Trade Unions in Multi-Stakeholder Initiatives : What Shapes Their Participation ?

There is a growing concern about the extent to which multi-stakeholder initiatives (MSIs), designed to improve social and environmental sustainability in global supply chains, give a meaningful voice to less powerful stakeholders. Trade unions are one particular civil society group whose participation in MSIs has received little scholarly attention so far. The objective of this paper is to examine the determinants that enable and constrain trade union participation in MSIs. Based on interviews, focus groups, observations and document analysis we determine local trade union participation in three MSIs, operating at company, national and transnational level respectively, in the Costa Rican pineapple industry. To explain the limited encountered trade union participation, an analytical framework is developed combining structural and agency dimensions, namely the MSI design and trade union’s power resources. The findings show shortcomings in the representativeness, procedural fairness and consensual orientation in the design and implementation of the MSIs. These are, however, not sufficient to explain weak trade union participation as trade union power resources also have an influence. Strong network embeddedness and improved infrastructural resources had a positive effect, whereas the lack of internal solidarity and unfavourable narrative resources constrained the unions’ participation.


Introduction
Multi-stakeholder initiatives (MSIs) aiming at improving environmental and social sustainability are omnipresent throughout different supply chains, from the Roundtable on Sustainable Palm Oil to the Alliance for Bangladesh Worker Safety in textiles.This form of governance is supposed to have greater legitimacy because they involve a diversity of stakeholders, including civil society, public and private actors [1][2][3][4].Most research on MSIs has examined their functioning and legitimacy [5][6][7], whereas only few studies assessed the participatory aspects of MSIs [8][9][10].Although MSIs aim to be inclusive and to reach the deliberative ideal where arguments overcome power dynamics, there is a considerable risk that existing power asymmetries are reproduced and that only powerful actors determine the course of action [11].This point of critique has been raised in several studies concluding that the voices of less powerful actors, such as small farmers or local communities, are often not heard in MSIs [1,12].
Trade unions, independent workers' organisations established through the principles of Freedom of Association (see ILO Convention No. 87 (1948)), are one particular civil society group whose involvement in MSIs received little scholarly attention so far.They are, however, relevant and legitimate participants when social concerns, especially labour rights issues, are considered in the initiatives.
Studies have demonstrated that certain procedures and mechanisms of MSIs can allow for more equal participation of all actors, in particular those that are traditionally less powerful [7,13,14].For example, an impartial facilitator who keeps dominant participants in check, encourages less vocal actors to share their opinion and maintains positive group dynamics [15].Other participatory procedures include setting clear goals and rules, providing access to information, working in small groups to build trust, using adequate materials adapted to the educational level and cultural background of (illiterate) participants, developing the technical capability of participants to meaningfully engage in the process, covering costs of participation and conducting stakeholder analysis to identify those relevant to the decision making process concerned.
The objective of this paper is to assess trade union participation in MSIs and explain the factors that enable and constrain their participation.Our contribution lies in a combination of two explanatory factors.First, the design and implementation of MSIs is analysed allowing an assessment of their deliberative potential.Second, the power resources of the relevant trade unions are described as they clarify the capacity trade unions have to participate in an MSI.By combining the importance of structure (i.e., MSI design) and agency (i.e., trade union power resources) in our analytical framework, we aim to contribute to a deeper understanding of the participatory processes of MSIs.
Trade union participation is examined through a within-case analysis of three MSIs existing in the European Union (EU)-Costa Rica pineapple supply chain.Each MSI operates at a different governance level, namely at company, national and transnational level.These are respectively the Ethical Trading Initiative (ETI), the National Platform for the responsible production and trade of pineapples (hereafter Platform) and the Civil Society Meetings (CSMs) organised within the EU-Central American Association Agreement.Based on this comprehensive overview, we found that the design and implementation of the MSIs did not enable notable substantial participation and that the union power resources also played an important role.Here, strong network embeddedness and better infrastructural resources had a positive impact, whereas fragmented internal solidarity and unfavourable narrative resources constrained trade union's participation.As such, the analysis confirms business domination and weak bargaining power of unions in the MSIs.
The paper is structured as follows.The rise of multi-stakeholderism is defined in the context of global labour governance and the basic concepts of deliberative governance are explored.Drawing on existing criteria of input legitimacy and power resource literature, an analytical framework is developed.Next, the research context, case study approach and methodology are presented before coming to the empirical findings.Here, trade union participation in the three MSIs is described and the explanatory factors-MSI design and implementation and trade union power resources-are analysed for each initiative.In the discussion and conclusion, we interpret the findings, make recommendations for improving MSIs and suggest avenues for further research.

Changing Labour Governance Landscape
Globalisation has brought about two trends worth considering in the light of labour rights protection.
First, a shift in global labour governance occurred in which labour rights are regulated through a combination of public (e.g., labour laws, international conventions, soft law initiatives), private (e.g., voluntary standards, codes of conducts) and hybrid (i.e., combination of both private and public initiatives) forms of governance.These new forms often seek the inclusion of non-state actors.Labour rights were historically dealt with at governmental level, as states developed labour law in the 20th century to secure justice in employment relations [16].Through domestic labour law and the participation to dedicated international organisations such as the International Labour Organisation (ILO), states have traditionally been the drivers of labour regulation.
The expansion of supply chains around the globe beyond one state's jurisdiction gave rise to a number of governance deficits, which neither the domestic nor international institutions have been capable of governing appropriately [17].Even though Gereffi and Mayer refer to broad societal issues, this deficit has also had an impact on the governance of labour rights.As a result, the governance landscape has been moving away from the traditional regulatory role of the state to the inclusion of non-state actors in policy processes [4,[18][19][20].In doing so, non-state actors were involved both to more effectively address complex cross-border issues and to increase the legitimacy of global governance.Non-state actors comprise a variety of stakeholders including private entities such as business actors, multinational corporations, non-governmental organisations (NGOs), trade unions and academia.The involvement of these actors through hybrid public-private governance is also termed multi-stakeholder governance [21].Although there is no internationally agreed definition for multi-stakeholderism, it has been broadly conceptualised in the field of 'interactive governance' as: "the complex process through which a plurality of actors with diverging interests interact in order to formulate, promote and achieve common objectives by means of mobilizing, exchanging and deploying a range of ideas, rules and resources."[22] While MSIs bring together multiple actors, they can occur in different forms and sizes such as multi-stakeholder alliances, partnerships, standards and roundtables [1,4,23].MSIs can operate at different scales-from local to transnational-in diverse sectors, regions and topics [24].They can follow different procedural approaches, vary in duration and can evolve over time from a dialogue platform to an independent organisation with a well-established governance structure.There is also a great diversity in the range of purposes that MSIs seek to fulfil.While some MSIs aim to solve specific problems and find a common ground, others promote learning and awareness raising, foster stakeholder dialogue or focus on standard-setting and monitoring [7,25].
Second, although "Everyone has the right to form and to join trade unions for the protection of his interests" as stipulated in the Universal Declaration of Human Rights (Article 23, paragraph 4 (1948)), many workers are not allowed to organise themselves in independent organisations, and trade unionists are persecuted or discriminated against by hostile management [26].In general, trade union bargaining power has weakened and trade union density has declined as a result of global pressures, including increased informality and flexibilisation of labour markets [27,28].Flexible sourcing practices of retailers and brands have increased pressures on suppliers across the globe.These demand pressures are commonly transferred onto the weakest actors at the bottom of the chain, namely the workers, who need to cope with insecure contracts, low wages and excessive overtime [29,30].Despite these challenges to union organisation, integration into global value chains also created new opportunities for local trade unions to connect and build alliances with NGOs and international trade unions [31][32][33].Through these cross-border networks, private standards and brands have been criticised in campaigns addressing violations of workers' rights at supplier sites.
These two trends, proliferation of MSIs and weak(ening) of trade unions in producing countries, are relevant when reflecting on the potential and limits of the prevailing labour governance.However, before trying to answer questions on the impact or results of MSIs one must understand the participatory processes existing within them.Therefore, this article focuses on explaining participation.

Deliberation and Participation
The concept of deliberative democracy is often put forward as an appropriate approach to assess new forms of governance, such as MSIs [9,34].Deliberation-careful consideration or discussion and thoughtfully weighing options-is a central feature of MSIs since their outcome is the result of a participatory process [23,35].Indeed, in his book 'Foundations and Frontiers of Deliberative Governance', Dryzek [36] explains how deliberative principles apply to governance networks, such as MSIs.
A first central aspect of deliberative democracy theory is the idea that deliberation promotes a kind of collective communicative power which neutralises coercive forms of power such as domination [37,38].Common reasoning is essential in deliberative governance and deliberation is indispensable for collective decision making.Therefore, Hendricks [37] explains that deliberative procedures should be designed as such that debates are shaped by the 'force of the better argument' and not by the most powerful or dominant actor.This entails that existing power asymmetries are diminished or even neutralised during the debates and that the outcome of the deliberative process accommodates or balances the participating interests.
Evaluations on MSIs reaching this deliberative ideal vary.Positive assessments of MSIs demonstrate how powerless actors can express their voice and successfully manage to influence decision making in their favour, whereas more critical assessments point to the failure of MSIs to redress existing power imbalances, leading to uneven participation and outcomes that do not meet the needs of less powerful actors, such as small farmers and actors from the Global South [1,12,35].In general, authors are rather critical of the optimism surrounding MSIs.
A second fundamental element of deliberative governance is participation, as it is the precondition for any other development in the process.Moreover, the involvement of all affected stakeholders in the deliberation process is considered a source of legitimacy, assuming actors can equally share their opinion, concerns, interests and knowledge [8,23].This source of legitimacy is often referred to in terms of input legitimacy, which addresses the question of who is entitled to make decisions and who is to be represented in the decision-making process [39].Similarly, according to Dryzek [40], the deliberative quality of MSIs depends on, among others factors, inclusiveness.
Utting [12] found that trade union involvement in MSIs varies considerably from little or no formal involvement to significant and more extensive participation.Indeed, when delving into the participation of a stakeholder, it becomes clear that physical participation does not automatically entail that the participant's interest will be taken into account or that they can contribute in a meaningful way to the process and influence the decisions made [14,20,23].In general, two dimensions of participation reappear in multiple studies under different labels.Luttrell [41] labelled the dimensions 'nominal and meaningful participation', referring to the fact that 'physical involvement of marginalised actors and even verbal participation by them, does not guarantee their concerns will be heard'.Similarly, Fransen [4] distinguished between 'surface appearance' and 'actual involvement' of societal interest groups in decision-making.Brem-Wilson [10] talks about 'formal and substantive participation', Dingwerth [42] refers to the 'scope and quality of participation' and Reed [15] discusses different ladders of participation distinguishing degrees of engagement.
In light of deliberative democracy theory, it is necessary to understand the participatory processes in order to grasp the outcome of the process.To be able to assess the potential of MSIs for the improvement of labour rights, it is imperative to understand the participation of trade unions in these MSIs as they represent the voice of the affected stakeholders, namely the workers.The research question addressed in this paper is therefore: what shapes the participation of trade unions in MSIs?Literature on deliberative governance indicates the importance of the design of the participatory processes.In addition, when discussing the feasibility of deliberative governance, Hendrick's [37] recognises the need to look at the capacity of particular groups in civil society, especially powerless groups, to generate deliberation.Hence, these factors are integrated in the analytical framework presented in the next section.

Analytical Framework
This paper aims to contribute to a better understanding of the factors that enable or constrain trade union participation in MSIs.For this purpose, an analytical framework is developed that takes into account both the structure of the MSI (i.e., its design and implementation) as well as the agency of the participant (i.e., its power resources) (Figure 1).Building further on the distinctions in participation introduced above, we contrast 'procedural participation' and 'substantial participation' to describe the observed trade union participation.The former refers to the physical attendance and continuity of participation, whereas the latter specifies the actual contribution in shaping the content of the debate, such as actors expressing their opinions and negotiating between divergent interests [43].

Multi-Stakeholder Initiative Design and Implementation
The design of an MSI and its implementation should be examined as it will clarify whether participation is possible, desirable and effective from the point of view of the stakeholder.If the MSI is not designed to reduce power imbalances, the power asymmetry existing outside the MSI will be reproduced which might in turn influence the possibility, interest and willingness of the weaker actor to participate and consequently also the achievement of deliberation.
Various studies have shown that the set-up of an MSI has an impact on participation, especially if it foresees power-neutralising mechanisms [23], or what Luttrell [41] calls mechanisms to 'level the playing field'.This includes power-sharing rules that allow for equal input [41], establishment of working groups and public consultation [6,9], policies to ensure balanced resourcing [44], clear selection procedures and voting systems avoiding dominance of powerful actors [5,23,41,45], an impartial facilitator [15] and providing access to information, translation services and technology [23,46].
As we will concentrate on the participatory processes within MSIs, our criteria for analysing an MSI's design are derived from existing literature on deliberative democracy and input legitimacy of transnational governance and more specifically, MSIs [2,7,34,47].MSI design will be assessed through three criteria: representativeness, procedural fairness and consensual orientation.
First, representativeness concerns stakeholder selection and processes that guarantee the inclusion of all relevant stakeholders, namely those affected by the issue addressed by the MSI.To assess this criterion we should, therefore, look at the stakeholder selection process [44].In addition, one should also examine whether certain groups are excluded or if the system favours a special category of stakeholders [2].Finally, the categorisation of the stakeholders could also play a role.Bolström and Tamm Hallström [8] explained how differentiating stakeholders in separate categories could potentially constitute a principle of exclusion from crucial decision-making arenas.
Second, procedural fairness stands for measures that diminish or neutralise power differences in decision-making processes by giving each category of stakeholder an equal and valid voice [7].This does not only involve equal voting rights, equal status as members and access to information are also part of this criterion.This is assessed by analysing the decision-making procedures, whether they be explicit formalised rules of procedure or implicit principles.In addition to decision-making, arrangements guaranteeing the feasibility for all stakeholders to use the MSI's participation potential is considered.More concretely, this means funding and capacity building for those groups that cannot afford to participate in the meetings [1,11,48,49].
Third, a consensual orientation is pursued through a culture of cooperation and reasonable disagreement [7].According to the Habermasian ideal [50,51], consensus should be reached through open discussions, where reasoning trumps bargaining, in a non-coercive environment.However, as Mena and Palazzo [7] acknowledge, MSIs gather a multitude of actors with different backgrounds and conflicting objectives.Therefore, these authors deem consensus as highly unlikely and suggest reasonable disagreement.Similarly, Luttrell [41] suggests to acknowledge disparities of power, address sensitive issues head-on, and to discuss the extent to which participants can 'agree to disagree'.

Multi-Stakeholder Initiative Design and Implementation
The design of an MSI and its implementation should be examined as it will clarify whether participation is possible, desirable and effective from the point of view of the stakeholder.If the MSI is not designed to reduce power imbalances, the power asymmetry existing outside the MSI will be reproduced which might in turn influence the possibility, interest and willingness of the weaker actor to participate and consequently also the achievement of deliberation.
Various studies have shown that the set-up of an MSI has an impact on participation, especially if it foresees power-neutralising mechanisms [23], or what Luttrell [41] calls mechanisms to 'level the playing field'.This includes power-sharing rules that allow for equal input [41], establishment of working groups and public consultation [6,9], policies to ensure balanced resourcing [44], clear selection procedures and voting systems avoiding dominance of powerful actors [5,23,41,45], an impartial facilitator [15] and providing access to information, translation services and technology [23,46].
As we will concentrate on the participatory processes within MSIs, our criteria for analysing an MSI's design are derived from existing literature on deliberative democracy and input legitimacy of transnational governance and more specifically, MSIs [2,7,34,47].MSI design will be assessed through three criteria: representativeness, procedural fairness and consensual orientation.
First, representativeness concerns stakeholder selection and processes that guarantee the inclusion of all relevant stakeholders, namely those affected by the issue addressed by the MSI.To assess this criterion we should, therefore, look at the stakeholder selection process [44].In addition, one should also examine whether certain groups are excluded or if the system favours a special category of stakeholders [2].Finally, the categorisation of the stakeholders could also play a role.Bolström and Tamm Hallström [8] explained how differentiating stakeholders in separate categories could potentially constitute a principle of exclusion from crucial decision-making arenas.
Second, procedural fairness stands for measures that diminish or neutralise power differences in decision-making processes by giving each category of stakeholder an equal and valid voice [7].This does not only involve equal voting rights, equal status as members and access to information are also part of this criterion.This is assessed by analysing the decision-making procedures, whether they be explicit formalised rules of procedure or implicit principles.In addition to decision-making, arrangements guaranteeing the feasibility for all stakeholders to use the MSI's participation potential is considered.More concretely, this means funding and capacity building for those groups that cannot afford to participate in the meetings [1,11,48,49].
Third, a consensual orientation is pursued through a culture of cooperation and reasonable disagreement [7].According to the Habermasian ideal [50,51], consensus should be reached through open discussions, where reasoning trumps bargaining, in a non-coercive environment.However, as Mena and Palazzo [7] acknowledge, MSIs gather a multitude of actors with different backgrounds and conflicting objectives.Therefore, these authors deem consensus as highly unlikely and suggest reasonable disagreement.Similarly, Luttrell [41] suggests to acknowledge disparities of power, address sensitive issues head-on, and to discuss the extent to which participants can 'agree to disagree'.
In the context of MSI design, we will look at whether and how mutual agreement is promoted.In addition, the communication attitude of participants is considered.Are these inclined to be constructive towards other participants or rather conflictual?For multi-stakeholder processes to result in a shared initiative towards joint objectives, a constructive attitude is needed from all participants [25].Finally, consensual orientation obviously will depend on the trust that exists between the participants.Brouwer et al. [11] found that for stakeholders to be able to address power dynamics, a basis of trust is needed.Part of the complex context of MSIs is that the weaker actors can have experienced a long history of being excluded and treated poorly, and consequently distrust the MSI if it is dominated by more powerful actors.

Union Power Resources
An underestimated and often overlooked factor in the research on participation of a weaker stakeholder in an MSI is this actor's capacity.To shed light on this aspect we rely on the theoretical notion of power resources that affect the capacity to effectively participate [8] and use the typology developed by Lévesque and Murray [27] to analyse union capacity in particular.Four trade union power resources can be distinguished: internal solidarity, network embeddedness, infrastructural resources and narrative resources.
First, internal solidarity refers to the relationship between union members, the level of engagement of members, the strategies to recruit new members, the communication methods used between union members and leaders, the leadership structure and the level of cohesion and the presence of a collective identity.Second, network embeddedness, or external solidarity, refers to the degree to which unions have horizontal and vertical links with other unions and with community groups and social movements.
Trade unions act at different levels from local to global, cross borders and connect with different actors, providing different opportunities to pursue union objectives.While some unions might be isolated, others have stronger ties to (inter)national unions or other civil society actors.Such ties can be supportive [3], however, collaboration between international NGOs and trade unions can also create tensions [52].Third, narrative resources refer to the range of values and stories about trade unions that provide shared understandings and frame the way union members think and act [27].Brouwer et al. [11] confirm this power resource which is invisible and difficult to change.In some cases, these deeply rooted structures, culture, behaviour and norms can lead to conservative, entrenched positions.Fourth, infrastructural resources refer to the material (money, meeting rooms), human (time, expertise) and organisational resources (use of technologies, training) [27].
We expect that weak levels of these four resources will negatively affect trade union participation in MSIs.

Research Context: Labour Rights Issues in Costa Rica and Its Pineapple Industry
Costa Rica is currently the biggest exporter of fresh pineapples, exporting 90% of fresh pineapples in the world, which represents 8.4% of the country's total exports [53].From 2000 onwards the country's pineapple industry began to expand rapidly, from 11,000 ha to more than 44,000 ha in 2018 [54].According to the Costa Rican Chamber of Pineapple Producers (CANAPEP), the pineapple industry has generated 32,000 jobs directly and over 130,000 jobs indirectly throughout the country.However, the rapid expansion has had negative environmental and social impacts, including those upon working conditions and the protection of labour rights.
Concerning labour rights issues, Costa Rica is notorious for its anti-union culture in the private sector (see infra), where only 2% of the workforce is unionised.Labour struggles were most prevalent in the banana industry, where the level of unionisation dropped from 90% in 1982 to 5% in 1987 after a defamation campaign against trade unions [55].Perhaps the most frequent obstacle to collective labour rights in this country concerns the promotion of solidarist associations (known as solidarismo) and, specifically, the extent to which such associations prevent the development and functioning of effective and independent workers' organisations such as trade unions [56].These solidarist associations are partly financed by management and do not recognise the right to collective bargaining as formulated in ILO Convention No. 98 (1949) [57].While solidarist associations are on the rise, anti-union practices such as discrimination and dismissal because of trade union membership have been repeatedly reported to the ILO Committee of Experts and are partly responsible for the weakening of trade unions across the country [58,59].In the wake of anti-union campaigns and employers' preference for negotiating with solidarist associations, the pineapple industry is characterised by a very low unionisation rate and accordingly no collective bargaining agreement has been established in any pineapple plantation [60,61].

Case Study Selection and Description
The EU-Costa Rica pineapple supply chain is an interesting case because it demonstrates how the expansion of an export crop has affected local communities and workers and raises the issue of the effectiveness of current labour governance mechanisms.Moreover, the industry is confronted with many of the environmental and social challenges that booming export regions face around the world.During our field research, we came across three MSIs in the pineapple industry, each one operating at a different governance level (company, national, transnational).Although the MSIs' objective, scope, duration, origins and functioning differ (see Table 1), they have in common that they aim at gathering the relevant stakeholders to make businesses and trade more sustainable.Instead of a comparative case study, we opted for a comprehensive within-case analysis of three key initiatives allowing for an exhaustive examination of trade union participation in MSIs across the EU-Costa Rica pineapple supply chain, ranging from their involvement in specific company issues to broader trade-related discussions in civil society fora.By analysing three initiatives we intend to create an overall understanding of trade union participation in MSIs throughout one specific supply chain and to identify which constraining or enabling factors they have in common.In what follows, the general characteristics of the three initiatives are described.[62].ETI is governed by a tripartite board (comprising trade unions, NGOs and corporate members) which reviews the performance of companies, can hold them accountable in case of complaints following disciplinary procedures, and provides remedies [63].To examine how unions in suppliers' sites participate in the work of ETI, we do not evaluate the general functioning of ETI but single out one corporate member of ETI, namely Fyffes.This company has recently been accused of labour rights violations in its subsidiaries' pineapple (ANEXCO) and melon (Suragroh) plantations in, respectively, Costa Rica and Honduras [64,65].Fyffes is an Irish importer and distributor of tropical produce which was sold to the Japanese conglomerate Sumitomo in 2017 [66,67].It was mentioned in 2016 in the Make Fruit Fair awareness raising and advocacy campaign calling for "Freedom and fairness for Fyffes workers" [64].According to the campaign, the violations concern a disregard of freedom of association, as Fyffes seems not to recognise independent trade unions.At the time of writing, Fyffes has been suspended from ETI due to a lack of progress on the accused labour right violations in Honduras [68].

National Platform for Responsible Pineapple Production and Trade in Costa Rica
In 2011, the Platform was established with the support of the United Nations Development Programme's (UNDP) Green Commodities Programme and coordinated by the Costa Rican Second Vice-presidency, Ministry of Agriculture and the Ministry of Energy and Environment.It received funding from a Dutch NGO, the Interchurch Organisation for Development Cooperation (ICCO), in the first phase (2011-2014) and from the Sustainable Trade Initiative (IDH), a Dutch public-private initiative, in the second phase to ensure the continuity of the process.The UNDP was the driving force behind the creation of this multi-stakeholder platform and identified a range of environmental and social challenges that should be dealt with by the Costa Rican government in collaboration with relevant stakeholders.
Over the entire period, the Platform gathered about 900 participants from more than 50 organisations from business, NGOs, communities, academia and related national government institutions to improve the sustainability performance of pineapple production through the development of a national strategy with concrete actions [69].The Platform had to develop an action plan determining the responsibilities of the government and industry players [70].For this purpose, it organised four annual plenary meetings, thirty thematic working groups and several panel debates.In 2016, the action plan was finalised and the Costa Rican government adopted it in decree N • 39462.The Platform mechanism did not put the social dimension on equal footing with the environmental and economic considerations, nor was it able to ensure that demands of all parties were equally considered (see infra).

Civil Society Meetings of the European Union-Central America Association Agreement
The new generation EU trade agreements, starting from the EU-Korea trade agreement in 2011, contain chapters on trade and sustainable development.These chapters refer to labour and environmental standards that should be respected in the framework of the agreement.Civil society meetings (CSMs) are created to follow up on, advise and monitor the commitments made in these sustainable development chapters.Even though there is variation in the legal texts establishing these meetings [71], several foundational features reoccur.First, each party agrees to create or consult an independent domestic civil society mechanism (often called a Domestic Advisory Group (DAG)).Second, an annual transnational meeting should be organised.Here, members of the domestic mechanisms and/or other civil society actors meet.Third, some interaction is foreseen between these two meetings and the intergovernmental body (officials of the EU and its trading partner(s)) that meets annually to discuss the implementation of the trade and sustainable development chapter.
The EU and six Central American countries-Costa Rica, El Salvador, Guatemala, Honduras, Nicaragua and Panama-signed the EU-Central American Association Agreement which has been applied since 2013.Each Central American country (should have) assembled its own DAG.The Costa Rican DAG, together with that from Guatemala, is functioning relatively well in the sense that an independent DAG with civil society members actually exists, as other DAGs are characterised by governmental presence and/or a general lack of participants.In general, the meetings have, similar to those organised in the context of other EU trade agreements, experienced a slow start and have had difficulties generating an internal dynamic [72].
Following discussions on labour issues and corporate social responsibility (CSR) during the civil society and governmental meetings, two seminars, in Costa Rica (May 2017) and Guatemala (May 2018), were organised.Both events covered general CSR topics such as responsible value chains, international CSR instruments and sectoral case studies.In addition, the OECD and the ILO gave a workshop on a related issue [73,74].

Data Collection and Methodology
The interdisciplinary research presented in this paper is based on (1) 37 semi-structured interviews with various actors from different stakeholder categories, (2) three focus groups with trade union members, (3) eight nonparticipant observations during CSMs in the framework of the Association Agreement and (4) document analysis.Extensive field research was conducted in Costa Rica (2015-2016) with Costa Rican representatives and in Belgium (2015-2017) with EU representatives.These two rounds of data collection were followed-up by interviews in 2018 to collect additional information on the concerned multi-stakeholder initiatives for this case study (see Table A1 in Appendix A).To protect the identity of the respondents, we aggregated the respondents per region and type of actor.The following combination of letters and numbers were used to indicate to which group a respondent belongs (see Appendix A): Costa Rica is abbreviated by the letters CR, U refers to unions, G for government officials, and B for business representatives.The letters EU represent European officials, EUCS stands for European civil society actors and ILO officials carry the letter I.
In Costa Rica we targeted key informants, categorised in the following three groups to get a comprehensive sample of different stakeholder perspectives: business representatives (such as CANAPEP), trade unions (at national, sectoral and plantation level) and government officials (including the Ministry of External Trade, Agriculture and Labour).Through snowball sampling, we conducted face-to-face expert interviews [75] with representatives of the Costa Rican pineapple unions (6), government (8), business (5) and ILO officials (3).In addition, three focus groups (with respectively 6, 14 and 4 trade union members) were organised to become more acquainted with Costa Rican trade unionism and to gather more factual information.For the perspectives of EU civil society actors and officials involved in the CSMs, we conducted 15 expert interviews with key informants identified from the list of CSM participants as well as officials from the European Commission and the Delegation of the EU to Costa Rica.
An interview guide was developed for each group of respondents covering topics related to the perception and functioning of trade unions in Costa Rica, the challenges to improve labour rights in the pineapple industry, the existing governance mechanisms and regulatory framework, and a set of more specific questions on the MSIs discussed in this study, including the motives for participation, the design and participatory decision-making process and the results.
The document analysis concerns the content of public video footage, press releases, event reports, presentations, email correspondence, participant lists and meeting minutes of the three MSIs.In addition, one of the researchers observed two transnational meetings, two DAG-to-DAG meetings and four EU DAG meetings where she could listen to the discussions and gain insights on the participatory dynamics and methods.This qualitative data was triangulated through interviews with members of the MSIs.
The relevant parts of the interview and observation notes and transcripts were extracted in a qualitative content analysis [76,77].A cross table was constructed for each of the three MSIs where the extracted data was summarised and reformulated in a more general language and structured according to the analytical framework (see Appendix B).Even though this analytical framework draws heavily on existing literature, it was fine-tuned in an inductive manner.This analysis was then transformed in a thick description of trade union participation and a study of the structural and agency factors determining trade union participation in MSIs.
Two important limitations of the data collection need to be acknowledged.First, the presence of the researcher in the room during the CSMs could have potentially influenced the discussions, because the participants feel they are being "watched" (i.e., observer effect [78]).Second, the subgroup of business actors was underrepresented in our sample because of the sensitivity of labour issues and their limited willingness to meet for interviews [79].Concerning the case of Fyffes, ETI declined our interview request and only confirmed the latest status update over email to preserve the confidentiality of their members. .This is worrying because they were exactly the ones behind the anti-union campaign, discrimination, persecution and dismissals.The initiative was, therefore, practically born dead for the union members.Following the pressure exercised by our members, they appointed four facilitators of the union but without adequate material and knowledge to enable facilitation.Their participation was inconstant and in some activities they could practically not participate at all.[ . . .] They did not take union members into account in this capacity-building event, which could have served as a platform for the establishment of real social dialogue.During the capacity building, management impeded and threatened normal participation of trade union members."-Personalcommunication with CRU2

Procedural Participation
The participation of Costa Rican trade unions to ETI occurred indirectly through the support they received from the International Union of Food, Agricultural, Hotel, Restaurant, Catering, Tobacco and Allied Workers' Associations (IUF) and the NGO Bananalink, with whom they had long standing relationships.Bananalink is also a partner of the Make Fruit Fair campaign.Mediation attempts by ETI and the Costa Rican Ministry of Labour to bring the trade union SINTRAPEM (Sindicato Nacional de Trabajadores/as del Sector Privado Empresarial) and management of the subsidiary plantation ANEXCO together around a negotiation table in 2016 failed [80].In response to the urgent action call, ETI conducted a field visit to investigate the allegations and produced an internal report with recommendations.One of those was a capacity-building session for unions and management which took place in November 2016 [EUCS7].ETI financed the event and sent independent consultants to facilitate a dialogue between local management and trade union representatives to discuss the matter.The participation of most representatives was cancelled as they did not receive the permission from the management to be absent to attend the capacity-building session [CRU2].

Substantial Participation
Local unions provided evidence of violations to prepare the campaign and complaint.They determined the direction and did the legwork for the campaign whereas Bananalink supported them [EUCS7].The NGO wrote to Fyffes Chairman in November 2016 asking him to address the issues without response; later they sent a petition letter signed by more than 40,000 people [81].The local union was engaged and communicative, they hosted a visit that helped to gather documentation for the Fyffes campaign.The mediation by ETI allowed unions to express their concerns about the willingness of management to have dialogue and the persistent labour rights violations in the plantations.During the capacity-building event, the trade unionists' input was limited as they were in the minority.Although the local union leader denounced these practices, no further actions were taken by ETI in Costa Rica because the focus had moved to the complaint in Honduras [EUCS7].

Participation of Trade Unions in the Platform
"They [CANAPEP] never wanted us to be present in the Platform.We went to the launch of the platform and after a couple of months they proposed to organise working groups on soil and pesticide application among others, but none of the working groups covered labour issues.We went to the ILO to suggest a working group for the discussion of labour aspects, but CANAPEP did not want to sit together with us, they prefer to sit together with the solidarist associations only.The platform is a lie, they just waste resources."-CRU1

Procedural Participation
When assessing physical attendance over the entire period of the Platform, producers (27%) and government actors (36%) dominated the meetings while unions (1%) and NGOs (9%) were underrepresented [70].Other smaller categories of stakeholders were buyers, international organisations, communities and academics.There was resistance from the producers' side to include trade unions upon which Bananalink pressured the Dutch NGO ICCO, which co-financed the Platform, to insist on trade union participation [EUCS7].However, industry players refused to address any of the trade union issues and ultimately trade unions decided to withdraw their participation.Consequently, they were not involved in the final decision-making process of the action plan and were also not part of the follow-up committee monitoring its implementation [82].

Substantial Participation
Before the Platform, there was hardly space for dialogue between trade unions, business and government to find solutions for the social and environmental problems of pineapple production.Problems were mainly discussed through judicial avenues.Trade unions saw the Platform as a unique opportunity to share their viewpoint with business and government [70, 83,84].A trade union representative of SITRAP (Sindicato de Trabajadores de Plantaciones Agricolas) attempted to put freedom of association and collective bargaining on the table during the first annual meeting and drew a picture of the difficulties that they are facing.However, attempts to facilitate dialogue between government, employers and trade unions to discuss the working conditions in plantations failed because of the irreconcilable differences in opinions among the parties.
In a workshop organised to review the proposed action plan in February 2014-where unions were not present-business representatives requested the modification of the reference to worker's organisations in the action line dealing with national dialogue on labour rights [84].In addition, CANAPEP put an ultimatum to exclude any reference to freedom of association.Industry players argued that the issue of trade unions is not unique to pineapple, and should be addressed at national level through enforcement of existing labour laws.Instead, they suggested to include in the action plan that the government should promote the international recognition of alternative labour organisations (i.e., solidarismo) that, according to them, represent pineapple workers in Costa Rica [83].Consequently, promoting national dialogue on labour rights was replaced by a more general action for promoting dialogue spaces on environmental and social responsibility, omitting explicit actions related to freedom of association [70].

Participation of Trade Unions in Civil Society Meetings in EU-Central America
Trade Agreement "It's a new type of meeting that doesn't exist at national level.The agreement creates a space to discuss the issue of labour rights violations.But it is still a very limited and superficial dialogue.The format does not allow for an integral discussion, the debated topics are secondary, not fundamental.It is nothing more than a dialogue of the deaf.There is no interest in reaching agreement.Everyone simply states his position.That's it.If certain sectors put their veto and don't want to discuss further, the dialogue loses its meaning."-CRU5

Procedural Participation
The official list of the members of the Costa Rican DAG contains seven trade unions.However, only two of them participate actively, namely Central del Movimiento de Trabajadores Costarricenses (CMTC) and Bloque Unitario Sindical y Social Costarricense (BUSSCO) [85] [CRU5].All secretaries-general of the Costa Rican trade union federations were included in the list; however, most of them are not aware of their membership or even of the existence of the DAG [CRU1,CRU5].It must be said that limited trade union participation in the DAGs is not specific to Costa Rica; in all Central American DAGs trade unions are barely represented.
Regarding trade union physical attendance during the CSR seminars, a considerable difference was noted between both seminars.Whereas this stakeholder category was barely represented during the first event in Costa Rica, there were significantly more-mainly Guatemalan-trade unionists present during the second event [EUCS5,EU3].

Substantial Participation
The limited trade union representation in the Costa Rican DAG, and Central American DAGs in general, has severe consequences on the substantive work done as little or no input is given by Central American trade unionists in the domestic meetings.During the transnational meetings Central American (and European) trade unionists have made some denunciations about labour rights violations such as the limited freedom of association or violations in specific companies (e.g., Fyffes).Nevertheless, little is done in response to these statements as the documents summarising the discussions of these meetings, which are presented to the intergovernmental board, remain general, and do not include the input [EUCS11].
In 2017, the members of the different DAGs agreed to collaborate on four themes: CSR, Decent Work, Small and Medium Enterprises and Market Access.The objective was to write a commonly agreed two-pager on each topic and to submit it to the Board at the occasion of the next transnational meeting.Two rapporteurs were appointed for each paper, an EU and Central American DAG member.Although there were exchanges of views from both sides, in the end no Central American trade unions contributed to the content of the documents [EUCS9].Ultimately, the documents were not presented during the next meeting with the Board as two Central American business organisations opposed the content of the documents on Decent Work and CSR shortly before the meeting, even though they had been agreed upon in principle by the Central American rapporteurs and all EU DAG members [EUCS6].
Turning to the CSR seminars, we learned that the organisers (i.e., the European Commission, relevant EU delegations and respectively Costa Rica and Guatemala) had determined not to focus on labour rights violations as such.Instead, broader and more positive issues were put forward to address labour rights (e.g., social protection) [I3].In Guatemala, during the workshop given by the ILO, participants were divided into small roundtable groups.Each table had governmental, business and trade union representatives who discussed several case studies [I3].This resulted in dynamic dialogues between the participants and was evaluated as a positive experience [EU3,I3,EUCS5].
In sum, we find that local trade unions did not participate consistently or directly and were sometimes even excluded from meetings in all MSIs (Table A2 in Appendix B).We note that procedural participation of unions was intermediate in ETI-Fyffes, low in CSMs, and can even be considered very low in the Platform since unions dropped out.Regarding substantial participation, unions were somewhat able to provide information and raise concerns, whereas they were still constrained in the extent to which they could influence decision-making and their concerns were taken into account.In the case of ETI, intermediate substantial participation was observed as they gave more input compared to the low substantial participation in the CSMs.Again, this contrasts with the very low substantial participation found in the Platform where unions were not at all able to provide input and consequently were also not considered in the final decision-making of the action plan.In what follows, we explain what factors led to these low degrees of procedural and substantial participation, based on an assessment of the MSI design and implementation (structural factors) and trade union capacity (agency factors).

Representativeness
To evaluate the representativeness in the three MSIs, their selection procedures were analysed with attention for the exclusion or categorisation of certain stakeholders groups.
First, in the ETI case, ETI members IUF and Bananalink, respectively a global trade union and an NGO, played the role of bridge builder and gatekeeper as only member organisations can raise concerns or file a complaint within this MSI.However, during the implementation of ETI's recommendations, the participation of the local trade unions was left to the local management of the plantation.The local union leader complained about the vague communication on the organisation of the meetings because they received the invitation to join the meeting with ETI only two days in advance [CRU2].This resulted in exclusion and thus low representativeness.Second, in the context of the Platform, unclear selection criteria, active resistance against trade union participation by the business side and a preference for high-level participants resulted in very low representativeness of the unions.Third, the EU-Central America trade agreement specifies that members of the DAG should be independent representative organisations, and that economic, social and environmental stakeholders should be represented in a balanced way.However, the selection of the participants is left to the discretion of the governments, with no clarity on the criteria used.In Costa Rica, the invitation procedure has been faulty as some trade union representatives included in the members' list were not aware of their new role.Commercial interests had already been involved during the negotiation of the trade agreement and businesses had closer ties with the Ministry of External Trade.They were, therefore, better aware of the creation of the DAG.In addition, Costa Rica decided to subdivide its DAG in three separate groups, one for each stakeholder category (business, trade union and environment).This categorisation could potentially isolate less well coordinated actors, such as trade unions, and impede collaboration between the different interests.Trade unions were implicitly excluded from the first CSR seminar as the organisers had framed CSR as a business topic.During the preparation of the second seminar, trade unions were actively invited by the local EU delegations and a Guatemalan CSR association.

Procedural Fairness
This criterion is assessed through the examination of the way decision-making procedures allow for equal opportunities between participants to express their voice and be heard as well as funding and capacity building supporting the participation of weaker actors.
In the ETI case, rules of procedures regarding governance mechanisms and disciplinary measures are formalised for members and the implementation is strongly monitored by ETI members representing local trade unions.However, ETI does not cover the translation of relevant documents on the progress of the case in the language of the affected workers.In order to give the local trade unions the opportunity to be correctly informed, Bananalink dealt with this costly task [EUCS7].Trade unions are exempted from paying ETI membership fees, but other financial support measures for local trade unions are limited.ETI did fund an ad hoc session to be given by consultants in Costa Rica to trade unions which was then obstructed by the local management.Since ETI does not actively promote local trade union participation along with a considerable margin of manoeuvre for local management to disregard ETI's efforts, the procedural fairness within ETI is low.
Even though the Platform was established to design a common action plan through the involvement of numerous stakeholders, no moderator or agreed decision-making procedure was foreseen.In the end, the content of the action plan was reached by consensus.However, by that time, trade unions were no longer involved in the Platform entailing the consensus was agreed upon by likeminded business actors.The absence of an entire stakeholder category was possible because the participation of the business side was considered more essential for the existence of the Platform [CRG4].Funding of the Platform was used for the organisation of workshops, but the topics did not cover labour rights nor were transport costs reimbursed for union representatives coming from the pineapple-producing regions.It can therefore be concluded that no procedural fairness pursued vis-a-vis trade unions and that the procedures were in favour of the pineapple producers.
Turning to procedural fairness in the CSMs, it should be noted that the Central American DAGs have not agreed upon rules of procedures.This entails that there are no clear rules on how decisions are supposed to be taken.Even though in practice decisions have been taken by consensus (see infra), business is in the majority to overrule the voice of other interests.During the transnational meetings all participants can express their concerns by raising their hand and they will be given the floor by the moderator.Since the outset, funding has been a critical point for the CSMs.Whereas the European Economic and Social Committee has taken up the role of secretariat in the EU and the European Commission makes travel funds available for at least one participant per stakeholder category, there is no funding whatsoever foreseen in Central America.This has been criticised heavily by EU and Central American civil society, because travel distances are rather important (within the Central American region or to Brussels) and not-for-profit actors have limited financial resources.To address these shortcomings, the EU has created a three-year project of three million euros to support civil society participation in the implementation of EU trade agreements [86].At the time of writing the implementation of the project has not yet started.Regarding the CSR seminars, more EU funding was available for regional participation in the second event, which had a positive impact on local trade union participation [EU3,I3].

Consensual Orientation
To determine this final criterion, the pursuit of mutual agreement is considered as well as the communication attitude and trust among the different stakeholders.
Following the allegations against Fyffes, ETI facilitated a meeting between the unions and the local management.It was, however, not possible to reach a mutual agreement.In general, the communication attitude of Fyffes was perceived as rather hostile.Due to Fyffes unresponsiveness to the grievance in its Costa Rican pineapple plantation, Bananalink launched a public campaign against practices in Honduras.This was not appreciated by Fyffes, who claimed this was against the code of conduct of ETI members.Fyffes stated that they do not respond to public pressure and they "dug their heels in" concerning the recognition of trade unions [EUCS7].Although building trust relationships is a central element of ETI's approach, trust has been broken both by Bananalink and Fyffes through their communication.In sum, even though consensual orientation is an important objective of ETI, it was impeded by the communication attitude and damaged trust on both sides.
In the Platform, the ILO mediated between trade unions and producers to find mutual agreement between them.This attempt was unsuccessful and trade unions left the Platform.In addition to irreconcilable positions, the negative and aggressive communication attitude, considered 'emotional' language, of trade unions was also part of the issue as they wanted to make denunciations and discuss labour rights violations.This stood in the way of dialogue and the evolution of the Platform as producers claimed this was not the right place for it.Moreover, there was an overall distrust in the neutrality of the Platform and a deep mutual distrust between the trade unions and producers.To conclude, even though there had been a mediation attempt by the ILO, consensual orientation could not be fulfilled due to miscommunication and a lack of trust.
Finally, there is a consensus-based approach in the CSMs.It is, however, uncertain if this approach of mutual agreement can be sustained as experience has shown that business interests do not shy away from vetoing the inclusion of labour proposals in statements from the transnational CSMs [EUCS10] as well as in joint working documents [EUCS6].An internal reflection is being conducted, mainly by the EU side, to assess how collaboration should be continued [EUCS5].Regarding the communication attitude, trade unions have made denunciations at several occasions concerning labour rights violations in specific companies such as Fyffes [EUCS11].These efforts received little support, as other participants claimed that denunciations should not be done during the meetings as the DAGs are not supposed to deal with specific cases.Since the inception of the DAGs, members have been trying to build trust within and among the DAGs.The recent unexpected last minute veto by Central American business representatives against the submission of joint working documents to the Board was a serious blow to this trust.In addition, local trade unions-who have protested against the trade agreement during the negotiations-remain on their guard for co-optation as the CSMs and their participation could legitimise the agreement.In sum, the consensual orientation in the CSMs appeared to be rather vulnerable as mutual agreement is under pressure, there is a mismatch of communication attitudes, and trust has been damaged.Turning to the CSR seminars, both organisers and participants agreed that the format of the workshops invited the participants to collaborate constructively instead of the more traditional conflictive communication [EU3,EUCS5].
In sum, the three criteria explaining trade union participation through the MSI design and implantation score low to very low in each MSI (Table A3 in Appendix B).To have a more comprehensive picture on the enabling and constraining factors of this participation, we should also take more internal aspects of trade unions into account.In the following section the influence of their power resources on their participation will be examined.

Internal Solidarity
In general, plantation unions suffer from weak levels of member engagement, because not all members are willing to sacrifice time during weekends or after work for meetings [CRU4].For example, in the case of ETI, the local union leader is most engaged in following up on complaints as other members do not play a prominent role due to their limited experience with and knowledge of legal procedures.Mobile phones are the main means of communication between members and local leaders, and not all members are literate.There is not a dense network of union representatives in the workplace and members are isolated in different teams spread across the plantation fields.At sectoral level, a major challenge is to foster a collective identity among plantation workers.Although workers' problems are of the same nature, the majority of plantation workers are Nicaraguan migrants which impedes their potential participation and membership to unions.They work to earn an income for their household, do not want to risk losing their job by forming or joining a union to fight their cause and may only be on the plantation on a temporary basis.Unions experience difficulties to mobilise new members to join the union and often lack a clear strategy on how to do so.The negative perception that unions destroy the economy and cause trouble discourages pineapple workers to join unions out of fear for reprisals [I2].

Network Embeddedness
The fragmentation of trade unions in the Costa Rican private sector can be traced back to two events affecting the country's political economy.First, the Communist Party and their labour confederation, who had succeeded in organising the banana workers, were outlawed shortly after the civil war ended in 1948 [55].The loss of legal recognition of the Communist trade union left banana workers, who had had one of the strongest and most militant unions of Costa Rica, divided.This represented a first severe blow to trade unionism in the agricultural industry as the corporations did their utmost best to maintain the situation and prevented the emergence of another powerful union among its workers.Second, during the 1980s, an economic crisis and steep rise of foreign debt resulted in forced neoliberal policies promoting export sectors.This period had detrimental consequences for Costa Rican trade unions and the current anti-union culture stems from this context of deregulation and austerity measures in which private employers created evasive structures and adopted a strong anti-union attitude [87].
In the case of ETI, SINTRAPEM has strong ties with IUF and Bananalink who supported them in preparing the complaint and campaign against Fyffes.They are also connected to the Coordinating Body of Latin American Banana and Agro-industrial Unions (COLSIBA) to share experiences and strengthen their ability to fight violations internationally.In the context of the Platform, local unions did not leverage their ties with international unions.Moreover, the network of pineapple unions is spread over different plantations and decentralised without well-established ties to other unions, for example the stronger public sector unions.Unions act more independently, targeting specific companies with specific issues or demands in isolation.Costa Rican (and Central American) trade union participation during the CSMs is stimulated by EU trade unionists who have, on the one hand, together with EU NGOs, continuously been raising awareness about the CSMs in Central America in order to increase trade union participation [EUCS9].On the other hand, they have repeatedly complained to the European Commission about the lack of Central American trade unions participating in the CSMs, hoping for EU pressure on the Central American governments to stimulate trade union participation.However, it should be noted that when Central American trade unions participate in the CSMs, they are most likely high-level representatives.In Costa Rica, they are representing national federations and consequently disconnected from the realities in the plantations.
The findings show that it has been difficult to build coalitions between local unions, however, trade unions were able to connect to international actors (in the case of ETI and CSMs) and to benefit from capacity-building activities.

Narrative Resources
Given the heritage of labour struggles in the country (see supra), unions follow a defensive narrative in the way they act and think.For example, they used the Platform as a forum to denounce anti-union practices by referring to anecdotes of discrimination and persecution, because there was no direct line of communication between workers and management in the plantations.Unions represent only a very small share of workers due to the anti-union campaign and management support for solidarist associations.They have to compete with the discourse of solidarismo which tries to win members through offering tangible benefits (e.g., credit opportunities or rain jackets) and are more convincing in the eyes of workers.Yet, solidarist associations do not recognise the right to collective bargaining since they are not allowed to negotiate collective agreements by law.
The ideological trade union background also informed their actions in MSIs.For example, in the CSMs, a union decided not to participate in the MSI because it would have interpreted its participation as approving the agreement and feared being co-opted [CRU5].Some unions did not perceive MSIs as valuable channels to achieve their objectives.Unions also questioned the credibility of private voluntary standards and feared that it is 'big business', used as marketing strategy for companies and not to genuinely improve labour rights [CRU2].This stock of negative experiences with certification audits translated into a general mistrust and disinterest towards mediation efforts and dialogue opportunities foreseen in MSIs.

Infrastructural Resources
Human resources for the daily functioning of unions are limited since offices are often run by one person and a secretary.The union leader needs to divide his time between representing workers in court, visiting the Ministry of Labour for mediation in the capital city, attending workers in the field and organising training sessions and meetings with the members [CRU2].
In general, being part of an MSI requires time and preparation, while in many cases unions are not able to send a representative because they have other priorities connected to their grassroots trade union activities.They may also not be familiar with the professional language (often English) used in those MSIs and lack the negotiation or communication skills needed to foster consensus.It is also more interesting for unions to directly negotiate with management than to participate in dialogue platforms with no immediate, concrete results.
Turning to the material resources, the costs of transport to the place where the MSIs are organised, often the capital city or abroad, are too high for unions.While some unions benefit from donor funding, other unions have limited operating resources coming from membership fees.Unions in the field have also limited organisational resources (such as computers, meeting rooms or cars) to communicate and interact with members and other actors.
The analysis of power resources of trade unions (Table A4 in Appendix B) in addition to MSI design and implementation allows for a more complete view of the determinants of trade union participation.

Discussion
Our findings endorse the concern that MSIs are not as inclusive as they aspire or pretend to be.Physical attendance does not necessarily imply participants will be heard.This confirms the relevance of distinguishing between procedural and substantial dimensions of participation when assessing how a (weaker) stakeholder is involved.In what follows, our findings are discussed and used to articulate practical recommendations, relevant for MSI organisers, participants and decision makers alike.
The three criteria assessed for the design and implementation of the MSIs, show how the participatory processes remain far from the deliberative ideal of collective communication and inclusiveness.The three initiatives experience several challenges constraining representativeness, procedural fairness and consensual orientation which explain in part the overall limited participation of trade unions.To achieve a better quality of participation and deliberation, power inequalities among its participants should be addressed by improving MSI design.Recommendations on the design and implementation of MSIs are deduced from our findings and clustered around each criteria.
First, the selection criteria for participation to the MSI should be clearly predefined and ensure the representativeness of the participants.Stakeholders should be identified together with a mapping of potential conflicts, expectations and their power resources [13].These should then be taken into account by a neutral facilitator, for example an impartial secretariat, who can support the participation of a contested stakeholder category.
Second, regarding procedural fairness, clear goals and rules of decision-making should be set from the start and agreed by all stakeholders, stronger and weaker stakeholders alike [88].This is a first exercise in redressing power inequalities and confirmed in other research on MSIs [7,13].It should also be possible to reassess these rules at a later stage.Participants should feel that their input is considered and the use of vetoes that are not open for discussion, especially by powerful actors, should be refrained from.Practically, as also stated in Brown [14], resources should be made available for translation services and transport ensuring the participation of less resourceful stakeholders.Finally, as became clear in the ETI case as well as the CSMs, international actors play an important role as bridge builders pushing for local trade union participation and making their voice heard.This leverage effect has been confirmed in other studies [31].
Third, concerning the consensual orientation in MSIs, we see that building trust between actors with divergent positions is challenging.Therefore, the format of MSIs should recognise existing power differences and facilitate a rapprochement between business and unions, for example through small working groups and capacity-building events as suggested by Reed [15].Yet, reaching a consensual orientation requires a constructive communication attitude and especially willingness from all parties to listen to different positions.An impartial moderator can help to preserve positive group dynamics, control dominant voices and ensure that less powerful actors can give input.If agenda items suggested by weaker participants are deliberately ignored, the credibility of the MSI is threatened and these participants can vote with their feet and withdraw [14].Finally, the organisation of mediation and capacity-building events should be left to, or at least involve, a neutral actor to overcome entrenched positions between conflicting parties such as local management and unions.
In general, MSIs should be interpreted as a learning process and benefit from continuous internal evaluation and reflection on obstacles that impede the engagement of weaker actors.This need for iterative learning is also stressed in adaptive management approaches to long-term participatory processes as suggested by Stringer et al. [89].In addition, our analysis showed how the local context often hampers deliberation and the rebalancing of interests.This was particularly the case in the Platform, where hardly any efforts were made to ensure that weaker actors could give their input and were listened to.Since every MSI is influenced by the (local) context in which it operates, it requires a customised and localised strategy that incorporates feedback from participations and is sensitive to signals showing the need for redesigning participatory procedures or capacity building adapted to weaker stakeholders.Finally, even though implementing changes is time and resource consuming, MSIs can benefit from a more bottom-up involvement of workers and unions, contrary to high-level representation, as has been demonstrated in recent developments in worker-driven social responsibility [88].
The analysis also indicates that the design of an MSI is not all-decisive as insights on trade union power resources also helped to explain their participation.We found that strong network embeddedness and improved infrastructural resources enhanced trade union participation, whereas the lack of internal solidarity and unfavourable narrative resources had a negative effect on their participation.
When looking for ways to boost union's power resources, the most feasible improvements can be found in network embeddedness and infrastructural resources.Strengthening ties with international organisations are an important enabling factor for trade union participation to MSIs.They can demand involvement in MSIs, facilitate meetings, and support local unions' infrastructural resources.Yet, dependency on international alliances can also become a pitfall if the need for better internal union solidarity is neglected and the union relies exclusively on external support.
Even though internal solidarity and narrative resources should not be ignored, changing a highly fragmented trade union landscape, the negative reputation of trade unions and local norms cannot happen overnight [27].Therefore, it is essential that local unions join forces and build alliances or coalitions at sectoral level to deal with societal and managerial counter pressures and more actively engage in the dialogue spaces created by MSIs.
When linking the design and implementation of MSIs to the power resources of trade unions, two relevant interactions, implying practical takeaways, are observed.First, procedural fairness can be increased through network embeddedness and infrastructural resources.For instance, in the ETI case the international partners of the local trade unions strived for their involvement by assisting them in their communication and translating relevant documents.In the CSMs, EU trade unions have been very insistent upon encouraging the participation of their Central American counterpart.In addition, funding is being made available to ensure the participation of less affluent participants.Strong network embeddedness is missing in the Platform, which is also the initiative with the weakest trade union participation.Second, consensual orientation is very difficult to attain in a context of negative narrative resources where historical conflicts, conflictual communication and distrust between the stakeholders impede progress of the MSI.This was obvious in the Platform as well as the CSMs.In general, MSI design could not compensate for deficits concerning internal solidarity and narrative resources as they are deeply entrenched in the domestic context.

Conclusions
This article shed light on the participation of trade unions in MSIs by analysing three initiatives in the EU-Costa Rican pineapple supply chain.In line with previous studies on inclusiveness of weaker actors in MSIs, the three initiatives confirm that trade unions can be procedurally allowed to participate but often fail to contribute substantially.While there was no substantial participation of unions in the case of the Platform, unions could express their position in the ETI-case.The Central American trade unions' substantial participation to the CSMs has been very limited, however, their procedural participation is on the rise in less formalised meetings.
Our main contribution lies in explaining trade union's participation by combining two dimensions: MSIs design and trade union power resources.First, the analysis of the MSIs design and implementation demonstrates that selection procedures can lead to the exclusion of local trade unions (representativeness), decision making rules to enhance equality and funding to enable participation of unions are rare (procedural fairness), and conflictual communication and distrust can also obstruct substantial participation (consensual orientation).Second, the analysis of union power resources highlights, on the one hand, how strong network embeddedness and better infrastructural resources are complementary to the design of MSIs for trade union participation.On the other hand, the lack of internal solidarity and adverse narrative resources of Costa Rican (pineapple) trade unions have a negative effect on their participation.MSI design could not compensate for these deficits as they are embedded in the local context and difficult to change.
This implies for practitioners involved in MSIs that improving the design of MSIs is necessary (see infra for recommendations) but not sufficient to enhance trade union participation.It stands or falls with the ability of unions to mobilize their resources and the willingness and commitment of all participants in the MSI to meaningfully engage in deliberation.Our paper calls for a continuous evaluation of MSI participatory processes (regarding representativeness, procedural fairness and consensual orientation) and strengthening of union power resources (internal solidarity, network embeddedness, narrative and infrastructural resources).
This paper has concentrated on the participatory processes in MSIs, however, additional research on the outcomes of such MSIs, especially concerning the promotion of labour rights, should be conducted.These outcomes should be scrutinised to know whether the MSI has ultimately strengthened weaker actors.Another interesting avenue for further research is the role of public actors in MSIs.Even though MSIs are often considered to be solely private mechanisms, governments are often involved as a creator or sponsor.Additional research on the (potential) influence of public actors, both in importing and exporting countries, would be useful in the assessment of power struggles at play in MSIs.In this light, a cross-country comparison with MSIs operating in labour repressive states embedded in different political economies can give a fuller account of how the institutional and political context influences opportunities for trade union participation in MSIs and how unions and MSIs deal with these context-specific challenges.

Figure 1 .
Figure 1.Analytical framework explaining trade union participation in multi-stakeholder initiatives.

Figure 1 .
Figure 1.Analytical framework explaining trade union participation in multi-stakeholder initiatives.

1 .
Procedural and Substantial Participation of Trade Unions 6.1.1.Participation of Trade Unions in ETI-Fyffes "Six management assistants were assigned as facilitators [for the capacity building event]

Table 1 .
General characteristics of the three MSIs.
5.2.1.The Ethical Trading Initiative and Call for Action against Fyffes ETI is a membership-based MSI bringing together companies, trade unions and NGOs to improve working conditions in global value chains.It has been established in 1998 with the support of the UK government and has developed a Base Code for corporate members to support continuous improvement regarding decent work

Table A4 .
Criteria for union power resources.In general, difficulties mobilizing workers to join union, lack of a collective identity among workers -No dense network of trade union representatives, weak levels of member engagement, limited knowledge of rights Narrative resources -In general, historical heritage of struggle and frustration lie at the basis of the defensive narrative of trade unions -Competition with solidarist associations and employer repression feed into the conflictual stance of trade unions -Unions question the credibility of MSIs because of negative experiences and are cautious for co-optation Infrastructural resources -Human resources are limited, lack of communication and negotiation skills required in MSIs -Union leaders have different priorities and need to divide time between core union activities and preparations for MSIs -Unions depend on membership fees or donors for funding of transport, computers and organisational infrastructure