MPIfG Working Paper 06/3, May 2006


Lobbying Systems in the European Union: A Quantitative Study


Andreas Broscheid , University of North Carolina, Pembroke

David Coen , School of Public Policy, University College London






This paper presents and tests a micro-theoretical model of EU lobbying across policy domains. In particular, we focus on two questions: first, we want to know why the number of interest representatives differs across policy domains and, second, we investigate why we find institutionalized fora for interest representation in some policy domains but not in others. Our argument focuses on the Commission's need for expert information and its costs of managing contacts with a large number of interest representatives. Both factors provide incentives for the Commission to create restricted-access fora as the number of interest representatives increases. Using cross-sectional data on interest representation in a wide range of policy domains, we find some support for our hypotheses.






In diesem Artikel entwickeln und testen wir ein mikrotheoretisches Modell, das zur Erklärung der Interessenvertretung in verschiedenen Politikfeldern in der Europäischen Union beiträgt. Dabei stehen zwei erkenntnisleitende Fragen im Vordergrund: Was beeinflusst die Zahl der Interessenvertreter in verschiedenen Politikfeldern? Und weshalb richtet die EU-Kommission in einigen Politikfeldern Foren mit beschränktem Zugang für Interessenvertreter ein? Unsere Erklärung basiert auf der Beobachtung, dass der Expertisebedarf der Kommission hauptsächlich durch Interessenvertreter gedeckt wird, dass aber die Interaktion mit einer großen Zahl von Interessenvertretern der Kommission Kosten (unter anderem Informationskosten) verursacht. Beide Beobachtungen führen unseres Erachtens dazu, dass die Kommission Foren mit beschränktem Zugang einführt, wenn die Zahl der Interessenvertreter ein bestimmtes Maß überschreitet. Wir vollziehen einen ersten Test unserer Hypothesen mit Querschnittsdaten der Interessenvertretung in verschiedenen EU-Politikfeldern, und finden unsere Erwartungen zumindest teilweise erfüllt.






Micro and macro studies of lobbying

Why are there more lobbyists in some policy domains than in others?

Empirically testing the first two hypotheses

Why are there more interest representation fora in some policy domains than in others?

Empirically testing hypotheses three and four



Data Appendix





Introduction [1]


The study of interest representation has been characterized by an interesting micro/macro distinction. Many studies that emphasize a systemic or sub-systemic perspective are dominated by informal theories and mainly descriptive empirical approaches. In contrast, micro-level studies of lobbying processes emphasize causal theories, which are frequently developed in a formal mathematical framework. We do not find many macro studies that build on micro models of lobbying, and we do not find many micro models of lobbying that address systemic questions.


This paper intends to contribute to the debate connecting micro and macro approaches to the study of lobbying. Using predictions from a formal model that we have detailed elsewhere (Broscheid and Coen 2003), we present preliminary statistical tests of the implications of our theory. Specifically, we try to find answers to two questions:


1. Why does the number of interest representatives differ across policy domains?

2. Why do we find institutionalized fora for interest representation in some policy domains but not in others?


We believe that a focus on organized interests targeting the European Commission is a good starting point for this investigation. First, the currently dominant formal theories of interest representation emphasize the role of informational lobbying, which is usually acknowledged to be the predominant type of lobbying of the Commission. As a result, the Commission can serve as a good empirical test case. Second, the Commission's relatively independent and specialized directorates focus on fairly cohesive sets of policies. The patterns of interaction surrounding the directorates generally provide a useful operational definition of policy domains. Thus, it is possible to compare different policy domains with different lobbying costs, expertise requirements and the like. Third, the study of European Commission lobbying is important as it deals with the potential influence of European civil society on the institution that shapes and implements legislation.


The paper will start with a brief overview of those aspects of the macro and micro literature on interest representation that have motivated the present study. We then provide a brief informal summary of our theory connecting the two perspectives, followed by a discussion of our data and empirical results. In conclusion, we present some perspectives for the study of European lobbying.


Micro and macro studies of lobbying


The main focus of system-level studies of lobbying has been the equality and fairness of the representation of all social interests. Pluralist studies of interest representation usually conclude that representation is generally fair, as under-represented interests would become involved if their interests are not sufficiently taken into account by decision makers (Truman 1951). The government, from this perspective, appears mainly as a neutral broker between the different interests.


The pluralist approach first came under attack from Olson's micro-level theory of collective action, which argued that certain interests were less likely to organize and hence to be politically influential (1965). Among macro-level studies, this argument provided parts of a micro foundation of theories of elite pluralism (for example, Schattschneider 1960) or studies of the varying power of industry interests (among others, McFarland 1991). The most extreme counter-theory to the pluralist argument of equal representation was neo-corporatism, an approach that combined a descriptive account of interest representation with an applied argument about the effectiveness of different lobbying systems in such areas as social policy and labor relations. In contrast to pluralist systems of interest representation, neo-corporatism was characterized by a monopoly of representation through peak associations and national labor unions (Streeck and Schmitter 1991). The great benefit of many macro theories was the ability to describe and categorize different systems of interest representation, and to provide a basis for comparative studies of different political systems (see, for example, Wilson 1990). Their weakness was often on the explanatory side, as Olson's criticism of pluralist theories highlighted.


Pluralism, elite pluralism, neo-corporatism and their variants represent high-level theories of entire political systems. A number of meso-level approaches have focused on the inclusiveness of interest representation at the sub-system level, particularly focusing on the distinction between insider and outsider lobbyists (Broscheid and Coen 2003; Grant 2004). Depending on their level of exclusivity, such sub-systems have been characterized as iron triangles (Freeman 1965; McConnell 1966), subgovernments (McCool 1990), issue networks (Heclo 1978), or advocacy coalitions (Sabatier and Jenkins-Smith 1993). Insiders can be simply actors that are frequently consulted (as in the case of issue networks or advocacy coalitions), or they are actors actively involved in bargaining and policy negotiation or in the implementation of policy solutions (Maloney et al. 1994). The main focus in this literature is on the categorization and recognition of different types of policy networks and their role in the policy process, not the explanation of their existence (for an exception, see the studies in Marsh and Rhodes 1992).


The insider/outsider lobbyist distinction is of particular importance for the study of European Union lobbying. While the European Commission attempts to be open and transparent in its interaction with societal interests, nevertheless a core of insiders has been established. We see an elite pluralist system in the form of fora to which "access is generally restricted to a few policy players, for whom membership is competitive and strategically advisable" (Coen 1997: 98). The selection of lobbying insiders is managed and organized with a wide variety of committees, working groups, conferences and other policy fora (Pedler and Schaefer 1996). In this study, we investigate factors that may cause the differently structured lobbying systems surrounding the European Commission. This includes looking at the number of actors in areas dealing with different types of policy questions as well as the question of insider lobbying through European Commission fora.


In order to propose a causal explanation of lobbying sub-systems, it is important to look at existing causal explanations of lobbying. The predominant formal-theoretical approaches view lobbying as the strategic communication of specialized information (for a recent summary, see Grossman and Helpman 2001). Building on game-theoretic models developed throughout the 1980s (such as Kreps and Wilson 1982; Crawford and Sobel 1982), these approaches argue that interest representatives have policy-relevant information that policy makers need in order to make effective policy decisions. If the political goals of interest representatives and policy makers diverge, then information may be transmitted in a biased manner. Although the policy maker takes informational biases into account when interpreting lobbying signals, the informational advantage of interest representatives provides them with political influence.


There are several variants of informational lobbying models. They can be based on whether the information is about the interest representatives' constituents (Ainsworth 1993; Potters and van Winden 1990) or about the impact of the policy environment on policy outcomes (Grossman and Helpman 2001). Other variations may be based on the number of lobbyists and on their position towards each other (see, for example, the Austen-Smith/Wright model on counteractive lobbying, 1992) or on the nature of the signal (for example, on discrete versus continuous signals, see Grossman and Helpman 2001; on costless versus costly signals, see Austen-Smith 1995; Lohmann 1995). One common element of almost all informational models is the fact that they investigate the interaction of one or two lobbyists with a unified government. This makes it difficult to draw conclusions about lobbying systems, which usually consist of more than two lobbyists. The model that we use in this paper tries to tackle this problem.


The informational approach is particularly useful for the study of lobbying in the EU (Crombez 2002). In Brussels the key to successful lobbying is not political patronage or campaign contributions, but the provision of information. In this context, the Commission, with its executive instruments and directives, acts as the focal point in the early stages of the lobbying process. As a technical bureaucracy it does not seek funds for re-election, but rather looks for a policy community that may provide a source of grass-roots and European-level information (Bouwen 2002; Coen 1997). The demand for the two types of information may vary across policies. For example, if a policy deals with technical standards or the regulation of sophisticated products such as pharmaceuticals, substantive expertise is very important. On the other hand, for policy that has (or might acquire) a high level of political salience in the member states, the Commission requires information on the preferences of relevant actors in the several states.


For the purpose of this study, we do not focus on the distinction between different kinds of information. We argue that technological as well as preference information addresses the question of whether a policy proposal "works," that is, whether it has a desirable outcome and whether it will be acceptable to the actors involved in the political decision-making process. However, technical information may be more costly to obtain, whereas information on preferences can be obtained at low cost by some organizations. In both cases, the Commission has to rely on private actors to provide it with much of the information it needs; therefore, there is opportunity for interest representatives to influence policy.



Why are there more lobbyists in some policy domains than in others? [2]


The number of groups representing different interests has been one of the most important questions in the interest group literature; existing explanations have focused on the number of potential group members and the role of selective incentives, patrons, political entrepreneurs and other factors in mobilizing these potential group members. Interestingly, a comparison of the number of groups in different policy domains – the density of interest group populations – is far less common (see, for example, Heinz et al. 1990; Mahoney 2004), and a theoretical account of group density is missing. In order to provide such an account, we believe that it is important to analyze the interaction between groups and decision makers, as different policy domains may exhibit differences in this interaction.


It has been observed that the interaction between interest representatives and the European Commission (and other EU institutions) is based on information (Bouwen 2002; Coen 1998). The Commission's staffing levels are very low, compared to the extent of its tasks (van Schendelen 1996), and interest representatives are often needed to provide expert information. At the beginning of our theoretical investigation, we therefore ask under which circumstances lobbyists are willing to provide useful information.


Lobbyists present specific positions on issues, and they can always present their slant on a given issue. In the world of formal models of communication (such as Gilligan and Krehbiel 1987, 1989; Lupia and McCubbins 1997), this is called "babbling": the lobbyists (or the "senders" of information) provide a standard recommendation, independently of whether the information consumer (or "receiver", here the Commission) may agree with this information or not. We argue that lobbyists always "babble" unless they receive specific incentives – rewards for informative signals or punishments for non-informative signals – to provide better information.


Why do lobbyists always babble, in the absence of rewards or punishment? Consider a situation in which a lobbyist provides information that is damaging to her [3] interests. The recipient of such information will surely take it very seriously, as it is obviously not self-serving information. As a result, it is very likely that the recipient (here the Commission) will act on this information in a way that is detrimental to the lobbyist's interests. Therefore, the lobbyist has no reason to provide this kind of detrimental information – unless she receives a reward, essentially a compensation, for its provision or is punished for not providing it.


Does the European Commission provide rewards for useful information? We believe that it does, by providing privileged access to lobbyists and interest groups that consistently provide such information. Actors with privileged access are routinely consulted, invited to workshops, consultative fora, etc. that form part of the policy-making process and allow lobbyists to influence policy more effectively. Furthermore, access translates into knowledge about political and administrative developments at the EU level, which in turn can translate into advance knowledge about EU contracts or grants or into influence on the early stages of the policy process (Coen 1998, 1999). These are all things that are highly valued by lobbyists and interest representatives. By granting or denying access, the Commission can reward useful information or punish babbling.


How does the Commission know whether a lobbyist babbles or provides useful information? It learns it after the fact, once it has proposed a policy, implemented a regulation, taken action to enforce a directive, issued a ruling, etc. If its action turns out to be bad – a proposal fails in Council and Parliament, member states resist a directive, a decision results in unexpected, negative outcomes, and so on – the Commission has to conclude that the information on which it acted was erroneous. If the Commission concludes that it was "suckered" by self-serving information provided by lobbyists, these lobbyists may lose their access and have to invest resources to regain the Commission's confidence.


These considerations help us establish the incentives that can induce an actor to become a European Union lobbyist – to enter the political fray and try to influence the European Commission (and other institutions). First, a lobbyist will attempt to influence Commission decisions because she has policy interests. In political-economic terminology, a lobbyist expects policy utility gains if the policy she prefers is supported by the Commission. These policy gains can be ideological (if the Commission proposal conforms to the actor's political outlook) or they can be material (in terms of budgetary transfers to the region represented by the actor, for example, regulations that favor the actor's industry, and so on). In general, it seems to make sense that actors are more likely to become lobbyists if large policy benefits are at stake. However, our formal model shows that the situation is more complicated (Broscheid and Coen 2003: 176). If potential policy gains are large, lobbyists have a stronger incentive to babble, as the size of the policy gains outweighs the possibility of losing access. This means in turn that under some circumstances high policy stakes may lead to uninformative lobbying.


Second, the actor/lobbyist will receive non-policy benefits, such as information on policy developments, European Union grants and contracts, and so on. As we noted above, these benefits can be granted (and withheld) by the Commission as a reward for accurate information (or as punishment for babbling). One interesting property of these non-policy benefits is that they are divisible. Access is most valuable if few other actors have it. If many actors in a policy domain gain access to the Commission, however, its value decreases, as each actor receives only a smaller share of the time each Commission official can spend with interest representatives. In addition, the comparative advantage of inside information decreases, as many other actors in the domain obtain this information and can act on it. This has important consequences for the amount of babbling versus informative lobbying that takes place: as more actors become lobbyists, non-policy (access) incentives become diluted, and the incentives for informative lobbying become smaller. Consequently, more lobbyists will be tempted to present non-informative political propaganda instead of useful information.


Third, the actor considering whether or not to become a lobbyist has to consider the costs of lobbying. First, these costs consist of organizational costs – mobilizing potential members, perhaps establishing a Brussels office, and the like. Second, the actor has to incur informational costs. If she wants to be taken seriously as a lobbyist, she has to show that the information that she presents to the Commission is reliable and based on expert information. In some cases, expert information is comparatively easy to obtain for a lobbyist; if the information required by the Commission is about the preferences of the group of actors represented by a lobbyist, the lobbyist simply has to survey her members' preferences. In other cases, however, expert information is of a technical nature and more difficult to obtain. Lobbyists may have to pay for scientific and other expert studies to credibly provide the information demanded by the Commission.


Taking these three factors together, an actor will become a lobbyist in a policy domain if the expected policy and non-policy benefits outweigh the organizational and expertise costs. How does this argument help us explain why some policy domains have more lobbyists than other domains? Remember that non-policy benefits decrease as the number of lobbyists in a domain increases. As a result, there is an optimal number of lobbyists in any policy domain; if this number is reached, there are no non-policy incentives for additional lobbyists to incur the informational and organizational costs and join the lobbying population as informed interest representatives. As we argued above, without non-policy incentives the lobbyists would only babble if they joined and, as a result, would not influence Commission policy – which means that there are no policy benefits to be derived from lobbying.


But how do we determine whether the optimal number of lobbyists is larger or smaller in a given policy domain? Here, we can consider the factors that influence the size of non-policy benefits and organizational/expertise costs to deduce two hypotheses:


Hypothesis 1. If the lobbyists in a policy domain receive greater non-policy benefits from lobbying than in another domain, we can expect more lobbyists to be active in the first domain.


Hypothesis 2. In a domain in which lobbying is comparatively costly, we will find fewer lobbyists than in a domain in which lobbying is less costly.



Empirically testing the first two hypotheses


The choice of a viable unit of analysis constitutes a problem. Our theoretical model talks about the number of lobbyists involved in the making of a particular government decision. Therefore, a straightforward test of our theory would have to look at the numbers of lobbyists involved in many different decision-making processes. We do not have such data, and they are very costly to obtain in a systematic manner. However, instead of looking at individual Commission activities, we can investigate interest populations in different policy domains. The hypotheses that we developed with respect to individual Commission activities can be easily extended to policy domains:


Hypothesis 1A. If the lobbyists in a policy domain receive greater non-policy benefits from lobbying than in another domain, we can expect more lobbyists to be active in the first domain.


Hypothesis 2A. In a domain in which lobbying is comparatively costly, we will find fewer lobbyists than in a domain in which lobbying is less costly.


Once we settle on policy domains as units of analysis, however, the difficulties begin, since we have to determine the exact boundaries of policy domains at the level of the European Union. Intuitively, we have a clear sense of different domains – agriculture, health policy, chemicals, labor policy and so on, based on the subject matter of laws and regulations. But a closer consideration leads to difficult definitional questions, such as: Does fisheries policy belong to agricultural policy, or is it a separate policy domain? Does pharmaceuticals policy belong to chemicals policy or to health policy, or is it a domain on its own? The answer is structural: policy domains can be identified as patterns of actor networks. Unfortunately, this creates an empirical problem: since we want to explain the structure of policy domains, we cannot use such a structure to define our units of analysis and thus our dependent variable. We have to find an independent indicator to distinguish between policy domains.


Our solution to the problem is to rely on existing institutional boundaries provided by the European Commission. Each directorate general (DG) that is involved in policy making and policy implementation roughly conforms to a policy domain, or a set of closely related policy domains. The institutional structure of directorates general is based on, and creates, patterns of regular interaction between different groups of governmental and non-governmental actors, which approximate the shape of existing policy domains. Furthermore, the jurisdictional boundaries of directorates general are obviously not the results of our empirical analysis, thereby guaranteeing that we do not choose those domain boundaries that create empirical support for our hypotheses.


Comparing the number of lobbyists associated with each directorate general creates new problems, as DGs differ in range (the number of policy issues they deal with) and intensity (the amount of policy-making activity they engage in). The latter factor is not a serious problem for us since the intensity of Commission activity is implicitly part of our theoretical discussion – it leads to varying levels of policy and non-policy benefits for groups. Policy domains with equal levels of Commission activity would be useless for our analysis. In fact, our main independent variables encode information about the levels of DG activities.


Our dependent variable is the number of interest groups active in a policy domain, which we obtain from the European Commission's Conneccs database. This database contains listings of labor and employer organizations, business associations, NGOs, and community-based organizations. Conneccs entries are based on voluntary reports by interest groups, which also detail one or more predefined policy areas in which they are active. As the Conneccs policy areas closely conform to the jurisdictions of different directorates general, it is easy to obtain the dependent variable from these data.


Although Conneccs relies on self-reported entries, we believe that it provides valid data for our present inquiry. First, our theory refers to interest representatives who have incurred the expertise and organizational costs of credible lobbying. It is likely that the Conneccs database weeds out to some extent those groups that are not serious participants in EU lobbying. Second, our investigation focuses on the European Commission. Even though there are additional databases of EU interest representatives, we believe that Conneccs is useful for our purposes as the European Commission created it. Hence, it is likely to exclude those actors that do not interact with the Commission.


In order to test our hypotheses, we have to identify data that measure non-policy benefits and lobbying/organizational costs. Unfortunately, it is very difficult, if not impossible, to measure these types of costs and benefits directly. However, we can observe indirect indicators that provide us with a sense of whether such costs and benefits are higher or lower in different policy domains. First, we code the number of policy-related units of each DG as an independent variable. While the number of units records the complexity and number of policy issues that a DG deals with, it also provides a measure of the policy benefits provided by the DG. Second, we include the number of DG staff as an independent variable. This measure serves as a proxy of both policy and non-policy benefits of lobbying. The more staff a DG has, the more policy it can propose, implement and enforce. Furthermore, since non-policy benefits are to a large extent linked to contact with Commission officials, an increased number of staff will result in higher non-policy benefits. Since policy benefits are already controlled for by the number of units, the coefficient of the staff variable will reflect mainly the impact of non-policy benefits on the number of interests.


The organizational costs of groups are difficult to measure at the macro level. As an approximation, we suggest that broad differences between policy areas reflect organizational and lobbying costs. In regulatory policy domains, the Commission requires a comparatively high level of expertise input and, as a result, interest representatives are expected to provide such information. This increases the costs associated with lobbying. Conversely, in distributive policy domains,[4] lobbying involves to a larger extent the representation of group interests; the expertise required in such policy area is more of an administrative nature and not likely to be provided by lobbyists. As a result, we assume that organizational costs are lower in distributive policy domains than in regulatory policy domains and the number of groups are correspondingly larger. We code distributive policy domains with a dummy variable that is '1' for the DGs Agriculture, Education and Culture, Employment and Social Affairs, Fisheries, Regional Policy, and Research and '0' for all other directorates general. Our hypothesis predicts that the coefficient of this variable is positive. However, it is possible that the distinction between distributive and regulatory policy domains corresponds not only to expertise requirements but also other to factors. In particular, as the European Union is predominantly a regulatory policy maker, we can expect policy type to be associated with the intensity of Commission policy-making activity. By including the number of policy units in our regression model, we try to control for this factor. However, if policy units do not perfectly capture the amount of policy making, the estimated impact of policy type on the number of groups may be negative rather than positive.


We include three control variables in our analysis. First, the age of the policy domain is important. As time passes, more groups can be formed; conversely, in "young" policy domains, some groups may not yet have been formed. Lowery and Gray (1995), for example, make this argument in regard to the American states. On the other hand, it is possible that new policy domains have a larger number of groups: the transfer of authority to the European Union may be the result of increased interest representation. Also, the Commission may be particularly active in new policy domains and thereby trigger group activity. We operationalize the age of a policy domain with a dummy variable that marks Justice and Home Affairs, Humanitarian Aid, and Health and Safety. In terms of Commission authority, these policy areas are not older than the Amsterdam Treaty: Justice and Home Affairs, for example, was transferred to the First Pillar of the EU by the Amsterdam Treaty.


The second control variable marks policy domains in which national or sub-national governments play a dominant role. We suppose that in such policy domains we should find fewer societal interest groups because the main interests are represented by governments (and their organizations). We use a dummy variable that is '1' for Competition, Economic and Financial Affairs, Enlargement, External Relations, Justice and Home Affairs, Regional Policy, Taxation and Customs Union, and Trade.


The third control variable is the number of consultative fora for interest representation. In the second part of this paper, we will focus on this variable as a dependent variable. Here, we include it to investigate whether there is the possibility that there is mutual causation between the number of groups and the number of fora. Substantively, consultative fora could foster the formation and participation of groups, as they create an insider-outsider divide that provides incentives for outsiders to become insiders.


We use an OLS regression model to estimate the relationship between the variables. We are aware that such a model may not be, strictly speaking, appropriate, as we are dealing with count data. However, the dependent variable ranges from 10 groups (Fisheries) to 221 groups (Enterprise), and it can be treated as approximately continuous. Since the variable is bounded below by zero, we use the logarithmic transformation of the dependent variable for our estimation. Since the resulting regression is nonlinear, we use bootstrapped error estimates and confidence intervals based on the bootstrapped coefficient quantiles (for more information on bootstrapping, see Shikano 2006).





The results are summarized in table 1. Our main variables of interest provide mixed results: the number of personnel has a positive relationship to the number of groups, but the estimated coefficient is not significant (the confidence interval includes the value 0); the number of policy units has a negative coefficient, contrary to expectations, but that coefficient is also insignificant. Distributive policy domains, in contrast, differ significantly from other policy domains; however, they have fewer interest groups than regulatory policy domains, not more, as predicted by our theoretical model. Newer policy domains have more, not fewer actors, contrary to our presumption, even though this difference is not significant. As predicted, policy domains in which national government interests dominate have fewer groups, but the difference is not significant, either.


The main substantive result of our estimation is the impact of the number of fora, which has a significant and sizeable coefficient. Due to the logarithmic transformation of the dependent variable, the substantive impact of the number of fora depends on the value of the dependent variable and is not easily summarized. For example, at the mean value of the dependent variable (about 59 groups) an increase of the number of fora by three increases the number of groups by ten. How can we explain this relationship? First, the number of fora may simply be an indicator of the political activity of the Commission (see Mahoney 2004). More active directorates general, such as Agriculture, should have more fora than other DGs, and they will attract more groups. Second, fora may stimulate group participation by creating an insider-outsider dynamics. As Coen noted, interests groups compete to have seats at the inner policy tables and are willing to spend significant funds to develop positive European credentials for favored access (Coen 1997, 1998). In such a competitive environment it is possible to envisage that the Commission pump-primes lobbying activity by initially inviting insiders that have proven themselves in the Brussels environment and then, on occasion, funding the creation of new pan-European groups. As a result, potential insiders step up their EU lobbying activity to establish credibility and improve access in later rounds of policy making. Under such conditions the creation of fora and the emergence of policy insiders can ratchet up interest group activity in Brussels.


One possible reason for the insignificance of the personnel and policy-unit variables may be multicolinearity among these factors. Indeed, the two variables are highly correlated (0.91), which possibly leads to inflated standard errors and coefficients that are sensitive to small changes in the regression specification. We tested for this by estimating two versions of the regression model that removed either the personnel or the policy-unit variables; the results did not change substantially.[5]


Another problem that may lead us to underestimate the impact of our independent variables is the possibility that outliers and/or influential cases are responsible for some of the results. In order to check for this possibility, we ran several standard tests that identify outliers and influential observations. [6] The observations corresponding to Education and Culture and DG ECHO (Humanitarian Aid) were consistently marked by these methods. Regression results that exclude these two observations are summarized in table 2. Overall, the results are not much different from those that include the two influential observations, except that now the coefficient for new policy domains is significant and positive.





So far, the results are mixed. If we inspect the entire data, the number of fora and the difference between regulatory and distributive policy domains exhibit significant impacts on the number of groups. If we exclude two influential observations, the age of policy domains becomes significant. Except for the number of fora, the impact of the significant independent variables do not provide clear evidence for our theoretical model, even though they help us understand the factors that influence the number of interest groups in a policy domain.



Why are there more interest representation fora in some policy domains
than in others?


The second general question that this paper addresses deals with the conditions under which the Commission establishes fora for interest representation, thereby giving some interest representatives privileged access. We argue that the decision to establish fora is the result of a trade-off between the informational needs and the legitimacy needs of the Commission. As in our discussion of the number of groups in a policy domain, we provide a non-formal summary of arguments whose formal derivation has been presented elsewhere (Broscheid and Coen 2003).


We have argued above that the provision and possible withdrawal of non-policy incentives by the Commission constitutes an important incentive for lobbyists to provide accurate information. Furthermore, we have argued that, in the case of the Commission, these incentives are closely linked to the provision of access to decision-making processes: those actors that provide accurate information will be rewarded by continuous access, and those that are found to have provided inaccurate information will be excluded. The problem with such access-related incentives is that as more actors receive those incentives, the amount each individual actor receives decreases. As a result, "crowded" policy domains will provide smaller incentives for lobbyists to provide accurate information – they babble.


If the number of interest representatives is too large, we can talk about "access overload" (Coen 1997): the number of interest representatives dilutes non-policy incentives to such a degree that there is little informative lobbying and lots of babbling. We argue that in such a situation the Commission has incentives to select some interest representatives and provide them with privileged access, thereby increasing the non-policy incentives those representatives receive. [7] Since selected interest representatives can lose their privileged access if it turns out that they provide inaccurate information, the Commission thus creates incentives against babbling. One possible strategy of selecting insiders is the creation of fora for interest representation in which Commission officials regularly consult with a select group of societal actors. This argument leads us to our third hypothesis:


Hypothesis 3. The probability of observing Commission consultation fora for societal interests increases with the number of groups in a policy domain.


Why do we talk about the probability of observing fora, instead of stating that a large number of groups (deterministically) leads to the creation of fora? The answer is that there are several other important factors that we believe interfere with the informational rationale of granting privileged access to some lobbyists. The key to these interfering factors is legitimacy. As a non-majoritarian institution with comparatively little democratic oversight, the European Commission has to exercise a variety of politically charged duties, such as the proposal of European legislation or the formulation of administrative guidelines, opinions and the like. Without a democratic basis, the legitimacy of Commission decisions is fragile and has to rely on accountability, shared norms, broad-based support for policy output, and regular consultation with a wide range of societal actors.


Scharpf (1999) distinguishes between input-oriented and output-oriented legitimacy. As to input-oriented legitimacy, Scharpf states that "modern input-oriented theorists rarely derive legitimacy primarily from the belief that 'that people can do no wrong.' Instead, they insist that policy inputs should arise from public debates that have the qualities of truth-oriented deliberations and discourses" (269). If we view input legitimacy from this perspective, then the Commission has to solve a dilemma. As it does not derive its authority directly from the people, it has to rely on public, truth-oriented debates. On the one hand, this means that it has to foster "truth-oriented deliberations," which may require it to establish institutional structures that limit "babbling" – Commission fora. On the other hand, it has to establish public debate in policy areas that often are highly technical and of low public salience. As the Commission restricts access to a few privileged actors, it limits the breadth of public debate, while at the same time increasing the truth orientation of the debate. According to these considerations, Commission fora should be more likely if policies require reliable information rather than broad-based consultation; we suggest that it is technical, regulatory policy domains that will see the establishment of fora as the number of interest groups becomes too large.


With respect to output-oriented legitimacy, Scharpf argues that "collectively binding decisions should serve the common interests of the constituency" (268). Since the Commission has no direct majoritarian basis, it has to determine the common interest through consultation. For policies that affect a wide range of actors and that are fairly non-technical, this requires broad consultation of societal actors and a consideration of their political demands. In such policy domains, the selective restriction of access will lead to a lower degree of legitimacy for Commission decision making. For policies that are highly technical and of low salience, the quality of information that the Commission uses for policy making is more important. Hence, in such policy domains it is more likely that the Commission is willing to restrict access in order to improve the informational basis for its policies and hence improve their output legitimacy.


These considerations lead us to our fourth hypothesis:


Hypothesis 4. The Commission is more likely to establish fora in policy domains in which technical information is required to make good policy. In policy domains that are less technical and that affect a large number of societal actors, Commission fora are less likely.



Empirically testing hypotheses three and four


The empirical test of our explanation of fora creation is based on the same unit of analysis and the same data used in our analysis of the number of groups in different policy domains. The dependent variable is the number of fora in different policy domains (which coincide with Commission directorates general).


The first independent variable of interest, the number of groups, is our previous dependent variable and does not have to be discussed any further. More difficult is the creation of a measure for the level of technical information required in a policy domain, and the number of societal actors affected by it. We try to solve this difficulty by relying on the general distinction between regulatory and distributive policy domains. Distributive policy domains may be bureaucratically complicated, but this type of administrative expertise can be presumed to reside in the Commission bureaucracy that grew with these policy areas; the Commission will not demand administrative expertise from societal interests. Also, distributive policy domains such as agricultural or regional policy affect a wide range of actors and are of comparatively high public salience. Regulatory policy, on the other hand, is usually fairly technical, of low salience, and directly affects only a small number of social interests (because even though there may be indirect effects – for example, if regulation causes higher prices – affected actors tend not to attribute these effects to the policy). We measure the distinction between regulatory and distributive policy with the same dummy variable used in our analysis of the number of groups.


Since a large number of groups is expected to be associated with Commission fora in regulatory policy areas, we also include a cross-variable of the distributional policy dummy and the number of groups. If the coefficient of the groups variable is positive (which is expected), then the cross variable should be negative. As control variables, we include measures of the number of personnel (as more personnel may facilitate the organization of fora) and the age dummy used in the group analysis (as newer policy domains may not have established fora even though they may in the future).


The dependent variable – number of fora per policy domain/DG – is a count variable, ranging from zero to 30. Since 13 of the 21 cases are located at the lower bound (no fora or just one forum), the assumption of a continuous dependent variable, which is essential for OLS, is violated. Due to the small number of cases, it is not possible to estimate a model that is appropriate for count data with a large number of zero observations (such as a zero-inflated Poisson model). We make do with a transformation of the dependent variable: first, we project the data range to the unit interval, by dividing the variable by 30 (the largest value). Then, we take the log-odds ratio of this variable, and essentially estimate a logistic regression with OLS. The coefficients will be difficult to interpret substantively. However, since we are mainly interested in the directionality of the coefficients, this does not matter.





Table 3 summarizes the results. The number of groups in a policy domain is strongly related to the number of fora – domains with more groups tend to have more fora, too. This conforms to our expectation. What does not conform to our expectation is that the number of groups is no less important in distributive policy areas. In fact, the coefficient of the interaction variable is positive, not negative, and it is not significant. In fact, no variable besides the number of interest groups is significant.





Since there is no significant interaction effect between the type of policy area (distributive/regulatory) and the number of groups, we exclude this variable from the analysis. Table 4 contains the resulting estimates. We find that the interaction term had masked the impact of the distinction between regulatory and distributive policy domains. If we exclude the interaction, we find that distributive policy domains have significantly more fora than regulatory policy domains. This is a significant but unexpected result. We predicted that we would find more fora in regulatory policy domains, which rely more on expertise and less on input legitimacy, than in distributive policy domains. The actual results contradict our expectations.


We can only speculate about the reasons for this relationship. One possibility is that regulatory policy domains are less prone to suffer from access overflow. They may be highly complex, specialized policy domains, in which a large number of interest groups reflects the complexity of the subject matter. Also, it is possible that in these policy domains we may find informal ways to distinguish insiders from outsiders. For example, many groups which, in the Conneccs database, claim to be active in a regulatory policy domain may in fact not participate in policy-making processes. In distributive policy domains, on the other hand, we may find that all groups that are interested in the policy are in fact lobbying the Commission, thereby creating access overflow. The reason for this increased willingness to lobby may be the greater technical simplicity of the subject matter under debate. These considerations are merely speculative. However, they point to research questions that might be profitably pursued in the future.


Overall, the results provide mixed support for our hypotheses. The main message is that there is a relationship between three factors: the number of groups in a policy domain, the number of fora in a policy domain, and the question of whether the domain deals with regulatory or distributive policy. More groups are associated with the number of fora, as our theoretical discussion predicted. However, in regulatory policy domains we find more groups but fewer fora, exactly contrary to our expectations.


Before we place too much weight on these results, we should offer a note of caution. We are dealing with a system of equations in which the dependent variables in both equations are endogenous: the dependent variable of one equation is an independent variable in the other equation, and vice versa. It can be shown that the error terms of such "non-recursive" systems of equations are correlated with each other, and with the independent variables, leading to potentially biased coefficient estimates (Achen 1986).





To what extent do our results support or contradict our hypotheses? Overall, there is evidence for two of our main contentions: first, that Commission activity influences group activity and, second, that group activity leads to the creation of fora for interest representation. To start with the second argument, our results demonstrate a strong relationship between the number of groups and the number of fora – the more groups, the more fora. This conforms to our argument that the creation of lobbying insiders is a reaction to lobbying overload. In the other direction, we find that distributive policy domains have fewer groups than regulatory domains. Since the European Union tends to be more active in regulatory policy domains, this relationship points to the supply of policy benefits as an incentive for group activity. In addition, new policy domains exhibit a larger group presence than older domains. This may partly be due to the fact that, in new policy domains, governments tend to engage in the production of new policies that attract attention by societal actors.


One of our more puzzling findings is the fact that distributive policy domains tend to have more fora than regulatory domains. Since distributive policy domains tend to have fewer groups, our lobbying-overload argument cannot quite capture this relationship. One possible explanation is that the relationship indicates that Commission fora can perform roles that our theoretical arguments do not capture. In distributive policy domains, for example, fora may not be used to generate expertise but to assure consultation with all stakeholders that may be affected by a policy – to generate input legitimacy, to use Scharpf's terminology (1999).


Another interesting – but methodologically worrisome – finding is the fact that the number of fora not only seems to be influenced by the number of groups, but also seems to influence the number of groups in a policy domain. On the one hand, this confirms our argument that government activism leads to group activism. Also, we can view fora as an additional source of private benefits supplied to lobbyists, thereby making their activities more profitable. However, methodologically, the apparent two-way relationship between groups and fora indicates a problem of mutual causation, which has been shown to lead to biased parameter and error estimates. Due to the small number of cases, cross-sectional solutions to this problem – such as the use of instrumental variables – are not viable.


Even though we cannot solve the methodological problem, we can use it to draw substantive implications. Possibly, the statistical relationship between the group and fora variables points to a substantive bi-directional relationship. On the one hand, there is the relationship discussed in the theoretical arguments presented in this paper: large numbers of lobbyists lead to uninformative lobbying signals, and the Commission reacts with the selection of lobbying insiders. On the other hand, the distinction between lobbying insiders and outsiders may create costs for lobbying outsiders, who may now be induced to invest resources to convince the Commission to select them as insiders, too. In other words, the creation of fora pump-primes group activism. This process may be reflected in the Conneccs data, as groups who in the past have been marginally involved react to the creation of fora with, among other things, creating entries in the database.


Substantively, then, the mutual relationship between groups and fora points to a dynamic process that cannot be captured by a cross-sectional sample. As a result, one of our main methodological insights is a call for studies that investigate the dynamic nature of European Union lobbying. In addition, we show that the study of group mobilization in the European Union should not be conducted without the simultaneous investigation of policy making that affects the studied groups, and vice versa.





The main theoretical concern of this study has been a micro-level foundation of macro-level characteristics of lobbying systems. Even though the empirical results that we present are preliminary and so far lack the necessary detail, they provide modest support for our arguments. The overall pattern that emerges is that factors associated with the level of European Commission policy making are related to the number of groups in a policy domain: there are more groups in new and in regulatory policy domains; the presence of Commission fora for interest representation also constitutes a predictor of the size of the lobbying population. In addition, one of our main theoretical arguments – that lobbying fora are a reaction to large numbers of lobbyists – seems to be supported by the fact that the policy domains with large numbers of interest groups also have more fora.


Besides providing initial support for our arguments, the results of our study indicate the direction future research has to take. In particular, the unclear directionality of the relationship between the number of interest groups in a policy domain and the number of fora indicates that the dynamic nature of lobbying has to be taken into account in the future. This contention receives support from the literature on lobbying in the United States, which has pointed to the presence of policy-making and lobbying cycles, in which interest group pressure, policy-making and interest group reaction to policy change alternate (see, for example, Vogel 1989; McFarland 1991). In addition, a dynamic perspective on European Union lobbying also requires an extended theoretical focus that takes account of the interaction between interest group pressure, the resulting institutional structuring of the interaction between EU and lobbyists, and the resulting incentive changes for lobbyists. Recent initiatives to increase the transparency of EU lobbying serve as a reminder that institutional change in interest representation is a continuing presence in European Union politics.




Data Appendix









Ainsworth, Scott. 1993. Regulating Lobbyists and Interest Group Influence. Journal of Politics 55 (1):41-56.

Austen-Smith, David. 1995. Campaign Contributions and Access. American Political Science Review 89:566-581.

Austen-Smith, David, and John R. Wright. 1992. Competitive Lobbying for Legislators' Votes. Social Choice and Welfare 9:229-257.

Bouwen, Pieter. 2002. Corporate Lobbying in the European Union: The Logic of Access. Journal of European Public Policy 9 (3):230-254.

Broscheid, Andreas, and David Coen. 2003. Insider and Outsider Lobbying of the European Commission. An Informational Model of Forum Politics. European Union Politics 4 (2):165-189.

Coen, David. 1997. The Evolution of the Large Firm as a Political Actor in the European Union. Journal of European Public Policy 4 (1):91-108.

Coen, David. 1998. The European Business Interest and the Nation State: Large-firm Lobbying in the European Union and Member States. Journal of Public Policy 18 (1):75-100.

Crawford, V., and Joel Sobel. 1982. Strategic information transmission. Econometrica 50:1431-1451.

Crombez, Christophe. 2002. Information, Lobbying and the Legislative Process in the European Union. European Union Politics 3 (1):7-32.

Falke, Josef. 1996. Comitology and Other Committees: A Preliminary Empirical Assessment. In Shaping European Law and Policy: The Role of Committees and Comitology in the Political Process, edited by R. H. Pedler and G. F. Schaefer. Maastricht: European Institute of Public Administration.

Freeman, J.L. 1965. The Political Process: Executive Bureau-Legislative Committee Relations. New York: Random House.

Gilligan, Thomas W., and Keith Krehbiel. 1987. Collective Decisionmaking and Standing Committees: An Informational Rationale for Restrictive Amendment Procedures. Journal of Law, Economics, and Organization 3 (2):287-334.

Gilligan, Thomas W., and Keith Krehbiel. 1989. Asymmetric Information and Legislative Rules with a Heterogeneous Committee. American Journal of Political Science 33 (2):459-490.
Grossman, Gene M., and Elhanan Helpman. 2001. Special Interest Politics. Cambridge, MA: The MIT Press.

Heclo, Hugh. 1978. Issue Networks and the Executive Establishment. In The New American Political System, edited by Anthony King. Washington, DC: American Enterprise Institute.

Heinz, John P., Edward O. Laumann, Robert H. Salisbury, and Robert L. Nelson. 1990. Inner circles or hollow cores? Elite networks in national policy systems. Journal of Politics 52 (2):356-390.

Kreps, David M., and Robert Wilson. 1982. Sequential Equilibria. Econometrica 50:863-894.

Lohmann, Susanne. 1995. Information, access, and contributions: a signaling model of lobbying. Public Choice 85:267-284.

Lupia, Arthur, and Mathew D. McCubbins. 1997. Knowledge, Information, Rationality and Choice. Cambridge etc.: Cambridge University Press.

Mahoney, Christine. 2004. The power of institutions. State and interest group activity in the European Union. European Union Politics 5 (4):441-466.

Maloney, William A., Grant Jordan, and Andrew M. McLaughlin. 1994. Journal of Public Policy 14 (1):17-38.

Marsh, David, and R.A.W. Rhodes (editors). 1992. Policy Networks in British Government. Oxford: Clarendon Press.

McConnell, Grant. 1966. Private Power and American Democracy. New York: Knopf.
McCool, Daniel. 1990. Subgovernments: Determinants of Political Viability. Political Science Quarterly 105: 269-93.

McFarland, Andrew S. 1991. Interest Groups and Political Time: Cycles in America. British Journal of Political Science 21:257-284.

McLaughlin, Andrew, and Grant Jordan. 1993. The Rationality of Lobbying in Europe: Why are Euro-Groups so Numerous and so Weak? Some Evidence from the Car Industry. In Lobbying in the European Community, edited by S. Mazey and J. Richardson. Oxford: Oxford University Press.

Olson, Mancur. 1965. The Logic of Collective Action. Cambridge: Harvard University Press.
Pedler, Robin H., and Guenther F. Schaefer (editors). 1996. Shaping European Law and policy: The Role of Committees and Comitology in the Political Process. Maastricht: European Institute of Public Administration.

Potters, Jan, and Frans van Winden. 1990. Modelling Political Pressure as Transmission of Information. European Journal of Political Economy 6:61-88.

Sabatier, Paul A., and Hank C. Jenkins-Smith (editors). 1993. Policy Change and Learning. An Advocacy Coalition Approach. Boulder: Westview Press.

Sargent, Jane A. 1993. The Corporate Benefits of Lobbying: The British Case and its Relevance to the European Community. In Lobbying in the European Community, edited by S. Mazey and J. Richardson. Oxford: Oxford University Press.

Scharpf, Fritz W. 1999. Legitimacy in the Multi-Actor European Polity. In Organizing Political Institutions. Essays for Johan P. Olsen, edited by M. Egeberg and P. Laegreid. Oslo: Scandinavian University Press.

Schattschneider, Elmer E. 1960. The Semisovereign People. A Realist's View of Democracy in America. New York: Holt, Rinehart & Winston.

Streeck, Wolfgang, and Philippe C. Schmitter. 1991. From national Corporatism to Transnational Pluralism: Organizing Interests in the Single European Market. Politics and Society 19 (2):133-164.

Truman, David B. 1951. The Governmental Process. Political interests and public opinion. New York: Knopf.

van Schendelen, M.P.C.M. 1996. EC Committees: Influence Counts More Than Legal Powers. In Shaping European Law and Policy: The Role of Committees and Comitology in the Political Process, edited by R. H. Pedler and G. F. Schaefer. Maastricht: European Institue of Public Administration.

Wilson, Graham K. 1990. Business and Politics. A Comparative Introduction. 2nd Edition. Chatham, NJ: Chatham House.







We would like to thank Pieter Bouwen, John Constantelos, Fabio Franchino and Christine Ingebritsen for their comments on earlier versions of this paper, and Jürgen Feick, Jörg Teuber and Cornelia Woll for their constructive criticism that has made this working paper possible. Nevertheless, all errors and omissions are the fault of the authors.



We provide a formalization of our theory in Broscheid and Coen (2003); here, we present an intuitive summary of our argument.



Following conventions common in game theory, we denote the first mover in an interaction with the female pronoun. As a result, lobbyists are female in our presentation, and politicians are male.



Lowi and Wilson distinguish between distributive and redistributive policies. As both types of policies deal with the distribution of material values, it is not necessary for our purposes to distinguish between them.



These results are not reported here, but the authors will provide the results upon request.
Specifically, we inspected the hat matrix, dfits and dfbetas. For an explanation of these indicators, see Bollen and Jackman (1985).



An alternative reaction to access overload is the consolidation of interest representatives into larger organizations. We do not pursue this possibility in the present study.



Copyright © 2006 Andreas Broscheid, David Coen

No part of this publication may be reproduced or transmitted without permission in writing from the author.

Jegliche Vervielfältigung und Verbreitung, auch auszugsweise, bedarf der Zustimmung des Autors.

MPI für Gesellschaftsforschung, Paulstr. 3, 50676 Köln, Germany





MPIfG: MPIfG Working Paper 06/3

[Zuletzt geändert am 29.03.2007 10:59]