(Inequality in) Interest Group Involvement and the Legitimacy of Policy Making

Abstract While interest groups are consulted at different stages of policy making to provide expertise and legitimacy, their influence is often criticized as being undemocratic. Yet, we know little about how their participation in policy making affects citizen perceptions of the legitimacy of governance. Based on survey experiments conducted in the UK, the United States and Germany, our study shows that unequal participation between group types reduces the benefits of interest group consultation for citizens' perceived legitimacy of decision-making processes. Importantly, these legitimacy losses cannot be compensated for by policies that represent the opinion of the under-represented groups and are even greater when policy decisions favour the over-represented groups. Moreover, we show that citizen perceptions of how economically powerful and representative of society different types of interest groups are act as important drivers of legitimacy evaluations. Our results provide important new theoretical and empirical insights into when and why interest groups affect democratic legitimacy.

In democratic systems, the key actors who shape policy are elected as representatives of the people.However, interest groups, who claim to represent certain parts of society but lack the democratic legitimacy bestowed through elections, also participate in it.They come in many forms, including business associations, firms, non-governmental organizations (NGOs), trade unions and think tanks (see, for example, Baroni et al. 2014).The fact that interest groups act as 'self-appointed representatives' (Montanaro 2012) makes their role in democratic governance ambiguous.On the one hand, they offer valuable technical expertise and can bring attention to constituency interests that might otherwise be overlooked (see, for example, Bouwen 2004;Flöthe 2020;Klüver 2013b).On the other hand, the interest group system is frequently argued to be biased and dominated by representatives of powerful business or smaller sections of society, rather than representing the public good or the majority of citizens (Flöthe and Rasmussen 2019;Schattschneider 1948).This creates the risk that interest groups affect democratic representation by pulling policy not only towards, but also away from, the popular will (see, for example, Becher and Stegmueller 2021;Giger and Klüver 2016;Lax and Phillips 2012;Rasmussen, Binderkrantz and Klüver 2021).
While there is no lack of discussion among academics and journalists about whether interest groups are a force that weakens or strengthens democracy, we still have little systematic knowledge about how legitimate citizens perceive interest group involvement in decision making.As a New York Times op-ed forcefully argued, the dominance of economic elites and business interest groups in policy making is something that citizens are aware of and that contributes to 'disillusionment with democracy'. 1 An example is the 'lobbying blitz' around the US COVID-19 relief package, with powerful industries and companies aiming at securing benefits for themselves.It has been compared to a 'gold rush' and seen as providing 'a lens into the part of Washington that most Americans despise'. 2These public perceptions matter because the effectiveness of democratic policy making and, ultimately, the stability of democratic systems hinge on citizens' sustained support for the political processes that produce policy and their acceptance of the resulting decisions as legitimate, even if they do not agree with them (Norris 1999).
Our objective is to examine when and why (unequal) involvement of different types of interest groups in a formal, closed form of policy consultation, specificallyhearings in national parliamentsaffects citizens' perceptions of the legitimacy of the policy-making process. 3As decision makers regularly consult interest groups on legislation that affects the daily life of millions of citizens, it is important to know how such efforts are perceived by their constituents, who ultimately control their re-election.
Since a key point of criticism of the influence of interest groups concerns the imbalance of power among them, we argue that procedures where policy makers prioritize obtaining input from some stakeholders over others may be particularly at risk of being seen as illegitimate.However, we expect that citizens distinguish between different types of groups: they are likely to be more sceptical of dominance by interest groups that are economically powerful and not representative of society as a whole.Moreover, we expect citizens to pay attention to not only who is consulted, but also whose preferences are reflected in the final policy decisions.Specifically, we examine whether citizens are concerned about unequal interest group involvement in decisionmaking processes even if the policy outcome is in line with the preferences of under-represented interest groupswhich might signal that these types of interests were nevertheless heard.Similarly, we test whether the negative consequences of unequal involvement are aggravated when the decision favours the type of interests that are invited to participate in greater numbers but goes against the opinion of the types of interests that are under-represented.
To test the hypotheses, we conducted survey experiments with a conjoint design among representative samples of the populations of three democracies: the UK, the United States and Germany.While we do not expect cross-national differences in the explanatory power of our theoretical framework, this allows us to test the robustness of our results to the idiosyncrasies of specific countries.Our surveys ask respondents to evaluate the legitimacy of two hypothetical policy-making scenarios in which equal, unequal and no involvement of two types of interest groups, as well as the policy decision and the groups' preference attainment, are randomly assigned.We find that citizens perceive policy-making processes as more legitimate when the interest group types that are key stakeholders on an issue are involved, but the benefit is smaller when the group types are not equally involved.
Moreover, citizens regard the under-representation of cause groups representing broader societal interestsin this case, environmental and consumer groupsas more problematic than that of business interests.Shedding light on the mechanisms behind these judgements, we demonstrate that citizen perceptions of business groups as being more economically powerful but less representative of society play a strong role in explaining this pattern.We also find that losses in legitimacy resulting from unequal involvement in consultations can generally not be compensated for by policy outcomes that reflect the views of the under-represented groups.Interest group participation and policy attainment interact only rarely, and when they do, they aggravate the negative consequences of unequal representation on legitimacy.Specifically, the negative effect 1 See: https://www.nytimes.com/2015/09/15/opinion/across-the-globe-a-growing-disillusionment-with-democracy.html?searchResultPosition=1 (accessed 12 August 2021).
on procedural legitimacy of business groups being over-represented is even stronger if the policy output satisfies them at the expense of the under-represented cause groups.
The study offers important and novel insights into citizens' views of the role of interest groups in democracy, which has been the subject of both scholarly and wider public debate for a long time.While citizens generally see benefits in the consultation of groups, they are sensitive to inequalities in the process and potential risks of influence from (economically) powerful groups and those that represent only small pockets of society.Therefore, decisions about the inclusion of interest groups in policy making have implications for citizens' willingness to accept the process, and most likely its outcomes too.In addition, interest group scholars prominently distinguish between interest groups defending business interests and those promoting societal causes, and this study sheds a first light on the relevance and meaning of this distinction in the eyes of citizens.The findings speak to a broader literature on how representative diversity affects perceptions of legitimacy, which has looked, for example, at gender (see, for example, Clayton, O'Brien and Piscopo 2019) and race (Scherer and Curry 2010).
(Unequal) Interest Group Participation and Legitimacy Perceptions For citizens to support and defend the institutions that govern their society, and to accept and comply with the outcomes of political decision-making processes, it is crucial that they perceive these institutions and processes as legitimate (see, for example, De Fine Licht et al. 2014;Dellmuth and Schlipphak 2020;Tyler 1994).Recent years have witnessed a number of important studiesmostly survey experimentson how citizens judge the legitimacy of various decisionmaking procedures and institutions (see, for example, Clayton, O'Brien and Piscopo 2019;Esaiasson, Gilljam and Persson 2017;Esaiasson et al. 2019).For instance, some studies examine citizen participation (De Fine Licht et al. 2014;Esaiasson, Gilljam and Persson 2012), while others have looked at how the characteristics of referendums and representative bodies affect how willing citizens are to accept political decisions (Arnesen and Peters 2018;Arnesen et al. 2019).
The ways in which interest groups participate in policy-making processes is likely to play an important role in how legitimate citizens perceive these processes too.After all, the rationale for involving interest groups is precisely their assumed potential to improve the policy-making process and its outcomes.In the European Union (EU), for example, they are often seen as a means to help citizens exert influence on decision making and legitimize policies (Héritier 1999).However, in national contexts too, group involvement in policy making is commonly assumed to boost legitimacy.The literature on public consultations has even argued that, in some respects, interest groups can serve as a surrogate for the public in the policy process (Lundberg and Hysing 2016).According to Mikuli and Kuca (2016: 5): 'the role of interest groups is sometimes perceived as being even more important for the concept of a participatory democracy than that of individual citizens'.We therefore hypothesize: Hypothesis 1: Citizens' perceptions of legitimacy are higher when the key stakeholder interest groups on a policy issue are consulted than when they are not.
At the same time, other work casts a much more sceptical view.Hibbing and Theiss-Morse (2004) documented that the vast majority of US citizens find that special interests have too much control and that campaign spending should be limited.Almost half of them even support preventing organized interests from getting in touch with members of Congress.One of the key topics in the literature on interest representation is bias or inequality in the representation of different group types (see, for example, Gray and Lowery 2000;Schattschneider 1960;Schlozman and Tierney 1986).The main concern is that certain types of groups, especially those representing powerful actors in society, may be over-represented in policy-making processes and have a disproportionate influence on political decisions.Indeed, Gilens and Page's (2014) work shows that there is a strong connection between the preferences of both business interest groups and economic elites and policy at the expense of mass-based interest groups and the average citizen.According to Helboe Pedersen, Halpin and Rasmussen (2015: 411; see also Bochel and Berthier 2020): '[r]epresentation is of obvious importance when committees decide whether and who to involve in their work … and it would reflect badly on the committees, committee members as well as the Parliament if committees refrained from taking evidence from the main stakeholders closely related to the topic being discussed'.
In line with this, we argue that citizens view policy-making processes that involve different interest group types with a stake in the issue to unequal degrees as less legitimate.There are several ways in which such inequality may distort the (perceived) benefits that interest group participation can have on policy making.By consulting a diverse and balanced range of stakeholders, parliaments can send a signal to citizens that they take an interest in and listen to a broad range of views (Bochel and Berthier 2020).If the interest groups representing different groups in society participate in unequal degrees, it might suggest to citizens that their indirect influence might also be unequal.Furthermore, putting important stakeholders into a minority position might lower both the problem-solving capacity and efficiency of decision making (see, for example, Bernauer and Gampfer 2013;De Bruycker 2016;Grossmann 2012;Mikuli and Kuca 2016;Wright 1996), as the information supplied becomes one-sided and the selected solutions may be less than ideal for solving the challenges at hand.According to Beswick and Elstub (2019: 948-9), 'a greater diversity in evidence can enhance the epistemic quality of committee inquiries as it can increase the amount of relevant information and views that committees would receive' and 'a broader range of views can also encourage the committee members to focus more on the common good'.
Similarly, when the involvement of groups representing different interests and constituencies in consultations is imbalanced, the transaction costs of policy making might increase due to a need to brief the excluded groups about the process and possibly accommodate them at a later point.Finally, interest groups are often argued to act as transmission belts between members of the public and decision makers.This helps decision makers learn about the interest of the public while making it easier for citizens to hold politicians accountable (see, for example, Albareda 2018; Bochel and Berthier 2020;De Bruycker and Rasmussen 2021;Furlong and Kerwin 2005;Rasmussen and Reher 2019;Rasmussen, Carroll and Lowery 2014).Unequal involvement of the key stakeholders could threaten the equity of these transparency and accountability mechanisms.
For these reasons, we expect that citizens perceive policy-making processes where the participation of different types of interest groups is unequal as less legitimate than processes with equal representation.Since our research design uses an example of a policy issue with groups belonging to two types of interest groups as key stakeholders, we hypothesize: Hypothesis 2 (unequal representation): Citizens' perceived legitimacy of the policy-making process on an issue concerning two types of key stakeholder interest groups is higher when their involvement in the consultation process is equal as compared to representatives of one group type dominating.
Despite the prominent role of interest groups in the formulation of policy in many political systems, as well as the widespread concern about their unequal influence, only a few studies have investigated how the public perceives their involvement.Most of them use experimental methods, and many are conducted in the field of international relations.Bernauer and Gämpfer (2013) and Bernauer et al. (2016) find that changes to existing practices of involving civil society groups in international negotiations affect the perceived legitimacy of global environmental governance processes in the United States, India and China.Bernauer, Mohrenberg and Koubi (2020) provide indicative, though not statistically significant, evidence that involving business groups in negotiations on transboundary air pollution reduces their perceived legitimacy in Germany and the UK, whereas involving environmental groups and experts strengthens it.Focusing on regulatory agencies in the EU, Beyers and Arras's (2021) findings support our expectation that citizens consider processes that consult both business and citizen interest groups as more legitimate than those that exclude a group type.Finally, Terwel et al.'s (2010) lab experiments with student samples find a positive effect of interest group involvement in a fictional advisory body on the acceptance of policy decisions on carbon dioxide capture and storage technology.Again, decision acceptance was significantly lower when only one of two group types were involved.
We expand this research by conducting two large-scale survey experiments among representative samples of the population in three democracies, using examples of real domestic consultation procedures, namely, parliamentary hearings.We focus on evaluations of procedural legitimacy, which have been shown to mediate the effects of procedures on decision acceptance (Esaiasson et al. 2019).Finally, as we elaborate in the next section, our study advances existing knowledge by examining the sources of variation in perceptions of different group types, as well as how evaluations of inequality in group involvement are affected by the groups' preference attainment.

How Perceptions of Interest Groups May Drive Assessments of Unequal Representation
The interest group community consists of different types of groups, where a key distinction is typically drawn between 'sectional' and 'cause' interests (Binderkrantz, Christiansen and Pedersen 2015;Giger and Klüver 2016;cf. Olson 1971).Business groups are often seen as prominent examples of the first category.They represent particular constituencies and provide concentrated benefits to their members, in contrast to cause groups, which represent ideas and (broader) societal interests, for example, environmental and consumer groups involved in the provision of public goods.Both the public and academic debates about interest group bias often centre on the risk that groups representing powerful industries dominate and disproportionately influence policy, making it less responsive to the interests or opinions of the wider public (see, for examples Flöthe and Rasmussen 2019;Hanegraaff and Berkhout 2019).
This concern is supported by empirical evidence.First, the policy positions of public interest groups, which form an important part of the cause group category, have been shown to be more frequently aligned with the majority of citizens than the positions of business associations and firms (Flöthe andRasmussen 2019). Secondly, Hertel-Fernandez, Mildenberger andStokes (2019) found that more contact with mass-based interest groups relative to corporate groups is associated with more accurate perceptions of public opinion among congressional staffers.Not surprisingly, the prominence of business interests is therefore often a key component in how bias in interest representation is understood and measured (see, for example, Hanegraaff and Berkhout 2019;Rasmussen and Carroll 2014;Schattschneider 1960;Schlozman and Tierney 1986).
On this basis, we expect that citizens generally consider a numerical over-representation of business interests in policy-making processes as more problematic than an over-representation of cause groups.Previous research offers some support.Beyers and Arras (2021) show that citizens consider procedures of regulatory agencies that consult only business groups as less legitimate than those that consult only consumer organizations and no different from processes consulting neither.Bernauer and Gämpfer (2013) show that citizens favour including environmental NGOs over business groups in international climate negotiations.Including only business groups decreases legitimacy compared to including neither, while including environmental NGOs has no effect.In a later study, Bernauer, Mohrenberg and Koubi (2020) find that involving environmental groups in international negotiations increases public support for the outcome, while involving business groups decreases it, though the effects are not statistically significant.By contrast, Terwel et al. (2010) show that while providing environmental NGOs and industrial associations with equal voice leads to greater decision acceptance than unequal voice procedures, there is no significant difference based on which group type dominates.
We take this discussion a step further by trying to uncover the sources of such potential differences in citizen attitudes towards dominance by business and cause groups in consultation procedures.First, we expect citizens to be more concerned about the under-representation of a group type in the consultation if they perceive it to have fewer economic resources than the overrepresented group type.While many different types of interest groups engage in mutual exchanges with decision makers (Bouwen 2004;Klüver 2013b), the character of these exchanges is thought to vary as a result of differences in the 'exchange goods' that different group types bring to the table.Business interest groups are assumed to have a competitive advantage in terms of economic and financial resources (Dür, Bernhagen and Marshall 2015;Mahoney 2004;Schlozman 1984).This allows them to supply decision makers with technical information, for example, about scientific aspects or the feasibility and effectiveness of specific policy proposals (Broscheid and Coen 2003;De Bruycker 2016).At the same time, there is a concern that business interest groups can use these resources to reap disproportionate benefits for their members at the expense of the general public (Dornhoff 2013;Gilens and Page 2014).There is also evidence that corporate campaign contributions are associated with less accurate perceptions of public policy preferences among congressional staffers (Hertel-Fernandez, Mildenberger and Stokes 2019).We expect that citizens widely share the impression of business groups having more economic resources and the concern that economic resources allow for greater influence in policy-making processes.In addition, citizens might suspect that more economically powerful groups are also more capable of voicing their concerns through other avenues than formal consultations.
Cause groups, on the other hand, which represent consumer, humanitarian, environmental and other interests, are seen as having fewer economic resources at their disposal but a stronger representative potential (Baumgartner and Leech 2001;Flöthe and Rasmussen 2019;Grant 1989;Klüver 2013a;Kohler-Koch 1994).They typically speak on behalf of a broader set of societal interests, giving them a stronger potential to act as a transmission belt between wider sections of the public and the political system (see, for example, Bevan and Rasmussen 2020;Giger and Klüver 2016).Again, we expect that many citizens share this view and therefore consider the overrepresentation of cause groups in policy-making processes as less problematic than a dominance of business groups.Hence, as a second mechanism explaining differences in public attitudes towards unequal interest group access to policy making, we hypothesize that the stronger a group type's perceived ability to represent society as a whole, the weaker the negative effect of its over-representation on the perceived legitimacy of the process.In sum, we expect that citizens' perceptions of the economic resources and representativeness of interest group types largely explains individual-level differences in the perceived legitimacy of the involvement of business and cause groups in policy consultations.As such, we hypothesize: Hypothesis 3a (economic resource moderation): The negative effect on perceived legitimacy of representatives from one interest group type dominating participation is greater the stronger citizens perceive their relative possession of economic resources.Hypothesis 3b (societal representativeness moderation): The negative effect on perceived legitimacy of representatives from one interest group type dominating participation is weaker the stronger citizens perceive their relative ability to represent society.

How Policy Attainment of Interest Groups May Moderate Assessments of Unequal Representation
As explained earlier, we expect that citizens generally consider policy-making processes with unequal participation of the key types of interest groups on an issue as less legitimate than those with equal participation.However, citizens might judge the fairness of the process based on not only the presence of different interest groups, but also the policy decision resulting from the process.A key reason for consulting interest groups in policy making is not only to send a signal that those with a stake in a policy issue have been heard, but also to ensure that the particular needs and concerns of those whom they represent are taken into account and reflected in the policy outcome (see, for example, Mikuli and Kuca 2016;Saurugger 2008).Primarily hearing one type of group might therefore not be regarded as particularly problematic in cases where the different types of groups hold the same opinion on an issue, as the views of the less-well-represented group type may still be 'coincidentally' represented if the dominating groups persuade policy makers to follow their position.
The situation is different when the two types of stakeholders disagree.Here, citizens' evaluations of a numerical imbalance in who is invited to consultations may depend on whose preferences are reflected in the policy outcome.We would expect that the negative effect on legitimacy of either cause or business groups dominating is stronger when the policy decision goes against the position of the groups in the minority, as compared to scenarios where both groups attain their preferences.In this case, the numerical under-representation of representatives from one group type appears to have actual consequences for who attains their policy preference, which could aggravate how problematic citizens find unequal numerical representation.In other scenarios, preference attainment might mitigate the negative effect of numerical imbalances in who is invitednamely, when the adopted policy satisfies the policy position of the group type that is under-represented.In other words, citizens might see it as less problematic that decision makers decided to include fewer groups of a particular type if it appears that they made sure to accommodate their views in the policy decision.The implementation of the interests of a group type might even compensate for their underrepresentation in the consultation.As such, we hypothesize: Hypothesis 4a (aggravation of under-representation): The negative effect on perceived legitimacy of representatives from one interest group type dominating participation is stronger when the policy output goes against the opinion of the group type in the minority.
Hypothesis 4b (mitigation of under-representation): The negative effect on perceived legitimacy of representatives from one interest group type dominating participation is weaker when the policy output is in line with the opinion of the group type in the minority.

Data and Methods
We test our hypotheses through survey experiments with conjoint designs, where respondents are shown vignettes with descriptions of policy-making scenarios.The survey was fielded through Qualtrics in Germany (N = 3,130; 27 September-10 October 2019), the UK (N = 3,048; 1 August-30 September 2019) and the United States (N = 3,179; 31 July-30 September 2019).The quota samples are representative of the populations in terms of region, gender and age.We conducted two separate experiments with different policy issues to increase our confidence that the findings are generalizable to different policy areas.Each respondent participated in both experiments in a randomized order. 4In each experiment, respondents are presented with a short text explaining a realistic scenario where a new policy has been debated and either passed or rejected by parliament.The text identifies two different types of interest groups as the main stakeholders on the policy issue and states whether they were invited to participate in a consultation process and whether or not they were invited in equal numbers.Before seeing the vignettes, respondents are provided with background information about the consultation process (see Figure S1 in the Online Supplementary Information [SI]).
Governments or parliaments can consult stakeholders in different ways in the policy-making process (Bishop and Davis 2002;Fraussen, Albareda and Braun 2020;Helboe Pedersen, Halpin and Rasmussen 2015).In open strategies, everyone who wishes to can participate, for instance, in online consultations where all interested stakeholders can submit written input (see, for example, Rasmussen and Carroll 2014).In closed strategies, a selected set of stakeholders are invited who are considered to have expertise and/or represent the constituents affected by the issue, for example, as members of government advisory committees (see, for example, Gornitzka and Sverdrup 2008).We focus on closed strategies, more specifically, on procedures through which parliaments can ask specific stakeholders to provide oral evidence (Eising and Spohr 2017;Helboe Pedersen, Halpin and Rasmussen 2015;Leyden 1995;Mikuli and Kuca 2016).These procedures play a key role for legislators to obtain technical information about potential new policies, as well as information about public preferences and the views of affected parties.As Eising and Spohr (2017: 319) explain in the German context: 'committee hearings are the most established and visible form of obtaining such policy advice and are also meant to enhance the legitimacy and transparency of decisions'.Such procedures are also regarded as more effective for stakeholders to influence policy makers than open procedures for submitting written evidence (Beswick and Elstub 2019).This focus allows us to study how citizens evaluate interest group involvement in political processes when policy makers have control over whom they include and exclude from the consultation.
In the United States, we refer to congressional hearings, which are used in both the House of Representatives and Senate, where most bill discussions start with a hearing.In the UK, we refer to oral evidence provided to Parliament, and in Germany, we refer to public hearings in the Bundestag, which are generally used on important pieces of legislation.The legal basis for the hearing is typically laid out in the rules of procedures of the relevant parliament (Mikuli and Kuca 2016).While they are adapted to their political systems, they share commonalities critical to our design (Beswick and Elstub 2019;Eising and Spohr 2017;Loewenberg 2006;Mikuli and Kuca 2016).First, they represent closed forms of consultation, where access to give evidence is determined by parliamentary committees.Secondly, they can all be used in the lawmaking process to collect evidence from interest groups, as well as other stakeholders, such as (academic) experts and citizens.
The hearings regularly get coverage in the media, meaning that it is reasonable to expect that citizens have heard of them.In some cases, the focus is precisely the selection of actors who are invited to give evidence.For example, the UK Parliament's Transport Committee was criticized in The Guardian for inviting an interest group whose 'membership is "tiny" as a proportion of drivers' and that denies human-created global warming to give oral evidence on road safety. 5The Washington Post published a critique of a hearing by the US Senate Committee on Commerce, Science and Transportation on college athletes' right to monetize their 'name, image and likeness', stating '[t]his is a witness list outrageously stacked with pro-NCAA [National Collegiate Athletic Association] advocates', with the athletes themselves being left out. 6 The cross-national aspect of the design is motivated not by an attempt to explain potential differences between countries, but rather to test the robustness of our results in different national contexts. 7The three countries are all advanced democratic societies with similar levels of economic development and established systems of interest representation.They are selected so as to differ with respect to state-society structures, with the UK and the United States resembling a pluralist model and Germany a neo-corporatist one, where it is more common to enter into privileged relationships with subsets of stakeholders.At the same time, we do not have specific expectations regarding differences in the effects of interest group involvement on citizens' legitimacy perceptions between the three countries.While differences in more general public perceptions of the role of interest groups in politics might exist given the variation in the structures in 5 See: https://www.theguardian.com/world/2020/oct/21/climate-science-deniers-to-give-road-safety-evidence-to-mps(accessed 10 October 2021).6 See: https://www.washingtonpost.com/sports/2021/06/08/congress-ncaa-nil-senate-commerce-hearing/(accessed 9 October 2021).
which they operate, we purposefully hold the nature of the consultation process constant across the three cases.Moreover, we have no particular reason to expect that citizens judge inequality in representation differently across the countries.
The two policy issues that we describe in the vignettes are: (1) tax cuts for producers of hybrid cars; and (2) restrictions on the sugar content of beverages.The issues fall into different types of policy areas and are likely to be of varying importance to citizens.Importantly, both are examples of issues that attract attention from both cause and business interest groups, allowing us to study the impact of potential biases in representation.The key types of interest groups consulted are representatives from business and environmental groups in the hybrid cars example and from business and consumer groups in the beverage example.We ask respondents to centre their attention on the relative representation of these two types of interests, as this is of key relevance to our study.However, we also state in the introduction that in addition to these actors, policy makers may also have consulted other actors, such as (academic) experts and citizens.
The issues are also selected to allow separating the effects of inequality in the numerical presence and policy attainment.On both issues, it is plausible that both the business and cause groups either support or oppose the debated policy.For instance, environmental groups might be in favour of policies promoting the production of hybrid cars over combustion-engine cars, or they might be against financial incentives for any car manufacturers.Business groups might support or oppose the tax cuts depending on whether their members would benefit or suffer from them.This allows us to manipulate the groups' positions in the experiment in realistic ways in order to separate the effects of unequal representation in the consultation process from the alignment of the decision with the preferences of the groups, as well as to test whether they interact.
In conjoint experiments, the values of each attributein this study, the inclusion of representatives belonging to the two group types, the outcome, groups' policy attainment and public opinion (see Figure 1)are randomly assigned to respondents.In our design, the assignment of values of each attribute is independent from the values of the other attributes, allowing us to treat each attribute as a random variable.A key advantage of this approach over simpler factorial designs where fewer dimensions are varied is that we can identify the effect of one attribute averaged over the values of the other attributes: this is the average marginal component effect (AMCE).For example, the policy positions of the group types are assigned independently from the public majority opinion to ensure that citizens do not rely on assumptions about public opinion and how the group types align with it, since they might otherwise assume public opinion to be more likely to correspond to the positions of cause groups.In addition, we can compare the strength of the AMCEs of the different attributes.Finally, we can estimate the effects of one attribute at specific values of another, that is, conditional AMCEs, to test our hypotheses positing interaction effects (see Hainmueller, Hopkins and Yamamoto 2014).
Figure 1 shows the introduction to the experiment and the vignette text of the hybrid car experiment, including the possible values of the randomly varied attributes (pages 2-5 in the SI contain further details on the vignettes and experimental treatments).The first is the inclusion of interest groups in the policy-making process through inviting them to provide oral evidence/ testify at a hearing in parliament.We focus on three stylized cases, where: (1) equal numbers of representatives of both types of interests are included in the consultation; (2) one of the types dominates; or (3) neither of them is invited.The other attributes are the decision, that is, whether the policy was adopted or rejected, and whether this decision was in line with the positions of each of the group types.We also state whether the majority of the public was in favour or against the decision and the size of the majority (55 per cent or 70 per cent).
Our outcome of interest is individuals' perception of the procedural legitimacy of the decisions (see Clayton, O'Brien and Piscopo 2019), measured by respondents' agreement with three statements: 'The process that led to the policy decision was fair'; 'When making the decision, policy makers took the views of all relevant actors into account'; and 'The process that led to the decision was democratic.' Agreement with each statement is measured on a five-point scale, and we construct a summary measure by taking the mean of the items. 8

Independent and Control Variables
We measure the numerical representation of the interest group types in the consultation process and whether the policy decision reflects their positions with categorical variables indicating the values of the attributes shown in Figure 1 (and Table S1 in the SI).Before the experiment, respondents were asked about their perceptions of the economic resources and representativeness of society of business, environmental and consumer organizations (only in the UK and the US).We measure the relative perceived levels of these resources as the difference between the score (0-4) that respondents gave to business organizations and the respective types of cause group used in the experiment (that is, environmental or consumer groups).The resulting measures The three items are presented to the respondents in a battery of questions alongside three items measuring substantive legitimacy, that is, legitimacy of the outcome, where items measuring the two types of legitimacy alternate.The scale reliability is high in both cases, with Cronbach's alpha = 0.85 for procedural legitimacy and 0.90 for substantive legitimacy.In the SI (Table S2, p. 6), we also show the results of an exploratory factor analysis, showing that the items load on two factors in line with our indices.If a respondent has a missing value on one item, the remaining items are used to calculate the scale value.As a robustness check, we also estimated Model 1 separately for each of the three components of the legitimacy scale, which yielded estimates in line with the results for the scale (see Table S9, p. 16).
of relative perceived economic resources and relative perceived representativeness thus range from −4 (high for cause; low for business groups) to 4 (high for business; low for cause groups).
We also include the other manipulated attributes of the scenarios in the models, including the degree of public support for the policy indicated in the vignette.Citizens might prefer decisions aligned with the public majority if they expect collective choices to be superior to choices made by individuals or out of a social desire to conform (Arnesen et al. 2019).We also include a measure of outcome favourability, that is, whether the policy decision is aligned with the respondent's policy preference (see De Fine Licht et al. 2014), which is an important factor in individuals' perceptions of legitimacy (Arnesen et al. 2019;Esaiasson et al. 2019).We asked respondents before the experiment what their position is on giving tax cuts to companies that produce hybrid cars and on state regulation of the amount of sugar that beverages may contain, measuring their views on five-point scales from 'strongly oppose' to 'strongly in favour'.Depending on the decision outcome that was randomly assigned to the respondent in the respective vignette, outcome favourability is measured by this scale or its inverse to indicate respondents' agreement with the decision.

Results
We merge the data from both experiments into a stacked format so that the unit of analysis is the respondent-policy issue (two observations per respondent).We cluster the standard errors by respondent and include issue fixed effects. 9We start by investigating whether the composition of the set of interest groups that are invited to provide oral evidence and the extent to which the policy outcome aligns with the preferences of the different group types influence how legitimate individuals perceive the decision-making process.Figure 2 shows the coefficients of unequal representation and policy attainment from Model 1 in Table 1.Equal representation and policy attainment of both group types are the respective reference categories.
In line with Hypothesis 1, perceived legitimacy is by far the lowest when neither cause nor business interests are invitedalmost one point on a five-point scale, or 20 percentage points on the scale, lower than if equal numbers from both group types are invited.It is highest when both group types are consulted in equal numbers.When more cause groups than business groups are consulted, perceived legitimacy drops 6.6 percentage points (0.33 scale points), and Notes: Coefficients with 95 per cent confidence intervals from Model 1 of Table 1.

9
We exclude respondents who failed an attention check or incorrectly answered more than half of the manipulation check questions, which asked about the key pieces of information in the vignettes.Analyses run on the full dataset in the SI (Table S4, pp.10-11) find very similar effects to those presented here.
when more business groups are consulted, it is 9.8 percentage points (0.49 scale points) lower, supporting Hypothesis 2.
The bottom half of the figure shows the effects of policy attainment, that is, whether the group types attained their preferred policy outcome.The pattern is similar to that for numerical representation: citizens view the process as most legitimate when both group types attain their preferred outcome.Legitimacy perceptions are lower when the policy only reflects the views of cause groups and lower again when only the preferences of business groups are implemented.Decisions that reflect the position of neither group type are considered the least legitimate, but the difference to the other scenarios is much smaller than in the case of numerical representation.
The overall smaller effects of policy attainment are plausible considering that respondents evaluated the legitimacy of the process rather than the decision to which it led.

Sources of Differential Legitimacy of Cause and Business Groups
Citizens clearly perceive an over-representation of business groups as less legitimate than an overrepresentation of groups representing cause interests.Are these differences rooted in beliefs about the group types' economic and representative resources?Figure 3 shows respondents' mean agreement that each group type has economic resources and represents society.Business associations are perceived as having higher levels of economic resources than the cause groups, with environmental organizations seen as least economically powerful.Meanwhile, as expected, consumer and environmental organizations are perceived as more representative of society than business groups, with consumer organizations coming ahead of environmental groups.
To find out whether these differences explain why citizens tend to perceive the overrepresentation of business groups as less legitimate than a dominance of cause groups, we interact numerical representation with the indicators of perceived group characteristics (see Models 2 and 3 of Table 1).Figure 4 shows the predicted levels of legitimacy for the different scenarios of interest group involvement along the scales of relative perceived economic resources (on the left) and representativeness (on the right).
Individuals who perceive business and cause groups to have similar amounts of economic resources and to be similarly representative of society, that is, around 0 on the scales, judge processes with an over-representation of business interests as only slightly less legitimate than processes where cause groups are over-representedin other words, they are fairly indifferent over which group type is better represented.By contrast, individuals who perceive business groups as much more economically powerful (at the positive extreme of the scale in the left-hand figure) and as much less representative of society (at the negative extreme of the scale in the right-hand figure) than cause groups clearly consider an over-representation of business groups as much less legitimate than an over-representation of cause groups, by around 0.8-1.0 points on the five-point scale.
Thus, in line with Hypotheses 3a and 3b, we can indeed conclude that the reason why citizens tend to see business groups' dominance over cause groups in policy-making processes as more problematic for democracy than the reverse is that they consider cause groups better at representing society while having less influence through alternative channels that involve transmitting economic resources.A related explanation is that many citizens feel that business groups have a stronger interest in acquiring resources for themselves than in representing the public and therefore prefer them to have less influence on policy.
Only few respondents are located at the other ends of the scales, meaning that they perceive cause groups as much more economically resourced and less representative of society than business groups.At these extremes, an over-representation of business groups has higher legitimacy levels as compared to cause groups dominating, but these differences are not statistically significant.Furthermore, it is also clear that individuals' perceptions of the legitimacy of business group dominance and of the absence of both group types strongly varies with their perceptions of the economic resources and representativeness of group types: both slopes are rather steep.By contrast, the perceived legitimacy of equal representation and of cause groups dominating does not vary as much across the economic resources and representativeness scales. 10

Moderating Effects of Policy Attainment
The next question we seek to answer is whether policy attainment conditions the impact of unequal representation on legitimacy.To test this, we estimate interactions between numerical representation and policy attainment (see Model 4 of Table 1).For the sake of clarity, we illustrate only the key estimates pertaining to our hypotheses in Figure 5.It shows the conditional effects of cause and business groups, respectively, dominating in hearings as opposed to equal numbers for three different scenarios of policy attainment, where the policy outcome is in line with only business interests in the first scenario, with both interests in the second and with only cause group interests in the third.
We find that the effect of an over-representation of cause groups versus equal representation on legitimacy perceptions does not differ statistically significantly between the three policy attainment scenarios, even though the coefficients change slightly in line with our expectation.This Notes: Positive (negative) values on the scales of relative perceived differences in economic resources and representativeness of society indicate higher levels for business groups (cause groups); 0 indicates equal levels.Predictions are based on Models 2 and 3 of Table 1.

10
We estimated equivalent models that control for interaction terms between citizens' ideological position and group representation, as ideological views might be associated with perceptions of economic resources and the representativeness of groups.The moderating effects presented here remain statistically significant when including these controls (see Table S10 in the SI, pp.17-18).means that the negative effect of more cause groups being present is neither mitigated (Hypothesis 4b) when the preferences of the numerically under-represented business groups are implemented in policy nor aggravated (Hypothesis 4a) in cases where the outcome is in line with the dominating cause groups.
By contrast, the negative effect of business interests outnumbering cause groups is statistically significantly larger when the outcome reflects only the interests of business groups, as compared to when it reflects the preferences of both group types.In other words, the negative effect of business groups' over-representation is further aggravated by their policy attainment.Meanwhile, although the negative effect of business groups dominating seems smaller in cases where the interests of cause groups are implemented in policy as compared to policy being in line with both, this mitigation effect is not statistically significant.Yet, we do observe that when the decision reflects only the position of cause groups, citizens no longer perceive processes in which business groups outnumber cause groups as less legitimate than those where cause groups dominate.In sum, these findings suggest that citizens care about equality in the participation of different group types, so much so that the substantive representation of the preferences of the under-represented groups in policy cannot compensate for their exclusion or underrepresentation in consultations.At the same time, inviting fewer representatives from one group type can decrease citizens' legitimacy perceptions even more when decision makers also fail to implement their views in policy.

Conclusion
One of the key rationales for including interest groups in policy making is the assumption that it increases the legitimacy of the process (see, for example, Bouwen 2004; Flöthe 2020; Klüver 2013b).At the same time, there is a long-standing concern that interest representation is biased in favour of narrow and powerful interests, posing a risk to the representation of wider society (Lowery et al. 2015;Schattschneider 1960).A substantial amount of empirical research focuses on whether such bias exists and what explains it (see, for example, Hanegraaff and Berkhout 2019;Rasmussen and Carroll 2014;Schlozman and Tierney 1986).Yet, there have been few attempts to analyse citizens' views on these issues, specifically, when and why (unequal) interest group involvement affects their perceptions of the legitimacy of policy making.
To explore this question, we studied how inequality in interest representation affects citizen perceptions of democratic legitimacy through original survey experiments.We contribute to a  broader literature that explores how various aspects of decision making affect perceived legitimacy (see, for example, Arnesen and Peters 2018;Arnesen et al. 2019;Clayton, O'Brien and Piscopo 2019;De Fine Licht et al. 2014;Esaiasson, Gilljam and Persson 2017;Esaiasson et al. 2019;Strebel, Kübler and Marcinkowski 2019;Tyler 1994).Although previous research has investigated how diversity in representation (for example, in terms of gender and race) affects legitimacy, it has rarely analysed how citizens perceive interest group involvement in policy processes, apart from a small number of studies, many of which focus on international negotiations (Bernauer and Gampfer 2013;Bernauer, Mohrenberg and Koubi 2020;Bernauer et al. 2016;Beyers and Arras 2021;Terwel et al. 2010).Moreover, we advance existing research by investigating the sources of variation in citizen perceptions of different group types and by paying attention to not only whether interest groups are numerically represented, but also whether their policy positions are accommodated.
We used realistic scenarios of policy-making processes and policy issues to conduct experiments with large representative samples of the population in three countries.Our findings clearly show that citizens consider interest groups as important actors in the democratic system: they see policy-making processesillustrated with the example of parliamentary hearingsas more legitimate if interest groups are involved.This suggests that citizens value the benefits that interest groups can provide to the process, including the expertise and voices of those affected by the policy, and are not necessarily sceptical of organized interests and their role in politics per se.
However, in line with our expectations, we show that citizens care strongly about (in)equality in interest representation.They perceive policy-making processes as less legitimate if some key stakeholders are prioritized over others in consultations.At the same time, not all types of biases are equally problematic.Citizens tend to be less concerned when cause groups are numerically over-represented or attain their policy preferences at the expense of business interests.We show that, to a large degree, these views can be explained by citizens' beliefs that business groups have more economic resources than cause groups and are simultaneously less representative of society.Thus, citizens generally share the notion of the interest group literature that cause groups defend broader societal interests, while business groups speak for narrower constituencies (Flöthe and Rasmussen 2019).
At the same time, there is also considerable variation in how individual citizens rate the resources and representativeness of the two types of groups.Individuals who perceive their levels to be equal assessed the legitimacy of processes favouring either of the group types very similarly.Overall, these findings indicate that judgements of bias are primarily related to how citizens perceive the resources and representativeness of different groups, rather than group type per se.For some interest groups this might be good news, as it suggests that they should ultimately be able to shape how citizens judge their involvement in policy making if they manage to engage in image building and increase their representativeness in the eyes of citizens.At the same time, groups might use such framing strategies to not only boost their own image, but also weaken that of their opponents, for example, by questioning their representativeness or underlining their resource advantages.
Finally, we show that losses in procedural legitimacy due to the unequal involvement of different types of groups cannot be compensated for by adopting decisions that reflect the preferences of under-represented groups.On the other hand, policy makers can actually exacerbate negative assessments of unequal representation by ignoring the views of excluded types of interests.If one group type has more opportunities to make its views heardin this case, business groupscitizens' perceived legitimacy of the process further decreases if the decision reflects the views of that group only.
The implications of this finding reach beyond the realm of interest groups and concern citizens' justice beliefs more generally.Clayton, O'Brien and Piscopo (2019: 127) concluded that '[e]ven in cases in which all-male panels advance feminist policies, citizens report lower average levels of procedural fairness, institutional trust, and acquiescence'.Hayes and Hibbing (2017) observe a similar pattern when black people are numerically under-represented in the process but the decision is in their favour.Our study provides concurring evidence from a different context, suggesting that citizens generally believe that denying a constituency their seat at the decision-making table cannot be remedied by outcomes that satisfy the group's demands.In other words, citizens value descriptive representation independently from substantive representationalso when it comes to interest groups.These findings suggests that decision makers should be cautious when considering whom to involve in the process, as inequalities might reduce citizens' respect for the process and, ultimately, their willingness to accept the decision.Given that the outcome of the process, which may not be clear a priori, can further decrease citizens' acceptance of it, designing the process in a way that is most likely to be considered legitimate by citizens from the start seems all the more crucial.
Our main findings hold in three settings where different cultures and institutions shape the involvement of interest groups in policy making: Germany, the United States and the UK.By studying a process that exists in similar forms across the three countriesoral consultations in parliamentwe are able to show that inequality in group involvement affects citizen perceptions in the three countries similarly.Yet, these similar effects might have different implications for the overall levels of public perceptions of the legitimacy of governance across contexts.This will depend on at least three factors, which can vary between contexts: (1) public perceptions of the economic resources and societal representativeness of different group types; (2) the degree of inequality in interest group involvement in policy-making processes; and (3) the prominence of oral consultations in parliament among the various channels through which interest groups may influence policy.
Such questions may be addressed in future research.Moreover, it should look more closely at citizens' political views and ties to groups, which might affect their judgement of bias in interest representation.Studies across larger numbers of countries or time points will also be able to examine whether citizen reactions to different types of organized interests are affected by the ideology of the government in power.Our study also encourages future research to examine whether the patterns apply to other types of interest groups, policy issues and (in)formal consultation processes, with a view to extending our knowledge about the potential of different stakeholders to strengthen the legitimacy of parliamentary processes and democratic governance more broadly.Supplementary Material.Online appendices are available at: https://doi.org/10.1017/S0007123422000242

Figure 1 .
Figure 1.Introduction to the survey experiment and UK version of vignette on tax cuts for hybrid car producers.

Figure 2 .
Figure 2. Effects of unequal representation and policy attainment on legitimacy.

Figure 3 .
Figure 3. Perceived resources and representativeness of interest group types.

Figure 4 .
Figure 4. Moderating effects of perceptions of group resources on effects of unequal representation on legitimacy perceptions.

Figure 5 .
Figure 5. Effects on legitimacy by numerical representation moderated by policy attainment.Notes: The figure shows selected coefficients (with 95 per cent CIs) of unequal representation (equal representation as reference category) at selected values of policy attainment from estimated interactions of these variables.The full sets of estimates are reported inTable 1, Model 4.

Table 1 .
Ordinary least squares regressions of legitimacy on representation and moderating variables