Conceptualizing consultation approaches: identifying combinations of consultation tools and analyzing their implications for stakeholder diversity

Contemporary governance is increasingly characterized by the consultation of different types of stakeholders, such as interest groups representing economic and citizen interests, as well as public and private institutions, such as public authorities and firms. Previous research has demonstrated that public officials use a variety of tools to involve these actors in policymaking. Yet, we have limited knowledge on how particular consultation approaches relate to stakeholder participation. To what extent do open, closed and hybrid consultation approaches, with the first two, respectively, referring to the use of public and targeted tools, and the third one implying a combination of both of them, relate to the policy engagement of a different set of stakeholders? In this paper, we identify the different tools used by the European Commission to engage stakeholders in policymaking and assess how variation in consultation approaches relates to stakeholder participation via a descriptive and multivariate analysis. We rely on two datasets: a regulatory database that contains detailed information on 41 EU regulations and a stakeholder database that comprises 2617 stakeholders that were involved in these regulations through different consultation tools. Our main finding is that implementing different consultation approaches affects stakeholder diversity. Specifically, closed consultation approaches lead to a lower level of business dominance than hybrid approaches that combine open and targeted consultation tools.


Introduction
Academics and public officials agree that insight into systems of stakeholder consultation is imperative to understanding dynamics and outcomes of public governance. Scholars have highlighted that consultation practices affect political participation and shape the capacity of governments to solve societal problems in an effective and legitimate way (Bingham et al. 2005). Likewise, supranational institutions such as the OECD and the European Commission have underlined the importance of stakeholder consultation to collect information for evidence-based policymaking and to ensure that regulations are aligned with the public interest (European Commission 2017a;OECD 2012). While the engagement of a variety of stakeholders is considered an important characteristic of good governance, there are also possible instrumental benefits. As argued by Nabatchi (2012, p. 704), if public officials involve a more diverse set of stakeholders, "they give voice to multiple perspectives and different interests, allowing for more thoughtful decisions that take a broader view of those who will benefit or be harmed by an action (Beierle and Cayford 2002;Sirianni 2009)". From the perspective of public officials, inclusive participation processes also generate other benefits, such as the potential to "generate better projects and policies, secure buy-in for decisions, and limit delays, mistakes and lawsuits (Burby 2003)," especially if these processes become more inclusive (Bryson et al. 2013, p. 28; see also, Feldman and Quick 2009).
To capture the possible benefits of inclusive consultation processes, public officials can use and combine a diverse set of consultation tools when developing a new legislation. The literature on (public) participation processes (e.g., Bryson et al. 2013;Fung 2006;Nabatchi 2012), for instance, emphasizes that "specific stakeholders may be involved in different ways at different steps or phases of the processes" and that particular types of stakeholder engagement, such as an approach focused on informing versus one of involving, require distinct tools (Bryson et al. 2013: 27). Furthermore, internal documents of public authorities often state that different consultation tools are used for specific purposes, thus hinting at the possible strategic or instrumental nature of these processes. For instance, the European Commission indicates that online consultations enable them to reach "a broad range and large number of stakeholders," whereas, for instance, expert groups and workshops are better suited to tap expertise, collect more detailed input, cover information gaps and make sure that underserved groups are not excluded (European Commission 2017a, p. 385). This statement also indicates that public authorities aim to ensure the engagement of a broad audience and are well aware of the risk to overlook certain constituencies.
Notwithstanding this awareness, one recurring concern regarding the consultation of stakeholders is the dominance of business interests, as has been demonstrated in various recent studies at the national, EU and international level (Binderkrantz et al. 2015;Fraussen et al. 2015;Hanegraaff and Berkhout 2018;Hanegraaff et al. 2016;Yackee and Yackee 2006;Yackee 2015). Particularly in administrative venues, business interests appear to demonstrate higher levels of engagement compared to citizen interests. It is often argued that bureaucrats mainly require technical information and that business groups are generally better equipped to provide this kind of policy input. For instance, Yackee and Yackee (2006) clarify that business groups can rely on their own technical data as well as expert insights from members, while they also have more capacity to rely on the services of lawyers and commercial lobbyists, for instance, to turn their particular demand in concrete policy proposals. A high level of business dominance might be problematic, as it may imply that organizations representing other constituencies with an important stake in a policy issue and capable of providing relevant (and different types of) viewpoints and expertise are to a (much) lesser extent being heard by policymakers.
Importantly, these studies often only consider one particular consultation tool. To more accurately assess how consultation practices shape stakeholder diversity, we need to consider how the combination of different stakeholder tools within a similar policy process affects stakeholder diversity. In that regard, this paper makes two important and innovative contributions to the existing literature. First, it conceptualizes and empirically examines the occurrence of different consultation approaches at the European Union (EU) level, focusing on the policy formulation stage. Rather than studying one particular form of consultation, we focus on the use of different consultation tools, and specifically the way in which they are combined, which we refer to as a "consultation approach." Second, we assess the relation between consultation approaches and stakeholder diversity, focusing on business dominance. Much previous work has addressed these questions by focusing on variation in stakeholder engagement and business dominance via a single and specific consultation tool, such as expert groups Sverdrup 2008, 2011;Rasmussen and Gross 2015) or online consultation (Quittkat 2011;Quittkat and Kotzian 2011;Røed and Wøien Hansen 2018). These studies, however, do not consider the possible implications of combining several consultation tools (but see Pedersen et al. 2015), even though that is a common practice in specific policy processes. We focus on the numerical dominance of business interests as it enables us to distinguish between organized stakeholders who advocate for distinct constituencies. Business dominance therefore provides a useful quantitative indicator of the extent which policymakers manage to reach out to different types of constituencies and consider various viewpoints when developing policy proposals.
In the next section of this paper, we clarify how a focus on consultation approaches enables a more comprehensive understanding of stakeholder consultation, and why such a perspective is both valuable and necessary to fully understand how the choice and combination of consultation tools shapes stakeholder participation. Subsequently, our research design clarifies our focus on stakeholder consultation in regulatory governance at the EU level. Here, we introduce two new datasets: a regulatory database and a stakeholder database, and explain how we conceptualize and operationalize our independent variable (i.e., consultation approaches), the dependent variable (i.e., business dominance), as well as three key issue characteristics that we include as control variables as it is commonly assumed in the interest group literature that the particular nature of a policy issues shapes the mobilization and hence also diversity of stakeholders (salience, political uncertainty and policy domain). The analysis first provides a descriptive overview of the variety of consultation approaches used by the Commission. The second part of the analysis includes a multivariate analysis of how these distinct approaches shape actor diversity. We find that regulatory issues that are characterized by a hybrid consultation approach are significantly more dominated by business interests, compared to those issues characterized by closed consultation approaches. The concluding section of the paper discusses our key findings and their broader implications, clarifies the limitations of our approach and formulates suggestions for future research.

Consultation approaches: a framework to examine stakeholder engagement
Research has generated mixed findings when studying the effects of consultation tools on stakeholder engagement and the dominance of particular (economic) interests. While some research has highlighted how consultation regimes reduce inequalities between policy insiders and policy outsiders (Bunea 2017), other work has demonstrated that rather than reduce inequalities in the political representation of different societal interests, consultation practices often reinforce bias as public officials are more likely to consult with established actors and organizations (Coen 2009;Eising 2007;. Rather than involving a more diverse and representative set of stakeholders, participatory practices "often end up including the 'usual suspects,' people who are easily recruited, vocal, and reasonably comfortable in public arenas" (Bryson et al. 2013, p. 29). Consultation tools are often biased toward similar organizations that often advocate for business interests. Consequently, consultation processes frequently act as a "primary mechanism of policy reproduction," as they result in policy outcomes that reflect the status quo (Howlett and Cashore 2009).
To more accurately assess the effects of consultation tools on stakeholder diversity, this paper develops a novel framework that accounts for the combination of tools. Specifically, we complement earlier work on policy consultation by examining how variation in the combination of consultation tools, that is, the use of distinct consultation approaches, relates to the engagement of a different set of stakeholders. In this way, we acknowledge that these consultation tools generally are not mutually exclusive options, and consider their joint effect on the policy engagement of different stakeholders.
This study focuses on formalized consultation tools, and examines the engagement of organized stakeholders, which we define as any type of organization that has a stake in a particular policy issues, such as firms, business associations, NGOs, civil society organizations and public authorities. 1 This focus on organized stakeholders implies that we exclude individuals. In this way, our contribution primarily aims to clarify how consultation approaches could remedy or reinforce a business bias among organized stakeholders.
Our conceptualization of consultation approaches emphasizes their inclusive nature, a feature that is key to improve our understanding of how variation in consultation approaches relates to stakeholder participation and analyzing the possible democratic implications (see also Coen and Katsaitis 2018; and see Rasmussen and Toshkov 2013 for a similar approach, assessing the impact of different forms of consultation on legislative duration). Specifically, we distinguish three types of consultation approaches, depending on the specific combination of stakeholder tools within a specific policy process: open, closed and hybrid approaches ( Table 1).
An open approach involves the utilization of tools that provide unlimited "self-selected" involvement to everybody who wishes to contribute-from private citizens to interest groups, firms and public institutions. The most common tool is a (web-based) consultation questionnaire. The goal of an open consultation approach is mostly to acquire input from a broad range of stakeholders as regards the issue at stake, such as information on possible solutions for societal problems or insight into the impact of particular policy measures. A closed consultation approach is rather different in nature. It involves tools such as expert groups, workshops and seminars, or combinations of these. In closed or targeted consultation approaches, public officials address specific and well-defined stakeholder groups and invite them to participate. Hybrid consultation approaches relate to consultation processes that make use of both open and closed consultation tools.
As these three different consultation approaches-open, closed and hybrid-vary in their inclusive nature, we expect that the application of a particular consultation approach will have important implications for the dominance of business interests within the set of engaged stakeholders. Specifically, considering the absence of gatekeepers and the objective of policymakers to reach as broad an audience as possible, and provide equal participation opportunities for all interested actors to foster input legitimacy, one might assume that open consultation approaches generally attract a set of stakeholders that is characterized by higher levels of diversity, compared to closed consultation approaches. However, as business interests face less collective action problems and often are politically organized in greater numbers (Olson 1965), they might also dominate open consultation approaches (e.g., Hanegraaff et al. 2011).
In closed consultation approaches, stakeholders are selected or invited, a process that could have two rather different implications in terms of diversity. First, if policymakers are keen to involve different types of interests and, for instance, seek to ensure that both environmental and economic concerns are taken into account, the set of engaged stakeholders might be relatively diverse. Second, if the aim is to fine-tune details of policy initiatives, only stakeholders who have similar preferences or significant expertise might be invited to contribute, which could reduce stakeholder diversity. While this approach limits the capacity to identify multiple policy dimensions and understand policy issues from different perspectives (Bryson et al. 2013;Fung 2006;Nabatchi 2012), it reduces the amount of (potential) conflict. Considering these contrasting expectations, it is difficult to draw clear expectations on how a hybrid approach, that combines closed and open consultation approaches, will affect the set of stakeholders. In the following section on our research design and data, we clarify the consultation tools that we focus on in this paper and explain our operationalization of our central concepts and variables in more detail.

Research design and data
In this section, we present the datasets used in this paper and discuss the operationalization of our independent variable (i.e., consultation approach), dependent variable (i.e., business dominance) and control variables. Before moving to the datasets, we clarify our focus on the EU and provide more background on the consultation processes of the European Commission. Open approach Approach that only includes the use of (public) consultation tools, such as online consultations, in which all interested stakeholders can get involved Closed approach Approach that only includes the use of targeted consultation tools, such as expert groups and workshops, where stakeholders are granted access Hybrid approach A mix of open and closed approaches, for instance, online consultation and expert groups

The case: European Commission consultation process
We use the EU as a case to empirically explore the occurrence of different consultation approaches and assess their relationship with business dominance as an important indicator of stakeholder diversity. More specifically, we focus on the European Commission, as this institution has exclusive right to initiate legislation. During the formative stage, that is before the Commission issues a legislative proposal that will subsequently be discussed in the European Parliament and the Council of the European Union, public officials within the Commission consult internal and external stakeholders to obtain political and expert information about the content and possible implications of the proposed legislation.
Since the early 2000s, starting with the White Paper on European Governance (European Commission 2001), the Commission has progressively developed a toolbox to implement consultation processes to effectively interact with stakeholders, leading to "one of the most elaborate and ambitious consultative regimes" (Bunea 2017, p. 47). The Better Regulation agenda has further developed the EU consultation regime by specifying in detail the different consultation tools available at the EU level and setting specific rules and recommendations (European Commission 2017b). A key objective is to realize a more balanced engagement of different societal stakeholders, to ensure public officials obtain information and insights from a variety of stakeholders. While the Commission highlights that it is "strongly committed to engaging with stakeholders and citizens throughout the policy cycle and has invested significantly in a range of tools to engage better about its activities," it also acknowledges that "despite the availability of these tools, it is clear that the level of participation has not reached its full potential and some stakeholders are still unwilling or able to engage" (European Commission 2017b, p. 13).
Public officials of the Commission have certain degree of leeway in choosing their consultation approach. That is, they can complement different types of tools to "engage all relevant stakeholders and to target potential information gaps" (European Commission 2017a, p. 391). This statement clearly indicates that the Commission is well aware of the possible limitation of particular consultation tools and therefore recommends a combination of them, rather than reliance on one particular tool. This flexibility might be limited by legal requirements (such as the obligation to organize a public consultation), as well as institutional norms (such as a strong emphasis on expert-driven or evidence-based policymaking, which would stimulate the formalized involvement of experts). Still, even within this framework, public officials make choices regarding which (additional) tools are being used, their respective timing and (possibly) the open or closed nature of these initiatives.
In what follows, we present two datasets that enable us to examine to what extent and how public officials combine different consultation tools and the implications of these choices for stakeholder diversity.

Datasets
We use two datasets collected in the context of a broader research project. First, our regulatory database includes the consultation tools related to 64 regulatory issues (Regulations and Directives) passed between 2015 and 2016 that followed the Ordinary Legislative Procedure-the standard decision-making process used for adopting EU legislation (European Union 2012). We decided to exclude distributive policy issues and focus on issues with a regulatory character that represent the dominant type of legislation passed at the EU level, (see also Majone 1999). 2 We clarify the implications of our focus on EU regulatory issues for the generalization of our findings in the conclusion of this paper.
To account for variation across policy areas (Van Ballaert 2017), the paper focuses on six different policy areas in which the EU has exclusive or shared competences with the member states: (1) Finance, banking, pensions, securities, insurances; (2) State aids, commercial policies; (3) Health; (4) Sustainability, energy, environment; (5) Transport, telecommunications; (6) Agriculture and fisheries.
The second dataset is the stakeholder database, which includes all the actors and organizations who mobilized in the 64 regulations through different consultation tools. Even though the regulatory database contains detailed information on 64 regulatory issues, our focus is on 41 issues that had stakeholder engagement through one of the three consultation tools discussed below (i.e., open consultations, workshops and public hearings, and expert groups). 3 The stakeholder database includes detailed information for each of the 2617 stakeholders that engaged in the policy processes related to the selected regulatory issues. The stakeholders involved in the regulations were manually collected by analyzing several EU documents. More specifically, the Impact Assessment of the Commission and the Commission Proposal were thoroughly reviewed to identify the different consultation tools used and the stakeholders engaged in each of them. If these documents did not provide the detailed list of stakeholders and only mentioned the consultation tool, a search within EU websites was conducted to obtain detailed information about the actors involved. If the list of stakeholders was not publicly available in the website, the responsible DG was contacted to request the list of actors that participated in the consultation tool mentioned in the official documents.
In total, 2617 stakeholders have been identified. This includes interest groups as well as firms, institutions and public authorities, but excludes private citizens and anonymous responses to Commission's consultations. We coded each identified stakeholder following a similar method and codebook as the INTEREURO project, a project designed to examine

Explanatory variable: consultation approaches
The use and combination of different consultation tools is mostly decided by the public officials responsible for the dossier (see European Commission 2017a). The Commission distinguishes eight consultation tools. In alphabetical order, these are: (1) Conferences, public hearings and events; (2) Eurobarometer surveys; (3) Expert groups of the Commission; (4) Focus groups; (5) Interviews; (6) Public consultations; (7) Consultations targeting SME's-SME panel; and (8) Workshops, meetings and seminars. In the context of this paper, we focus on a more limited set of consultation tools. In this study, we consider three different types of consultation tools: (1) "Open (online) consultations," 4 (2) "Conferences, public hearings and events," which have been merged with "Workshops, meetings and seminars" due to their similar formats and (3) "Expert groups of the Commission." 5 These consultation tools are included to be consistent with the objective and scope of the study. Firstly, these are the formal consultation tools that were applied in the regulations included in our sample and for which data were publicly available. Secondly, we focus on those consultation tools that target organized stakeholders, instead of private citizens-such as interviews with experts and Eurobarometer surveys. Table 2 presents the consultation tools included in our study.

Dependent variable: business dominance
To examine how consultation approaches affect the diversity of stakeholders, we focus on the percentage of business organizations in each regulation (Berkhout et al. 2017;. That is, we operationalize stakeholder diversity by considering business dominance, as it enables us to distinguish between organized societal stakeholders who advocate for distinct constituencies. This variable was constructed by dichotomizing a coded variable that categorized stakeholders in one of the following groups: professional associations, business associations, firms, citizen groups, labor unions, institutions, public authorities and others. 6 More specifically, when a stakeholder was categorized as Business (i.e., professional associations, business associations or firms), it was coded as "1," and "0" 4 Since 2017, legislative initiatives that entail an Impact Assessment should include a wide-ranging, open web-based consultation. According to the Commission, this "ensures transparency and accountability," while it also provides "any stakeholder the possibility to contribute" (European Commission 2017a, p. 391). However, the consultations approaches implemented for the set of regulations included in our sample were developed between 2009 and 2014, when the "Better Regulation Toolbox" was not yet approved. Consequently, we expect to find certain degree of variation in the use of this consultation tools across regulations. 5 We do not include interviews and consultations targeting SMEs because they focus on particular target group, nor do we include Eurobarometer survey as they are mainly used to assess the opinions of European citizens, rather than stakeholders on a particular issue. Focus groups were not found among the sampled regulations. Lastly, direct meetings have been excluded as they are not considered among the main formal consultation tools listed by the European Commission. 6 An inter-coder reliability check was run on 100 observations chosen at random from the complete sample list of stakeholders collected. Two coders independently coded six variables for these 100 observations. The Krippendorff's inter-coder reliability test for variable used to construct the variable 'business dominance' was α = 0.83, which confirms the reliability of the data (Krippendorff 2004). when otherwise (Hanegraaff and Berkhout 2018). The final variable, at the level of regulation, has a value that ranges from 0-if no businesses are present-to 1-if only business participated.

Control variables: regulatory issue characteristics
When assessing the effects of consultation approaches on our dependent variable, we also control for the potential effect that particular issue features may have on business dominance. As argued by Klüver et al. (2015), we need to account for the contextual nature of (EU) politics to better understand stakeholder mobilization and diversity. Policy issues differ considerably in the level of attention they receive, the conflict and uncertainty they cause among political actors, and the constituencies that are affected by them. In this study, we therefore include salience, political uncertainty and policy domain as control factors.
The first characteristic considered here is the salience of the regulation, which can be captured in different ways (see Beyers et al. 2018 for a discussion). We focus on media coverage, which is considered a valid and often-used indicator of public salience as it is independent from the perception of the actors that are directly involved in the decision-making process (De Bruycker et al. 2019). Understood as the degree of attention that a regulation attracted in the media, salience is calculated by counting the number of relevant articles published in five EU-wide outlets. In the analyses, we use the log of this count variable, given a skewed distribution with high outliers. The time frame we use ranges from 2 years before the first consultation tool was implemented up to December 31, 2017. 7 Previous research has proved the importance of salience of policy issues in explaining mobilization (for a discussion, see Klüver et al. 2015). In this case, we expect that media salience will increase the visibility of the policy issue among constituents, leading to higher levels of mobilization, particularly among non-economic groups (see De Bruycker et al. 2019, 296).
Secondly, political uncertainty is defined as the political contestation regarding a legislative proposal and is operationalized by quantifying and combining the voting results in the Council of the European Union and the European Parliament regarding the regulations. 8 Following a similar logic as Gray et al. (2004), we assume that higher levels of political uncertainty will trigger the mobilization of a larger and more diverse set of stakeholders. This variable ranges from 0 to 1, where 0 is the lowest level of political uncertainty and 1 is the highest level of political uncertainty.
Lastly, we account for the policy domain of the regulation, as previous research demonstrated considerable variation in consultation approaches across policy areas (e.g., Quittkat 7 The outlets included in the Factiva search are: Financial Times, Politico Europe, Agence Europe, EurActive, EUObserver and European Voice. For each regulation, a search code was developed including key terms related to the regulation and general terms related to European Union decision-making process. Subsequently, human coders examined the articles obtained via the Factiva search and excluded those that were not deemed as relevant. That is, only those that specifically discussed the regulation or the regulatory process leading to the regulation included in our sample were coded as relevant. 8 We calculate our political uncertainty score with the following formula: is the squared percentage of yes-votes within each legislative body involved in the decision-making process (for the EU: (V European parliament ) 2 + (V Council ) 2 ); n is the total number of legislative bodies involved; (1−) is used to transform the output value in a score from 0 to 1 (1 being the highest possible degree of political uncertainty). See Julio and Yook (2012) for a similar approach to capture political uncertainty. and Finke 2008; Rasmussen and Toshkov 2013;Van Ballaert 2017). More precisely, we distinguish regulations developed under "core" economic Directorate-Generals (DGs) of the Commission from those developed in non-economic DGs. 9 In this case, we expect that issues under economic DGs are more likely to have higher incidence of business organizations when compared to those developed under non-economic DGs. Table 6 in "Appendix" provides a summary of the descriptive statistics and the correlations coefficients among all the variables presented above.

Analysis
The analysis proceeds in two steps. Firstly, we present a descriptive overview of different consultation tools used by the Commission and the frequency of open, closed and hybrid consultation approaches. Secondly, we examine the effects of consultation approaches on business dominance, while controlling for variation in issue characteristics. Table 3 presents how the Commission uses and combines the different consultation tools in each of the 41 regulations included in the study. In total, we identified seven distinct combinations of consultation tools, each of them related to one of the three consultation approaches presented in Table 1  exclusively relied on hearings and workshops to obtain input from stakeholders; in nine cases, stakeholders were only involved through expert groups; and four regulations used a combination of closed consultation tools. This means that nineteen regulations exclusively relied on closed consultation approaches. An example in that regard is Regulation 2016/792 on harmonized indices of consumer prices and the house price index, which was discussed in expert meetings composed of representatives of national public authorities. The last three rows of Table 3 present the different combinations of tools that lead to hybrid consultation approaches. That is, they combine open consultation tools with at least one closed consultation tool, or the other way around. One regulation using such an approach was Directive 2016/97 on insurance distribution, which relied on input from stakeholders involved through an open consultation conducted from November 26, 2010, until February 28, 2011, followed by a workshop and an expert group meeting. As noted by the Commission, and illustrated by our findings, there is no one size-size-fits-all approach to stakeholder engagement at the EU level. As Table 3 demonstrates, there is a lot of variation in terms of which consultation approaches are being used to gather input from stakeholders.

Descriptive overview of consultation approaches
In Table 4, we provide an aggregate picture that collapses the different combinations of consultation tools into three consultation approaches: open, closed and hybrid. The two most frequently used consultation approaches are closed and hybrid. Unsurprisingly, compared to an open approach (which is used on only three regulations in our sample), hybrid approaches on average attract more stakeholders, while closed approaches on average engage a more limited set of actors. It is important to note that the standard deviations presented in Table 4 are large relative to their means, indicating highly skewed patterns of interest groups engagement. This is consistent with previous research studying interest group mobilization that showed that although a limited set of policy issues attract a (very) large number of stakeholders, the majority of issues are characterized by the engagement of a rather limited amount of stakeholders (Baumgartner et al. 2009;Halpin and Thomas 2012). Yet, the difference between the mobilization potential of open and hybrid approaches is quite remarkable. While an open approach theoretically enables all interested stakeholders to participate, these findings show that combining open tools with targeted consultation tools results in a larger set of engaged stakeholders. However, whereas the difference between the number of stakeholders involved between hybrid and closed approaches is statistically significant (t(36) = (− 4.893), p < 0.001), a t test does not lead to significant differences when comparing hybrid and open consultation approaches.

Business dominance across consultation approaches
To what extent do different consultation approaches lead to a more diverse set of stakeholders? As specified before, we focus on a single yet key dimension of diversity, i.e., business dominance (cf., Berkhout et al. 2017). The analyses presented below explore the relationship between the two most frequent consultation approaches, namely closed and hybrid, and business dominance. Open approaches have been excluded from this part of the analysis due to the very low number of cases encountered, which could generate biased results. As clarified above, more than 90% of the included cases are characterized by either a "closed" or a "hybrid" consultation approach. It is worth noting that the average proportion of business interests in the 38 regulations using either closed or hybrid approaches is 49%. Yet, as shown in Fig. 1, this distribution varies significantly across consultation approaches. The regulations that used closed consultation approaches (i.e., hearings and workshops and/or expert groups) have significantly lower percentages of business groups (M = 0.362, SD = 0.321), compared to those that implemented hybrid approaches (M = 0.613, SD = 0.122; t(36) = (− 3.173), p = 0.003).
To further explore the relationship between consultation approaches and business bias, we conduct a multivariate analysis. Table 5 presents the results of a fractional regression  model. 10 This is an appropriate method considering that the dependent variable is a percentage that ranges from 0 (i.e., no business presence) to 1 (i.e., complete business dominance) (see Papke and Wooldridge 1996). According to the results, hybrid consultation approaches lead to a significantly higher level of business presence even when controlling for issue characteristics. More specifically, the marginal effects of the model presented in Table 5 indicate that implementing a hybrid approach increases business dominance by 18% (SE = 0.087), compared to regulations that use closed approaches. Remarkably, none of the control variables related to the specific character of the regulations (i.e., salience, political uncertainty and policy domain) seem to be significantly related to business dominance. This finding underlines that variation in stakeholder diversity is primarily driven by variation in consultation approaches.
It is worth highlighting the important role and impact of the inclusion of closed consultation tools within hybrid consultation approaches. For instance, "Directive 2016/97 on insurance distribution," related to requirement for purchasing insurance products, used a hybrid consultation approach by combining an open consultation with an expert group and a workshop. In this case, 113 business organizations out of 210 stakeholders participated through the different consultation tools. More specifically, 59% of the stakeholders were businesses-a similar percentage as the one presented in Fig. 1. However, if we unpack the distribution of stakeholders for each consultation tool, we observe important differences. Whereas 77% of the stakeholders involved through the open consultation were businesses, for the workshop and expert group the proportion of business interests is, respectively, 42% and 12%. A similar example is "Regulation 2016/426 on appliances burning gaseous fuels," which established rules for commercializing burning gaseous fuels appliances and dealt with the energy efficiency of these products. The regulation also relied on a hybrid approach: an open consultation that mostly attracted business interests (i.e., 81% of the stakeholders involved), and an expert group in which only 34% of the 50 organizations involved were businesses. The final percentage of businesses in this regulation was 53%. These examples illustrate that in some cases that apply a hybrid consultation tools, business dominance was decreased by using closed consultation tools, rather than the open consultation tools.
Finally, if we examine closed consultation approaches, we observe a striking dominance of public authorities. As shown in Fig. 2 (which offers a detailed categorization of group type for each type of consultation approach), the lower levels of business organizations in closed approaches are explained by a smaller proportion of firms, yet also a much higher prevalence of public authorities. In this vein, we can state that, particularly among closed consultation tools of the European Commission, "much political 'lobbying' is government lobbying government" (Halpin and Thomas 2012, p. 595). Importantly, though, we find variation in the prevalence of public authorities among closed approaches. For instance, two regulations about fisheries using closed approaches ("Regulation 2016/2094" and "Regulations 2016/1139" establishing a long-term and multiannual plan for certain fisheries) organized large expert groups with an important presence of business. Yet in both cases the percentage was around 50%, which is lower than the incidence of business organizations found in an average case using a hybrid approach. Another example of an issue characterized by a closed approach (including a workshop and an expert groups) is "Regulation 2016/424 on cableway installations"-aimed at providing high levels protection and safety to cableway installations designed to carry persons. In total, 30 different stakeholders participated in the consultation process, of which only 6 (i.e., 20%) were business organizations. More specifically, these 6 business groups participated in the workshops organized by the Commission, while the expert group was exclusively formed by representatives of public authorities. Similarly, only 2 out of the 35 organizations consulted to developed "Directive 2016/2102 on the accessibility of the websites and mobile applications of public sector bodies" were business groups. Again, a workshop and an expert group were organized to consult stakeholders when developing this directive and the two business groups participated in the workshop, whereas the expert groups were exclusively composed of public authorities and institutions such as universities, hospitals or schools.
An important difference between the two approaches as depicted in Fig. 2 is the high incidence of individual firms in hybrid approaches, which is mostly explained by their high level of mobilization in public consultation tools. The low entry barriers of public consultations facilitate the involvement of multiple firms that are not that frequently invited in closed approaches. An additional interesting observation derived from Fig. 2 is the limited prevalence of citizen groups in both closed and hybrid approaches. That is, public officials do not engage a more diverse set of actors when applying closed consultation approaches. Instead, they appear to favor the technical knowledge and expertise that is supplied by other public authorities.

Discussion
Our results indicate that public officials mostly use either closed or hybrid consultation approaches and that there is a lot of variation in how consultation tools are being combined in closed and hybrid approaches. Our descriptive overview of how consultation tools are combined does not reveal a clear logic that policymakers follow when making decisions on which consultation approach to apply. On the one hand, this raises questions about the rationale behind these choices, and the apparent absence of clear bureaucratic guidelines on how to combine consultation tools. Yet, the ability to address complex policy problems might not be increased by implementing more streamlined procedures and instead necessitates some "redundancies." This alternative point of view was recently highlighted by Baumgartner and Jones, who argue that we cannot expect a "single ideological structure, nor a single hierarchically controlled bureaucratic process, to generate useful solutions to social problems that we do not even yet fully understand. Only messy, overlapping, entropic information collection processes are likely to do this" (2018, p. 8).
On the other hand, ensuring the participation of a more diverse set of stakeholders might not be the (only) priority of Commission officials. In that sense, as rightly underlined by Van Ballaert, the chosen consultation approach might be part of larger political battle and strategic considerations from policymakers (that might vary from issue to issue) and could also inform the chosen consultation approach. For instance, he clarifies that the Commission might also be "using recommendations from online consultation to pressure stakeholders thereafter in other fora or by granting privileged access to supportive actors" (Van Ballaert 2017, 13-14).
Importantly, our results suggest that to avoid business dominance it is necessary to strategically select certain consultation approaches. More specifically, our study shows that approaches combining targeted and open tools do not necessarily lower the dominance of business interests. On the contrary, our results indicate that closed approaches are significantly less dominated by business interests. Hence, from a normative point of view, it is important to highlight that closed consultation approaches, which entail higher levels of interaction and discussion between public officials and stakeholders (Birnbaum et al. 2015), are characterized by lower presence of business organizations. This does not necessarily imply that closed approaches are more "balanced" in nature compared to hybrid ones. Rather, while hybrid approaches are often biased in favor of business organizations, closed ones are frequently dominated by public authorities, as demonstrated in Fig. 2.
Our findings also imply that public officials have a more active role to play. They clearly have some leeway in deciding which tools will be combined, which as we demonstrated has important implications for the type of stakeholders that become involved, and the societal voices that are being heard. Most importantly, our results suggest that a more inclusive approach requires an active role of public officials in carefully applying and combining different consultation tools. While such combinations require a substantial investment in time and resources of both public officials and stakeholders, the engagement of a broader and more diverse set of non-state actors could lead to more effective and legitimate policy outcomes, a result that would make such efforts very worthwhile from a democratic perspective. Yet, the participation of a broader set of stakeholders could also lead to increased conflicts and more disputed policy outcomes. Future research would surely benefit from examining these relations in more detail.
As for stakeholders, it raises questions about the extent to which they should participate in all these different tools or rather focus their attention on one particular consultation channel (because of strategic considerations in order to be more influential, or as they do not enjoy access to other channels). In that regard, it should be noted that although academics often studied consultation tools as channels for influence (e.g., Judge and Thomson 2019), they are not the only nor the most important or effective one (Lundberg and Hysing 2016). More fundamentally, we concur with Baldwin that the value of consultation processes extends beyond avenues for influencing public policy. Specifically, she argues for a broader conceptualization of (possible) influence, stating that "stakeholder influence includes not only measurable change to policy decisions but also regular participation and the ability to raise issues and participate in deliberation with decision makers" (Baldwin 2018, p. 2).

Conclusion
This paper conceptualizes and analyzes the occurrence of consultation approaches and their effects on the diversity of stakeholder engagement. To so identify variation in consultation approaches, we have considered different consultation tools and examined the way they are combined by public officials. This comprehensive approach to understanding the relationship between consultation approaches and stakeholder participation has important normative and practical implications. As stated by Bunea (2017, p. 46) "Consultations constitute a direct communication link between decision makers and affected actors and represent an important channel through which policy feedback is received in the policy process about the feasibility of policy choices, the legitimacy of adopted measures and the potential challenges in policy implementation ." A better understanding of these processes is highly important in a context where public officials are increasingly dependent on stakeholders to gain information and expertise, and high investments are made in consultation processes in order to improve the effectiveness of policy measures and reinforce the legitimacy of public institutions (Craft and Halligan 2017; van den Berg 2017). Considering this trend and these objectives, it seems important that consultations do not reinforce existing inequalities in the political representation of societal stakeholders. Moreover, in practical terms, our findings shed new light on the actual implications of using different consultation approaches for the set of participating stakeholders and in this way provide a unique contribution to the existing literature and valuable insights for public officials.
Specifically, we find that hybrid approaches, that combine open and targeted consultation tools, lead to a significantly higher dominance of business interests among organized stakeholders, compared to closed consultation approaches, which are mostly dominated by public authorities. The dominance of businesses and public authorities in consultation approaches has important effects on the legitimacy of the Commission, as we find a very limited participation of citizen groups, whose participation in the process may "by itself, be of normative democratic value" (Yackee and Yackee 2006, 137). Furthermore, their higher level of mobilization might reduce the diversity of viewpoints and expertise provided to policymakers, and ultimately result in policy decisions that are less attentive to the public interest (Yackee and Yackee 2006, 137). Yet, we concur with Klüver that "the sheer number of groups does not necessarily say anything about whether a political system is biased toward a certain group type" (Lowery et al. 2015(Lowery et al. , 1219. Thus, additional research is necessary to examine whether business dominance in hybrid consultation also implies higher levels of policy influence. The interpretation and generalization of the findings are subject to four important limitations. Firstly, due to our focus on the consultation approaches used by the European Commission, studies focusing on national or subnational policy venues might lead to other results given different consultation approaches. Importantly, many OECD countries implement consultation tools that are rather similar to the ones considered in this paper (Rodrigo and Amo 2006). In that regard, we believe our framework can also be applied at the national level or subnational level and encourage work that identifies and compares the occurrence of the different consultation approaches in multiple countries, and assesses their implications for stakeholder diversity.
Secondly, we focus on regulatory issues. Due to the technical and complex nature of many of these issues, they tend to attract a higher proportion of business actors, even though this varies considerably across the regulations we studied. Future research would benefit from including (re)distributive issues as well, in order to examine whether there are significant differences for these two types of policies. For instance, even if EU policymaking is highly technical and public officials especially look for policy expertise, it seems likely that distributive issues will lead to the engagement of more citizen groups, as well as more public authorities (e.g., Beyers et al. 2015), and are therefore characterized by a lower level of business dominance.
Thirdly, we have focused our attention on organized stakeholders and excluded individual citizens. While individuals participate in consultation processes at the EU level, it is very difficult to establish which specific constituency or societal interests these people represent or advocate for, as they might provide input on a scientific basis (e.g., as academic expert), a citizen concerned about a particular cause, or a professional who has specific economic stakes and interests in the issue at the hand. Still, future work would benefit from including both organized societal stakeholders and citizens who participate on an individual basis, especially considering that the EU and several national governments aim to increase the involvement of citizens in policymaking (e.g., Hendriks and Lees-Marshment 2019).
Finally, we acknowledge that this is a first and partly exploratory study based on the consultation approaches used in the policy formulation stage for a limited set of regulations. This is a suitable approach for demonstrating the value of analyzing consultation approaches through a novel and more comprehensive conceptual framework. Future work is necessary to examine whether our findings vary across stages of policy cycle, between government departments, or across political systems. In addition, to further improve our understanding of the relationship between issue-level features, consultation approaches and stakeholder participation, a greater variety of consultation tools might be included, ideally with large-n datasets that consider an even longer period, fully aligned with the slow pace of the legislative process. At the same time, to gain insight into the consultation choices made by public officials, and the extent to which they are aware and act upon the trade-offs of prioritizing certain consultation approaches, indepth case studies, surveys and experiments are equally important, as they will provide more insight into considerations public officials make when selecting one consultation approach over another. While such choices will be shaped by habits and routines, specific objectives could also play a part here. One can imagine that different tools are used or prioritized and combined when the aim is to gather expertise or establish legitimacy, or when the goal is to open up the process or rather limit the engagement with stakeholder to a narrower set of actors given the technical nature of the policy issue.
In conclusion, our central argument here is relatively simple but highly significant. If we aim to fully capture and comprehend processes of stakeholder engagement, we should not limit our attention to one specific tool, but rather consider all the different tools that have been applied by public officials in the context of singular policy process. If we instead only study one tool, our assessment of important questions, such as stakeholder bias and corporate capture, is likely to be inaccurate, as it is based on a partial picture of the stakeholder engagement process. Furthermore, if our standard assumption is that within a singular policy process multiple tools will be applied, this has important implications for studying the representation of societal stakeholders and the role of public officials.