Multi­stakeholder initiatives through the lens of labour regimes: towards a heuristic analytical framework

.


Introduction
This paper applies the concept of labour regimes to better understand the workings of multi-stakeholder initiatives (MSIs) in global production networks. Labour regimes are "spatially and temporally contingent systems of contested but co-constituted relations among and between global production dynamics and territorialised economic, political, and social formations" (Wickramasingha and Coe, 2022, 73). In understanding employment outcomes at the workplace, the labour regimes concept pays attention to the intersection of labour (workers and workers representatives), capital (global lead firms and national manufacturers), and the state constituted across multiple sclaes from the global to the workplace. I explain this concept in more detail in Sections 2 and 4.
The concept of labour regimes speaks well to the composition of MSIs as MSIs are also coordinated efforts between public, private, and social actors constituted across multiple scales (Gereffi and Lee, 2016) often representing labour, capital, and the state. The rational of the approach promoted by this paper then is that it integrates both global production dynamics and territorialised economic, political, and social formations in understanding the progress of MSIs. As I show with my empirics, connecting labour regimes with MSIs has three advantages. First, such a connection helps better understand the contested, but often obscured relations among and between stakeholders configured across multiple scales form the global to the workplace. Second, looking through the lens of labour regimes enables a closer look at the stakeholder composition, in particular, the interplay of politics and power relations among and between actors who are included as well as those who are excluded from MSIs. Third, labour regimes are inherently temporal, thus, this concept can help understand the dynamic nature of MSIs as they evolve across time and space. These latter two points have received less attention in the existing works on MSIs. electrical safety of the Bangladeshi garment factory buildings in the wake of the Rana Plaza building collapse in April 2013. The Accord was hailed as a promising initiative grounded in the principles of industrial democracy due to its inclusion of both consumption based and production based actors (Anner et al., 2013;Reinecke and Donaghey, 2015;Donaghey and Reinecke, 2018). Notwithstanding the merits of these characteristics of its stakeholder composition, the Accord was primarily driven by global actors with a notable absence of key national stakeholders -the state and manufacturers -from its Steering Committee. Yet, inadequate attention has been paid to this absence of the state and manufacturers. Now that the Accord has completed its tenure, I focus on this absence of the state and manufacturers from the Accord. Applying the labour regimes concept, I ask to what extent this absence of the state and manufacturers affected the Accord's ability to deliver its objectives. Using insights derived from qualitative inquiry and extensive fieldwork in Bangladesh, I demonstrate that this absence of the state and manufacturers from the Accord engendered paradoxical outcomes. On the one hand, the inability of the state and manufacturers to influence the work of the Accord, and the operational independence granted to it, meant the Accord was able to successfully implement its building safety program while it was in operation. On the other hand, the historically sedimented strong state-manufacturer alliances, coupled with the lack of commitment of the state and manufacturers to the Accord's mandate, affected the transition of the Accord as its tenure ended in 2019, and thus the long-term sustainability of the building safety program.
The rest of this paper is structured around seven sections. Section two discusses the emerging works on labour regimes followed by a review of MSIs in global production networks. In section four I describe my analytical framework connecting the labour regimes concept with MSIs. Section five offers a description of the Bangladeshi apparel industry and the Accord. Section six explains my methodology followed by a detailed analysis of empirics. Empirics are presented across three stages spanning the labour-capital-state intersection: creation of the Accord; the Accord in operation; and the transition of the Accord.

Labour regimes in global production networks
The labour regimes concept has been developed by scholars of labour and global production to gain a nuanced understanding of how labour process dynamics at the point of production are connected to broader political economic dynamics beyond the workplace. This labour regimes concept seeks to bring different facets of labour -labour agency, labour governance, and the labour process -into an integrated multi-scalar, analytical framework to understand employment outcomes. The concept of labour regimes has an intellectual lineage back to the 1970 s, when sociologists (Braverman, 1974;Burawoy, 1979;1985) understood labour regimes as the mechanisms through which labour is socially reproduced, organised, and mobilised at the workplace. While these earlier notions of labour regimes were largely underpinned by managerial control mechanisms at the workplace, labour regimes can also be strongly linked to reciprocal relations of production and reproduction of a labour force between the workplace and the local scale (Jonas, 1996). At the local scale, labour regimes are conditioned by historically sedimented social, class, and cultural relations (Peck, 1996;Pattenden 2016), worker resistance strategies (Jonas, 1996;Smith et al., 2018), and control mechanisms at lived in spaces (Lee, 1999;Pun and Smith, 2007).
The utility of the labour regimes concept for understanding employment outcomes however expands beyond the local scale. At the national scale, labour regimes approach can be deployed to examine the interplay of labour control mechanisms and resistance strategies (Anner, 2015). Importantly, labour regimes approach is innately multi-scalar, thus, is capable of incorporating the variegated political, economic, and sociocultural relations that produce and reproduce global production networks and their constituent workers (Carswell and De Neve, 2013;Selwyn, 2013;Pattenden, 2016;Mezzadri, 2017;Baglioni, 2018;Smith et al., 2018;Campling et al., 2021;López, 2021;Baglioni et al., 2022;Wickramasingha and Coe, 2022). Of these, three timely and important dynamics that shape labour regimes are worth noting: contract manufacturing relations such as price, delivery times and ethical codes; historical legacy of labour practices and labour movements; and national labour regulations (Smith et al., 2018). The labour regimes concept also allows the space to incorporate labour agency more fully into the analysis of interplay between control and resistance, nuance out variegated relations and dynamics at different scales, and integrate the interventions of international civil society organisations (ICSOs) in shaping governance and agency of labour (Wickramasingha and Coe, 2022). The labour regimes approach thus has the potential to theorise the complex intersections of global production dynamics and territorialised economic, political, and social formations that produce specific employment outcomes in a particular place of production (Coe and Yeung, 2019).
The labour regimes approach has thus far been developed and referred to both as an institutional framework for accumulation and labour regulation constructed around specific labour market reciprocities; and as fluid and dynamic sets of social relations and power structures constituted at multiple scales from the workplace through to national and global levels. In these works, the concept of labour regimes has been applied more broadly to either a sectoral context (i.e. apparel labour regimes - Smith et al., 2018;Wickramasingha andCoe, 2022, automotive labour regimes -Campling et al., 2021) or in terms of a specific geographical context (i.e. national labour control regimes -Anner, 2015, local labour control regimes -Jonas, 1996;Pattenden, 2016) to understand employment outcomes. In this article, I take the labour regimes concept in a new direction. I exploit its unique ability to frame the interplay of global production dynamics and territorialised formations to explain specific labour governance initiatives in global production networks that are both spatial and temporal in nature, and comprise multiple stakeholders. Because labour regimes are systems of labour control and resistance contingent upon fluid social relations between labour, capital, and the state, this concept can be extended to examine labour governance initiatives which are also institutionalised based on a set of fluid social relations between such actors straddling both global and national.

Multi-stakeholder initiatives in global production networks
The term multi-stakeholder initiatives (MSIs) is often reserved, and used almost exclusively to refer to ethical sourcing initiatives of the Global North such as the Ethical Trading Initiative (ETI) of the UK and Fair Labor Association (FLA) in the USA. However, there is a variety of different models of MSIs that represent global and nationally based initiatives and often a combination of both. The terminology 'MSI' thus is not and should not be exclusive to the ethical trading initiatives originating from the Global North. Drawing on the Stakeholder Theory (Freeman, 1984), I argue that the term MSI can be productively used to identify and analyse alliances between businesses, governments, and civil society, which are seen as promising ways to realise development goals. In the governance of labour in global production networks, such MSIs can often be a broad-based coalition -of global or national in nature -and can consist different types of global and national actors such as lead firms, trade unions, labour campaigns, manufacturers, state institutions, and workers who work together in standard setting, implementation, monitoring, sanctioning, and capacity building on a particular labour issue (Piore and Schrank, 2006;Mayer, 2014;Gereffi and Lee, 2016). One example is the Apple-Foxconn case in China (Mayer, 2014), where in the early 2010 s, the governance regime of Foxconn transformed into a new form of joint labour governance that involved a wide range of stakeholders including Apple Inc, Foxconn, the Chinese Government, national trade unions, and international and national civil society organisations. Another example is the Horticultural Ethical Business Initiative (HEBI) in the Kenyan cut flower industry (Dolan and Opondo, 2005). HEBI illustrates how the cornerstones of MSIs -diversity and multivocality -shape the partnerships and capacity to achieve workers' rights. In this paper, I refer to such, a more fluid concept of MSIs.
In global production networks where 'governance deficits' prevail in national regulatory arenas (Newell, 2001;Kabeer, 2002;Gereffi and Mayer, 2006;Mezzadri, 2016;Mayer and Phillips, 2017;Ruwanpura, 2018;Hewamanne, 2020;Saxena 2020) and lead firms' private governance mechanisms (born out of ETIs and FLAs above) are criticised for their inability to affect meaningful change (O'Rourke, 2003;Stevens, 2008;De Neve, 2009;Locke et al., 2009;Erwin, 2011;Goger 2013;Locke, 2013;Ruwanpura, 2016, MSIs have been touted as more effective in delivering meaningful labour outcomes at the workplace (Bartley, 2005(Bartley, , 2011O'Rourke, 2006;Amengual, 2010;Locke et al., 2013;Mayer, 2014;Gereffi and Lee, 2016). Yet MSIs can be a challenge to successfully navigate. Authors noted that while MSIs have potential to develop into credible drivers of decent work, their effectiveness and long-term viability are in question due to the conflicts of interests as well as politics and tensions among and between stakeholders (Dolan and Opondo, 2005;García-López and Arizpe, 2010;Schouten et al., 2012;Moog et al., 2015;LeBaron, 2018LeBaron, , 2021Bair et al., 2020). Mapping these contested stakeholder relations however is often a challenge given: a) these stakeholders are varied and configured across multiple scales from the global to the workplace (Gereffi and Lee, 2016); and b) the complexity of relations among and between them which are fluid, dynamic, and evolve constantly across time and space (Wickramasingha, 2022).
The labour regimes concept speaks well to the stakeholder composition of MSIs due to its ability to map the contested and obscured relations constituted at different scales from the global to the workplace under one analytical domain. In particular, connecting the labour regimes concept with MSIs is helpful in three ways. First, engaging the labour regimes concept as an analytical framework means having the ability to seek multiple perspectives from labour, capital, and the state in understanding the workings of MSIs (Wickramasingha and Coe, 2022). This helps establish direct links between decisions made at the global level and the national and workplace scales, so that the 'cause and effect drivers' of MSIs can be properly connected, understood, and explained. This also allows a closer scrutiny at the intersection of 'global' and 'local', where currently, MSIs are noted as being driven mostly by global actors and are not adequately localised (Dolan and Opondo, 2005;Kang, 2021). The multi-scalar analytical frameworks as offered by existing works on labour regimes (Pattenden, 2016;Smith et al., 2018;Wickramasingha and Coe, 2022) are particularly of use here with their insightful and nuanced treatment to both territorialised and global production dynamics as I explain in detail in the next section.
Second, the labour regimes concept is well equipped to deal with stakeholders who are included in MSIs as well as stakeholders who are excluded. Indeed, as scholars have pointed out, some stakeholders find reasons to either exclude other stakeholders or marginalise their voices in the dialogues and decision making process. In other cases, some stakeholders remain absent (Locke et al., 2009;Dolan and Opondo, 2005;Schouten et al., 2012;Kang, 2021). When this happens, as Soundararajan et al. (2019) noted, the functions of MSIs are hampered by major structural governance barriers. In this respect, who participates in MSIs, who are excluded, and who have the authority to speak for those who are not represented, have a direct bearing on the likely outcomes of MSIs and their capacity to eliminate exploitation at the workplace (Dolan and Opondo, 2005). Examining the nexus of inclusion and exclusion can thus yield important yet, often obscured insights that help explain the potential of MSIs to deliver decent work, or otherwise, the factors that might hinder this potential (Sharma and Bansal, 2017). Because of its ability to offer an integrated and nuanced treatment to different stakeholders -both present and absent -across multiple scales, the labour regimes concept has the capacity to map the interplay of these 'inclusive' and 'exclusive' dynamics, the tensions and politics they generate, and their implications on the outcomes of MSIs.
Third, the multi-scalar concept of labour regimes is sensitive to the temporal nature of ever changing relations between labour, capital, and the state (Smith et al., 2018;Wickramasingha and Coe, 2022). Precisely because of the dynamic nature of these relations, MSIs are also inherently temporal in nature, as they most often evolve across time and space. This evolution defines how MSIs unfold in a specific place of production, in particular, their objectives, stakeholder composition, and their capacity and commitment to effect meaningful change (Wickramasingha, 2022). This is however an element that has not been adequately captured in the existing analyses of MSIs when attempting to examine their role in improving working conditions in supply chains. As I show later in my empirics, the labour regimes concept can serve as a tool to examine the dynamic nature of MSIs and their drivers of change. It can reveal how the effectiveness and the sustainability of MSIs can be considerably defined by their temporal nature both in terms of their stakeholder composition and time and space across which they exist, reproduce, and evolve.

Connecting labour regimes with multi-stakeholder initiatives: Towards an analytical framework
In the multi-scalar conceptualisations, labour regimes are theorised as composed of nested scales institutionalised formally in organisations with rules and regulations but also informally in relational norms and habits (Baglioni, 2018;Smith et al., 2018;Campling et al., 2021;López, 2021;Wickramasingha and Coe, 2022). As these works note, labour regimes more broadly emerge from the interplay of global production dynamics including contract manufacturing relations of lead firms, strategies of international civil society organisations, and the influence of international regulatory bodies; and the territorialised state policies, capitalist agendas, trade union positionality, labour politics, local social relations, and the labour process. This interplay underpinning multiscalar approaches to labour regimes help grasp the variegated links and interdependencies connecting the actors and processes involved in the formation and functions of MSIs.
As Fig. 1 illustrates, in applying the labour regimes concept to MSIs, I have identified three notable scales across which MSIs are configured and evolve. At the global scale, lead firms, international regulatory bodies (i.e. ILO), and ICSOs (i.e. trade unions, civil society organisations, and labour campaigns) play a defining role in initiating and institutionalising MSIs. Such interventions can take the form of consumer campaigns led by ICSOs in response to a particular labour incident, forcing lead firms to take more responsibility for labour outcomes in their supply chains. Such pressure can lead to the formation and sustenance of MSIs, an example is the Bangladeshi Accord (Reinecke and Donaghey, 2015). Lead firms and ICSOs can also be influenced by 'bottom-up' movements lead by trade unions and workers from the grassroots, where Apple-Foxconn case in China is an example (Mayer, 2014). Moreover, as Nelson and Tallontire (2014) pointed out, most often, executive governance aspects of MSIs remain mostly with global actors, in a mode of transnational co-regulation. This means, MSIs are often driven by global actors, where the explicit intention of the state is to retract the state's direct control over particular arenas of regulation (Abbott and Snidal, 2009;Mayer and Gereffi, 2010;Esbenshade, 2012;LeBaron and Lister, 2015). Thus, the very existence of MSIs is usually defined by these global, private and non-governmental stakeholders. Stakeholder relations and politics at the global scale and between the global scale and national/sub national scales thus remain important to understand the formation, progress, and the sustainability of MSIs.
Next, as Fig. 1 demonstrates, both the formation and success of MSIs can be notably influenced by four dynamics at the national scale. First is the legacy of the national historical trajectory in terms of its political economy, class relations, and social movements (Hadjimichalis and Hudson, 2006;Ruwanpura, 2012;Saxena, 2020). They define the institutional structures and power relations that exist in the national context when MSIs are institutionalised and operationalised. This legacy can take various forms such as the involvement and the support (or lack thereof) of national stakeholders, and the degree to which social actors such as trade unions and workers can play a role within MSIs (Dolan and Opondo, 2005). Second, class struggles at the national level between labour, capital, and the state play an important role. They define MSIs ability to successfully institutionalise to the local context, enforce standards, and their long-term sustainability (Mayer 2014;Bair et al., 2020;Kang, 2021). Third, national labour regulatory frameworks play a defining role in MSIs. In some countries, weak national regulations can trigger the formation of MSIs to drive decent work in supply chains (Abbott and Snidal, 2009;Mayer and Gereffi, 2010;Vogel, 2010;Esbenshade, 2012;Donaghey and Reinecke, 2018). In others, strong national regulations can complement MSIs capacity to deliver decent work (Amengual, 2010;Locke et al., 2013;Ruwanpura and Wrigley, 2011;Ruwanpura, 2016). Fourth the ability of national trade unions to play an inclusive role, the degree of involvement of trade unions in MSIs, and the unions' capacities and structural make-up can define how workers interests are represented in MSIs and their capacity to eliminate exploitation at the workplace (Dolan and Opondo, 2005;Zajak 2017;Donaghey and Reinecke 2018;Ashwin et al., 2020;Wickramasingha, 2022).
Finally, MSIs are defined by the workplace dynamics and the labour process. The workplace is instrumental in understanding MSIs, as ultimately it is the workplace labour regimes with significant governance gaps that mostly give rise to MSIs (Mayer, 2014;Gereffi and Lee, 2016;Donaghey and Reinecke, 2018;Gunawardana, 2020;LeBaron, 2021). Three elements are particularly worth noting here. One, while MSIs seek to regulate workplaces, the successful implementation always depends on the degree of support from managers. While in some cases support can be forthcoming, in others the support is forced through pressure from lead firms, ICSOs, and market sanctions as was the case with the Bangladeshi Accord (Donaghey and Reinecke, 2018). In countries with weak regulatory capacities, governance gaps, and strong state-capital alliances, support to implement and enforce MSI standards can be a significant challenge. Two, potential of global initiatives to address structural problems of exploitation at the workplace is always a question (Alamgir and Banerjee, 2019), with intended and unintended consequences of compliance and global pressures on workplace standards (Sinkovics et al., 2016). Three, the space for workers and their representatives to get actively engaged in MSI activities is instrumental for the successful execution (Dolan and Opondo, 2005;Reinecke and Donaghey, 2015). Overall, the dynamics represented in these three scalesglobal, national, and the workplacedo not exist in isolation, but remain co-constituents that are constantly being re-shaped and transformed by each other as they evolve. Hence, the temporalities of MSIs. In the remainder of this paper, I take the Accord on Fire and Building Safety in the Bangladeshi apparel industry to demonstrate the merits of this approach.

The bangladeshi apparel industry and the rise of the accord
As of 2021, Bangladesh was the third largest apparel manufacturer in the world with a 6 % share of the world market. In the financial year 2019-2020, the industry was valued at $34 billions and accounted for 83 % of the country's total export earnings (BGMEA, 2020; World Bank, 2020). With over 4,000 export oriented garment factories, the industry directly employed approximately 3.6 million workers in 2019 (The Asia Foundation, 2020). The industry is administered by the Bangladesh Garment Manufacturers and Exporters Association (BGMEA). The economic growth notwithstanding, the Bangladeshi apparel industry has continuously come under pressure for its exploitative labour regimes. With a national minimum wage of $95 per month in 2019 (Wickramasinghe, 2020), Bangladesh has long had the reputation for paying the lowest wages among major producers in the industry -Cambodia ($182), China ($326), Indonesia ($280), and Vietnam ($180) (AFWA, 2020). Workers, especially women, often face wage differentials, insecurity, and discrimination (Saxena, 2014;Anner, 2020). As both Anner and Saxena found, the industry is known for excessive working hours, on which workers often do not have a choice. Additionally outsourcing, part-time work, and informality within the industry contribute to workers sense of extreme vulnerability (Anner, 2015). This labour market vulnerability has eroded Bangladeshi workers structural power and their ability to demand greater social protection at the workplace (Anner, 2015;Wickramasingha and Coe, 2022). Health and safety remain a serious concern in the industry with mostly ill-structured and overcrowded factory buildings. Between 2005 and 2018, the industry saw 18 fatal factory fires and building collapses (Wickramasinghe, 2020). Of these, the Rana Plaza building collapse was the catalyst that gave rise to the Accord.
The Accord came into existence as a result of widespread social movements, led by global trade unions and labour campaigns (ICSOs) calling for lead firms to take responsibility for the safety of workers in their supply chains (Anner et al. 2013;Donaghey and Reinecke, 2018). This led to the creation of the Accord in 2013, a legally binding agreement between 190 lead firms primarily from Europe and international trade unions, namely, IndustriAll, UNI Global Union, and the IndustriAll Bangladesh Council that included 7 Bangladeshi trade unions. The Agreement was witnessed by the Clean Clothes Campaign, Workers Rights Consortium, International Labour Rights Forum, and Maquila Solidarity Network. The Accord focused on the electrical, fire, and structural safety of around 1,500 garment factories supplying mainly to European brands. Its mandate included: independent and regular inspection of building safety; facilitating and monitoring the progress of the remediation work; safety trainings and establishing safety committees at workplaces; and establishing a complaint and grievance resolution mechanism. The Accord was hailed as a 'new paradigm' in the enforcement of global labour standards and human rights due to its involvement of both production and consumption based actors (Anner et al. 2013;Donaghey and Reinecke, 2018).
The Accord is identified as a Global Framework Agreement (GFA) because it had 'legal teeth' and was a collective agreement between trade unions and brands. Having noted that, I argue that the Accord can also be characterised as a MSI due to two reasons. First, the Accord is a coalition of multiple stakeholders -international trade unions, international civil society organisations, national trade unions, international regulatory bodies, lead firms, the state, and manufacturers (the latter two were largely absent, which was the focus of this paper). Second is the institutional features of the Accord. The institutional features of MSIs typically include an executive board, an assembly or council representing different stakeholder interests, a technical advisory committee, and an executive director with a secretariat to handle the daily operations (Fransen, 2011;Ponte, 2013;Riisgaard et al. 2020). The Accord satisfied all these criteria (Fig. 2).
The Accord depicted a complex configuration of how global and national regulatory mechanisms were entwined with one another. It was governed by a Steering Committee (SC) with equal representation of six members each (three alternative) from signatory brands and trade unions. The SC was responsible for the selection, contracting, compensation, and review of the safety programme. The signatories to the Accord appointed an Advisory Board comprising brands and retailers, manufacturers, Bangladeshi government, trade unions, and NGOs. The Advisory Board's role was to ensure all stakeholders can engage in constructive dialogues with each other and provide feedback and input to the SC, thereby enhancing quality, efficiency, credibility and synergy of the program. Further, the Accord had several technical committees encompassing engineering, remediation, complaints, training, and administration. The day-to-day management was handled by the Accord Secretariat based in Dhaka, Bangladesh. Indeed, the Accord is well documented in the existing literature. Of these, scholars have significantly advanced the knowledge of the Accord with their detailed treatment to its governance model built on the principles of industrial democracy, joint liability in global supply chains, and the intersection of public and private labour governance regimes (Anner et al., 2013;Reinecke and Donaghey, 2015;Donaghey and Reinecke, 2018;Bair et al. 2020;Reinecke and Donaghey, 2021). Notably, others discussed the merits of union inclusive governance models, labour networked interactions with global actors, and increased strategic capacities of trade unions, all facilitated by the Accord (Scheper, 2017;Zajak, 2017;Ashwin et al., 2020). Notwithstanding the merits, the Accord was also critiqued by scholars for its potential to address structural problems of exploitation at the workplace and the intended and unintended consequences of global pressures and compliance on workplace standards (Sinkovics et al., 2016;Alamgir and Banerjee, 2019;Kabeer et al. 2020).
I contribute to this growing literature on the Accord by exploring an element that has received less attention. Although technically, the state and manufacturers had seats on the Advisory Board, they were not part of the Steering Committee and had little to no decision making powers in the Accord. As I show later, this had created tensions between the Accord, and the state and manufacturers, which resulted in the latter two withdrawing and largely remaining 'absent'. Yet, this absence of the state and manufacturers, the tensions and politics this absence generated, and how they affected the ability of the Accord to deliver its objectives have received less scrutiny with few exceptions (i.e. Bartley, 2018;Bair et al., 2020;Kang, 2021). With the Accord now having completed its tenure, I ask to what extent this absence of the state and manufacturers defined the Accord's progress and its long-term sustainability. First however, I briefly explain my research methodology in the next section. Fig. 2. Governance structure of the Accord (developed by author based on: Accord, 2020).

Research methods
I have adopted case study approach with the Accord taken as a 'case' example. Case studies approach enables the incorporation of all aspects of research from data collection to data analysis while allowing space to deploy conceptual frameworks to analyse the generated data (Yin 1994;Hanson and Pratt 1995;Gibson-Graham 1996), which was useful for this study. The data emerge from a larger study I conducted in South Asia on apparel labour regimes in global production networks for my doctoral work between 2016 and 2020. The data used in this paper was collected during four months of field work (January to April 2019) in and around the Dhaka suburbs -Gazipur, Mirpur, and Narayanganj garment factory clusters. I employed snowball and purposive sampling methods. I recruited participants for my research based on the contacts I already had in the industry through previous work and research as well as from the new contacts I was making while conducting initial rounds of interviews. I employed qualitative research methods and carried out 91 semi-structured and unstructured interviews (including 7 repeated interviews) among lead firms, auditors, international trade unions, international regulatory bodies, non-governmental organisations, state authorities, manufacturers associations, national trade unions, national civil societies, policy institutes, and factory executives. Among the interview participants were two members of the Accord Steering Committee, five Bangladeshi union federations affiliated to the Accord, a former Director of the Accord, and a corporate social responsibility consultant (CSR) to major lead firms. The CSR consultant was involved in the creation of the Accord and took part in the initial meetings in Geneva. The Accord Secretariat was going through a legal battle with the Bangladeshi government during my fieldwork, thus declined to be part of the study.
My interview protocol included questions related to: a) the creation of the Accord; b) operational aspects of the Accord including factory inspections and remediation work; c) relations between the Accord, the state, manufacturers, and trade unions; d) outcomes of the Accord; e) extension of the tenure of the Accord; and f) the future of the safety program beyond the Accord. Along these themes, the interviews helped understand the contested relations between labour (workers and workers representatives), capital (lead firms and manufacturers), and the state in negotiating the progress of the Accord. The interviews lasted between 60 and 90 minutes. Each interview was calibrated with responses from other participants and secondary data where possible. Fieldwork included visits to 14 garment factories that included seven large scale (over 2,500 workers), three medium scale (between 600 and 2,500 workers), and four small scale (below 600 workers), as per BGMEA criteria. The Accord did not work with small scale factories as confirmed by the ones I visited. During those visits, I spoke to factory management who gave me accompanied tours across the production floors. My findings are also informed by the Panel Discussion organised by the Action Aid Bangladesh in April 2019 to mark the sixth anniversary of the Rana Plaza accident. In order to maintain anonymity, I provide no identifiable information pertaining to the people, offices, and factories visited. The face-to-face interviews were complimented with a review of literature and secondary data pertaining to discourses of labour governance in global production networks, published corporate and governmental records, and media outputs.

Understanding multi-stakeholder initiatives through labour regimes
While the scope of labour regimes can generate manifold insights when applied to MSIs, I focus on one question that has received less attention in the existing works on the Accord: "To what extent the absence of the state and manufacturers influenced the effectiveness of the Accord's building safety program and its long-term sustainability?" The analysis is divided into three subsections. First, I explain the historically sedimented class relations at the national context, labour regimes, and their influence on the Accord's inception. In the next two subsections I explain the paradoxical outcomes of the Accord influenced by the absence of the state and manufacturers: a) how this absence has helped the Accord successfully implement the building safety program in the short term; and yet, b) how the lack of commitment of the state and manufacturers affected its transition. In applying the labour regimes concept to the Accord, the discussion is underpinned by the contested intersection of labour (workers and workers representatives including national unions, ICSOs, and regulatory bodies), capital (lead firms and manufacturers) and the state.

Creation of the Accord and the 'absence' of the state and manufacturers
When applying the labour regimes concept, the 'absence' of the state and manufacturers from the Accord can be explained by three elements at the national context: historically sedimented production regimes; powerful state-manufacturer alliances; and labour politics.
When initial trade liberalisation took place in 1977, Bangladesh was struggling with a deteriorating economy and a corrupt and ineffective government (Sobhan, 2004;Lewis, 2011). With a growing, unskilled population living in abject poverty, the country was heavily dependent on foreign aid (Wood, 1997). Thus, when the apparel industry took off in the early 1980 s "the priority was to ensure economic growth", which pushed social welfare standards to the margins (BD-Government/2). By that time, the state had considerably weakened and or politicised the previously strong trade unions (Lewis, 2011). Collectively, these events paved the way for a new class of capitalists to emerge, jointly driving the industrialisation process with the state (Saxena, 2020). The state-capital alliance structured a largely deregulated industry and labour market from the outset, with manufacturers given considerable freedom to shape sourcing and production practices. State-manufacturer alliances continued to grow in the ensuing four decades, with manufacturers enjoying great power in the industry (Tighe, 2015;Wickramasinghe, 2020). Indeed, this was a situation compounded by the fact that by 2019, between 60 and 70 % of the Bangladeshi parliament members had business interests in the apparel industry (BD-ICSO/6; BD-CSR Consultant/1; BD-Union/5; Wickramasingha and Coe, 2022). Due to these close relations, one labour leader noted that: "Earlier, only feelings were in favour of capital, but now the capital itself dominates the government" (BD-Union/5).
The strong state-manufacturer alliances at the national context resulted in workplace labour regimes that were mostly devoid of state protection and meaningful representation of workers. At the workplace, most Bangladeshi manufacturers fostered labour regimes that did not conform to decent work but rather benefitted themselves (Kabeer, 2002;Anner, 2015Anner, , 2020Tighe, 2015). State institutions largely failed to enforce labour regulations, thus the state was often accused of turning a blind eye to loopholes in the system that enabled illegal forms of accumulation (Sobhan, 2004; BD-Union/7). In fact, prior to the collapse, the Rana Plaza building -the owner of which was a member of the ruling party -had consistently received the building approval to operate from the government, despite its structural issues (BBC, 2013).
Unionisation in the industry was either legally prohibited -as was in the case of Export Processing Zones (EPZs) -or actively discouraged in factories outside EPZs, with harsh punishments for disobedience (BD-CSO/1/5; BD-Union/3/4). Indeed, since the decade of 2010, around 20 independent union federations and civil societies actively worked in the garment industry with most of them emerging from the grassroots (e.g. BD-Union/2/3/4/7/9/13; BD-Civil Society/1). Of these, four of the unions I interviewed were headed by female labour leaders (BD-Union/ 4/9/13; BD-Civil Society/1). Three female labour leaders were former child labourers, while 'BD-Union/4 ′ worked as a machine operator in the industry. 'BD-Civil Society/1 ′ was instrumental in the process of creating the Accord. Their struggles were successful in giving leadership to grassroots movements where they individually and collectively raised workers issues in Bangladesh (Siddiqi 2017;Ashraf and Prentice 2019;Wickramasingha 2022). Still, as unions claimed, the progress was slow, and in some cases took years to successfully establish unions in factories, mainly due to anti-union practices of employers.
In discouraging unions, manufacturers, together with the department of labour, promoted alternative mechanisms of organisation in the form of Participation Committees (PCs). PCs represented collective efforts on the part of the state-manufacturer alliance in containing workplace bargaining power, thereby driving down the number of unions at the workplace (BD-Trade Union/1/2; BD-Civil Society/1; Bair et al. 2020;Wickramasingha and Coe, 2022). On national policy platforms, trade unions were marginalised and often omitted from policy discussions. Consequently, social welfare standards featured in national industrial agendas with little to no representation of labour. Labour regimes in most Bangladeshi workplaces thus had led to degradation of work, leaving the majority of apparel workers vulnerable to highly exploitative accumulation practices (Kabeer, 2002;Saxena, 2014Saxena, , 2020Anner, 2015Anner, , 2020. It was against this backdrop the Rana Plaza accident happened. From the field accounts, it was evident that the stakeholder composition of the Accord was influenced by this pre-existing labour regimes and the strong state-manufacturer alliances. As per unions, manufacturers, and government representatives, the state and manufacturers were intentionally left out of the Steering Committee of the Accord by lead firms and ICSOs (BD-Union/1/7; BD-Civil Society/1). As another participanta previous President of a manufacturers' association quoted: "From the entrepreneurs or government side, there was no involvement and no consultation at all" (BD-Technocrat/1). While the government and BGMEA were offered seats on the Advisory Committee of the Accord, 'BD-Technocrat/1' said that they declined due to the lack of executive power in the Advisory Committee. This was corroborated by a former Director of the Accord: "The government was a part of the Advisory Committee but they did not contribute" (BD-ICSO/1). Similarly, he revealed that: BGMEA was asked to be on board as a member of the Advisory Committee, but BGMEA rejected. BGMEA wanted to be a member of the Steering Committee. But on what basis? I blame BGMEA [for the Rana Plaza accident] because their governance is terrible" (BD-ICSO/1).
BGMEA's conflicts of interest were also pointed out by a lead firm who argued that the Accord was funded by brands to make factories safe, which could not be entrusted to BGMEA (Lead Firm/2). Bangladesh Knitwear Manufacturers and Exporters Association (BKMEA) took up a seat on the Advisory Committee, but later stepped down for the same reasons (BD-Technocrat/1). According to a civil society organisation, it did not help when in the wake of the Rana Plaza accident, the state and BGMEA hid information about the events that led to the accident and the enormity of its impact on workers (BD-Civil Society/1). As a trade union corroborated: "When the meeting was taking place in Geneva, the government was trying to hide information" (BD-Union/1). Further, it was evident from the field that the decision to exclude the state and manufacturers from executive powers was influenced by the objections of national trade unions indicating severely fractured relations between the state, manufacturers, and labour (BD-Civil Society/1; BD-Union/1/ 3/4/7). When the initial discussions on the Accord were held in Geneva, national trade unions had strongly opposed the inclusion of the state and BGMEA. As a CSR Consultant who was involved in the process quoted: "Labour groups [national trade unions] were going to boycott the Accord and the meetings in Geneva if the government and employers were invited to those meetings" (BD-Consultant/1).
From these accounts, it appears that this exclusion of the state and manufacturers from the Steering Committee happened due to the lack of confidence global stakeholders and national trade unions had on the ability of the state and manufacturers to contribute in any meaningful manner to the Accord's objectives. These accounts indicate that both the government and BGMEA wanted to play a decisive role in the Accord. However, neither was granted a seat on the Steering Committee, and were only offered a role on the Advisory Board, which did not have decision making powers. In the next two subsections I explain that this 'absence' of the state and manufacturers resulted in paradoxical outcomes: while it helped the Accord successfully execute its building safety program in the short-term, it nevertheless affected its long-term sustainability.

The Accord in operation: A success in the short-term
During its tenure of five years between 2013 and 2018 (which later extended to 2019), the Accord completed 90 % of initial remediation at 1,215 factories, institutionalised 1,243 safety committees, and resolved 688 safety complaints (Accord, 2020). When applying the labour regimes framework, it was evident that the Accord's success with the building safety program was significantly owed to its unique stakeholder composition, and in particular the presence of ICSOs and lead firms and the absence of the state and manufacturers. Through the presence of ICSOs, lead firms were made accountable to the coalition, who were capable of leading large consumer-based campaigns that may have potentially damaged the reputation of lead firms: "If buyers did not adhere to the requirements, the [global] trade unions would have had people on the roads" (BD-ICSO/6). It is important to note that this specific leverage ICSOs had was highly temporal, as it was tied to the Rana Plaza accident, and the outrage it created among consumers, resulting in the urgency to regulate the industry in Bangladesh (Anner et al. 2013;Reinecke and Donaghey, 2015;Donaghey and Reinecke, 2018). As Anner (2015: 29) noted, the potential to utilise the "symbolic power" was ripe during this time, as consumers across the world were shocked by the level of the "human horror" and its deeply disturbing images. Enforcement of the Accord was thus grounded on a chain of market sanctions, which first sought to influence consumer purchasing decisions and apply pressure on lead firms (Donaghey and Reinecke, 2018). Lead firms then transferred this pressure to manufacturers through their purchasing decisions. As an ICSO noted if the Accord deemed a factory unsafe, lead firms immediately withdrew orders: The strongest tool the Accord held in their hands was to cut factories off from business. If a factory is not complying, the Accord will inform their member buyers that the factory is not safe, and the buyers will listen to them and discontinue business. With this torturous instrument, the Accord made the factories follow their instructions (BD-ICSO/6).
Indeed, in this equation, the absence of the state and manufacturers from the Accord's Steering Committee, or rather, the inability of these two stakeholders to influence the decisions of the Accord should not be overlooked. The government and BGMEA did not have much choice in this particular instance, as they were dealing with the massive backlash Bangladesh had to endure in the aftermath of the Rana Plaza accident: Buyers and development partners said they won't accept manufacturing goods produced by us as our workers are not safe… They said, 'goods you produce have the blood of your workers'… after hearing this, the country tensed up (BD-Government/2).
Because the government faced the risk of losing investment after the Rana Plaza accident, the government let the Accord have operational independence from 2013 to 2018 (BD-Union/1; BD-ICSO/1; BD-Auditor/6). Many manufacturers initially had rejected the recommendations of the Accord, but eventually complied when they realised that they could not rely on their alliances with the state in this instance (BD-Union/2/4). As a major auditing firm working for several lead firms corroborated: "Factories had no other choice, but had to agree with everything at that point" (BD-Auditor/6). Thus, it appeared that in the wake of the Rana Plaza accident, the Bangladeshi government and BGMEA took a back seat and allowed the Accord a free rein in governing the building safety standards in the factories that they took over. Consequently, safety standards of over 1,200 export-oriented garment factories were significantly improved between 2013 and 2019 due to the strict approach of the Accord (BD-ICSO/1) as was also found by other scholars (Donaghey and Reinecke, 2018;Bair et al., 2020;Kang, 2021).
Importantly, the inability of the state and manufacturers to influence the Accord meant that national trade unions and workers were able to play a meaningful role in the factory inspection and remediation program, a finding consistent with existing works (Reinecke and Donaghey, 2015;Zajak, 2017;Ashwin et al., 2020). The Accord created 'safety teams' with workers trained as safety champions to monitor the building safety conditions and report back to national unions. National unions were directly involved in routine monitoring of the factory safety programs where they served as the connection line between the workplace and the Accord (BD-Union/3; BD-Civil Society/1). At this point, it is important to note, that when the Accord started working in Bangladesh in 2013, only about 32 factories were unionised (BD-Government/2). However, the global pressure in the wake of the Rana Plaza accident compelled the state and manufacturers to accept unions in garment factories. This pressure had increased the number of unionised factories to 715 by 2019 encompassing 10 % of the workforce (BD-Union/1; BD-Civil Society/1; BD-Government/2). Essentially, this meant, that most of the factories the Accord worked with were not unionised. Still, national trade union signatories to the Accord were able to access garment factories for safety inspection, as four of them confirmed (BD-Union/1/2/ 7/14). Within this multi-scalar collaboration stretching from the workplace to the global, workers played a central role in identifying workplace safety issues and escalating them to national trade unions. Trade unions in turn brought them up with the Accord. A form of social governance (Gereffi and Lee, 2016), McCubbins and Schwartz (1984) earlier referred to this system as 'fire alarm model of regulation'. Through this channel of communication unions and workers were able to hold manufacturers accountable for non-compliance and escalate them to global platforms in order to resolve labour issues at the workplace.
The Accord thus created space for trade unions and workers to be involved in negotiating workplace issues, a space that did not exist in the pre-Rana Plaza context in Bangladesh as was also documented by the collaborations of Donaghey and Reinecke (2015;. Indeed as Zajak (2017Zajak ( , 1015 argued, as well as providing a "shadow protection" for national trade unions and workers under the umbrella of the Accord, the accessibility to and interactions with global actors provided national unions with new opportunities for developing strategic capabilities in securing solutions to labour issues and confronting dominant domestic actors. Yet, this success was temporally contingent. As Zajak also noted, even though in the immediate aftermath of the Rana Plaza accident, the affiliation with the Accord provided protection and legitimacy for unions in Bangladesh, as the time passed and the urgency to regulate the industry diminished, this affiliation provoked counter-reactions by the state-manufacturer alliance to contain the emerging power of unions and the Accord. And the trade unions and their newfound leverage were still too weak to keep pace with the "ever stronger opposition from management and political-economic elites" (Zajak, 2017(Zajak, , 1020. As I show in the next subsection, these contested relations among and between global and national stakeholders affected the transition of the Accord when its tenure ended.

Termination of the Accord: End of the road and a messy transition
The Accord started as a temporary coalition with a five-year tenure (2013)(2014)(2015)(2016)(2017)(2018). The agreement between the Accord and the Bangladeshi government was that at the end of the stipulated five-year term the government and manufacturers will take over the building safety program. This process however turned out to be highly contested. By employing the labour regimes concept and looking at the multi-scalar relations among and between global and national stakeholders, these developments can be understood along two notable dynamics.
First, the impending termination gave the opportunity to the state and manufacturers to look for alternatives to replace the Accord. Some participants noted that this was partly due to the way the Accord "bypassed the national government" and "intruded on national sovereignty" (BD-CSR Consultant/1; BD-Manufacturer/6/8/9) as was also noted by other scholars (Bartley, 2018;Bair et al. 2020;Kang, 2021). To quote 'BD-Manufacturer/6 ′ : "The Accord behaved like a dictator". Consequently, as early as 2015, the Remediation Control Cell (RCC) was institutionalised at the Department for Inspection of Factories and Establishments (DIFE) to take over from the Accord (BD-ICSO/1). Yet, as above participants noted, by 2018, the RCC was nowhere near ready to take over from the Accord in terms of the capacity, resources, and 'knowhow'. Moreover, given the long-standing neglect of labour rights in Bangladesh and the refusal of the Bangladeshi state and manufacturers to implement basic labour standards (Sobhan, 2004;Kabeer, 2002;Lewis, 2011;Anner, 2020), the ICSOs, lead firms, and other industry stakeholders had little confidence on RCC (BD-ICSO/1; BD-Civil Society/1; BD-Union/1/2; Lead firm/3). Consequently, in 2017, ICSOs and lead firms sought an extension of the Accord up to 2021. Lead firms agreed to a new Accord that put greater emphasis on the rights of workers to organise a union, recognising that worker empowerment is fundamental to assuring workplace safety (Lead firm/3). As was also noted by Bair et al. (2020, 19) in their insightful narratives of "the state strikes back" the state and manufacturers strongly objected to the extension (BD-Civil Society/1).
The case for the extension was then appealed at the Bangladeshi High Court by the Accord. In early 2019, the High Court issued the verdict in favour of the Bangladeshi government and directed that the Accord closed operations and handed over the remediation work to the government by 7 April 2019 (Business and Human Rights Resources Center, 2019). This decision was met with disappointment by lead firms who complained that the government was not showing serious commitment towards improving workplace standards: "The minister of commerce [a manufacturer himself] was so outspoken and said that they can manage their own factories, and that Bangladesh does not need outside parties…. We said fine, one more Rana Plaza and we will be out of here" (Lead firm/3). The Court's decision was resented by unions as well, who claimed that the government and BGMEA did not want the Accord to stay due to the strict stance of the Accord in compliance and monitoring (BD-Civil Society/1). "Our government and manufacturers are disturbing the Accord's work because the Accord is a neutral body. When the Accord visits factories they are not fooled" (BD-Union/7).
Second, it is important to recognise the orchestrating role played by the International Labour Organisation (ILO) in the attempts of the national stakeholders to replace the Accord thereby providing legitimacy to the efforts of the state-manufacturer alliance. Categorised under 'global regulatory bodies' (Fig. 1) in the labour regimes framework, ILO had a neutral Chair of the Accord's Steering Committee (Fig. 2). Yet, ILO did not have decision making capacity within the Accord. In fact, as a promoter of tripartism (Prentice, 2021), ILO saw the Accord as "bypassing and undermining" (BD-INGO/2) the national government due to the independent manner in which the Accord operated: There are limits… They [the Accord] have antagonised so many people… They have put their own little system, like 'Vatican'. Their attitudes are unacceptable if you are a pro-Bangladeshi working with the sector… It takes decades to establish standards. And there is a role that everyone has to play to help establish the proper institutional structure, so one day, there will be a self-regulating industry (BD-INGO/2).
Indeed, ILO took a leading role in the development of RCC by providing resources, technical assistance, and strategic direction (BD-ICSO/1). ILO had its own international consultants based at RCC overseeing and facilitating the transition of the Accord. Thus, although ILO had long acknowledged the labour rights violations in Bangladesh, and the suppression of organised labour and the basic human rights of workers (Prentice, 2021), ILO Country office in Bangladesh essentially planned for the replacement of the Accord as early as 2015 with RCC (BD-Union/7). In referring to this, a prominent civil society organisation accused ILO of siding with the state-manufacturer alliance: They [ILO] are funding the government and BGMEA to defeat the Accord…. Relations between ILO, the government and workers are not equal. Because the collusion between factory owners and the government is strong, ILO leans towards them [the statemanufacture alliance] (BD-Civil Society1).
It appeared then that the alternative model provided in the form of RCC with the legitimacy of ILO was a welcome intervention for the statemanufacturer alliance, which deeply resented the interventions of the Accord. ILO facilitated a way out of the Accord for the Bangladeshi state and manufacturers which provided a disincentive for them to commit to the Accord and an incentive to oppose the Accord, which became viable strategic pathways.
Amid these developments, in May 2019, the Accord's global union signatories, BGMEA, and the government signed a Memorandum of Understanding that allowed the Accord to remain in the country for additional 281 days to process the handovers (BD-Technocrat/3). At this point, given the lack of confidence global stakeholders had on RCC, the agreement was to transfer the operations of the Accord not to RCC, but to a newly formed organisation -the Ready-Made Garment Industry Sustainability Council (RMG-SC). Ironically however, as also noted by Bair et al. (2020), BGMEA established from the outset that disciplinary actions for non-compliant factories cannot be taken by RMG-SC without the prior approval of BGMEA. This was a significant departure from the original Accord that did not require such approval, and thus, was a defining ingredient of its autonomy. As Reuters (2019) quoted a union: "The deal is sure to compromise the safety and security of garment workers given there will be no independent decision-making by the Accord''. Moreover, Bangladeshi unions affiliated with IndustriAll criticised the agreement, noting that they had not been involved in the negotiations (Bair et al. 2020). Thus, as of 2019, the future of the building safety programme remained uncertain.
The evolution of the Accord in the post-2019 context, in particular its re-configuration to RMG-SC is demonstrative of the dynamic nature of MSIs. The urgency to regulate Bangladeshi workplaces came with several factory fires, followed by the collapse of the Rana Plaza building soon after. Their impact was vividly visible and emotionally disturbing, connecting with millions of people, and especially consumers and activists in the West. As Bair et al. (2020) also noted, this was a moment when the future of the industry in Bangladesh looked uncertain. It was in this context that the Bangladeshi government and BGMEA took a backseat and allowed a free-hand to the Accord. By 2018, this urgency and the backlash was subsiding with the efforts of the building safety programme reducing the risk of sourcing from Bangladesh (Schuessler et al., 2019). The industry also prospered, with exports increasing from $21 billion in 2013 to $32.9 billion in 2018 (Bair et al. 2020, Fig. 2). Bair et al. noted, that this lack of economic fallout from Rana Plaza encouraged manufacturers and the state to take an increasingly aggressive stance against the Accord. The state-manufacture alliance thus retaliated by rejecting the Accord's request for an extension. In this respect, as Bartley (2018) also noted, although the Accord sought to substitute for the Bangladeshi state in the governance of the building safety standards in the short-term, it was unable to push the Bangladeshi government to enforce its own labour laws or build state capacities that could outlast the Accord's tenure and transform labour governance more broadly in the industry.

Conclusions
In this article I demonstrated the utility of applying the labour regimes concept to understand the workings of MSIs in global production networks. In so doing, this paper contributes to the existing debates on both labour regimes and MSIs. From the perspective of labour regimes, it pushes the analytical purchase of labour regimes beyond a sectoral/ geographical context. While the scope of the existing works on labour regimes are much larger, this paper extracts the main concepts used in labour regimesthe interplay of labour, capital, and the state across global production and territorialised dynamicsto craft an analytical tool to apply to MSIs in global production networks. The rationale for this exercise is that MSIs are also coordinated efforts between labour (workers and workers representatives), capital (lead firms and manufacturers), and the state. Moreover, MSIs are usually configured across multiple scales from the global to the workplace. Thus, the configuration of MSIs well align with the concept of labour regimes.
From MSIs perspective, linking them with labour regimes has three notable advantages. First, such a connection can help better understand the contested, but often obscured relations between labour, capital, and the state configured across multiple scales. As I have demonstrated, this helps isolate specific events, relations, and interactions between stakeholders and examine how they shape MSIs and their ability to deliver decent work. Most often, this means establishing direct links between decisions made at the global level and the national and workplace scales, so that the cause and effect drivers of MSIs can be properly connected, understood, and explained. This also allows a closer scrutiny at the intersection of 'global' and 'local', where currently, MSIsincluding the current case studyare noted as being driven mostly by global actors and are not adequately localised. Second, looking through the lens of labour regimes enables a closer look at the stakeholder composition itself, in particular, the actors who are included as well as those who are excluded from MSIs. As I have shown, this is highly relevant in a context where the interplay of politics and tensions between actors who participate in MSIs, who are excluded, and who have the authority to speak for those who are not represented, directly influence the likely outcomes. Third, labour regimes concept can reveal the temporal nature of MSIs as they evolve across time and space. On the one hand this helps understand the events that shape the existence and the progress of MSIs. On the other, this allows space to examine material as well as qualitative changes to MSIs at different points in time such as the stakeholder composition, objectives, and the outcomes. As I have shown in the current example, such a scrutiny can reveal how and why MSIs unfold across production networks and the dynamic nature of their effectiveness. The labour regimes concept thus can be an effectively tool to understand MSIs in global production networks.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Research Scholarship and the Graduate Research Support Scheme.

Research involving human participants and/or animals
The study involved interviewing human participants. All ethical approvals for this was obtained prior to the fieldwork from the Institutional Review Board of the National University of Singapore.

Informed consent
Each interview participant was informed of the project in detail and the explicit consent of the participants were obtained prior to the interview.