Firm participation in voluntary regulatory initiatives: The Accord, Alliance, and US garment importers from Bangladesh

Abstract

Most research on private governance examines the design and negotiation of particular initiatives or their operation and effectiveness once established, with relatively little work on why firms join in the first place. We contribute to this literature by exploring firms’ willingness to participate in two recent, high-profile private initiatives established in the aftermath of the Rana Plaza disaster in the Bangladesh ready-made garment (RMG) sector: the Accord on Building and Fire Safety and the Alliance for Worker Safety in Bangladesh. Using novel shipment-level data from U.S. customs declarations, we generate a set of firms that were “eligible” to join these remediation initiatives. We are able to positively attribute only a minority of US RMG imports from Bangladesh to Accord and Alliance signatories. Firms with consumer-facing brands, publicly-traded firms, and those importing more RMG product from Bangladesh were more likely to sign up for the Accord and Alliance. Firms headquartered in the USA were much less likely to sign onto remediation plans, especially the Accord.

Various efforts at transnational private regulation have accompanied the rise of global supply chain production. Activist and consumer pressure, often in the wake of “attention focusing events” (Lohmeyer and Schuessler 2018), frequently provokes such initiatives. Firms seek to improve their social and environmental performance, protect their brand names, and, perhaps, forestall other forms of regulation. These programs take varying forms: industry-wide codes of conduct, multi-stakeholder initiatives, certification efforts and accreditation schemes. Private sector auditors and NGOs typically take on monitoring and some enforcement roles.

Scholars across several disciplines have sought to understand the conditions under which these private governance initiatives emerge; their effects on labor, environmental, and sustainability outcomes; and the ways in which private governance interacts with public authority, in both home and host jurisdictions (Bartley 2018, Bartley et al. 2015, Berliner et al. 2015a, Berliner and Prakash 2015, King and Toffel 2009, Knudsen and Moon 2017, Locke 2013). While large, publicly-facing firms that sign on to initiatives often attract significant attention, we know much less about those firms—large and small—that do not participate. This is not surprising as it is difficult to identify the set of firms that might be implicated by a particular event or “at risk” of signing, especially when those firms operate across jurisdictions in developing countries. As a result, we know less about what motivates some firms to participate in private regulatory efforts, and what motivates other firms to remain on the sidelines.

In this article, we attempt to shrink this blind spot by focusing specifically on a single industry in a single, but significant, country. We examine how firms sourcing apparel and textiles in Bangladesh for the US market responded to a series of industrial accidents in Bangladesh’s ready-made garment (RMG) industry. In the wake of the 2013 Rana Plaza disaster—the largest ever in the garment industry—consortia of firms, labor groups, and NGOs established two private remediation and governance efforts: the Accord on Building and Fire Safety (hereafter, the Accord) and the Alliance for Worker Safety in Bangladesh (hereafter, the Alliance). Taken together, the Accord and Alliance present a useful case. First, they were arguably the most high-profile and ambitious private governance initiatives to emerge in the 2010s. Some activists and global brands touted these initiatives as new models for transnational private regulation, one that (in the case of the Accord) included binding commitments to building and fire safety, as well as a formalized role for local labor unions and Global Union Federations (GUFs). Second, they were bounded in space (Bangladesh), time (five years), and industry (RMG), making the identification of eligible firms far easier. Third, the initiatives emerged quickly after the Rana Plaza disaster and ensuing public outcry. Most signatories joined soon thereafter, mitigating concerns about diffusion or incremental learning across firms.

We theorize about what leads firms to participate in the Alliance and the Accord, and we test our claims empirically, using novel data from U.S. customs bills-of-lading to identify firms sourcing apparel from Bangladesh for the U.S. market. We merge this information with publicly-available data on firms’ ownership and managerial structure. While these data have their own challenges and limitations, we use them to construct a far better assessment of the set of firms who were “eligible” to sign the Accord and the Alliance. These data allow us to assess several theoretical expectations regarding the conditions under which eligible firms subscribe to private governance initiatives. We hypothesize, and find support for, a pattern in which “consumer-facing” firms (retailers and global brands) and firms controlled by a publicly-traded (versus privately held) parent are more likely to sign either the Accord or the Alliance. We also find, consistent with our expectations regarding firms’ reputational risk, that firms sourcing more product from Bangladesh are more likely to participate. And perhaps unsurprisingly, given the debate at the time of its creation, US-based firms were less likely to sign the Accord. Moreover, our results show that the volume of shipments that we can definitively trace to signatories of either the Accord or the Alliance represent a minority of the overall volume of American apparel imports from Bangladesh.

Taken together, our findings provide support for several existing arguments. Firms subject to broad public pressure—whether through consumers or shareholders—are more likely to participate in initiatives which seek to avoid industrial accidents or, more modestly, which provide firms with cover (in the form of noting participation in efforts to improve working conditions) in the face of pressure from activists or additional industrial accidents. Moreover, there appears to be distinctly different “corporate cultures” at play; US-based firms seem less willing to join these initiatives, particularly the more stringent of the two—the Accord (Knudsen 2017). Our findings also suggest, however, that if the firms sourcing from a given location vary in their ownership structure, scale of production and brand identity, private regulatory initiatives may find less success. To the extent, for instance, that the apparel sector in a given developing country is populated by privately-held firms producing mass market goods, perhaps with geographically diversified sourcing strategies, our findings suggest limited participation in, and success of, regulatory initiatives like the Alliance and the Accord. Moreover, as we note in the concluding section, our analyses of firm participation in the Bangladesh initiatives brings into question the effectiveness of labor-related governance efforts aimed at single producer countries, versus those aimed at an industry more broadly, or those focused on lead firms and brands (regardless of their production locations).

In what follows we briefly summarize transnational private regulation, especially as it relates to labor rights and working conditions in the apparel sector. We then describe the Accord and the Alliance. Next, we draw on existing literature to develop a set of expectations regarding firm-level decisions to participate in these initiatives. We test these hypotheses using the data derived from RMG shipments from Bangladesh to the United States. In the concluding section, we discuss the implications of our findings for broader questions of private sector labor rights governance.

1 Apparel production and the private governance of worker rights

Many consumer goods sold in Europe and the United States are produced in low- and middle-income countries, in factories that are part of multi-layered and complex supply chains. While “lead firms” vary in their sourcing behavior and differ in the control they exert over suppliers (Bartley and Child 2014; Bartley et al. 2015; Gereffi et al. 2005; Gereffi and Mayer 2010), retailers and brands are usually distant from production sites. Lead firms typically source from multiple factories, often in different countries. At the same time, individual factories often produce goods for several different lead firms and consumer markets (Anner 2018; Locke 2013; Osgood 2018). In the apparel sector, retailers and brands rely on market-based transactions to source inputs, cut and sew fabrics and complete final assembly (Dallas 2015; Gereffi 2014). Most garment production is low-skill, labor intensive, with modest capital equipment requirements. Berliner et al. (2015b) describe the garment industry as “a vast, widely extended and decentralized international network of suppliers, producers, and retailers combined with centralized brand control over design, quality and manufacturing” (p. 7).

The structure of apparel production generates immense pressures for factories to reduce costs, maintain quality, and decrease the turn-around time on orders (Bartley et al. 2015). The Multifibre Arrangement (1974–2004), which set country quotas for apparel and textile exports from developing to developed nations, limited some of these competitive pressures. The agreement also facilitated the initial growth of apparel production in countries – including Bangladesh – with favorable quota allocations (Knudsen and Moon 2017). The end of the agreement in 2004 heralded an era of increased competition (World Bank 2005). Brands', consumers' and retailers’ leverage over garment suppliers increased, as threats of exit became more credible (Anner 2018; Silver 2003).

Employment in the apparel sector can generate gains for relatively low-skilled workers; factory jobs offer an alternative to work in the rural or informal sector, especially for women (but see Blattman and Dercon 2016). Yet apparel production also creates incentives for factory owners and managers to ignore safety and structural problems and limit compensation to a bare minimum (Barrientos et al. 2011; Bartley et al. 2015; Berliner et al. 2015b; Locke 2013). When a surplus of low-skilled workers exists (Rudra 2002; Milner and Rudra 2015), or when economic downturns lead to reduced demand in key consumer markets (Lim and Prakash 2017), the consequences for workers may be even more severe. These perils often are exacerbated by the domestic political situations in apparel-producing countries (Adolph et al. 2017; Mosley 2011). Countries with weak state capacity lack adequately-staffed labor or health and safety inspectorates (Piore and Schrank 2008; Schrank 2013). Often, host country governments and business owners deny workers the legal or practical rights to form unions and bargain collectively (Ahlquist 2017, Berliner et al. 2015a, Zajak 2017). The politically privileged position of factory owners therefore allows them to capture many of the gains from global production.

Since the 1990s, activists have directed their spotlight at facilities owned by, or producing for, global brands. Through “naming and shaming,” activists brought consumer and shareholder pressure to bear on lead firms such as Apple, the Gap, and Nike (Bartley 2007, Bartley and Child 2014, Seidman 2007, Vogel 2010; also see Peterson et al. 2016). In response, many implemented corporate social responsibility (CSR) programs, either via their own codes of conduct or participation in industry-wide standards and certification schemes (Bartley 2018; Fransen and Burgoon 2014; Locke 2013; Marx 2008). These private governance initiatives allowed firms, NGOs and developed country governments to address some of the negative consequences of economic integration, without significantly limiting its scope (Bartley 2007; Fransen 2011; Knudsen and Moon 2017).¹

Some brands, including Knights Apparel, Apple, Levi Strauss & Co. and Nike, sought to define themselves as leaders in responsible production (Bartley et al. 2015; Berliner et al. 2015b). Firms based in developing countries used participation in global voluntary schemes to signal commitments to protecting workers and the environment (Berliner and Prakash 2014; Cao and Prakash 2011; Potoski and Prakash 2009). Indeed, Distelhorst and Locke (2018) find that manufacturing firms’ compliance with labor and environmental standards is associated with a significant increase in purchases by supply-chain partners (also see Görg et al. 2016).²

Research on effectiveness of these voluntary private governance schemes typically finds that they fall short in protecting workers, for a variety of reasons (Applebaum and Lichtenstein 2016, Berliner and Prakash 2015). These include a lack of direct monitoring of member firms (Lim and Tsutsui 2012; Berliner and Prakash 2015); corporate structures that privilege sourcing and strategy imperatives at the expense of CSR goals and fail to hold buyers directly accountable for factory conditions (Anner et al. 2013; Fransen 2011); and “forum shopping” among codes of conduct as well as among aspects of individual schemes (Anner 2012).³ Perhaps the greatest obstacle to the success of private governance schemes are local political institutions. Private governance appears most effective when it complements, rather than substitutes for, public sector governance (Amengual and Chirot 2016; Locke 2013; Locke et al. 2013). Bartley (2018) argues convincingly that many private governance schemes assume that it is possible to transcend local conditions when, in fact, these codes interact with, and are affected by, political interests and institutions in each host country. We return to this point – that domestic governance often changes the meaning and effect of transnational standards – below.

Existing analyses of voluntary private regulation, however, tend to focus on the broad design and effectiveness of such schemes. There is a relatively small literature on the determinants of firm-level participation in these schemes, particularly with respect to labor rights. On environmental governance, some consider how country-level factors, including public sector regulatory quality as well as levels of participation among trade and investment partners, affect national participation rates in voluntary environmental schemes (e.g. Berliner and Prakash 2014; Prakash and Potoski 2006; Prakash and Potoski 2007). Other analyses account for firms’ selection into voluntary programs, but they address firm selection to avoid bias in their assessment of program outcomes, rather than treating selection as of direct theoretical interest (e.g. Berliner and Prakash 2015; Potoski and Prakash 2005). In the realm of environmental programs, Hsueh (2017) finds that larger firms, as well as firms that already have a chief sustainability officer, are more likely to participate in voluntary climate initiatives. Similarly, when consumers demand attention to environmental concerns, or when other firms in the industry adopt environmental initiatives, firms are more likely to commit to voluntary environmental programs (Delmas and Toffel 2008). NGO campaigns may serve to heighten pressure from consumers and supply chain partners (Chrun et al. 2016; Innes and Sam 2008; Nikolaeva and Bicho 2011). Firms also may commit to voluntary regulatory initiatives as a means of avoiding attention or inspections from public regulators (Toffel and Short 2011; also see Videras and Alberini 2000).

With respect to labor-related initiatives, Fransen and Burgoon (2014) confine their analysis to apparel sector firms in Europe; they attribute these firms’ choices among private governance initiatives to their national corporate cultures and regulatory institutions. In an analysis of which footwear companies choose to join the Fair Labor Association, Marx (2008) highlights the role of NGO pressure, as well as public ownership status. The analysis of firm choices, especially in the apparel sector, is thus far, quite limited. Yet Hsueh’s (2017) analysis points out that firm-level attributes tend to be important at explaining variation in firms’ adoption of voluntary standards. We therefore focus our analysis on understanding why certain firms choose to join building and fire safety initiatives in Bangladesh – both in terms of why firms join any initiative, and why they choose to join either the Accord or the Alliance. This analysis, which is made possible by our use of shipments-level data, allows us to investigate program emergence as the product as firm-level decisions. Firm participation also affects the overall efficacy of private governance initiatives, an issue we return to in the conclusion.

2 The RMG sector in Bangladesh: Reformulating private governance

Bangladesh is the world’s second-largest exporter of garments, with over 7000 garment factories (Labowitz and Baumann-Pauly 2015). Figure 1 displays the growth of the RMG sector in Bangladesh, its importance to the overall economy, and the importance of the United States and the Eurozone as export destinations. Bangladesh’s apparel sector expanded significantly after the expiration of the Multifibre Agreement (MFA), which had allocated export quotas across a broad set of developing countries. Since the latter half of the 2000s, RMG earnings have accounted for approximatley 15% of Bangladesh’s GDP and over 80% of its export earnings. The US remains the single largest export market, but, taken together, the Eurozone countries—with Germany leading the way—overtook the US in the early 2000s. Lead firms, brands, and retailers in the industry are highly cost-sensitive, as garment consumers have come to expect inexpensive, rapidly-changing fashion. Part of Bangladesh’s growth as a garment exporter reflects the benefits of industrial agglomeration. But the government’s willingness to overlook sweatshop conditions and repress workers’ collective rights also have played an important role in attracting orders from demanding multinational customers (Berliner et al. 2015b; Fransen and Burgoon 2014).

Fig. 1

The Bangladeshi RMG sector (UNCTAD 2018)

The Bangladesh garment industry has been the site of numerous industrial disasters. A November 2012 fire at the Tazreen Fashion factory, in the outskirts of Dhaka, killed at least 117 people. The Tazreen facility subcontracted production for several foreign brands, including Netherlands-based C & A, Hong Kong-based Li & Fung, and US-based Wal-Mart. While Tazreen was the deadliest fire in Bangladesh’s history, it was only one of many factory fires in the RMG sector that year. Six months later, the eight-story Rana Plaza building in Sevar, another industrial area in Dhaka, collapsed, resulting in 1129 deaths and approximately 2500 injuries. Rana Plaza was designed as an office building, but several garment factories operated on its upper floors. These factories produced for a range of global brands, including Benetton, Bon Marché, The Children’s Place and Primark. While workers had reported their concerns over the building’s integrity to management, factory owners demanded that workers enter the building on the morning of the collapse. The April 2013 disaster attracted global attention, but again, it was only one among many industrial accidents in Bangladesh that year, albeit the deadliest.

Activists hoped that, like the 1911 Triangle Shirtwaist factory fire in New York, Rana Plaza might serve as a “focusing event,” leading to improvements in working conditions in the garment sector (Bair et al. 2017, Lohmeyer and Schuesser 2018, Schuessler et al. 2019). But activists also worried that the private governance tools of the 1990s and early 2000s were not adequate to protect RMG workers in Bangladesh. Indeed, the Tazreen Fashion factory had been audited only a few weeks before its fire.⁴ Tight connections between factory owners and politicians and policy makers rendered effective government regulation unlikely. Workers in Bangladesh lacked both the meaningful legal rights to form labor unions and the organizational capacity to voice concerns to management or inspectors.

Domestic union leaders and transnational labor activists therefore seized upon global media coverage of the Rana Plaza collapse as an opportunity to advocate for a new type of private sector regulation, one focused on a single set of problems (building and fire safety), in a single country, and that involved many overseas firms. A coalition of global unions and NGOs worked with retailers and brands to create the Accord. The creators of the Accord hoped that broad participation would assuage local firms’ concerns that improving worker health and safety would raise costs, thereby shifting production and orders away from Bangladesh.⁵

The Accord, signed in May 2013, and in place through May 2018, was based on a previously-launched initiative, signed by only two companies (US-based PVH and German-based Tchibo). It focused on building and fire safety, with an aim to involve international brands and retailers in the process of ensuring worker safety in Bangladesh’s RMG sector. As a tripartite initiative involving labor unions, NGOs and brands/retailers, the Accord required signatory firms to disclose subcontractor facilities in Bangladesh; fund (via firms’ membership fees) building safety and fire inspections of those facilities; and develop “corrective action plans” to remediate health and safety issues.

The Accord was a binding instrument, allowing signatory firms to be held legally accountable in their home countries for breaches of its terms (Knudsen and Moon 2017). Signatory firms committed to maintain their sourcing relationships with current supplier factories, at least during the first 2 years of the Accord’s five-year existence (Bartley 2018). The Accord did not necessarily prevent signatory firms from reducing the prices they paid to suppliers; but signatories were expected to set prices at a level which allowed suppliers to make fire and safety-related repairs, where necessary, and to generally operate safely. While the precise funding arrangements for remediation were not fully addressed at the Accord’s founding,^, brands and retailers agreed to share the costs of repairs with factory owners.⁶ Another component of the Accord sought to empower workers, providing them with training, a complaints mechanism, and elected factory-level health and safety committees. And given the weakness of local labor unions in Bangladesh, the Accord offered a voice for organized labor via the participation of IndustriALL and UNI Global Union, both global federations of unions (see Houssart et al. 2018, for instance). The Ethical Trading Initiative, a UK-based organization including firms, labor unions and NGOs, formed in 1998 to advocate for worker rights in global supply chains, urged its garment sector members to participate in the Accord (Knudsen and Moon 2017, Knudsen 2018).

Over 190 apparel companies from 20 countries signed the Accord, administered by the Bangladesh Accord Foundation, based in the Netherlands.⁷ The Accord’s signatories included 63 German firms, 33 British firms and 22 U.S. firms. Not all retailers and brands, however, were keen to participate in a legally-binding initiative. A smaller group of companies, largely based in North America, pursued a different program, the Alliance.⁸ The Alliance was launched in 2013, with the same five-year duration. Its signatories included Gap, Target, VF Corporation and Wal-Mart; NGOs and organized labor groups also participated in the Alliance’s development. While the Alliance included many of the same elements as the Accord, such as inspections, remediation and worker empowerment,⁹ its critics noted that the Alliance did not create legal obligations for its signatories; did not make its inspection reports public (it later changed this practice), reducing transparency; and did not prevent participating firms from setting lower prices for suppliers – effectively making it impossible for them to both improve worker conditions and continue as supply chain partners.¹⁰ Additionally, the “worker empowerment” components of the Alliance focused mainly on individual-level worker training and access to reporting mechanisms, rather than collective level (e.g. union-oriented) activities. Other observers worried that the creation of two separate initiatives would dilute the capacity of foreign brands and retailers to effect changes in the RMG sector.¹¹

When the Accord’s five-year term ended in May 2018, the organization had inspected more than 2000 factories, identifying over 150,000 fire and building safety hazards.¹² Each participating factory was inspected every three to four months, with more frequent inspections for those facilities with more severe safety issues. The Accord published 1571 corrective action plans for these facilities. Of the safety issues identified in initial inspections, nearly 90% were fixed. The Accord initiated a training program for workers at over 1000 factories and completed the program at 301 facilities by October 2018. Among other things, the program introduced workers to the Accord’s Safety and Health Complaints Mechanism. Via this mechanism, the Accord received over 900 complaints. Meanwhile, the Alliance completed over 900 inspections, some in factories also covered by the Accord.¹³ In its final report, in December 2018, the Alliance noted that 428 Alliance-affiliated factories had completed all material elements of their corrective action plans. Remediation included structural retrofitting and the installation of sprinkler systems. The Alliance offered its Fire Safety Training program in over 1000 facilities, covering over 1.5 million workers; and it oversaw a confidential worker hotline (the Amader Kotha Helpline)– accepting calls from workers in any factory, in any industry – which received approximately 5200 calls per month in 2017 and 2018.¹⁴ The Alliance reported that 181 factory-level worker safety committees had been formed, facilitating the reporting of workers’ concerns (although, notably, via a mechanism that did not involve labor unions).

While some observers note that many facilities did not address all critical items identified in inspections, others pointed to the significant number of factories that participated in the inspection and remediation process. Observers also frequently raised concerns about the gap between the costs of proposed repairs and the funds available to factory owners.¹⁵ Perhaps the biggest challenge, in terms of the longer-term situation in Bangladesh, concerned the 2018 expiration of both initiatives. After lengthy negotiations, the Accord’s stakeholders agreed in June 2017 to extend the program for a second five-year term after the initial program expired.¹⁶ The “Transition Accord” was intended to operate for three (and possibly four) years. In May 2018, however, a Bangladeshi High Court ordered the Accord to end its work by November 2018. While the Supreme Court issued a stay against this order, the Accord’s long-term prospects in Bangladesh remain clouded. The government asserts that its regulators are now prepared to assume the Accord’s role as enforcer of building and worker safety; some RMG factory owners assert that the Accord’s members never provided sufficient funds for factory upgrades.

Meanwhile, the Alliance ended its operations in Bangladesh on December 31, 2018, stating that it was transferring its worker helpline and training programs to local organizations in Bangladesh, and that its regulatory functions would be assumed by a Bangladesh-based body. In its final months, the Alliance noted that it hoped to redirect some of its activities to the government of Bangladesh, perhaps in conjunction with the Bangladesh Garment Manufacturers and Exporters Association. In March 2019, several former Alliance members and other brands formed the Nirapon initiative, a locally-managed organization with a mandate to assume some of the factory safety monitoring and training duties previously performed by the Alliance.¹⁷ The extent to which brands and retailers will participate, and to which Nirapon will have sufficient financial or regulatory resources, remains unclear. One might well imagine that factory owners will be hesitant to address worker safety issues, given fewer pressures from global brands or from the Bangladeshi government, (e.g. Anner et al. 2018, James et al. 2018). Indeed, a fire in a residential area of Dhaka in February 2019 resulted in over 70 deaths; it was the result of the illegal storage of highly combustible industrial chemicals in a residential building.

Although the Accord and the Alliance differed in some important ways,¹⁸ both asked foreign firms to take some responsibility for working conditions in supplier factories in Bangladesh. Both created systems to inspect and (at least partly) fund repairs in those factories (also see Anner et al. 2013, James et al. 2018). Both initiatives aimed to involve garment sector workers, albeit in different ways: the Accord envisioned a role for labor unions (despite the barriers they face in Bangladesh), while the Alliance focused on training workers on the agreements’ provisions and encouraging them to directly report violations. Like Better Work, a joint ILO/IFC initiative, the Accord and the Alliance brought together a range of stakeholders, including brands and retailers; NGOs; intergovernmental organizations and (in the case of the Accord) global labor union federations.¹⁹

The Accord and the Alliance focused on a narrow set of issues, rather than on core labor rights, such as freedom of association and collective bargaining. Indeed, the initiatives did not even include some elements of building safety, such as the integrity of a facility’s boiler room.²⁰ Notably, the Accord and the Alliance focus on a single (critical) industry in a single country. Rather than focus on multinational firms across a range of industries and countries, like the Ethical Trading Initiative or the UN Global Compact (see Bartley et al. 2015; Berliner and Prakash 2015; Bernhagen and Mitchell 2010), the Accord and the Alliance focused on export-oriented suppliers only in Bangladesh’s RMG sector. Inspection reports as well as interviews indicate that these initiatives have – despite resistance from some factory owners and government officials – improved factory conditions in Bangladesh overall (James et al. 2018).

The initiatives exemplify a “retailers to the rescue” model (Bartley 2018), in which lead firms and brands attempt to use their leverage as buyers to affect conditions in their supplier factories. As an example of “non-state market-driven” CSR (Auld et al. 2008), these initiatives offer reputational benefits (or the avoidance of reputational costs) to participating firms (Brookes 2019). These programs – especially the Accord, given its legally binding nature – could be viewed as substituting for public governance. That is, inspection and enforcement functions become the domain of the initiatives, rather than of the state. Bartley (2018) suggests that, while such programs might be a more effective means of influencing labor-related outcomes, they also remove pressure for public bodies to become more accountable, or for the government to improve the content and enforcement of domestic labor laws.

While we share Bartley’s concerns about the prospects for longer-term changes in developing country workers’ conditions, we focus on a more immediate question: which brands and retailers elect to participate in these programs? Firm participation in these initiatives is a necessary, albeit not sufficient, condition for their success in improving worker safety. Once the Accord and the Alliance were launched, some firms producing in Bangladesh joined the Accord; others turned to the Alliance; and still others eschewed participation in either program. Firms signing on to the Accord or the Alliance trumpeted their participation loudly; yet we know little about firms who had sourcing relationships in the Bangladesh RMG sector, but chose not to take part in the Accord or the Alliance.

3 Who signs the Accord and the Alliance?

The effectiveness of private regulation in Bangladesh depends not only on the extent to which factory inspections identify health and safety issues and on global brands’ willingness to contribute toward remediation, but also on the earlier decision by brands and buyers to participate in the initiatives. The private regulation model assumes that brands and buyers participate in the Accord and the Alliance: brands are central not only as sources of financing for remediation, but also as sites of pressure on factories. If brands sourcing from Bangladesh do not participate in the Accord or the Alliance, the incentives for factory owners to make factory improvements are significantly reduced (Anner et al. 2013). Yet analyses of private governance have more often focused on their effects on firm-level behavior, rather than on the determinants of membership. Even when studies address selection into a governance program, they treat such analyses as a requirement for identifying program effects, rather than as a question of central interest (e.g. Berliner and Prakash 2015).²¹

We therefore focus on explaining which firms are most inclined to participate in private governance initiatives. Our customs data (see below) suggest that buyers who did not sign the Accord or the Alliance account for a significant proportion of the RMG product imported into the US from Bangladesh. We assume that the overall motivation for private governance is material, akin to Vogel’s (2005) “market for virtue.” Some consumers, supply chain partners and investors reward firms for “beyond compliance” behavior, and they punish firms for appearing to violate worker rights or to degrade the environment (Auld et al. 2008; Distelhorst and Locke 2018; Locke 2013). Participating firms might therefore avoid the reputational harm that results from industrial accidents or reports of specific labor rights violations, even several nodes removed in the supply chain (Brookes 2019); as such, membership offers “club goods” to participating firms.

Moreover, consumers and shareholders may assess an industry in its entirety, rather than evaluating firms individually. They may hold a view on the general working conditions associated with clothing labeled “Made in Bangladesh.” To the extent that this is the case, apparel firms – especially those with a high proportion of their production in the specific location associated with private governance or with recent industrial accidents – may have reason to encourage other firms to take part in private regulatory initiatives. Indeed, those firms with stronger individual incentives to participate (see below) may worry that, if others in their industry sourcing from a given country also do not commit to voluntary private regulation, participation in private governance initiatives will leave them at a competitive disadvantage (Chrun et al. 2016; Fransen 2011; Nikolaeva and Bicho 2011; Vogel 2010). The Bangladesh initiatives therefore may have some “public goods” features as well: if they succeed at making industrial accidents less likely, they offer protection from reputational harm to all firms sourcing from Bangladesh. But because these benefits accrue to the industry broadly, there are significant incentives to free ride. The costs of the Accord and the Alliance – membership contributions, potential expenses related to building repairs, commitments not to relocate production, and exposure to legal action in the case of the Accord – are borne only by participating firms, but the reputational benefits could accrue more broadly. We expect that the incentives to free ride will be most likely to prevail for firms not facing public or shareholder pressure, and for firms that source smaller amounts from Bangladesh.

With these general club and public good motivations in mind, we also theorize about firm-level variation in the incentives to participate in private regulatory initiatives. Which RMG lead firms and retailers are therefore most likely to perceive net material benefits from joining the Accord and/or the Alliance? We hypothesize that three characteristics are relevant to explaining firm participation: the degree to which the firm has a strong brand identity; the firm’s ownership structure (publicly traded versus privately held); and the location (country) of the firm’s headquarters.²² We also expect that firms with more activity in Bangladesh will be more likely, all else equal, to participate in private governance initiatives: these firms are more exposed to country-level risk and, therefore, are less likely to engage in free-riding.

First, public pressure from NGOs and activists can enhance firms’ incentives to participate in private regulatory schemes designed to protect or improve labor rights (Bartley and Child 2014). Firms with stronger brand identities, especially among more affluent consumers in the developed world, should gain more from “beyond compliance” activities (Bartley 2018; Gereffi et al. 2005; Gereffi and Mayer 2010). RMG production in Bangladesh usually involves final (rather than intermediate) goods,²³ but apparel sector firms nonetheless vary in the extent to which they have strong brand identities. Fransen’s (2011) survey of the CSR preferences of Western European firms, for instance, reveals that firms that fall into lower consumer market segments (e.g. mass merchandisers and discount stores) are significantly less likely to support stringent private regulation of labor conditions. While most apparel consumers remain price-motivated (Berliner et al. 2015b), those consumers who pay a premium for branded products may be willing to bear the additional costs that come with better working conditions (Vogel 2010). For instance, Nike – which has long faced pressure from labor rights activists as well as university licensees – became the first company, in 2005, to disclose publicly a complete list of its factories (Berliner et al. 2015b). While Nike had previously led a movement against factory disclosure,²⁴ claiming that making such information public would undercut their competitiveness, they came to view “responsibility” as central to their brand identity. We can contrast branded, higher-end firms with mass market retailers; the latter serve consumers who are motivated mostly by price. All else equal, consumers who purchase branded and luxury, rather than mass market, products tend to express greater concerns with labor and environmental conditions (Bartley and Child 2014, Gereffi 1994, Vogel 2010; also see Hainmueller et al. 2014, Hainmueller and Hiscox 2015); they are more likely to punish firms in response to industrial accidents (O’Rourke 2003).

Note that human and labor rights activists play a key role in alerting consumers (and shareholders) to violations in far-away production locations (e.g. Keck and Sikkink 1998, Peterson et al. 2017). Given the assumed differences in consumers’ receptivity to rights-related campaigns, activists are more likely to engage in “naming and shaming” of branded, consumer-facing firms, especially those with positive corporate reputations (Bartley and Child 2014; Marx 2008).²⁵ Such firms will most appreciate the “boycott shield” offered by voluntary initiatives (Auld et al. 2008). We therefore expect that firms' strong brand identities, which typically occupy higher-tier segments of consumer markets, will have the strongest material incentives to commit to private regulations (Arora and Cason 1996, Bartley 2018, Bartley et al. 2015, Fransen 2011, Hsueh 2017), including the Bangladesh initiatives.

Second, we anticipate that firms accountable to shareholders will generally perceive greater benefits from participation in social compliance programs; these benefits include protection from reputational risk (O’Rourke 2003; Vogel 2010). Publicly traded firms are more visible, and often larger, than their privately-held counterparts. The revelation of rights-related violations can have immediate consequences for equity market valuations (see, for instance, Bartley 2018, Marx 2008). Participation in private governance initiatives offer at least plausible deniability in the face of industrial accidents. For instance, in January 2018, the Bangladesh Investor Initiative, a group of socially responsible investors including shareholders in several apparel retailers, urged major retailers to sign on to the Transition Accord (the post-May 2018 scheme). The group expressed concern that, at the time, only sixty Accord signatories had joined the Transition Accord. This call came soon after two global retailers had reached arbitration settlements at The Hague, which underscored the importance of sector-wide participation, as well as the role of GUFs, in the Accord’s effectiveness (Houssart et al. 2018). Additionally, listed firms which participate CSR initiatives become eligible for inclusion in various indices and funds with “sustainability” mandates (Fransen 2011; Vogel 2010). Marx’s (2008) qualitative study of footwear companies’ participation in the US-based Fair Labor Association (FLA) concludes that publicly-traded status is a necessary condition for participation.

This said, it is worth acknowledging the possibility of countervailing pressures related to ownership structures: shareholders’ expectations of strong quarterly returns combined with the liquidity of equity and debt instruments could render publicly-traded firms more sensitive to the costs – in terms, for instance, of contributions to remediating safety hazards, or maintaining commitments to suppliers in Bangladesh – of social compliance programs. Indeed, Fransen and Burgoon’s analysis (2014) of European apparel firms reveals no significant relationship between firm participation in voluntary governance initiatives and their ownership structure. On balance, though, we expect that, given anticipated pressure from shareholders, publicly traded firms will be more likely to embrace efforts to improve conditions in Bangladesh.

Among privately-held firms, we expect greater variation. If firm owners are strongly committed to the protection of labor – as, for instance, Levi Strauss & Co. and Knights Apparel have been (Berliner et al. 2015b) – then private ownership makes such practices easier to implement, even when those practices reduce firm earnings. If, on the other hand, privately held firms’ owners have little regard for working conditions, they will face no contending pressure from shareholders. We anticipate significant variation among privately-held firms’ motivations, leading to a less systematic relationship between private ownership and Accord or Alliance participation.

Third, and despite the globalization of production and consumption, we anticipate that firm nationality affects participation in the Accord and the Alliance. While we acknowledge differences among European countries, we expect a more pronounced difference between firms headquartered in the US and firms headquartered in Europe, for several reasons. Corporate cultures emerging from continental European countries anticipate a greater role for organized labor and a more equitable division of profits between workers and management. These cultures often reflect national laws and institutions (Auld et al. 2008; Fransen and Burgoon 2014; Schuessler et al. 2019). Although some European firms may hope to flee from these higher standards when sourcing product abroad, European shareholders, consumers and regulators are inclined to hold their firms to higher standards (Nikolaeva and Bicho 2011; Rathert 2016). The same logic that implies that trade with higher-standards countries can improve labor standards abroad (Cao et al. 2013; Greenhill et al. 2009) also suggests that firms from countries with higher labor-standards should be more likely to participate in labor-protecting arrangements when sourcing from overseas. European firms may be more inclined than their US-based counterparts to view their relationships with NGOs as cooperative rather than adversarial; Sasser et al. (2006) cite this difference as helping to explain forestry sector firms’ choices between the NGO-sponsored and industry-backed private governance efforts.

More broadly, the European “regulatory state” (Bach and Newman 2007) is more protective in its approach to workers (as well as it its approach to privacy and to health and safety). These protective efforts frequently include reliance on voluntary private standards, especially multi-stakeholder arrangements, as sources of governance (Knudsen 2018, Vogel 2010). This approach may offer additional incentives for firms to participate in voluntary private regulation: some European governments require publicly-traded companies to issue annual reports on their social and environmental practices. In other cases, public pension funds are required or at least encouraged to consider social compliance outcomes when making asset allocation decisions. And, in 2017, France became the first country to pass legislation requiring large firms to mitigate social and environmental risks (including human rights violations) in their supply chains, or face legal action (Evans 2020).

Therefore, we expect that firms based in Europe are more likely than their US-based counterparts to participate in governance initiatives in Bangladesh. This is particularly true for the Accord, which creates greater legal obligations, and which emphasizes collective (versus individual) worker empowerment. This is consistent with Sotorrio et al’s (2008) analysis, which finds that even among the best-reputed firms, European firms are on average more socially responsible than their North American counterparts.

Finally, in the specific cases of the Alliance and the Accord, firms’ decisions regarding participation also relate to the initiatives’ history. It was a small group of U.S. brands and retailers that promoted the formation of the Alliance, as it provided a non-legally binding alternative to the Accord. But this is not the only causal pathway by which firm nationality – operationalized below as the country in which a firm’s administrative headquarters are located – affects participation. We expect that U.S. firms are less likely than their European counterparts to sign on to either of the Bangladesh initiatives. Note that we also could consider variation among European countries: analyses of CSR programs more generally point to the Nordic countries, as well as the Netherlands, Switzerland and Denmark, as leaders in ethical consumption and corporate social responsibility efforts. Other continental economies tend to be “followers,” while Mediterranean states often are last to join such efforts (Knudsen and Moon 2017; also see Fransen and Burgoon 2014). Our empirical strategy, however, prevents us from differentiating among European countries, so we leave intra-European differences as a topic for future research.

4 Empirical analyses

4.1 Bill-of-lading data

We test our expectations regarding the correlates of firm participation using a unique dataset. Existing work on the Accord and the Alliance has taken the form of research reports (e.g. Labowitz and Baumann-Pauly 2014, 2015) and business school case studies (AlBhadily 2015, Subramanian 2016) designed to tease out whether the Accord or Alliance has managed to identify dangerous factories and secure improvements in working conditions (also see James et al. 2018). No systematic work has attempted to explain why some RMG importers responded to Rana Plaza by joining a private governance initiative, while others did not.²⁶

To evaluate our hypotheses regarding participation, we first must identify firms that could have signed the Accord or Alliance. We take the relevant universe of firms to be those who were sourcing RMG product in Bangladesh at the time of the Rana Plaza collapse. While the ideal dataset would measure firms exporting from Bangladesh to all destinations, customs data are collected by importing jurisdictions. We have access to data via bills of lading from US Customs, but not from other countries’ customs authorities. Given the size of the US apparel market, however, many non-US firms also do business in the US, so these data allow us to identify many RMG firms sourcing from Bangladesh, with notable variation in ownership structure, firm nationality, and brand identity.

A bill of lading is required for any shipment arriving in the US whether by land, sea, or air. We contracted with the firm Import Genius to acquire this data. We provided Import Genius with a list of keywords that could appear anywhere in the bill of lading (see the supplemental materials). Import Genius then provided us with data for each shipment with a bill of lading containing one or more of our search terms, originating in Bangladesh, and arriving in US ports.²⁷ Given the cost of the data and our primary goal of identifying the universe of firms importing RMGs into the USA, we purchased data for three time windows: 15 August - 31 October 2011, 1 July-30 November 2012, and 1 March 2013–31 August 2015. These windows reflect the periods of peak arrivals for the key US holiday retail season, allowing us to capture both the most active periods as well as variation over time. The data contained in the bills of lading include the date and port of arrival, the size of the shipment (in kilograms and number of containers), as well as the consignee name and address.

Our data allow us to answer questions about firm participation that would otherwise be intractable. Nevertheless, we are careful to acknowledge the limitations of our customs-based data. First, our dataset relies of the set of search terms we provided. In an effort to err on the side of expansiveness, we intentionally selected terms that may pick up more than just RMG shipments. Some shipments in our dataset, for instance, included sheets or towels, rather than garments. Given our search terms, we assume that we pick up some shipments that are not strictly RMG; but we may miss other shipments that belong in the RMG category.²⁸ In assembling the Import Genius data, we uncovered 2192 shipments that failed to conform to or parameters (out of 99,883 total shipments); they were not, in fact originating in Bangladesh or the bill of lading information indicated that the goods were not, in fact RMG products. We remove these shipments and the recipient firms associated with them from our dataset, if these were the only shipments to them in the data.²⁹

Second, some firms are not eager to be identified in US Customs filings, especially as third parties such as Import Genius have started making bills of lading data more accessible to industry competitors. Firms may take pains to camouflage their identities. As a result, there may be shipments going to Accord or Alliance firms that are not attributed to them, and there may be US importers of Bangladesh-sourced RMG that we miss entirely. Moreover, our dataset includes some shipments that could not be attributed to any firm by name. We therefore interpret our findings as representing a lower bound on the volume of Bangladeshi RMG shipments attributable to Accord or Alliance signatories.

Third, the Import Genius data are limited to firms with customers or operations in the United States. Firms that do not import into the USA do not appear in these data. Firms operating in several countries may have made different sourcing and shipping decisions for different markets. We are only able to observe what they did with their American operations. Because our data only include firms that were importing RMG product into the USA, non-American firms operating in the USA are, by construction, likely to be larger, transnational “core” brands whereas the set of American firms will include logistics firms and wholesalers as well as retailers. For this reason, we condition our analyses on both the volume of RMG imports into the USA and a firm’s nationality. Doing so allows us to better distinguish “size effects” from attributes related to firm nationality, corporate structure or brand identity.

Notwithstanding these shortcomings, our data provide the best extant documentation of the universe of firms “exposed” to the Rana Plaza shock, especially among those with operations in the United States. While there surely are drawbacks to using the bill of lading approach, our search term and tools are constant over time, allowing us to evaluate changes using a common benchmark. The overall picture painted by our data (discussed around Appendix Fig. 1) is one in which there is little systematic effect of the Rana Plaza disaster on shipments of RMG from Bangladesh to the US. Rather, consistent with the goals of both the Accord and the Alliance, global brands continued to source product from Bangladesh after the Rana Plaza event.

4.2 Firm participation in remediation plans

To test our hypotheses, we take the consignee (listed in the shipment data) to be the firm of interest. Using a combination of text analysis tools and subsequent hand coding, we identified a set of 1356 unique entities receiving RMG shipments to the US. For each of these firms, we search public databases to determine the firm’s business type (a brand, a retailer, a logistics firm, a manufacturer, a wholesaler, or other/unknown, which might include a bank or other financial owner); the country in which the firm (and corporate parents, if any) sites its administrative headquarters; whether the firm (or its corporate parent, if any) is publicly traded; and whether the firm signed the Accord, Alliance, both or neither. We also code whether a firm is a subsidiary or a stand-alone firm. We consider a firm to be “consumer-facing” if its business is coded as either a “brand” or a “retailer”.

To measure each firm’s involvement in Bangladesh, we aggregate shipments up to the recipient firm level for the entire period of our data. We identify 24 entities that either signed the Accord or are subsidiaries of firms that did. These 24 entities represent 22 different Accord signatories. We identify 29 entities that either signed the Alliance or are subsidiaries of Alliance signatories. These 29 entities represent 23 different Alliance signatories. Of these firms, one (Fruit of the Loom, a subsidiary of Berkshire Hathaway) signed both. The Alliance reports that, of firms signing by March 2017, 24 were based in the USA. Of all firms signing the Accord by March 2017, 23 were based in the USA. This indicates that our US-centered dataset does a reasonable job of picking up the signatories operating in the USA, particularly for the Alliance.

With respect to our dependent variable, firms could choose to sign the Accord or the Alliance, but these choices are not necessarily mutually exclusive. Additionally, the underlying number of signatories to each plan is relatively small. For these reasons, we fit three sets of logistic regression models rather than a multinomial model.³⁰ The outcome variable for the first set of models is an indicator variable for whether a firm signed on to either initiative. The second set uses an indicator of whether the firm signed the Accord as the outcome of interest, while the third set of models describes whether the firm signed the Alliance. We use the same set of covariates for all regressions. We include shipping volumes (in log kilograms) as an indicator of the firm’s relative level of engagement with the Bangladeshi RMG sector. We also include dummy variables for whether the firm (or its corporate parent) is headquartered in the USA. As displayed in Table 1, there is significant variation in firm nationality. Sixty-two percent of firms identified as Accord signatories are headquartered in countries other than the US; relative to the overall composition of Accord signatories, U.S. based firms are overrepresented. Among the Alliance signatories in our dataset, 79% are US-based. This approximates the composition of Alliance signatories overall.³¹

Table 1 Distribution of firms’ HQ nationality by remediation affiliation

Our models also include a measure of whether a firm is “consumer facing,” i.e., a global brand or retailer. We view our “consumer facing” indicator as describing (in coarse terms) a firm’s visibility to the broader public and therefore its attractiveness as target for activists and others seeking to induce participation in private governance initiatives. We also include an indicator for whether a firm (or its corporate parent) is publicly traded. Given the small number of firms signing, we are concerned with the problems with “rare events” in our data. We therefore include penalized (Firth) regression model in each set of regressions.³² Tables 2 and 3 display results.

Table 2 Which RMG Importers to the USA joined any private remediation plan?

Table 3 Which importers to the USA signed the Accord and Alliance?

The regression results in Table 2 are consistent with our hypotheses. Consumer-facing and publicly-traded firms, as well as those sourcing more product from Bangladesh, are more likely to join either agreement. These results imply that branded firms face greater reputational risks from industrial accidents such as Rana Plaza, and that these reputational concerns are on average more pronounced for publicly-held firms. Consistent with the idea that there are important cross-national differences in corporate attitudes toward CSR initiatives, firms based in the USA are less likely to sign on to any remediation plan; this also holds after controlling for a firm’s volume of imports into the USA.

In Table 3 we examine Accord and Alliance signing separately. Our findings are largely similar to those reported in Table 2 with one noteworthy exception: among those importing to the USA, firms that are also US-based are significantly less willing to sign on to the Accord. This is consistent with some U.S. firms’ concerns, stated publicly in the wake of Rana Plaza, that the Accord created excessive legal liability for brands (also see Houssart et al. 2018). At the same time, there is no discernable relationship between firm nationality and Alliance participation. As before, it is important to control for import volumes because the non-US firms in our sample tend to be larger, transnational firms. Consumer-facing and publicly traded firms remain more prone to join the private remediation initiatives, but the magnitude of these relationships are reversed for Accord and Alliance signatories, reflecting the differences in the mean levels of these indicators for Alliance and Accord signatories in our data. Further analysis does not show any evidence that public listing and consumer-facing status interact in any meaningful way (see, e.g., Appendix Table 4). In the appendix we also report analyses including only those firms with ultimate administrative headquarters in the USA. These regressions largely confirm that consumer-facing and publicly traded firms are more likely to sign on to remediation programs.

The results in Table 3 reveal important differences in corporate practices: even after accounting for size and public visibility in both retail and financial markets, US-based firms are less likely to sign on to remediation plans, especially the Accord. By way of interpretation, Fig. 2 displays the relative risks for Accord and Alliance signing for each covariate. Because the underlying rate of signing either agreement is low, we focus on the (predicted) relative risk of signing the Accord and the Alliance (respectively), holding the other covariates at their central tendencies in the observed data.³³ For the shipping volume variable we compare the predicted probability at its first and third quartile values; dummy variables are toggled. All interpretation quantities are derived from the logit models with the full slate of covariates.

Fig. 2

Relative risk of signing the Accord or Alliance for a change in covariates Note: Solid points have 95% confidence intervals that do not contain 1

In Fig. 2, we use circles to denote the Accord and triangles to signify Alliance. The vertical grey bar is at a relative risk of one. Because confidence intervals are wide, making it difficult to see the plot, we omit them and instead use filled points to denote estimated relative risks in which the 95% confidence intervals do not contain one. The figure indicates that a firm at the third quartile in its sourcing of RMG product from Bangladesh is about twice as likely to sign the Accord compared to a similar firm at the first quartile; that firm is over three times more likely to sign the Alliance. Firms headquartered in the US are 85% less likely to sign the Accord than those headquartered elsewhere. But US-based firms are no more likely than others to sign the Alliance, once we condition on the scale of the firm’s connection to Bangladesh, ownership, and consumer exposure. Consumer-facing firms are also more likely to sign these agreements. Although the point estimates for the Accord regressions are about twice as large as for the Alliance, the difference between the two is not statistically significant at conventional thresholds. Public companies are much more likely to sign either agreement, but the effect is far larger for the Alliance (fifteen times more likely to sign than a privately held one) than for the Accord (three times more likely) among the population of firms we study here. These results again confirm our expectations: ownership, brand identity and nationality have important independent influences on firms’ willingness to participate in private sector efforts to govern working conditions.

4.3 Accord and Alliance signatories and import volumes

Now that we have identified the predictors of firm participation, how much of the shipping volume in our data can we positively attribute to identified Accord and Alliance signatories? The answer is a relatively small minority. Shipping volume to Accord and Alliance signatories was 4% and 13% of total volume, respectively, regardless of whether we aggregate shipments over the entire observation period or average monthly. Even recognizing that these values should be treated as a lower bound, we nevertheless consider this modest proportion inconsistent with claims that Accord and Alliance signatories account for a large proportion of US RMG imports from Bangladesh.

While additional work will be needed to confirm this finding, there is an important substantive caveat as well. We focus on product shipped to Accord or Alliance signatory firms, as distinct from product manufactured in a factory that counts Accord or Alliance signatories as customers. Many factories produce for a variety of customers, so it is quite possible that a significant proportion Bangladeshi RMG factories are subject to inspections and improvement under the Accord or Alliance even if Accord and Alliance signatories represent a minority of overall RMG volume exported to the US. The bill-of-lading data say little to nothing about the factory that produced the product in any particular shipment, so we are not able to speak directly to this issue. An important question for future work involves identifying what proportion of Bangladeshi RMG production emerged from factories that were inspected under the Accord or Alliance; preliminary reports from the organizations themselves suggest that it could be quite substantial.

5 Conclusion

What leads firms to sign on to potentially costly private governance initiatives? The influence of firm-level characteristics on this decision has received little empirical attention, especially in the domain of workplace safety and labor rights. We examine the Accord and the Alliance, two private governance initiatives that emerged among global RMG brands and retailers in the aftermath of the Rana Plaza disaster. In order to identify which firms are “at risk” for signing the Accord and Alliance, we employ novel US Customs bill-of-lading data. We therefore are able to test a set of expectations regarding how firms in the RMG sector reacted to the Rana Plaza disaster, at least among those selling to the US market. We find that consumer-facing, publicly-trade and European-based firms are more likely, all else equal, to participate in these initiatives. We also report evidence that participating firms account for only a small percentage of RMG exports to the United States.

Given that the success of private regulatory initiatives rests on participation by brands and retailers (e.g. Anner 2013), identifying the determinants of membership also improves our broader understanding of voluntary private regulation. We find that numerous firms refrained from participation, perhaps because many firms share production facilities and some could free ride on the efforts of the few. Or perhaps those firms that are privately-held, US-based, and without clear brand identities have very limited material incentives to participate. At the same time, when “brand” considerations involve both consumers and shareholders, we can expect the incentives to participate to be relatively strong.

With respect to lower levels of participation by privately-held, non-branded firms, we might worry that programs like the Accord and the Alliance may not cover a large enough share of the country’s industrial production to move the needle on worker safety. In our data, we can link firms signing either of the remediation plans to only a minority of all the RMG imports from Bangladesh. Over the April 2013 to August 2015 period, shipments in the “neither” category averaged 82% of the US RMG imports from Bangladesh.³⁴ Our analysis also suggests that rights activists might consider whether there exists an effective means of drawing attention to the behaviors of privately-held, mass market retailers. Such firms are likely to be very cost-motivated; only if sourcing from problematic factories leads to higher product costs might these firms elect to participate directly in private governance. Perhaps, then, these firms are more affected by country-level changes to market access – through labor-related conditions in preferential trade agreements and unilateral preference programs, for instance – than by private governance schemes.

Moreover, a reasonable concern with single-country private governance initiatives is that they might encourage firms to shift their sourcing to other jurisdictions with a dimmer spotlight, leaving Bangladeshi workers worse off while failing to improve overall labor safety in the industry. Establishing causality in this regard, even when using the bill-of-lading data, is quite challenging. Nevertheless, our preliminary conclusion is that this appears unlikely to have occurred. During the post-Rana Plaza period covered by our data, shipping volumes to the US (in millions of kilograms and as a percent of all apparel shipments) appear quite stable; there does not seem to be any shift away from Bangladesh (although there still may have been foregone investment or capacity expansion). More rigorous evaluation of causality in this case will require additional empirical evidence.

Related to the concern about shifts in production to other countries is the question of whether a more effective approach would focus at the industry-level, across producer countries (e.g. Brookes 2019). One might imagine, for instance, that targeting large lead firms no matter where they produce, or holding lead firms accountable in their home countries (Evans 2020; Knudsen and Moon 2017) could be more effective than attempting to regulate in a single country, especially one in which national government actors are not necessarily interested in labor-related upgrading (Anner 2019; Ashwin et al. 2019). Furthermore, it is worth noting that the success of all private governance initiatives requires not only a de jure commitment to participation, but also de facto compliance – for instance, aligning the commitments made by the social compliance division of a lead firm with the decisions made by its purchasing and sourcing departments (Amengual and Distelhorst 2019; Anner 2019). As such, firm membership in private voluntary initiatives is a necessary, but not sufficient, condition for labor-related upgrading.