THE TWO FACES OF INTEREST GROUP INFLUENCE AND THE DIRECTIVE BANNING TOBACCO ADVERTISEMENT

Transcription

1 THE TWO FACES OF INTEREST GROUP INFLUENCE AND THE DIRECTIVE BANNING TOBACCO ADVERTISEMENT Introduction Interest group influence has become what the Higgs boson represented for physics until very recently: something which is crucial for the discipline, about whose existence researchers are sure but which they do not know how to detect. Although the study of power is one of the core tenets of social sciences attention on this topic has recently revamped in the interest representation literature: the recent increase in manuscripts, special issues in academic journals, workshops and conferences dedicated to this issue is telling in this respect. Nonetheless, scholars have hitherto struggled in progressing research in this field mainly due to epistemological, theoretical and methodological confusion. In other words, researchers do not agree on what influence is, on how can they measure it and with what. Different conceptualizations of interest group influence have been analysed with different approaches and the discussion in the literature has prematurely moved to the methodological level. In this vein, this paper brings epistemological and theoretical consciousness into the analysis of interest group influence as a precondition for a methodological debate. The literature has, indeed, slowly but ineluctably leaned towards preference attainment as the main theoretical approach to measure influence in the past years neglecting though the epistemological and theoretical premises which underpin this approach. In discussing its premises, I argue that they can be relaxed in order to make it accountable for two types of power which have hitherto received separate treatment: the first face of power and the second face of power. In fact, the unspecified spatial and temporal conceptualization of the political space embraced by studies using this approach has hitherto significantly limited research. This in turn allows the literature to move the discussion to the methodological level and employ more sophisticated methods. The first section starts by introducing preference attainment as a theoretical approach along with its premises. The discussion proceeds to the presentation of the two faces of power and how they can be analysed from a preference attainment approach only if its theoretical premises are relaxed. The modifications to be made to preference attainment are discussed in the ensuing section: the temporal and spatial aspects of the political space represent the focus of the discussion. The following section concludes by introducing data reduction through scaling as the most appropriate methodology for preference attainment. The theoretically and methodologically rejuvenated preference attainment 1

2 approach is then applied to a case: the lobbying dynamics in the EU directive banning tobacco advertisement What is Preference Attainment and What is Good for? Preference attainment has become the mainstream theoretical approach to detect and measure interest group influence. Scholars address interest group influence by embracing preference attainment in several domains: US regulatory agencies (Yackee, 2004, Yackee and Yackee, 2006, Yackee, 2006), US legislature (Mahoney, 2007a, Baumgartner et al., 2009), UK (Bernhagen, 2012) and the EU (Mahoney, 2007a, Klüver, 2013). This approach has undoubtedly become mainstream in the last years and researchers have more and more focused on how to methodologically advance it rather than granting attention to its theoretical assumptions. Nonetheless, a discussion on the latter is logically prior and functional to a methodological one. In this vein, in this section I will present preference attainment along with its theoretical premises, namely the focus on issue as political space and how this affects the conceptualization of actors preferences and policy change. The ensuing section relates preference attainment to the two faces of power introduced by Bachrach and Baratz (1962) arguing that this approach can account for both if precise theoretical choices are made by the researcher. This term was first introduced in relation to interest group influence by Dür (2008b) as one of the main three methods to measure influence along with attributed influence and process tracing. Nevertheless, preference attainment cannot be considered a method, such as process tracing 1 which is a qualitative analytical tool used to study within-case variation (Gerring, 2004, Gerring, 2007) by focusing on causal mechanisms, counterevidence and the weighing of the evidence in the light of the theory chosen (Gerring, 2008, Vennesson, 2008, Hancké, 2009, Brady and Collier, 2010, Collier et al., 2010, Collier, 2011). Dür (2008b) uses as example of process tracing the analysis of tobacco industry lobbying on the EU directive banning tobacco advertisement conducted by Neuman et al. (2002) 2. By relying on a thorough analysis of hundreds of tobacco industry documents the authors demonstrate (with a good level of confidence, what Collier, 2010, would call a smoking gun test) the influence tobacco industry has exerted on the EU decision-making and its outcome. Although using process tracing as analytical method the authors address the issue from a preference attainment perspective. Indeed, the drawing of tobacco industry s position on the regulation of tobacco advertisement and sponsorship from various documents as well as the focus on how the 1 Brady and Collier (2010) and Collier et al. (2010) use also the term causal process observation. 2 For a longer and more detailed version of this work see Bitton et al. (2002). 2

3 Commission proposal moved throughout the process towards those preferences indicate the theoretical assumptions underpinned by preference attainment, which will be illustrated below 3. Attributed influence on its part does not arguably deserve a separate discussion as it is common practice in the literature (Dür, 2008b, Dür, 2012, Klüver, 2013). The latter conceives attributed influence as based on actors or experts perception on power relations. This does not refer to either a theoretical approach nor a method. Indeed, perceptions on power relations can underpin different theoretical assumptions on power: theoretical choices on how to conceive influence and how to measure it are simply moved from the researcher to experts or actors themselves. On top of that, this can be done through various methods spanning from qualitative ones, such as interviews and archival analysis, to quantitative ones, such as large-n survey analysis. Again, the categorization proposed by the literature (Dür, 2008b, Dür, 2012, Klüver, 2013) is not internally consistent and categories are not mutually exclusive. For instance, Neuman et al. (2002) process tracing heavily relies on attributed influence: Philip Morris claimed success in securing German opposition to the advertising directive on the basis of subsidiarity in a 1993 memo (Neuman et al., 2002 p.1325) and BASP also exposed the tobacco industry s success in circumventing existing EC labeling laws (Neuman et al., 2002 p.1328) 4. As demonstrated, preference attainment is not a method but a theoretical approach to interest group influence: it analyses whether and to what extent the final policy output moves towards societal actors preferences. In doing so, researcher conceives the political space as determined by individual issues and this bears also consequences on how they perceive actor s preferences and policy change. Studies of interest group influence which use preference attainment assume a space of political competition defined by the issue at stake. Accordingly, power as control over events/outcomes considers actors with different preferences on the event, which in turn bears different outcomes for different actors competing with each other with the aim of increasing their net utility (Hart, 1976). The process is as follows. First of all, the researcher identifies the political space which is usually represented by a piece of legislation (this is the event in Hart, 1976, terminology) which bears costs and benefits for actors: policy proposals issued by Westminster (Bernhagen, 2012) or the European Commission (Klüver, 2013) and US bureaucratic rules (Yackee, 2004, Yackee and Yackee, 2006, Yackee, 2006). Then, the researcher reduces the political space into dimensions which in turn are divided into factors. The number and substance of factors and 3 Panel 2 p in Neuman et al. (2002) shows a comparison between tobacco industry s preferences on the advertisement regulation and the compromise proposal introduced by Germany in the Council, which can be considered as the policy outcome. This panel provides an example of this work addresses interest group influence from a preference attainment perspective. 4 I t should be noticed that if conceived as a method attributed influence would be a method of data gathering not of data analysis in Schwartz-Shea and Yanow (2002) terminology, differently from process tracing, for instance. 3

4 dimensions as well as how to derive the latter from the former is the topic of the ensuing section. Actors preferences on those factors/dimensions are either a priori assumed (Dür, 2008a) or deducted from data through interviews (Mahoney, 2007a, Baumgartner et al., 2009) and various documents (Yackee, 2004, Yackee and Yackee, 2006, Yackee, 2006, Kluver, 2011, Klüver, 2013) and the same is done for the policy outputs before and after lobbying 5. Before proceeding to the discussion on how actors preferences and policy change are conceived it should be noticed that the fact that each issue represents a political space potentially secluded from the others makes findings difficultly generalizable, especially single case studies. Although the literature has tried to overcome this problem by using large N samples the selection of the cases is still a problem (Lowery, 2013): for instance, should researchers focus on a random sample as in Baumgartner et al. (2009) or on politically salient issues likely to create awareness in the society as in Bernhagen (2012)? Last but not least, even in those theoretically driven samples the issue of the comparability of cases is still open 6. Nonetheless, this is a congenital limitation of this theoretical approach which can be attenuated by conceiving the political space in a broad manner, as suggested below, but which cannot be overcome. The focus on issues as political spaces bears several implications on how researchers consider actors preferences and policy change. Interest group influence is conceived as the fulfillment of preferences as a result of a policy output. Actors preferences are (either explicitly or implicitly) coded with respect to the issue at stake and if the policy outcome moves towards an actor s preferences that actor is deemed to be influential. For instance, Bernhagen (2005) measures actors policy preferences with respect to whether they are against, concerned about, neutral to or in favour a policy proposal. Similarly, Yackee and Yackee (2006) measure actors preferences with respect to their agreement on the modifications made to US regulatory agencies rules. In doing so, researchers assume that what they measure, namely usually policy positions, represent a valid proxy for preferences and that those preferences are stable and constant throughout the political space. As seen above, several data gathering methods can be used to retrieve actors preferences but each of them measures positions vis-à-vis a policy proposal, not actors preferences themselves. Researchers assume also that those preferences (or better, policy positions) are constant and stable 5 Documents providing information for actors preferences and policy outcomes can in turn be analysed through various methods, such as (either manual or computer assisted) content analysis and discourse analysis. Each of those methods bears specific advantages and disadvantages especially with respect to validity and reliability which fall outside the scope of this work: see Neuendorf (2002) and Fairclough and Fairclough (2013), for instance. It should be noticed that the methodological aspects of this debate should be taken into more consideration by the literature on interest group influence, which has uncritically leaned towards the use of content analysis (Yackee, 2004, Kluver, 2011, Kluver, 2013). 6 Unfortunately, sometimes methodological choices drive the construction of the sample, as in Kluver (2013) and Yackee (2004, 2006). In fact, in those cases the use of content analysis (manual in latter and computer assisted in the former) constrains researchers on which cases to select. For instance, Kluver (2013) selects a non random sample: the sample includes only legislative proposals that were preceded by a non-standardized publicly accessible consultation (Kluver, 2011 p.493). 4

5 across the political space. Again, this is a inherent limitation of the preference attainment approach due to its focus on a single issue as the main political space and thus it cannot be overcome but the validity of these assumptions rests on the conceptualization of the political space, which will be discussed below. Preference attainment relies on another theoretical assumption: policy change. Indeed, the focus is on change in public policy: a policy proposal turned into a legislative act by Westminster (Bernhagen, 2012) or a change in content of the Commission proposal after the online consultations (Kluver, 2011, Klüver, 2013), for instance. This approach addresses interest group influence from a comparative statics perspective comparing the situation before and after policy change (Simon, 1953, Baldwin, 1971). In line with Simon (1953 p.504) preference attainment looks at unilateral coupling, namely to determine the influence of A upon B, we simply observe a number of situations in which the behavior of A varies, and note what is the concomitant variation in B's behavior. The variation in B s behavior refers to a variation in decision-makers behavior and, more specifically, to policy change. The assumption made by researchers is that policy change is determined by interest groups when it moves towards the latter preferences and not by luck, for instance, or other factors external to the political space (Barry, 1980). This thermostatic conceptualization of policy change (Howlett and Cashore, 2009) is a consequence on the definition of the political space in terms of individual issues and the considerations raised above hold also for this aspect. As seen, preference attainment is not a method but a theoretical approach which underpins clear premises. Yet, what is it good for? In line with what claimed by Kluver (2013 p. 62) that this approach offers several advantages [ ] it covers all channels of influence, preference attainment accounts for both direct and indirect lobbying (e.g. access and voice in Beyers, 2004, terms). Direct lobbying can take the form of exchange of campaign contributions, information and so on and so forth: everything which falls into the traditional categories of exchange and persuasion (Hall and Deardorff, 2006, Lowery, 2007). Indirect lobbying refers to an action on the part of interest groups whereby instead of directly tackling decision-makers they try to raise awareness in the media and public opinion, for instance. This type of power is rather subtle and usually it propels through what Simon (1953) calls anticipated reactions or feedback processes 7. The latter refers to the eagerness 7 The concept of anticipated reactions is common in congressional voting behavior theories (Denzau and Munger, 1986, Arnold, 1992, Verdier, 1995, Bailey, 2001). The main argument is that when latent support on an issue is present amongst voters and someone is eager to mobilize that support (Bailey, 2001) legislators will preemptively further those interests. It should be noticed how indirect lobbying has always been common practice in the traditional interest representation literature. Apart from those works explicitly focused on influence, the classics, both pluralists (Bentley, 1908, Truman, 1959) and the Olsonian tradition (Olson, 1965, Olson 1982), have always perceived the action of interest groups as a side effect of external shocks or their membership dynamics. Conversely, those works more focus on decision-makers, namely the public choice strand, both in its private (Peltzman, 1976, 5

6 of decision-makers to further an interest with the perspective of future gains (in electoral terms, for instance). In conclusion, preference attainment is suitable for any form of lobbying in that it purposely black-boxes the dynamics of interest representation being exclusively focused on actors preferences and policy change. Finally, although the researcher should pay attention to what Lowery (2013) terms the other guy did it null hypothesis (Lowery, 2013 p.10), namely the difficulty of attributing success to individual interest groups in an advocacy coalition, this is also called metering problem in lobbying coalitions (Gray and Lowery, 1998), preference attainment can account for lobbying in coalitions. More generally, just like the researcher can affirm that the interest group A has been more influential than the interest group B since the A s preferences have resulted to be closer to the policy outcome than B s ones it can also test whether and to what extent A, B and C have been more successful than D, E and F. Indeed, lobbying through advocacy coalitions have been the focus of several works using a preference attainment perspective (Mahoney, 2007b, Mahoney, 2007a, Klüver, 2010, Klüver, 2012, Klüver, 2013). It should be noticed that the theoretical and methodological premises hitherto embraced by researchers and thoroughly discussed below bear issues on the analysis of lobbying coalitions. Not only do the theoretical and methodological novelties proposed in this work allow the researcher to account for both the first and the second face of power but they also provided more valid tools to analyse lobbying coalitions. Two Faces, One Approach In the literature two main conceptualizations of power 8 may be individuated: a behaviorist and a structural one. These two may in turn be traced back to two separate traditions in the American social sciences in the aftermath of WWII: the pluralist and the structural tradition. Although being typical of different disciplines, being the latter more sociological and the former more related to political science, these two strands have shaped contemporary perceptions on power and in turn the debate on interest group influence. This section introduces those two types of power arguing that they both may be analysed from a preference attainment perspective if precise theoretical premises are embraced, which will be discussed in the following section. Pluralists conceptualization of influence may be legitimately termed as power (Macht) insofar it is conceived in comparison with the term rule (Herrschaft). The latter is nothing but a species of the genus power, namely it is the institutionalized form of power (Popitz, 2001). Power, as employed Peltzman, 1984, Posner, 1974, Stigler, 1971) and public interest claim (Mitnick, 1980, Levine and Forrence, 1990) versions, have embraced an indirect type of influence similar to the one discussed in this section. 8 Although I acknowledge that traditional sociology perceives power and influence differently in this work I used them interchangeably. 6

7 by pluralists in the Weberian meaning, refers to the possibility to impose an actor s will in a social relationship (Dahl, 1957, Popitz, 2001, Swedberg and Agevall, 2005) with the aim of controlling another actor or an outcome (Hart, 1976) 9. Two elements are noteworthy for the discussion developed in this study. Firstly, power is always something potential or latent, namely it hides a base. According to the base on which power relies, the latter may assume different forms. For instance, in the EU interest groups can employ different resources: from information provision (Chalmers, 2011) through support to an institution vis-à-vis others (Woll, 2007, Woll, 2009) to popular support (Caporaso and Tarrow, 2009). The second and more important connotation of power in its behaviorist conception is its omnipresence, as argued by Popitz (2001) with the result of the disembeddedness of power from the context. As Polsby (1960 p. 475) notices, nothing categorical can be assumed about power in any community [...] if anything, there seems to be an unspoken notion among pluralist researchers that at bottom nobody dominates in a town. Conversely, elitists concentrate on what they term power structures (Hunter, 1953), which may be associated to the concept of rule. The latter, as mentioned above, is the institutionalized form of power: the result of the processes of depersonalization, formalization and integration (Popitz, 2001). In other words, where power relationships are no longer based on the individuals, but on the role they perform (i.e. the value positions in Simon, 1953), where they are bound by norms, rituals and procedures 10 and where they are embedded in the social structure which, in turn, bolsters them, they may be conceived as rule. Worthy of note for this discussion is the process of integration. In fact, power relationships come to be embedded in the social and political system where they are at work deriving also legitimacy from them. The result is straightforward: power structures are perceived as stable over time and highly centralized, as elitists (Hunter, 1953, Schulze, 1958) assume in their works. Bachrach and Baratz (1962) provide a Janus representation of power discussing its two faces 11. The first face of power can arguably be associated with the behaviorist conceptualization of power introduced above whilst the second face to the structural one. The former accounts for the phenomenology of power as occurs in opens negotiations between actors during formal decisionmaking when the political agenda is already set. In other words, it is more focused on the action and conflict side of power (Woll, 2007). The second face refers to a structural conception of power 9 Hart (1976) argues that power in IR theories can be conceived in a threefold manner: as control over resources, as control over actors and as control over events/outcomes. Nonetheless, in his discussion Hart (1976) himself notices that often one type of control is functional to another. 10 In the quest for the political equivalent of money Baldwin (1971) identifies the set of symbols associated with political power. 11 Although acknowledging that other faces of power have been identified in the last decades (Digeser, 1992, Lukes, 2005, Giddens, 2013) this work aims to provide the literature with an analytically feasible theoretical approach to what is beyond the strictly behaviorist approach to influence. In this vein, limiting the analysis to the second face of power is reasonable. 7

8 which is deeply embedded in the institutional arrangements and in the predominant social and political values. This type of power grants particular attention to institutions, claiming that all forms of political organization have a bias in favour of the exploitation of some kinds of conflict and the suppression of others because organization is the mobilitisation of bias (Schattschneider, 1975 p.71). As emphasised above, the study of power must start from the analysis of dominant values, social norms and institutional procedures: it focuses on the rules of the game which in turns determine the dimensions on which the political discussion is based on. Those two faces of power have hitherto been analysed with different approaches. The literature on interest group influence, especially those works using a preference attainment approach, has hitherto focused on the first face of power. Conversely, more structuralist/institutionalist approaches have tried to address the second face of power, explicitly (Lowery, 2007, Woll, 2007, Woll, 2009) or implicitly (Bouwen, 2002, Bouwen, 2004, Mahoney, 2004, Eising, 2007). This work argues that albeit different causal mechanisms are in place both faces of power may be analysed from a preference attainment perspective if precise theoretical choices are made by the researcher in terms of the political space under analysis. As in the words of the founders of the two faces of power (Bachrach and Baratz, 1962 p.948), the two faces of power have rather different phenomenology but they both can be accounted for by a preference attainment perspective: Of course power is exercised when A participates in the making of decisions that affect B. But power is also exercised when A devotes his energies to creating or reinforcing social and political values and institutional practices that limit the scope of the political process to public consideration of only those issues which are comparatively innocuous to A. To the extent that A succeeds in doing this, B is prevented, for all practical purposes, from bringing to the fore any issues that might in their resolution be seriously detrimental to A's set of preferences Preference Attainment and the Political Space Preference attainment can arguably be used to analyse both types of power. This section introduces the theoretical choices on the conceptualization of the political space the researcher has to make in order to do so. As seen above, one of the main characteristics of the preference attainment approach is its focus on issues (events in Hart, 1974, terminology) as units of analysis. An issue based approach to power can be associated to both types of power thanks to a broader specification of the political space. Indeed, researchers using this approach make theoretical choices on the political space which are not theoretically driven but rather due to methodological paucity. This section 8

9 overviews such theoretical choices making them explicit and, then, the following section introduces multidimensional scaling as the optimal methodology to account for the more valid conceptualization of the political space introduced in this work. This methodology allows to bridge the gap between the two faces of power. The final section illustrates an application of this methodology to a highly discussed case of lobbying, namely the EU directive banning tobacco advertisement. The Temporal Aspect and the Policy Cycle These two sections discuss two crucial limitations preference attainment has hitherto self imposed: a relaxation of these assumptions is essential to improve the validity of the analysis of interest group influence in its two faces. The first limitation concerns the temporal aspect of the political space and the second one the spatial aspect. The literature using preference attainment has hitherto addressed the political space in an uncritical manner. In Klüver (2009), Kluver (2011), Klüver (2013) as well as in Yackee (2004), Yackee and Yackee (2006), Yackee (2006) the temporality of the political space is extremely limited: in the former it starts and it ends with the Commission proposal before and after the online consultations and in the latter the political space is limited to the rules promulgated by various US federal agencies and how they change according to the notice and comment rulemaking. In Bernhagen (2012) the time span underlying the political space is wider but still unspecified and the same applies to those works using a preference attainment approach in single case study analysis, such as Bitton et al. (2002), Neuman et al. (2002). I argue that a wider conceptualization is needed for preference attainment to account for both faces of power. In line with works on political competition (Thomson, 2011) I use the concept of policy episode, which is a slice of history, centered on a policy issue s career as affected by agendasetting, alternative specification, and decision making (Barzelay and Gallego, 2010 p. 211). As argue above, preference attainment is theoretically focused on policy change. Nevertheless, two aspects have been neglected by the literature: when the policy change occurs and the alternative specification. The two concepts of power introduced above can be reconciled in a preference attainment approach if the latter conceives a political space as composed by an agenda setting stage and a decision-making one. The first face of power is theoretically more appropriate for the latter whilst the second face for the former. Indeed, the behavioral conceptualization of power may arguably be associated with the decision-making stage where interest groups compete against each others to move the decision making outcome towards their preferences. Conversely, the second face 9

10 of power is more appropriate for the agenda setting stage where directly as well as indirectly interest groups determine the political agenda 12. In line with the literature on spatial decision-making and the critics of the preference attainment approach, the investigation of the non-decision making should be central in the analysis of agenda setting: what Lowery (2013) terms influence as a non-event. The analysis of interest group influence should also take into consideration the reference point, namely the negative policy change: what happens if no policy change occurs. The reference point is defined by the literature on spatial decision-making as the decision outcome that would occur if the legislative proposal were not adopted: in other words, the decision outcome in the event of failure to agree (Thomson, 2011 p.41). This is theoretically and (sometimes) empirically different from the status quo ante. As for the empirical difference, the reference point does not always coincide with the status quo ante, especially for distributive and redistributive issues. If a proposal sets out higher regulatory standards on environmental protection at EU level, as in the case of Figure 1 13, and that proposal is not enacted the reference point likely coincides with the status quo ante, namely the situation previous to the proposal where lower regulatory standards were in place at EU level. Yet, if a proposal setting out a new budgetary allocation fails to be implemented it may be the case that a provisional rule enters into place: this is the case of the EU budget, for instance. In this instance, the budget allocation rules before the proposal (the status quo ante) and the ones in place in case of failure of the proposal (the reference point) are fundamentally different. Yet this does not only apply to budget allocation. Use again as example Figure 1 and now assume that a firm is generally against environmental protection because it negatively affects its economic interests but it is favorable to the regulation of environmental protection at EU level in that regulation at MS level (what the result of a failure to regulate at EU level could be, namely the reference point) would be worse for its interests being stricter in the firm s MS or leading to distortion in competition among MSs. In this example two conclusions may be drawn: as seen above, more than one dimension is usually needed to represent the political space in a valid way and the reference point does not always coincide with the status quo ante. 12 The public choice theory in its variants has always focused on agenda setting: from direct ways of capture such as bribes, revolving doors (Dal Bo, 2006), career benefits (Dunleavy, 1991, Wilson, 1980) and budget allocation (Niskanen, 1971), to more indirect one such as the exchange of information (Carpenter, 2004), campaign contributions (Denzau and Munger, 1986, Grossman and Helpman, 1996), votes (Lohmann, 2003, Lohmann, 1998) and shared values (Levine and Forrence, 1990). 13 Figure 1 illustrates the analysis Klüver (2009) conducts on the online consultations carried out by the Commission for the CO2 emission directive. The points in the plot represent the preferences of the interest groups involved and they are placed along a continuum pro- v. anti-environment protection. Furthermore, also the Commission s positions both before and after the online consultations is placed therein. 10

11 The reference point is also theoretically different from the status quo ante and arguably more suitable for the first face of power. Indeed, the reference point represents a hypothetical outcome which bears hypothetical actors preferences which are theoretically comparable with the actual ones. The dimensions as determined by the issue at stake are the same for both the actual outcome and the reference point. When the proposal is on the table interest groups can also lobby for a nondecision in the same way they can lobby for a particular decision. In this vein, the reference point is conceptually more suitable for the first face of power. Contrariwise, the status quo ante is theoretically more appropriate for the second face of power and its focus on agenda setting. In fact, the difference between the status quo ante and the proposal represents the political agenda with respect to that particular issue: some aspects of the issue are left out and others are inserted into the agenda as a reaction to the interest group influence (in its second face version). The Spatial Aspect and Dimensionality Another theoretical assumption which the literature using preference attainment has hitherto embraced but which should be relaxed in order to increase the validity of the analysis of interest groups influence in its two faces is unidimensionality. The strength of the spatial assumption varies greatly among studies moving from those considering space in what Benoit and Laver (2012) call an orientational manner to explicitly spatial models of political competition. Nonetheless, even in the latter the political space is considered as one-dimensional, being more or less regulation (Yackee, 2004, Yackee and Yackee, 2006, Yackee, 2006, Kluver, 2011, Klüver, 2013) or the degree of approval on a proposal (Bernhagen, 2012). Political science, especially those strands focusing on party politics and voting behavior, has slowly but ineluctably come to the conclusion that it is very common to need more than one dimension to describe key political differences (Benoit and Laver, 2012 p.195). The multi-dimensional spatial conception of political competition dates back to Lipset and Rokkan (1967) and it has become standard practice both in the US with the works on (roll call) voting behavior in Congress (Poole et al., 1990, Poole and Rosenthal, 1997, Poole, 2005, McCarty et al., 2008) and in Europe both at national level with the Comparative Manifesto Project (Gabel and Huber, 2000, Budge, 2001, Gabel and Hix, 2002, Klingemann, 2006) and at EU level with the project Resolving Controversy in the EU (Thomson, 2011, Thomson et al., 2012) and the study of European Parliament party system (Hix et al., 2006, Bakker et al., 2012). The literature on interest group influence should follow these steps. Take as example Figure 1 again: it should be noticed that the vertical axis is for illustration purposes and it has not substantive meaning. The political space is conceived as composed by a single dimension: EU regulation on environmental protection. Although being parsimonious the 11

12 one-dimensional representation of political competition conceals crucial information 14. In fact, in this case the dimension identified by the researcher can arguably refer to two potentially distinct dimensions: regulation at EU level and regulation on environmental protection. These two dimensions have already been identified by the literature on European party space as pro-european integration v. anti-european integration and social left v. social right (Hooghe et al., 2002, Bakker et al., 2012). In other words, the former refers to whether regulation should be enacted at EU level or at Member State level and the latter to whether regulation should weigh more materialist or postmaterialist values, namely whether environment protection should be considered more than economic interests, such as potential losses in competition and thus income on the part of some industries and potential unemployment. If agreement prevails in the literature that those two dimensions are crucial in the European party space as well as in the European public sphere (Stimson et al., 2012) why should this not be the case for interest groups? Two approaches to dimensionality may be found. The methodological approach refers to the number and the substance of the dimensions representing a political space, which can be identified by following two procedures: a priori and ex post (Benoit and Laver, 2012, De Vries and Marks, 2012). The theoretical approach refers to the source of dimensionality, which can be twofold, as emphasised by De Vries and Marks (2012): top down and bottom up. The discussion in this section follows a logical order illustrating thus firstly the theoretical approach and then the methodological one in the ensuing section. The literature on party politics can be divided into two theoretical approaches to dimensionality may be found: a sociological bottom-up and a strategic top-down (De Vries and Marks, 2012). The former identifies the society as source of dimensionality, as in the theory of cleavages (Lipset and Rokkan, 1967), and the latter the political system, as in the contemporary approached to party politics, for instance the Comparative Manifesto Project (Gabel and Huber, 2000, Budge, 2001, Gabel and Hix, 2002, Klingemann, 2006). If applied to interest group influence in its first face of power dimensionality assumes a top-down character: the issue at stake determines the dimensions of the political space. More specifically, public policy determines the political space in an exogenously manner. Conversely, the second face of power has bottom up conceptualization of the political space: interest groups determine the political agenda thus determining the dimensions of the political space. In this case dimensionality is endogenously driven. I label these two approaches to dimensionality as respectively homeostatic and thermostatic, close in spirit to the literature on policy change (Howlett and Cashore, 2009). 14 Benoit and Laver (2012) illustrate this loss of information when the researcher uses a one-dimensional space with the map of the Circle line in the London Tube. 12

13 Nonetheless, these two different conceptualizations of dimensionality can be reconciled thanks to the temporally broad concept political space comprising both agenda setting and decision making, as advocated above. The one-dimensional assumption bears also consequences on how actors preferences are assumed. The focus on preferences is central in this approach, as seen above. Nonetheless, interest groups preferences vis-à-vis a piece of legislation are assumed to be expressed in absolute and not in relative terms as a consequence of unidimensionality (Weisberg, 1974). As seen above for the example in Figure 1, interest groups preferences are assumed to refer to EU regulation on environmental protection without taking into consideration that an interest group may have different preferences on EU regulation and environmental protection as different dimensions and that maybe preferences on those dimensions are related, as demonstrated above. As a result, actors utility function comes to be conceived as single peaked and convex (Dür, 2012). Preferences are, indeed, measured with respect to an ideal point the dimension and the closer the policy outcome is to the ideal point the higher is the utility gain. Nonetheless, as discussed above, it is likely that the political space requires two dimensions to be validly described and in that case actors preferences must be considered as relative and not absolute 15. Two Faces, One Approach, One Methodology This section introduces data reduction as the main ex post methodology to dimensionality. It arguably allows the researcher to identify and justify dimensions in terms of consistency and exhaustiveness. Then, this section argues that multidimensional scaling as compared to factors analysis represents the optimal method to be used in a preference attainment approach being the former more able to account for dimensionality. This works concludes by applying the theoretical and methodological innovations in preference attainment to a case study, namely how to detect and measure interest group influence in the directive banning tobacco advertisement. As mentioned above, two methodological approaches to dimensionality may be identified: a priori and ex post (Benoit and Laver, 2012, De Vries and Marks, 2012). Although both are ultimately theoretically driven (De Vries and Marks, 2012), namely that some previous knowledge on the political space is always indispensable (Benoit and Laver, 2012), they underpin different conceptualizations of dimensionality, different research designs and often different results. The 15 It should be noticed how in political science different assumptions on actors policy preferences may be found. For instance, in two different works dealing with interest group influence legislators preferences (not interest group) are assumed to be cumulative and single-caved (Denzau and Munger, 1986, Hall and Deardoff, 2006). Although it is not unreasonable to conceive also interest groups preferences by assuming a minimum level of regulation they would accept and thus assuming that all those superior levels would fall within their preferences this transcends the rationale of the preference attainment approach. 13

14 literature on interest group influence has hitherto addressed dimensionality in a theoretically uncritical manner. Those works using small-n studies, such as process tracing to detect causality in single case study, have used spatiality in a mere metaphorical manner. As for large N studies, the methodological approach used has so far been an a priori one following methodological needs rather than theoretical guidance. Although dimensions such as more v. less regulation are theoretically justified they have been more the result of the need for reliability for the comparability of cases and methodological paucity than driven by theoretical considerations. Indeed, research has hitherto followed this process. The researcher identifies a dimension which is as externally valid as possible, for instance more regulation v. less regulation or the level of support for a proposal, and, then, he or she either implicitly (Bernhagen, 2012) or explicitly (Yackee, 2004, Yackee and Yackee, 2006, Yackee, 2006, Kluver, 2011, Klüver, 2013) chops that dimension into factors. Take as example Figure 1 again. By relying on the process used in the Comparative Manifesto Project Klüver (2009) identifies 41 factors constituting the dimension pro v. anti environmental control. Then, she codes the online consultation documents sent by various interest groups and the Commission proposals according to those factors thus being able to place each actor s preferences as well as policy outputs onto that dimension. An a priori methodological approach to dimensionality bears a major issues: the justification of dimensions in terms of internal consistency and exhaustiveness (Benoit and Laver, 2012). In other words, by using an a priori methodology the researcher cannot know whether the dimensions selected are exhaustive 16. If handled with care (especially in the interpretation of dimensions) the ex post methodological approach to dimensionality can solve this issue. Data reduction analysis is the main approach to find dimensions of a political space ex post (Kruskal, 1964, Weisberg, 1974, Rabinowitz, 1975, Benoit and Laver, 2012). This methodology allows the researcher to estimate the number of dimensions by conceiving dimensions as clusters of factors on which actor s preferences are inter-correlated (Benoit and Laver, 2012). In this vein, it provides two fundamental types of information: whether a dimension is internally consistent and whether the number of dimensions is exhaustive. It should be noticed how these issues are separate. Indeed, as argued by Weisberg (1964) an apparently multi-dimensional political space can be in reality one-dimensional if the single dimension is exhaustive. Conversely, there are political spaces which necessarily need a multidimensional representation. 16 It should be noticed that by following this methodology although the researcher cannot know whether dimensions are exhaustive the researcher can know whether they are internally consistent. Nonetheless, the process whereby the researcher can know whether dimensions are internally consistent is rather inefficient: he or she can measure the internal consistency of dimensions by investigating if data for the factors are correlated between each other (Benoit and Laver, 2012) but this would require an iterated modification of the number and substance of factors before and after the data gathering process until the dimensions are consistent. 14

15 Two main methods of scaling are identified by the literature: multidimensional scaling (MDS) and factor analysis (FA). Albeit both methods have been used to spatially represent data, multidimensional scaling (MDS) is arguably the most appropriate for preference attainment. Indeed, as suggested by Benoit and Laver (2012 p.213) for a distance matrix of unknown dimensionality, MDS is the most common inductive method for mapping the positions of the parties and the same arguably holds true for interest groups. Indeed, after the works on voting behavior in Congress have brought MDS into political science (Poole et al., 1990, Poole and Rosenthal, 1997, Poole, 2005, McCarty et al., 2008) it has started to be used in several domains, for instance the analysis of coalitions among (Wessels, 2004, Wessels, 2005, Hausermann, 2006, Thomson, 2011) 17. Furthermore, compared to FA MDS is better able to account for dimensionality and, more specifically, to discriminate between one-dimensionality and multi dimensionality (Brazill and Grofman, 2002). Indeed, in MDS correlated data on factors result in proximity in the space: a dimension is identified when the distance between actors preferences can be ordered onto it (Kruskal, 1964, Weisberg, 1974). Conversely, FA represents correlation through angles between vectors: a correlation of zero would be displayed by 90ᵒ separation between the lines connecting the variables' points with the origin [ ] A single dimension is found if variables covary perfectly, which is to say if the variables are identical to one another except for linear transformation (Weisberg, 1974 p.766). This in turn bears a side effect: FA uses a fictitious dimension to calculate angles (Weisberg, 1974) thus sometimes identifying more dimensions than those with substantive meanings 18. The Directive Banning Tobacco Advertisement and Preference Attainment This section applies both the theoretical and methodology innovations proposed in this work to a case: the decision-making process and the relative interest representation in the Directive banning tobacco advertisement. This case does not purport to be representative of interest representation dynamics nor of the determinants of interest group influence: different causal mechanisms may be in place when it comes to interest group influence. Nonetheless, how preference attainment in its rejuvenated version is applied to this case may be generalized to other cases. Furthermore, this case is also relevant for theoretical and methodological purposes. First of all, it represents one of the the most ferocious lobbying campaign seen in Europe (Dawe, 1992 in Bitton et al, 2002 p. 7) and thus, not by chance, it has been studied from an interest group influence perspective (often from a 17 Although the case study below does not consider the issue of coalitions being more focused on a demonstration of how to use this methodology to measure interest group influence MDS bears also this value added: it can better account for lobbying through advocacy coalitions. 18 Several types of MDS are present but the basic difference is between metric and non metric ones. A thorough discussion on this falls beyond the scope of this work but it should be noticed that the choice of the method depends on the type of data and, more specifically, whether correlation between data and the proximity in the political space moves linearly or simply monotonically. 15

16 preference attainment perspective) in various works (Bitton et al., 2002, Neuman et al., 2002, Boessen and Maarse, 2009, Adamini et al., 2011). Secondly, the vast amount of confidential tobacco industry documents provide insight into industry lobbying strategies in a way that usually remains concealed from political scientists (Gruning et al, 2008 p. 142) 19. This is crucially important, especially when using the preference attainment approach. Last but not least, this case is of particular interest in that it allows a thorough discussion on the two theoretical premises of preference attainment on which this work focuses: the temporal and spatial aspects of the political space. A Brief Overview In 1984 the European Council invited the Commission to use its right of initiative to legislate in the tobacco control area leading thus to the 1987 Europe Against Cancer (EAC) program (Boessen and Maarse, 2009). After two years of negations with the European Parliament (EP) the Commission adopted a proposal in 1991 envisaging a total ban on cross-border advertisement and sponsorship of tobacco products. The following year the EP voted in favour of this proposal, but the Health Council blocked it. In the Council a blocking minority consisting of Germany, UK, Netherlands and periodically Denmark and Greece (Bitton et al, 2002) continued to strike down this proposal until The 1995 and 1997 were crucial years in that the EU enlargement to Austria, Finland and Sweden and the shift in the UK government changed the internal dynamics within the Council. Accordingly, a blocking minority was no longer reachable and in 1998 the directive passed including also all forms of advertisement and not only those cross-border (as in the 1992 proposal) as result of amendments from the EP (Neumann et al, 2002; Bitton et al, 2002). Lobbying on the Directive Banning Tobacco Advertisement The differences between the two interests involved in this case were abysmal. The tobacco industry employed one of the most aggressive and complex lobbying campaigns ever seen in Europe, following the strategies it had already applied against tobacco control legislation in other contexts (e.g. US) (Bitton et al., 2002, Neuman et al., 2002). Conversely, the health interests suffered from a serious lack of coordination at EU level: only with the support from other actors did they manage to activate and lobby. The tobacco industry shopped across lobbying venues using both direct and indirect strategies. At national level it strategically targeted those Member States (MSs) which were crucial to form a blocking minority in the Council with direct as well as indirect strategies (PM, 1993 p.5). In doing 19 The 1998 US Master Settlement Agreement made tobacco industry documents publicly available. 16

17 so, it employed a wide array of strategies (Bitton et al, 2002; Boessen and Maarse, 2009). The direct contact with ministers was the prevailing strategy in UK, Germany and Greece. Conversely, the use of targeted voice with the support of supposedly independent think tanks was deemed to be the optimal strategy in Netherlands and Denmark. In the latter, indeed, there was a strong tradition of non-interference in the private consumption sphere (PRIMA, 1992 p.4; Neumann et al, 2002). At EU level the tobacco industry lobbied both indirectly and directly. It exploited the support of MSs as well as alliances with third parties. For instance, it managed to file a proposal written by the head of the Confederation of European Community Cigarette Manufacturers (CECCM) to the Council thanks to the support of Germany (Lepere, 1993 p.1). It lobbied crucial MSs to object the Commission s funding to the Bureau for Action on Smoking Prevention (BASP) (Bitton et al, 2002). Furthermore, it exploited the support of the European industrial associations (i.e. the then UNICE) and of the communication community. As for direct strategies, it lobbied extensively the European Parliament (EP) both in 1992 and in the second round in 1997 when the introduction of the codecision procedure gave veto power to it (Adamini et al., 2011). The tobacco industry managed to put into practice this impressive lobbying campaign given its organisational advantage along with the huge amount of resources it could deploy. All the premises for the victory of the tobacco industry were in place, as acknowledged by the head of the CECCM himself: it won't be easy but I firmly believe that we have a realistic prospect of success (Lepere, 1993 p.3). The action of the health lobby on its was late and uncoordinated. Only in 1990 when the first proposal had already been drafted the national leagues of cancer established the European Cancer League (ECL) with headquarters in Brussels. Furthermore, in 1992 the European Public Health Alliance (EPHA) was created thanks to the support of the European Citizen Action Service (ECAS). Even with the support of those EU umbrella associations the coordination between national health lobbies was extremely difficult given the low awareness of what was going on in Brussels and the differences in attitudes towards lobbying practices across countries (Boessen and Maarse, 2009). Only in 1997 when the possibility of a qualified majority in the Council emerged did health interests start to actively lobby (Boessen and Maarse, 2009). Despite the organisational deficiencies and the scarce resources the health lobby succeeded in representing its interests at the EU level thanks to the support and funding from the Commission. The latter set up in 1989 an ad hoc committee of cancer experts and started to fund the BASP, which became a relevant actor. The BASP, indeed, had a major role in providing technical information to the EP committees in 1992 (Neumann et al, 2002). Even when the pressure of the DG Agriculture and some MSs forced the Commission to stop funding the BASP in 1994 (Watson, 1995) the latter was replaced with two European networks on smoking prevention. 17

18 Measurement I extrapolated actors preferences from primary sources and the content of the legislative acts through a manual content analysis. In both cases I also relied on the extensive secondary sources on this case (Adamini et al, 2011; Boessen and Maarse, 2009; Bitton et al, 2002; Neumann et al, 2002). Then, I manually coded the preferences of the actors involved 20, the policy outputs, the reference point and the status quo ante according to five factors. Table 1 shows the coding procedure. The five factors are: EU intervention, strength of regulation (i.e. soft law or hard law), the forms of advertisement covered by regulation (i.e. any form of advertisement or only those with cross-border effects), sponsorship (i.e. included or not), advertisement targets (i.e. only people under 18 or everyone). Those factors represent the issues around which the debate orbited according to primary and secondary sources. The Temporal and Spatial Aspects of Dimensionality Two aspects deserve further discussion: the temporal and spatial aspects of the political space. As seen above, the political space is to be divided into agenda setting, decision-making and alternative specification. In the EU the agenda setting comprises what occurs before the Commission issues a proposal and the decision-making starts therefrom until a piece of legislation is issued (e.g. a directive). In this case the result of the agenda setting is represented by the 1991 Commission proposal and the result of the decision-making by the 1998 Directive. As for the alternative specification, the failure of the Commission proposal (e.g. due to a block in the Council) would have likely resulted into a soft law approach, such as the 1989 resolution on smoking in public places (Bitton et al., 2002). This scenario would have been different with respect to the status quo ante, which is represented by the absence of EU regulation. Not only these two concepts are theoretically different, as suggested above, but they also bear different empirical implications with respect to actors preferences. Indeed, the tobacco industry preferred regulation at EU level if in the form of soft law (PM, 1991): this would have sent a signal on the eagerness of the industry to regulate tobacco advertisement along with its good faith preventing also future stronger regulation at EU level 21. Conversely, the health lobby preferred the status quo ante to the lost occasion of 20 Future applications of this approach should consider individual interest groups and derive coalitions empirically with MDS and not assuming them as done in this section. Nonetheless, two points justify this choice in this work. Firstly, the simplicity of the analysis is functional to the illustration purposes of this work. Secondly, in this case the a priori assumption about broad coalitions hold. Indeed, the prominent role of Philip Morris and CECCM as pivot of the coalition led to an extremely coordinated action on the part of the tobacco industry. The same applies to the health lobby especially thanks to the role of the Commission. 21 Self regulation in tobacco advertisement and smoking in public places has been the rule until the beginning of the 2000s across advanced democratic countries. Several studies associate self regulation to a strategy put in place by the tobacco industry to prevent or at least delay top-down regulation (Baggott, 1988, Gilmore et al., 2002, Gruning et al., 2008, Cairney et al., 2012). 18

19 regulating advertisement at EU level with hard law 22 : a soft law approach would have constrained future regulation jeopardizing the possibility of stronger regulation 23. Before proceeding to the discussion of the results this part focuses on dimensionality. As illustrated in Figure 2, the political space assumes a two dimensional character. First of all, as shown by the Eigenvalues in Table 2 and the scree plot in Figure 3 the second dimension significantly increases the validity of the political space. Eigenvalues account for the extent to which the dimensions are exhaustive: as may be seen, when the second dimension is added the cumulative squared Eigenvalues increases significantly. The researcher can also gather this information from the scree plot in Figure 3, which plots Eigenvalues in descending order of variance explained [ ] this downward-sloping plot may exhibit an elbow as the Eigenvalues suddenly level off, implying that adding another latent dimension does not substantially increase the amount of explained variance in the set of indicator variables [ ] this elbow can be used to infer inductively the best lowdimensional representation (Benoit and Laver, 2012 p. 206). A related interesting aspect is how to interpret dimensions. As mentioned above, prior knowledge of the political space under analysis is a requirement also when using an ex post methodological approach to dimensionality. In this case the researcher should not be surprised to detect two dimensions which arguably can be reconciled to the form of regulation on tobacco advertisement (dimension 2) and EU intervention along with its strength (dimension 1). As seen above, these two dimensions are recurrent in European party politics as well as public opinion. Furthermore, this is also supported by the knowledge of the case under analysis. The debate between material and postmaterial values, namely what have been called the political economy and public health frames, represent the traditional dimension on which national debates on tobacco control has been based in the last decades (Frankenberg, 2004, Marmor and Lieberman, 2004, Studlar, 2007, Gruning et al., 2008, Cairney, 2009, Studlar, 2009, Studlar, 2010, Cairney et al., 2012). Nonetheless, this dimension has been associated with another one in Europe: whether EU should regulate on this area or not and how. Indeed, the tobacco industry has always challenged the legal capacity of the EU to legislate in this area favoring a soft law approach at EU level, such as the proposal introduced at the Council of Ministers on the part of the CECCM with the support of Germany (Bitton et al., 2002, Neuman et al., 2002). 22 This also shows how actors preferences are relative and not absolute in this example. 23 I acknowledge that a year later the directive was repealed by the ECJ and the Commission issued a new proposal in 2001 which led to the 2003 directive (the one dealt with in this work is usually termed TAD1 whereas the 2003 one TAD2). Nonetheless, they represent two different political spaces which deserve a separate analysis. 19

20 Interest Group Influence and the Two Faces This section measures interest group influence in the two connotations introduced above from a preference attainment perspective. The second face of power accounts for how much of the variation between the initial policy output and the status quo can be attributed to an interest group. In this vein, a measure is needed such that it increases the more the status quo is distant to the actor s preferences as well as the closer the initial output is to the actor s preferences. The first face of power follows a similar rationale. It accounts for to what extent actor s preferences are distant from the final policy output as well as from the reference point. Accordingly, a measure is needed such that it increases the more the reference point is distant from the actor s preferences as well as the closer the final output is to actor s preferences, in line with Dür (2012). The measures respectively for the first and second face are as follows: - First face of power (decision-making): A1 RP A1 P2 - Second face of power (agenda setting): A1 SQ A1 P1 - RP: reference point - SQ: status quo - P1 and P2: initial and final policy output - A1: actor s preferences The distance between tobacco industry preferences and the reference point is approximately which must be discounted for the distance between tobacco industry preferences and the final policy outcome, which is approximately The result is a negative number, namely : this means that tobacco industry did not exert any influence on the decision-making stage. Rather, the final output moved further away from its preferences. As for the second face of power, results show that the tobacco industry did not exert influence even in this case. The difference between the distance from tobacco industry preferences from the status quo and the initial policy output is almost null, namely Conversely, results show that the health lobby has been influential in both the agenda setting and decision making stage. In the latter instance health lobby preferences resulted to be distant from the reference point and almost coincidental with the final policy output: the difference between two distances is (equivalent to the first distance). The health lobby proved to be influential also in the agenda setting stage where the difference between the distance from its preferences to the status and to the initial policy output is In conclusion, the health lobby has been more influential than the tobacco industry in both agenda setting and decision-making. 20

21 Preference attainment as applied to this case reveals interesting results, especially from the perspective of the literature which almost unanimously argues for the victory of the tobacco industry in the case under analysis. Nonetheless, two caveats are needed. First of all, preference attainment bears the theoretical limitations introduced above: for instance, those related to perceiving an issue as a political space. In fact approach cannot reveal that the tobacco industry managed to block the directive for several years in the Health Council or that it succeeded in annulling it few years later thanks to the ECJ ruling mentioned above. Again, the temporal and spatial aspects of the political space and the limitations derived therefrom are congenital to this approach and they must be acknowledged by the researcher. Secondly, this section purports exclusively to illustrate the novel theoretical and methodological modifications to preference attainment introduced in this work as applied to a case. The process whereby that approach is applied to the case can be transposed to other cases. Yet, this does not apply to the substantive findings regarding interest representation dynamics and influence. Considerations on how interest groups lobby and how they manage to do so successfully fall outside the scope of this work. Conclusion This work provides an example on how to analyse interest group influence from a preference attainment perspective accounting for both the first and the second face of power. In doing so, it argues for a more theoretically sound conceptualization of the political space in its temporal as well as spatial aspect. Indeed, the researcher must take into consideration the whole policy cycle comprising agenda setting, decision-making and alternative specification. In other words, the researcher must investigate interest group influence when the issue enters the political agenda and when actors act to modify the piece of legislation taking also into consideration what was the situation before that issue was on the agenda as well as what would happen if the piece of legislation did not pass. Furthermore, attention needs to be reserved to dimensionality and the implications it bears on actors preferences. Indeed, political competition cannot be exhausted by a single dimension: often two dimensions are needed to validly represent reality. This also allows less restrictive assumptions on actors preferences. The scholarship has hitherto struggled in measuring interest group influence in valid and reliable manner and in its two main connotations. The works which so far have tried to analyse power in its structural form has used an institutionalist approach. The implicit or explicit rationale behind those works is that the more access or contacts an interest group has to or with decision-makers the more it is able to set the agenda. Yet, how can we compare access to the Commission with contacts with a US Congressman? Furthermore, the difference between access and influence is abysmal. On the 21

22 other side, those works focusing on the behaviorist conceptualization of power, usually from a preference attainment perspective, have focused either on individual issues or on samples of issues difficultly comparable and, on top of that, only on the decision-making stage. Furthermore, as discussed above, those studies lack in validity in representing the political space. The theoretically and methodologically rejuvenated preference attainment approach proposed in this work is far from being perfect but it allows to analyse two types of power which have hitherto received separate treatment in a valid manner. Furthermore, although at first sight it may appear to neglect reliability it provides bases for comparable findings thus hopefully leading research to a more cumulative progress. For instance, this work can pave the way for a research project close in spirit to the Resolving Controversy in the EU (Thomson, 2011, Thomson et al., 2012) where the researcher divides European politics into clearly defined political spaces on which interest groups act with the aim of systematically analyse the determinants of interest group influence in its two faces. 22

23 Figures and Tables Figure 1 Source: Klüver (2009) Table Commission Proposal 1998 Directive EU Hard Law v. Forms of Sponsorship Advertisement Intervention Soft Law Advertisement Targets Status Quo Reference Point Tobacco Industry Health Lobby

24 Figure 2 Table 2 Abs(Eigenvalues) (Eigenvalues)^2 Dimension Eigenvalues Percent Cumul. Percent Cumul

25 Figure 3 25