how we view the EU as a democratic and responsive political system.

Business bias

Contemporary research on EU interest representation commonly differentiates between citizen groups and business interest representation (Dür et al. 2015; Eising 2004, 2007; Klüver 2013; Rasmussen and Carroll 2013). Building on this research, we compare the presence of business and other interests in the EU and the national interest group populations. As indicated, several scholars suggest that the interest group system in the EU is more biased towards business interests than national political systems.

First, the EU’s geographical scale exacerbates collective action problems that typically affect non-business interests in a more severe way. To start with, groups vary in the extent to which they have a ‘local’ membership base. Groups that rely on national resources to maintain themselves, such as national government subsidies or the ‘local’ constituencies of certain unions, are less likely to mobilise at the EU level (Beyers and Kerremans 2007). In addition, the so-called ‘public sphere’ deficit (e.g. Schlesinger 1999) at the EU level, i.e. the absence of a common EU media environment hinders mobilisation of non-business groups that typically rely on national media sources to arouse salience of the issues they stand for and mobilise their constituency. This makes it more challenging for them to mobilise EU-wide through professional marketing, direct contacts with journalists, and joint activities nurturing solidary or purposive benefits for their constituencies. As a result, the EU interest system is seen by many as being biased towards business interests.

We contend, however, that assessing bias at the system level does not take into consideration that bias is structured at the level of policy domains. For instance, it is expected that we would see more citizen group activity on cultural or social issues, while the political goals of business groups lead them to be more active on economic issues. In other words, it could very well be that the overall bias we observe in the EU is not caused by the particular institutional EU context but is a function of the type of issues on the agenda of the EU institutions, given its explicit mandate to regulate economic, rather than social or cultural policies (Majone 1998). Despite these trepidations, most scholars still argue that the EU favours the input of business groups, regardless of the issue on the table (e.g. Bouwen 2004; Coen and Katsaitis 2013), spurred by the demand of EU policy-makers. Coen and Katsaitis (2013), for instance, note that the regulatory nature of most EU policies and the small number of the Commission’s staff, produces ‘demand’ on the part of policy-makers for technical information. Scholars also point out that different EU institutions, most notably the European Parliament and the European Commission, attract different types of lobbyists (Bouwen 2004). Combined, the literature would predict that the EU-level interest group population is more biased than the national-level interest group populations towards the mobilisation of business groups over other types of organisations. We therefore hypothesise that the EU is more biased towards business groups both at the system level and at the level of policy domains. Regarding the latter hypothesis, we take into consideration that it has not yet been tested directly but is based on aggregate assessments of bias in the EU. It could be that at the level of policy domains business bias is absent.

Hypothesis 1: The EU interest group population as a whole and per domain is more biased towards business than national interest group populations.

Representational capacity

Our second conceptualisation of diversity takes into account how interest groups balance their political orientation with their membership orientation. We refer to the combination of interest aggregation and interest articulation as representational capacity. We define representational capacity as the capability of groups to manage the membership involvement with the exercise of political influence. This directly builds upon the organisational tension identified by Schmitter and Streeck (1999: 21–54) between the logic of membership and the logic of influence: interest associations continually have to manage demands from both members and policy-makers. This is not an easy task. Members put other demands on interest organisations than policy-makers. Members prefer organisations with a clearly demarcated identity so that their interests align with those of other members and are distinct from other organisations. To ensure their support in the form of dues, participation in (voluntary) activities, and information on what members prefer, organisations have to provide selective benefits, such as services or collective insurances to their membership. Ensuring membership engagement with the organisation thus calls for relatively specialised organisations so that members perceive their interests to be properly represented. Policy-makers, on the other hand, require a distinct set of policy-relevant information, including constituency information, technical expertise, but also information about broad societal support. In other words, they prefer generalist organisations to have a guarantee on the legitimate nature of the policy input (see also Kohler et al. 2017). This tension can lead organisations to specialise in the provision of policy input (similar to think tanks) or the involvement of members (similar to social clubs).

Assuming some level of organisational specialisation, conceptualising diversity in terms of representational capacity is relevant in at least two important ways. First, it points our attention to the kind of policy information we can expect groups to offer to policy-makers. In broad terms, this may be information on the interest or preference of the members of the organisation or it may be information on the technical quality of certain proposals (Bouwen 2004; Braun 2012). Organisations with a strong focus on membership involvement are more likely to provide a different type of policy information compared to organisations with a well-developed policy orientation (Daugbjerg et al., 2017). Hence, organisational specialisation results in different kinds of interest groups capable of offering distinct types of policy information (Minkoff et al. 2008).

Second, the way in which interest groups strike a balance between policy influence and membership involvement determines the extent to which they are capable of performing a ‘transmission belt’ function. After all, attributing a transmission belt function assumes that interest groups simultaneously transmit societal interests to policy-makers and convey policy compromises to their membership (Braun 2015; Halpin and Fraussen 2017). This notion is especially relevant in neo-corporatist approaches to interest representation (Schmitter and Streeck 1999) and particularly applicable to the EU context as the European Commission explicitly requires such an intermediary function of the groups it reaches out to (European Commission 2001; Kohler-Koch and Finke 2007).

The representational capacity of interest groups and the different dimensions groups tend to prioritise are likely to vary across national- and EU-level interest group populations. First, the relative ‘weakness’ of EU-level policy-makers allows interest groups, even when relatively non-cohesive or specialist, to be involved in the political process. This should produce a relatively strong focus on interest articulation at the EU level compared to the national level. Second, as regards the membership environment of European umbrella groups, we expect multi-layered European federations to ‘outsource’ their membership-oriented activities to national associations or affiliates. We expect that these organisations have advanced internal structures for membership engagement, offering multiple types of membership and well-developed channels for engagement in EU public policy-making. By comparison, at the national level, such a specialised type of organisation would not exist and similar types of task division are also unlikely. As a result, we hypothesise:

Hypothesis 2: EU-level interest groups have stronger representational capacity than interest groups active at the national level.

Design

To test our hypotheses, we compare the composition of national-level and EU-level populations of politically interested membership organisations. This data collection strategy is in line with recent mapping studies of interest group populations (Berkhout et al., forthcoming; Halpin and Jordan 2011). Next to the EU level, we select four member states: France, United Kingdom, Germany and the Netherlands. As our expectations relate to differences between the EU and national-level political systems, we do not focus on systemic factors explaining variation across national-level interest group populations. At the same time, by selecting four countries considered to belong to different interest representation regime types (e.g. Balme and Chabanet 2008: 28) ‒ the UK as a pluralist system, the Netherlands and Germany as corporatist systems, and France as statist system ‒ we enhance the external validity of the observed differences between the EU and these different types of member states. We combine top-down (policy-oriented) and bottom-up (mobilisation-oriented) mapping strategies to construct representative samples of the organisational populations. The use of different types of sources reduces potential sampling bias (cf. Berkhout and Lowery 2008). These sources include, first, directories of associations, most notably the European directory of Public Affairs (2014), OECKL Deutschland (2011),¹ the Pyttersen Almanak (2014),² the Directory of British Associations (2012),³ and a combination of French lists of associations.⁴ The sections listing business, professional and trade associations, and the section on NGOs of the European Public Affairs Directory are comparable to the relevant sections of the national directories of associations. Secondly, from the ‘top down’, we include membership organisations active in the parliaments and sample these from: the Lobby register of the Bundestag (2014), the register of accredited lobbyists to the European Parliament as can be found in the EU Transparency Register (2014), and from the participants in hearings and roundtables in the French (2011–2014), Dutch (2011) and British (2011) parliaments. The Dutch and British parliamentary lists have been shared by researchers in the Interarena project (see Helboe Pedersen et al. 2015). The sample sizes vary between 309 interest associations in the Netherlands and 448 interest associations in the United Kingdom amounting to a total of 1785 interest associations.

These sampling procedures do not produce perfectly comparable samples of interest groups. First, as regards the directories, for the French case we could not rely on a general inclusive directory and used alternative lists instead. This makes the inclusion of established, professionalised associations in the French case more likely than in the other cases. The figures presenting data on membership involvement in the French case (Figure 3) should be read with this in mind. Second, as regards the parliamentary data, the Bundestag and European Parliament registers do not require ‘invitation’ on the part of members of the parliament whereas presence in the parliamentary hearings in the UK, France and the Netherlands does. We do not aspire to compare the individual member states with each other and these differences largely average out in the comparison with the EU.

Data collection

To examine our two concepts of diversity (group type and representational capacity), we rely on information provided on the websites of the interest groups. This information has been coded by six extensively trained master’s students. The coders are fluent in at least two languages and were assigned records from multiple countries and sources. The data sources were sampled randomly and the sample presented below includes a sufficient number of organisations for the aggregate-level analysis undertaken. The intercoder-reliability test shows that coders reached higher agreement on the policy interest scale (Krippendorff Alpha 0.61) than on the members-service scale (Krippendorff Alpha 0.38). Note that when accounting for co-variation, these scores need not be interpreted conservatively (Neuendorf 2002: 152). Records coded in the first months of the data collection were coded by two coders (25%) and only included when the coders agreed on the selection of the organisation and its types of members. This implies that for the data we use in this paper the intercoder-reliability scores are higher than for the test records. The operationalisation of the (in)dependent variables is summarised in Table 1.

Table 1. Variables, operationalisation and measurement.

Variable	Variable description	Details
Dependent variables
Business bias	Proportion of business interests in policy domain	Based on membership variable (see below)
Representational capacity 1 – policy interest dimension	Policy interest: the extent to which an organisation is involved in policy-making	1 to 6 scale that indicates no policy interest at all (1) and full policy interest (6). (see Appendix for full coding-scheme)
Representational capacity 2 – membership dimension	Membership involvement: the extent to which an organisation involves its membership in decision-making	1 to 6 scale that indicates membership involvement ranging from: involvement at all (1) to fully run by members (6) (see Appendix)
Representational capacity 3 – ‘ideal type transmission belt’	Combination of policy interest and membership dimension	Variable based on policy interest and membership involvement scales.
		Dummy variable: 1 = organisations with a score higher than 4 on both scales; 0 = organisations with lower scores on both scales
Independent variable
Institutional effect (H1 and H2)	Effect of different governance levels	Dummy variables that represent variation between: the Netherlands, the UK, France, Germany and the EU level
Control variables
Density	Number of interest groups per policy domain in member state/EU level	Proportions per policy domain based on the COFOG classification (see http://unstats.un.org/unsd/cr/registry/regcst.asp?Cl=4; and below)
Policy competence	Extent to which a policy domain is an EU competency; conversion table of COFOG coding into EU or national competencies	EU competency: 05: Environmental protection; 0.42: Agriculture
		Shared competency: 01: General public services; 02: Defence; 04: Economic affairs; 0.43: Energy; 0.44: Manufacturing and construction; 0.54: Landscape protection; 0.45: Transport
		National competency: 03: Public order and safety; 06: Housing and community amenities; 07: Health; 08: Recreation, culture and religion; 09: Education; 10: Social protection; 0.73: Hospital services; 0.82: Culture; 10.4: Family; 10.7: Social exclusion
Group type	Business vs other groups; based on type of members	Dummy variable: 1 = Members are commercial firms, 0 = Members are not firms. In Table 3: additional category used: members are individuals as professional
Membership structure	Presence of decentralised organisational units, such as geographical or functional chapters, or sub-associations	Does the organisation have an ‘indirect’ membership structure?
		1 = Yes, 0 = No
Social/leisure focus	Organisations that are predominantly focused on social or leisure goals	Domain of membership according to the International Classification of Non-Profit Organizations is:
		1) Group 1 ‘Culture and recreation’ or Group 10 ‘Religion’ 2) Not in any of these

We have two dependent variables. The first dependent variable, business bias, is the proportion of business interest associations per policy domain. We distinguish between business associations and other interests based on their types of members. Associations with commercial firms as members are business interest associations. As said, we focus on bias in policy domains because normative concerns about the ‘overrepresentation’ of business interests are mainly related to the presumably business-friendly implications for policy-making. To see whether or not policy-makers in the EU are more receptive to business groups than those at the national level, we need to compare bias at the level of policy-makers which are structured along policy domains (Pappi and Henning 1998: 554–6). We use the UN Classification of the Functions of Government (COFOG) to measure policy domains. Each organisation is coded in the most specific category applicable. We then aggregate these categories into 18 policy domains. Organisations may be active in multiple policy fields, but we rely only on the area which coders judged to be the most important. Overall, this provides us with the proportion of business groups per policy domain across the four countries and the EU.

We disaggregate our second dependent variable, representational capacity, into three distinct components. First, the variation in policy interest is the extent to which organisations provide policy input and position themselves in relation to specific public policy programmes. This is indicated in press or policy statements published on the website. Second, student coders classify the extent to which the organisation involves members in organisational decision-making and activities. This includes participatory structures such as committees, membership meetings and internal elections aimed at aggregating and representing the preferences of members. Third, we combine these measures into a single indicator of the organisations’ capability to represent interests. More specifically, we construct a dummy variable that identifies organisations with high levels of policy involvement and membership involvement (scores of 4 or higher). It follows that these interest groups have high representational capacities and are capable to perform the transmission belt function (see Appendix for full coding instructions).

The independent variable in both analyses is the variation across the five political systems (the EU and the four countries). We thereby aim at isolating the institutional effect that causes the variation in interest group diversity, in particular the distinction between EU and national-level political systems. We include country dummy variables in the analyses to see whether the countries and the EU differ regarding the proportion of business groups active and the representational capacity of individual organisations.

We also control for several factors. For our first analysis, on business bias, a first crucial control variable is the competency-level of policy domains. The main reason for doing so is the strong regulatory nature of the EU system, implying a large amount of economic competences. This means that any skewness towards business groups might be a result of the type of issues that are dealt with rather than a function of system-level institutional prioritisation of business interests. We use a conversion table (see Table 1) based on our own assessment of policy competence of policy domains and distinguish among domains of EU competence, shared and national competences. Secondly, we control for the density of the policy domain, as business bias could be a result of the crowdedness of policy domains, although previous findings are inconclusive in this regard. That is, at the level of issues, Baumgartner and Leech (2001) suggest that business interest organisations are more likely than others to maintain presence on issues with a low density, whereas Berkhout et al. (2016) found a higher relative presence of business interests on ‘crowded’ issues. To measure the relative density, we use the proportions of actor types per policy domain in each system.

We add three controls to the analyses on representational capacity. First, we control for group type (business or professional group). Firms, as members of associations, are already ‘organised’ and their participation is likely to be of a different kind than that of individual citizens (e.g. Offe and Wiesenthal 1980). Second, we control for the nature of the membership structure (presence of geographical or functional chapters or sub-associations). Such subsections are tools used to manage organisational consensus formation and should therefore be positively related to membership involvement. Finally, we also control for predominant social or leisure goals of organisations. Such goals are likely to reduce the policy orientation of groups and affect the nature of the membership involvement. They are likely to lower the representative capacity of groups.

Analysis

Our analysis proceeds in two steps. First we descriptively assess business bias across the different levels of governance. By means of a multivariate OLS-regression analysis, we subsequently examine the statistical significance of these differences between the EU and national levels while controlling for several alternative explanations of bias. Second, after describing differences regarding the representative capacity of organisations we follow up with regression analyses to account for the variation between the EU and national levels.

Figure 1 shows the proportion of associations representing firms as members per system. Using this measure, about half of the groups in the EU are business associations. This is consistent with findings in recent studies (see Berkhout et al., forthcoming; Lowery et al. 2015). The share of business interest groups in the EU is similar to that in France and higher than those in the Netherlands and the United Kingdom. Please note that we here operationalise bias in the form of proportions of group types per political system rather than per policy domain (as we will do in the regression analysis in Table 2). Moving to the domain level allows us to examine differences in bias across domains. This matters greatly for our understanding of the policy implications of bias because averages per system mask the strong bias that exists in certain policy domains. That is, as a crude indicator, 13 out of 91 policy domains have 100% business interest representation, whereas in seven of them we note a complete absence of business interests. There must be important sector-level dynamics that explain such variation. We will examine this in the regression models in Table 2.

Figure 1. Proportion of business interests per political system.

Table 2. OLS regression on the proportion of business interest associations per policy domain, n = 90 policy domains in four countries.

Country	The Netherlands	−0.15
Reference category: European Union		(‒1.27)
	France	0.18
		(1.53)
	United Kingdom	−0.15
		(‒1.33)
	Germany	0.10
		(0.88)
Density	Proportion of organisations in policy field	0.089
		(0.93)
Policy competence	National competence	−0.53^***
		(‒4.12)
	Mixed national/EU competence	−0.16
		(‒1.24)
	R^2	0.32

Standardised beta coefficients; t statistics in parentheses; n = 90.

* p < 0.10

** p < 0.05

*** p < 0.01.

To test whether business bias can be explained by systemic, institutional effects we ran a multivariate OLS-regression analysis on the proportion of business interests per policy field. The country coefficients restate the minor cross-national differences observed in Figure 1. We control for the organisational density of the policy domain and do not find a significant relationship. This means that business bias is unrelated to the number of organisations active within a given community. This is in contrast to findings at the level of issues (Baumgartner and Leech 2001; Berkhout et al. 2016), where business interest organisations were found to have lower (US), respectively higher (EU), relative presence on ‘crowded’ issues. The bottom rows of the table indicate that a larger proportion of business interest organisations is active in policy domains in which the EU – rather than the member states – enjoys exclusive competences. Areas of exclusive EU competences, such as trade and market regulation, attract business interest organisations to a stronger extent than other interest groups, regardless of this being at the EU or national level.

We now turn to the second part of the empirical analysis in which we focus on representational capacity, and for which we individually assess policy interest, membership involvement and subsequently examine the combination of a group’s policy interest and membership involvement. High scores on both dimensions indicate a strong representational capacity. We start by presenting the frequency distributions of the policy interest and membership involvement scales in Figures 2 and 3. The national distributions resemble each other. As can be seen in Figure 2, around 20% of the organisations is fully committed to influencing public policy, and, when adding up the proportions of the top categories on the x-axis (labelled 4, 5 and 6) one finds that around 40% of the organisations falls into any of these categories (ranging from 33% in the United Kingdom to 44% in Germany). The organisations in these national systems also are relatively similar regarding the extent to which they allow members to be involved in the organisation. Around 40% of the organisations provide substantial opportunities for members to engage within the organisation (Figure 3).

Figure 2. Organisational orientation per political system: frequency distributions of scores on policy interest.

Figure 3. Organisational orientation per political system: frequency distributions of scores on membership involvement.

The key difference is between the organisations active in the EU and those active at the national level. EU-level organisations are definitively more policy-oriented, with about 50% aiming to influence policy, and surely less open to involve their members in strategic decision-making. Less than 10% of them fully embrace membership control over the organisation. Around 30% (labelled 5) does allow substantial membership input. Consequently, the differences between the EU and the national systems in the sums of the top categories (labelled 4, 5 and 6) are quite substantial (ranging from 58% in the EU to 83% in Germany). These differences are in line with our expectations. They indicate that EU-level interest groups tend to be better at engaging with policy-makers rather than with their members when compared to national groups (see also Kohler-Koch and Buth 2013). Nonetheless, we note that EU organisations tend to score relatively high on both dimensions. Moreover, these aggregate differences do not tell us much about the specialisation of individual organisations and may no longer hold when controlling for organisational factors. We look at this in the discussion of Table 3.

Table 3. OLS and logistic regression on the policy interest, membership involvement and combined policy/membership scores (logistic regression on indicator when both scores are greater than 4: ‘Representative capacity’).

		Model 1			Model 2			Model 3
		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
		Policy interest	Membership involvement	Representative capacity	Policy interest	Membership involvement	Representative capacity	Policy interest	Membership involvement	Representative capacity
Country	The Netherlands	−1.69^**	0.30^*	−0.87^**	−1.46^**	0.35^**	−0.67^**	−1.44^**	0.57^**	−0.24
reference category: European Union
		(‒12.42)	(2.27)	(‒5.01)	(‒10.52)	(2.62)	(‒3.71)	(‒8.50)	(3.47)	(‒1.08)
	France	−1.55^**	0.68^**	−0.80^**	−1.58^**	0.65^**	−0.82^**	−1.55^**	0.90^**	−0.31
		(‒11.89)	(5.45)	(‒4.89)	(‒12.23)	(5.16)	(‒4.93)	(‒9.10)	(5.42)	(‒1.40)
	United Kingdom	−1.92^**	1.03^**	−1.07^**	−1.66^**	1.04^**	−0.88^**	−1.63^**	1.26^**	−0.42
		(‒15.23)	(8.48)	(‒6.57)	(‒12.86)	(8.28)	(‒5.17)	(‒10.01)	(7.96)	(‒1.92)
	Germany	−1.54^**	1.19^**	−0.65^**	−1.49^**	1.14^**	−0.63^**	−1.46^**	1.38^**	−0.14
		(‒11.75)	(9.42)	(‒4.03)	(‒11.46)	(9.04)	(‒3.79)	(‒8.73)	(8.48)	(‒0.66)
Interest	Culture and social clubs	−1.43^**	0.22^*	−1.29^**	−1.39^**	0.34^**	−1.22^**	−1.39^**	0.33^**	−1.21^**
		(‒13.57)	(2.16)	(‒7.18)	(‒13.10)	(3.25)	(‒6.64)	(‒13.10)	(3.24)	(‒6.63)
Members	Members are professionals				−0.012	0.46^**	0.21	−0.014	0.44^**	0.17
					(‒0.12)	(4.56)	(1.49)	(‒0.13)	(4.40)	(1.19)
	Members are firms				0.31^**	0.33^**	0.43^**	0.29^**	0.21^*	0.18
					(3.17)	(3.47)	(3.34)	(2.71)	(2.04)	(1.22)
	Multi-level membership structure				0.61^**	0.081	0.43^**	0.61^**	0.11	0.49^**
					(6.72)	(0.92)	(3.60)	(6.70)	(1.21)	(4.06)
Interaction	EU*Members are firms							0.054	0.48^*	0.97^**
								(0.25)	(2.31)	(3.47)
	Constant	4.94^**	3.75^**	0.15	4.55^**	3.49^**	−0.29	4.53^**	3.29^**	−0.68^**
		(51.96)	(40.93)	(1.27)	(39.55)	(31.22)	(−1.96)	(31.16)	(23.36)	(−3.61)
	Observations	1783	1777	1777	1771	1766	1766	1771	1766	1766
	R^2	0.23	0.072		0.26	0.084		0.26	0.087

t statistics in parentheses, OLS regressions, Models 1(3), 2(6) and 3(9) logistic regressions.

* p < 0.05

** p < 0.01.

Table 3 presents nine OLS and three logistic regression analyses on the policy involvement and member involvement of the individual interest organisations studied in the EU and the four member states. These are clustered in three models. Model 1 assesses whether EU-level groups have indeed a stronger representative capacity than national-level groups (hypothesis 2). We compare the countries and control for clubs interested in culture and social activities.⁵ These tend to be more numerous at the national level than at the European level. The nature of the participation of members in these organisations is, partially, of a different kind than the decision-making-oriented participation common in other types of organisations included in our study. The coefficients are significant in the expected directions (negative for policy interest, positive for membership involvement, negative for the combined measure). As already observed in Figures 2 and 3, the orientations towards polices and members differ between EU-level groups and national groups. As can be seen from the negative coefficients in Model 1(1), national interest organisations tend to be less policy interested than EU-level organisations. The reverse is the case for membership involvement, Model 1(2). The logistic regressions assess the likelihood that organisations have high scores on both policy interest and membership involvement (>4). The negative significant coefficients reported in Model 1(3) indicate that EU organisations generally are ‘stronger’ than their national counterparts. They are more likely to have high scores on both dimensions, and, in that way, should be able to develop stronger representational functions than national organisations.

Model 2 controls for the type and structure of membership. The positive and significant coefficients in all sub-models indicate that business, professional and multi-levelled associations compared to other organisations are more likely to be interested in public policy, engage members and combine these functions. Also note that the addition of these explanatory variables adds only little explanatory power to the models (compare Model 1 and Model 2). This indicates that when it comes to representational capacity, cross-national differences are more important than the differences among the types of interest represented. In Model 3 we include an interaction effect between EU and business interest associations. The coefficient of this variable is positive and significant for Models 3(8) and 3(9) on membership involvement and representative capacity. Business interest associations in the EU, compared to those active at the national level, are more likely to involve members and, as a consequence, more capable in representing interests in the policy process. We visually present the strength of this relationship in Figure 4 where we plot the predicted probabilities of the representative capacity of business interest associations versus other association in the EU and the member states studied.⁶ The figure demonstrates that business interest associations have a stronger representational capacity, in particular those at the EU level.

Figure 4. Predicted probabilities of the representative capacity of business interest associations versus other associations in the EU and four member states, adapted version of Model 3(9) from Table 3.

Conclusion

This paper is a first attempt to compare interest group populations across governance levels in the EU. More precisely, we compared the business bias and representational capacity of interest groups in the EU with those in France, the UK, the Netherlands and Germany. As regards business bias, we found no substantial differences between the interest group population at the European level and those at the national level. Our analyses suggest that system-level variation is not associated with this variation but that policy competences are and possibly also other policy domain-related factors. This is in line with the finding that interest group mobilisation is largely specific to policy domain or issue (Berkhout et al., forthcoming). Put more concretely, this means that, overall, there might be more business groups active in Europe, but at the level of policy issues, for instance as regards agriculture, the environment or construction policy, the share of business groups in the EU compares to their shares in the four countries. This finding supports the idea that business bias is related to the type of issues on the EU legislative agenda rather than the effect of institutional characteristics.

In terms of representational capacity we find a stronger policy engagement at the EU level as opposed to the national level and a stronger membership orientation at the national level compared to the EU level. Yet, surprisingly, our results also suggest that in particular EU-level organisations with either professionals or firms as members and a multi-level membership structure have the strongest representational capacity. So, although in terms of business bias the diversity of the interest group population at EU level does not seem to be as limited as is commonly argued, in terms of our second aspect of diversity, representational capacity, the EU-level system seems more biased to business interests. They seem, across the board, to perform better in terms of representational capacity than other types of interests. The bias towards business in the numbers of organisations active is consequently exacerbated by the organisational capacity of business interest associations to aggregate and articulate interests.

We would like to identify a number of venues for future research. First, business bias is a far more complex phenomenon than commonly understood. That is, business and non-business interests do not randomly distribute over policy domains. And various factors affect this process. Some country-specific institutions may encourage encompassing interest aggregation, certain policy domains may be very open to various groups, and certain interests are just not easy to organise. Second, assuming that we want to study implications for public policy or representation more broadly, business bias must be related to the varying capacities of organisations to exhibit meaningful activities in relation to politics. In other words, this means that the ‘transmission belt’ function of interest organisations is deeply related to questions about bias. That is, ‘transmission’ puts substantial organisational demands on groups that only relatively strong groups are willing and able to develop. Our finding that there is a strong correlation between business interests and representational capacity suggests that especially business interests are better transmission belts than others, and that this is particularly the case at the EU level. This is in line with previous research demonstrating that more consultation practices seem to reinforce existing biases (De Bièvre et al. 2015). This is normatively problematic in light of the Commission’s deliberate outreach to civil society organisations on precisely the grounds of their representational capacity (Perez-Solorzano Borragan and Smismans 2012). Third, the limited findings on difference between EU and national political systems also nuance arguments about the effect of political institutions for the formation of interest group communities. Institutional factors such as the legal competence of policy-makers matter at the level of policy domains rather than political systems. In addition, supply factors, i.e. the structure of the economy and salience of issues in a given geographical setting as well as the internal structure of interest groups, seem to outweigh attempts by government to structure the composition of interest group communities (Berkhout et al. 2015).

Our study is hopefully a stepping stone towards more comparative studies of interest group populations. Over the last decade students of interest groups have started to map interest group population in many countries. Our study shows the potential relevance of comparing the outcomes of these studies. We therefore urge other scholars to engage in comparative studies of interest group density and diversity across different political systems, but even more so across different policy domains within distinct political systems.

Funding

We would like to acknowledge the financial support of the Netherlands Organisation for Scientific Research project [grant number 461-10-519].

Notes on contributors

Joost Berkhout is Assistant Professor of Political Science at the University of Amsterdam. His main research focus is in the field of interest group politics. He currently conducts research on the differences in the numbers and types of interest groups per sector and per country. [[email protected]]

Marcel Hanegraaff is Assistant Professor in Political Science at the University of Amsterdam. His research is on the politics of interest representation in a transnational and EU context, as well as on the functioning of international organisations in the fields of climate change and global trade. [[email protected]]

Caelesta Braun is Associate Professor in Public Administration at Leiden University. Her main fields of research include interest representation and regulatory governance, with a focus on political behaviour of organised interests, stakeholder management of regulatory authorities and decision-making. [[email protected]]

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. We rely on the selection of 4830 ‘politically active’ associations by Jentges et al. (2012).

2. The Dutch Pyttersen’s Almanak (2014), section A, lists 9601 collective action organisations which are (primarily) non-profit and ‘of supra-local interest’.

3. The Directory of British Associations reports collective action organisations of ‘national’ interest.

4. There is no French directory of associations similar to the directories in the other countries. Similar to the EUROLOB II project (Kohler-Koch et al. 2017), we therefore use the members of the two major business interest federations. These are the 220 professional member-federations of the French association for small- and medium-sized companies Confédération Général des Petites et Moyennes Entreprises (CGPME) (see http://www.cgpme.fr/le-reseau/federations) and the 87 member-federations of the major employers’ federation Mouvement des Entreprises de France (MEDEF) (see http://www.medef.com/medef-corporate/le-medef/federations.html). The study also includes the 133 members (Titulaire and Suppleant) of the Le conseil national de la vie associative (2014) in order to guarantee the inclusion of French citizen groups (see http://www.associations.gouv.fr/112-le-conseil-national-de-la-vie.html).

5. Organisations classified under Group 1 and Group 10 of the International Classification of Non-Profit Organisations.

6. We calculated these probabilities on the basis of an adapted version of Model 3(9) from Table 3. We contrasted the EU-level groups to all national groups rather than including each country as a dummy variable.

Appendix.

Coding scheme concerning the dimensions of representational capacity

Table

Representational capacity	1. No mention of interest representation before government or through creating broader awareness
- Policy-interest scale	2. Interest representation is mentioned in mission statement or somewhere else on website, but no mention of policy-oriented activities
	3. Minimal policy-oriented activities reported, such as: ‘the association land owners find sustainable forest management important’
	4. One or two concrete policy-oriented activities found, including monitoring and information-oriented activities
	5. A separate section of the website reports on policy campaigns and positions, or throughout the website one finds references to public policy, including indirect strategies such as creating awareness
	6. The policy section is substantial and includes several different types of documents (policy statement, public campaigns, etc.)
Representational capacity	1. The members only transfer money (e.g. called ‘donors’ or ‘supporters’)
- Membership involvement scale	2. Members may have some official rights but there is no explicit reference to this
	3. Members are called ‘members’, and seem formally to have voting rights
	4. Members have some opportunity to participate in the organisation, such as through volunteering
	5. Members have several ways to participate and decision-makers have a track-record within the organisation/cause
	6. Members fully run the organisation, also paid staff seem to be selected primarily on the basis of their loyalty to the cause rather than management experience (e.g. association of doctors is managed by a doctor)