Chameleon pluralism in the EU: an empirical study of the European Commission interest group density and diversity across policy domains

Abstract

This paper contributes to the discussion surrounding interest groups in the European Commission. We inspect the Commission's lobbying register and assess the density and diversity of the interest group population per policy domain. The results suggest that while at the system level élite pluralism with its preponderance of business interests is a credible hypothesis, this is not the case at the sub-system level, where chameleon pluralism better conceptualizes variation of the interest group populations as a function of the age of the Directorate General (DG), capacity of the DG, nature of the policy domain, and involvement of member states. Bridging theoretical considerations on input/output legitimacy with informational approaches, we argue that different policy domains demand different types of legitimacy that are supported by the provision of different types of information (technical/political).

Keywords:

• Interest groups
• information exchange
• legitimacy
• lobbying
• pluralism

Today there is a vast European Union (EU) literature mapping the density of lobbying activity in the EU (Berhkhout and Lowery 2010; Wonka et al. 2010) and a number of policy studies that seek to assess the impact of lobbyists on EU policy formulation (Klüver 2011; Rasmussen 2012). While others follow the multi-level policy process (Beyers and Kerremans 2007; Dur and Mateo 2010), few have sought to map the variance within the European Commission and between Directorate Generals (DGs) across policy domains (Broscheid and Coen 2007; Mahoney 2004).

Drawing on US models of informational lobbying exchange and variation in lobbying activity across congressional policy committees (Baumgartner et al. 2009; Hall and Deardorff 2006), research on EU interest intermediation has begun to recognize that there is variance in interest group activity across policy domains (EU sub-systems) and European institutions (EU system). This paper aims to contribute to this line of research by first examining empirically the theoretical applicability of EU system-level conceptualizations of interest intermediation at the EU sub-system level (policy domains). Secondly, providing empirical evidence on the diversity of the interest group population across policy domains, we develop an informational demand side model and argue that there is variation in informational demands and legitimacy requirements by policy-makers across sub-systems. This in turn affects the diversity and density of interest groups within DGs.

Finally, while we note the existence of élite pluralism and preponderance of business interests in lobbying the European Commission (Coen 1997, 1998; Eising 2007), we also make an empirical argument that it is too simple to posit that business groups dominate across all policy domains. Consequently, we see interest group activity in the European Commission as a form of chameleon pluralism (Coen and Richardson 2009), where interest group type, density and activity is a function of the policy type, age and capacity of the DGs responsible for policy-making.

INFORMATIONAL LOBBYING EXCHANGES

We start from the assertion that different EU institutions demand different types of information supplied by different types of interest groups (Bouwen 2002). Specifically, we recognize that the Commission does not have all the relevant expertise or resources required for policy-making and that it demands expertise (information) from interest groups and provides benefits in exchange. The benefits provided can be: (i) direct, as in the case of policy impact; or (ii) indirect, as in the case of information exchange about policy.

At the system level there would appear to have been a correlation between the increasing informational demands of the EU policy process and increasing interest groups (Coen and Richardson 2009). Yet at the sub-system level it is also observable that resources vary across DGs, and the Commission can neither provide infinite benefits to, nor process the information of the entire EU lobbying population. Therefore the DGs form stronger (informal) ties with groups that provide relevant and reliable information over time and unique interest group clusters emerge across policy types (Broscheid and Coen 2003).

Moreover, recognizing that EU policy-making is more than a resources exchange, we note that in order for the Commission to remain a relevant political institution, and not just a quasi-regulatory agency, it needs to maintain both its output and input legitimacy (Majone 2002, 2009; Scharpf 2009). Output legitimacy is associated with the quality of policy and requires technical expertise. Technical information assists in the design and implementation of policy such as cost effectiveness. Input legitimacy is associated with the level of public consensus and participation over policy, and requires political expertise related to public consultations and public opinion. The more directly accountable the institution (or policy) is to the public, the greater the need for input legitimacy. The less directly accountable the institution (or policy) is to the public, the higher the need for output legitimacy. The Commission, a bureaucracy not directly accountable to the EU public through elections, will show overall greater demand for output legitimacy and technical expertise (Neyer 2010).

Informational approaches are also used to explain interest group diversity. Studies show that civil society groups such as non-governmental organizations (NGOs) or social movement organizations (SMOs) supply expertise that is perceived as political (Mahoney 2004; Warleigh 2001). Business interests, such as firms and professional associations, supply expertise that is perceived as technical (Coen 1997, 1998; Eising 2007). That is not to say that civil society groups do not or cannot provide technical information (Chalmers 2011, 2013), or business interests cannot or do not provide political information (Coen 2009). What we argue is that, assuming actors provide a bundle of information goods, business will be weighted towards technical provision and NGOs towards the political information.

However, this distinction is not mutually exclusive, for example NGOs providing technical information for Climate Network Europe or firms providing political information on wider public consultations to build reputations as information providers over the long run. The legitimacy each category of group brings is associated with the nature of the group (public vs. private) and the information they provide overall (political vs. technical). Thus EU policy-makers broadly seek out and encourage access from those groups they perceive to bring the type of information that will help to legitimize the policy they are seeking to formulate (Broscheid and Coen 2003; Chalmers 2013).

Accepting that the nature of the policy domain affects interest group density, it is likely that it affects the demand for different types of legitimacy and information. The absence of a detailed examination of policy domain interest group diversity raises three issues. First, we have not tested whether EU system-level theories apply to the sub-system level. Research assumes that élite pluralism applies across the entire system, including sub-systems. Second, this limits our understanding on interest groups and the reasons behind their clustering around specific policy domains. Third, the lack of empirical research looking at variation in demand for input/output legitimacy and interest group diversity across policy domains has allowed discussions on the risk of institutional capture by business to expand indiscriminately to the entire system.

The reality, as this paper attempts to empirically demonstrate, is that a new complex form of chameleon pluralism has emerged within EU institutions. In light of the increasing diversity of the interest group population in Brussels and the dominance of business groups as a percentage of total activity (see Figure 1), it is clear that the attempt to establish corporatism as the system of interest intermediation failed to transpire (Streeck and Schmitter 1991). Rather, proponents of élite pluralism argue that business interests are the largest group in Brussels and are the most successful at managing to form pan-European alliances across most policy domains (Coen 1997, 1998; Eising 2010). However, recognizing the diverse needs of policy domains and the complexity of the policy-making process at the EU level, different types and numbers of interests can focus on different domains and different groups can exert influence depending on the needs of the DGs.

Figure 1 Percentage of interest groups per type out of total

We argue that this sub-system activity is not a form of neo-corporatism with its formal access rules and hierarchy (Streeck and Schmitter 1991). Nor is it Meso corporatism, which is seen as a defensive lobbying reaction of business to structural adjustments (Grant 1987). Rather, we see a pluralist relationship where access is generally restricted to interest groups that can meet the input/output legitimacy requirements of the European Commission, and DGs filter interest groups that can help them to formulate policy in their respective policy domains (Broscheid and Coen 2003). Thus density, type and interest activity in different DGs are a competitive and strategically advisable interest group game constrained by the bureaucratic informational needs of the policy. Noting the policy variation, ‘EU pluralism might be best characterised as a kind of chameleon pluralism, capable of changing its appearance over time during the policy cycle for a given policy problem or within a sub-sector over a longer period of time’ (Richardson and Coen 2009: 348).

EU SUB-SYSTEM MODEL OF DIVERSITY AND DENSITY

In this section, we present variables that have an impact on informational diversity across policy domains and therefore the chameleon lobbying exchange between interests and institutions. We argue that these variables affect demand for type of legitimacy and therefore informational type. In turn, this has an effect on the policy domains' interest group diversity. As the previous section has discussed, the level of expertise, the length of time that interests and institutions have been dealing with one another, the resource dependency (capacity for exchange), the nature of the policy good, and the intergovernmental nature of the policy all impact the type of interests, the density and the lobbying strategies. As such, we have identified four key variables that may affect the density and type of interest within the Commission: (i) expertise and capacity as measured by the size of staff of the DG; (ii) the type of policy domain (regulatory or distributive); (iii) the involvement of national or sub-national government in the policy domain; and (iv) age, as captured by the date of establishment of the DG.

Expertise and capacity: staff size

Resources and in particular staff numbers are critical for bureaucracies to make policy. In order to produce policy, DGs require staff to do research and go over information supplied by interest groups. The more staff a DG has available, the more man hours it can dedicate to its own research in order to promote its agenda to other policy domains and/or institutions. In addition, larger staffed DGs are likely to have a greater variation of opinion over policy (Egeberg et al. 2003). In order to have a better understanding of public opinion that will lead to publicly legitimate policy proposals and streamline internal research, better staffed DGs demand more political information. Smaller DGs by default can do less research than their better staffed counterparts. The ability to streamline information and address fully technical expertise will be smaller since resources are limited. Therefore, smaller DGs are more likely to demand technical information that is more complex and requires more resources to produce.

	H1a: DGs with greater numbers of staff have a greater concentration of interest groups that supply political information.
	H1b: DGs with smaller numbers of staff have a greater concentration of interest groups that supply technical information.

Policy makes politics: nature of the policy domain

Broadly, policy domains can be divided into two categories, namely (i) regulatory and (ii) (re-)distributive. Following from Lowi (1972), (re-)distributive policy domains involve the transfer and redistribution of wealth and resources, while regulatory policy domains administer and govern the conduct of relevant actors such as business. The demand for technical information is higher for DGs dealing with regulatory policy. First, in order to produce policy, DGs assess available policy options that are not only complex but part of niche expertise. They raise demand for technical information in order to ensure they will collect as much available information as possible. Second, regulatory policy domains are less politicized than distributive ones; input legitimacy plays a smaller role than output legitimacy. Therefore demand for political expertise is lower than demand for technical expertise. Distributive policy is less complex but much more politicized (Christiansen 1997; Princen and Kerremans 2008; Scharpf 2009). This increases the need for consensus and political information but overall reduces demand for technical information.

	H2a: DGs managing regulatory policy domains have a greater concentration of interest groups that supply technical information.
	H2b: DGs managing distributive policy domains have a greater concentration of interest groups that supply political information.

National government involvement

The extent of formal and informal involvement of national (and sub-national) government in policy at the EU level impacts the informational demands of DGs. Member states' governments are either provided with expertise by interest groups largely at the national level and/or can rely on their own national agencies (ministries, bureaucracy) (Saurugger 2009). The additional formal/informal involvement of member states in Brussels automatically increases the politicization of the policy-making process; because more national perspectives are available, the DG has greater demand for political information (Scharpf 2009). Therefore, DGs where member states are involved both informally and formally show greater demand for political expertise. In policy domains where there is less national government involvement, there is less politicization. This reduces demand for input legitimacy and increases demand for output legitimacy, raising demand for technical expertise, as issues are resolved at the DG, the Committee of Permanent Representation (COREPER) or working group level (Fouilleux et al. 2005).

	H3a: DGs managing policy domains with greater member state involvement have a greater concentration of interest groups that supply political information.
	H3b: DGs managing policy domains with less member state involvement will have a greater concentration of interest groups that supply technical information.

Age and reputation building: date of establishment of DG

In order for a DG to become established and to establish expertise, it needs to focus primarily on its policy output and secondarily on its input legitimacy (Neyer 2010). We would expect older DGs to have established insider relationships, as they have gone through a number of repeated games with compliant and non-compliant interest groups and would have used their discretion to exclude those seen to be misleading or untrustworthy (Broscheid and Coen 2003). This raises the quality of policy and output legitimacy, which allows them to spend their resources on input legitimacy. Secondly, the strong presence of insiders may cause institutional capture (Coen 2009). Thus ‘older’ DGs demand political expertise and ‘access activity’ in order to defend against accusations of capture. Newer DGs have gone through the policy-making process fewer times and therefore the insider relationships established are not as strong, while the relevant interest groups are likely to be less experienced, which increases demand for technical expertise in order to compare, select and raise output legitimacy.

	H4a: Older DGs have a greater concentration of interest groups that supply political information.
	H4b: Newer DGs have a greater concentration of interest groups that supply technical information.

However, the relationship between date of establishment and concentration of interest groups is ambiguous; NGOs in the 1960s were neither in the same numbers nor as relevant in Brussels as they are today (Greenwood 2007). This has given business groups an advantage over NGOs in older DGs, as they have had the chance to go over more iterated games and establish insider relationships (Coen 2007). This allows for the earmarking of resources for output legitimacy. Newer DGs offer a growing population of NGOs the chance to enter the policy game more or less under equal conditions with in-house groups, thus making it more likely for demand input legitimacy to take place.

DATA

Information on the different types of interest groups and their policy domains of interest was collected from the European Commission's Register of Interest Representatives (RIR) database. The RIR database at the time of collection contained just over 4,000 registered interest groups (Commission of the European Communities 2011).

The RIR contains four main categories of interest groups and 15 sub-categories; we examine only the main categories:

i.	Professional Consultants/law firms; contains the above-mentioned actors.
ii.	In-House; contains business interests such as companies and associations.
iii.	NGO; contains civil society groups such as SMOs, NGOs and think tanks.
iv.	Other; contains miscellaneous, mainly religious groups and regions.

We use the established DGs in order to define and operationalize the policy domains of interest of the groups.

After accessing the RIR online, we collected information on the policy domain of preference per type of interest group. Interest groups self-register on the database as part of the Commission's push for further transparency in the policy-making process that takes place in Brussels (Commission of the European Communities 2008). The interest groups that register provide detailed information such as their name and address, select category of interest group type (and sub-category), and select among 33 available policy domains the ones they are interested in (lobby for). Interest groups can register as many policy domains of preference as they wish. Because we utilize DGs in order to operationalize policy domains and the number of available policy domain options for registration is more than the DGs, some policy domains were collected together. In most cases the available options are DGs broken down into more than one policy domain. For the policy domains that were put together, the data per type of group were collected as a combination of the fields. Moreover, following this process some DGs that were similar and contained too few interest groups were also grouped together; the data were collected in the same way. This ensures that all information is gathered in the same manner and guarantees the validity of the data. For our inferential analysis the diversity of type of interest group per policy domain is operationalized as the percentage of type of interest group, as available by the database, over the total number of interests per policy domain. Because the data provide only a snapshot of the interest group population, we are interested in the relative position of type of interest group with respect both to the policy domain and the independent variables. We present this information in more detail in the next section.

Next, we collected data on the DGs that are used to operationalize the independent variables: (i) staff and (ii) date of establishment of each DG. This information was available on the Commission's website. The variables are continuous. For the other two variables, (i) type of policy and (ii) involvement of national/sub-national government in policy domain, we have coded these variables as dummies: (i) type of policy (distributive 1, regulatory 0); (ii) national government (involvement 1, non-involvement 0). We recognize that in some cases it is difficult to allocate a DG to only one specific category, for example environment, but here we made a value judgement that the preponderance of activity was of a certain type. For the DGs that were put together, the number of staff was added, the date of establishment was either the same or averaged.

In coding the involvement of member states in policy domains, we drew on the original coding of Broscheid and Coen (2007), who in turn use qualified majority voting (QMV) in the policy domains as an indicator. Accepting that the Lisbon Treaty altered the degree of delegation and state involvement in some policy domains, the authors looked at the formal definitions of member state involvement in policy-making in Brussels, such as competences and modes of governance (Naurin and Wallace 2008) to create the final codes. We have coded member state involvement based on the relative additional informal discussions/interference taking place across policy domains in Brussels with member state interest. Following this, we cross-examined our positioning with the relevant literature (see Fouilleux et al. 2005; Pollack 2000; Reh et al. 2010). That is to say, the ‘involvement’ of member states in a policy domain expresses the additional interest and pressure provided by member states in policy domains.

EMPIRICAL ANALYSIS

Figure 1 illustrates that the EU system is dominated by two categories of interest groups: In-House and NGO. Out of the 4,007 interest groups registered, 1,903 or 47 per cent are In-House while 1,246 or 31 per cent belong to the NGO category; the remainder is divided between Other (15 per cent) and Professional Consultants (7 per cent). This reaffirms previous empirical observations about interest group diversity at the EU system level (Wonka et al. 2010). It also provides provisional support for the assertion that the Commission's utility of different types of information is contingent on input and output legitimacy. That is to say that the Commission has a higher demand for output legitimacy and therefore technical regulatory information which is supplied by In-House type groups. Conversely, the European Commission has a lower demand for input legitimacy as mirrored in the smaller number of NGO type groups.

Second, the data provide validity for system-level élite pluralism as a conceptualization of the interest intermediation in Brussels. Although the number of interest groups has increased over the years, business groups have fared much better than others. This becomes even more evident when we look further into the main two categories of interests. The In-House category involves mostly companies and professional associations (84.5 per cent; 1,566) of the category. The NGO category includes mostly NGOs (civil society groups) (78.8 per cent; 982).

The smaller number of other groups suggests that their information is in less demand and may be the consequence of being intermediary groups, rather than primary actors interfacing with the public and consumers. For example, law firms and professional consultancies essentially provide services for parts of the rest of the interest group population and, moreover, do not have information to provide to the Commission directly. Secondly, they have fewer incentives to register on the RIR. Remaining informal (outside of the database) empowers their insider/outsider status, making them harder to detect and regulate. In the Other category we find mostly regions and religious groups. With respect to regions, relevant informational demands can be covered by two different actors instead. Firstly, the Committee of Regions can streamline the information available from regions; secondly national governments are better in gathering information from regions and stronger lobbying targets for regions.

Nevertheless, the system includes a respectable amount of NGO type groups – nearly a third, a percentage higher than that in the late 1990s. Thus there are indications that input legitimacy has a role to play and it is increasing over time, although not enough to tip the balance at the EU system level (Bellamy 2010). The number of interests surrounding Brussels and the Commission has grown drastically but output legitimacy has higher utility overall and firms and associations have the expertise demanded. This leads us to our central point of analysis, as this is not the case at the sub-system level, as indicated across the different policy domains in Figure 2.

Figure 2 Number of interest groups per type per policy domain (starting from earliest established DG)

The snapshot changes when we look at Figure 2 in detail. First, arguments and empirical evidence on the system level do not apply across the sub-system level. This goes against previous studies that assumed that the dominance of business interests applies across all policy domains. The diversity of the interest group population varies, in some cases strongly. Moreover, this variation appears to come as a result of the variables we have identified to have an impact on the DGs' demand for different types of information. Second, in agreement with previous studies, the density of the population across policy domains varies (Broscheid and Coen 2007).

As expected, policy domains that largely demand more input legitimacy and political information have a larger percentage of NGO type groups than In-House. The DGs managing them are distributive, have more staff and have greater government involvement. For example, DG Development (DEVCO) density includes 44 per cent NGO types but 31.3 per cent In-House; similarly education and culture (EAC) has 41 per cent and 33 per cent respectively; in human rights (ECHO) more than half of its population (55 per cent) is NGO type groups, overrunning in absolute numbers all other type groups.

Conversely, policy domains that demand more output legitimacy and more technical information are regulatory, with less staff and member state involvement. They show a greater percentage of In-House groups. For example, in the Common Market (MARKT) field 70 per cent of the category are In-House interests; the economics (ECFIN) and energy (ENER) fields have 60 per cent and 54 per cent respectively. However, policy domains are complex; when the demand for different types of legitimacy becomes more mixed, the diversity of the population follows. For example, the agricultural field (AGRI) has nearly an equal concentration of In-House groups and NGOs, as it is not purely a distributive field but also includes a lot of regulatory policies.

The results do not support hypothesis 4, with In-House groups appearing in larger numbers in older DGs. Most of those established before the 1960s show a larger concentration of In-House groups than NGOs. This can be justified as a result of business interests having more chances to enter into repeated games with these DGs and become insiders, earmarking resources for output legitimacy and technical information. In addition, most of these DGs are regulatory, which by default demands more technical information. Distributive ones such as AGRI or EMPL show higher percentages of NGO groups in comparison to the others. Moreover, in DGs established after the 1960s the percentage of NGOs increases. This could be the result of a growing NGO population that could counterbalance the dominance of business interests, as well as a maturing EU polity where input legitimacy plays a growing role.

A closer examination of the variance of the groups shows that overall In-House groups have a more stable presence across policy domains. However, NGO groups drastically increase in policy domains where demand for political information is likely to be greater, while In-House groups are reduced. This could be explained by the fact that because resources (benefits) are limited, demand for one type of information can reduce demand for another. Moreover, the more stable presence of In-House interest groups suggests that output legitimacy has a higher and more stable utility for the Commission across policy domains. The demand for input legitimacy is more volatile across policy domains.

This, firstly, shows that EU system-level theories need to be adapted to the evolving EU sub-system. We see an interchangeable form of pluralism across the EU sub-systems, with variation depending on legitimacy and informational demands. Thus it is no longer applicable to argue that interest representation in Brussels is a form of élite pluralism dominated by business; rather in the 2010s the conceptualization of chameleon pluralism is more relevant and shows the maturing Brussels polity.

Secondly, in accepting the greater variety of interest styles in the policy process, we add to the understanding of the EU formulation and agenda process. Recognizing that the EU regulator/regulatee relationship is more complex and evolving thus moves us to a more nuanced understanding of the concerns of business institutional capture in the EU policy-making process. As civil society groups provide input legitimacy, and may actually outnumber business interests in some cases, it is possible that in these policy domains discussions on business capture are inflated.

In the following section we move into our inferential analysis in order to examine the causal relations identified in more detail. Specifically we regress the dependent variable (percentage of type of group out of the total per policy domain) on the independent variables. Parameter estimates obtained via OLS are presented in Tables 1 and 2. We focus on the largest groups of the population, namely In-House and NGO. The causal relationships show in the data established that policy domains that have greater demand for output legitimacy, operationalized as DGs with: i) national government involvement; ii) are distributive; iii) are established for longer; iv) have more staff, will have a greater clustering of NGO type groups and therefore positive correlation. In-House type groups will cluster less around them and more around DGs with the opposite characteristics and therefore show a negative correlation

Table 1 OLS regression of percentage of In House interests

Variable	Coefficient	Std. Error	t-Statistic	p-values.
C	0.493794	0.052251	9.450394	0.0000
National Government	−0.076987	0.029823	−2.581419	0.0178
Distributive Policy Domain	−0.080790	0.034553	−2.338178	0.0299
DG Establishment	0.004721	0.001063	4.440582	0.0003
DG Staff Size	−0.000218	9.28E-05	−2.346164	0.0294
R-squared	0.588118	Mean dependent var		0.488287
Adjusted R-squared	0.505742	S.D. dependent var		0.127499

Table 2 OLS regression of percentage of NGO interests

Variable	Coefficient	Std. Error	t-Statistic	p-values.
C	0.275193	0.048192	5.710333	0.0000
National Government	0.055875	0.024212	2.307731	0.0318
Distributive Policy Domain	0.065641	0.027608	2.377622	0.0275
DG Establishment	−0.003282	0.000885	−3.708478	0.0014
DG Staff Size	0.000161	8.65E-05	1.857195	0.0781
R-squared	0.511286	Mean dependent var		0.290268
Adjusted R-squared	0.413544	S.D. dependent var		0.100978

The explanatory power of the model is good and the coefficients are statistically significant. As the p-values indicate, all coefficients are statistically significant at the 97 per cent level or above, except for DG staff size at the regression of NGO interests which is significant at the 92 per cent level. The magnitude of the coefficients depends on the units of measurement, therefore their quantitative impact should be interpreted accordingly. Consider, for example, the last regression (percentage of NGO interests). National Government and Distributive Policy Domain are dummy variables, so their quantitative impact is given by the coefficients. DG Staff Size is the actual number of staff, therefore an increase of staff by 100 people increases the percentage of NGOs by 0.016 (0.016 = 0.00016*100). DG Establishment is measured as 2011 minus year of establishment. Therefore, a DG that has been established for ten years reduces the percentage of NGOs by 0.03.

The following broad observations can be made from the two tables. The parameter estimates for National Government indicate that the involvement of national (and sub-national) government in the policy process of specific policy domains reduces the clustering of groups that provide technical information (In-House), but it increases that of groups that provide political information (NGO). This can be attributed to the fact that member states supply their own technical information, which they have either acquired from business interests at the national level or their bureaucracies, causing limited demand for technical expertise in Brussels. At the same time the higher politicization of the policy domain requires more political information in order to establish consensus.

Distributive policy appears to drive away In-House lobbyists but attract NGO group types. In highly politicized domains the Commission demands political information in order to build up input legitimacy. Moreover, distributive policy is less complex than regulatory policy, which explains the reduction in In-House groups and the increase in NGO type groups that can provide more political information. Finally, it suggests that contrary to supply side arguments, NGOs will lobby not only within the national realm but also at the EU level.

As we have discussed above, the relationship between date of establishment of a DG and concentration of interest group is ambiguous. The positive relationship for In-House interests and the negative for NGOs can be justified considering the population of both groups in Brussels in the 1960s. In-House groups have had a head start in comparison to NGOs. Through repeated games, In-House groups managed to gain trust and ring-fence older DGs when NGOs were in smaller numbers. Older, more experienced DGs already have insider circles that provide them with expertise on technical information and have earmarked more resources for them.

It also appears that better staffed DGs, operationalized as number of staff per DG (logged), attract more NGO type groups while smaller ones attract In-House groups. Better staffed DGs are better able to manage and even produce their own technical information, such as reports, studies and so on. This allows them to produce high-quality policy, which increases their output legitimacy, and spend resources in order to increase their input legitimacy. This encourages the clustering of NGO type groups. On the opposite side, In-House type groups will avoid well-staffed DGs and go after those with less staff as they require more technical expertise in order to retain output legitimacy.

CONCLUSIONS

The results largely confirmed our hypotheses; policy domains that place higher utility on output legitimacy demand more technical expertise and as a result show a stronger presence of business interests and associations. On the other hand, policy domains that place higher utility on input legitimacy demand more political information and show a stronger presence of civil society type groups. Nevertheless, our results confirm previous studies' conceptualizations of interest group intermediation and population diversity at the system level (Wonka et al. 2010), as well as density at the sub-system level (Broscheid and Coen 2007).

This paper has three main implications. First, from a theoretical perspective we have linked legitimacy considerations with informational approaches. This connects ‘what’ type of information and ‘why’ it is demanded with ‘who’ supplies it. Second, the paper provides empirical evidence that shows the diversity of the interest group population across policy domains and variables that affect it. Third, it shows that EU system-level conceptualizations such as élite pluralism, although still applicable at the system level, do not apply at the sub-system level. Thus the reality that we observe is a form of chameleon pluralism and the flexibility it offers defines much better the existing interest representation model for the EU sub-system level. Thirdly, we have a more fine-grained analysis of the regulatory state relationship between regulator and regulatee that perhaps allows us to create a more nuanced micro contribution to the institutional capture debates. However, further research is needed in order to clarify the ‘active’ representativeness of these groups and the diversity of sub-systems across different institutions.

ACKNOWLEDGEMENTS

The authors are grateful for insightful comments from Richard Bellamy, Michelle Egan, Wyn Grant, Neil Mitchell, Christine Reh, Nino Majone and Albert Weale and participants at the IPSA June 2012 conference, as well as the JEPP referees. The research for this article was funded by a Nuffield Foundation Grant number SGS/3915.