Governance gaps in eradicating forced labor: From global to domestic supply chains *

A growing body of scholarship analyzes the emergence and resilience of forced labor in developing countries within global value chains. However, little is known about how forced labor arises within domestic supply chains concentrated within national borders, producing products for domestic consumption. We conduct one of the ﬁ rst studies of forced labor in domestic supply chains, through a cross-industry comparison of the regulatory gaps surrounding forced labor in the United Kingdom. We ﬁ nd that understanding the dynamics of forced labor in domestic supply chains requires us to conceptually modify the global value chain framework to understand similarities and differences across these contexts. We conclude that addressing the governance gaps that surround forced labor will require scholars and policymakers to carefully re ﬁ ne their thinking about how we might design operative governance that effectively engages with local variation.


Introduction
Over the last two decades, the global value chain (GVC) theoretical framework has been deployed to analyze the drivers of labor exploitation in the global economy and the governance gaps that facilitate it. 1 Beginning with Hopkins and Wallerstein's (1977) early use of the global commodity chain concept to theorize the changing international division of labor in the late 1970s, the GVC framework has been at the heart of scholarly efforts to understand the impact of economic globalization on labor standards. An important blind spot in the literature, however, concerns the dynamics of labor exploitation in domestic supply chains. The prevailing emphasis has been on the reorganization of production into supply chains feeding worldwide markets, and particularly on the challenges and opportunities created for workers in export-oriented sectors in the developing world. However, shifting global relations of production and the rise of buyer-driven chains have also impacted working conditions in supply chains geared toward domestic consumption and local markets. This is true in both developed and developing country contexts.
This lacuna is notable in the recent body of GVC scholarship tackling contemporary forced labor. Indeed, one of the most urgent problems in research on supply chains today concerns the question of how far and in what ways the reorganization of production is fueling the use of forced labor by industry. In this article, we utilize the internationally recognized definition of forced labor, first put forward by the International Labour Organisation (ILO) in its 1930 Forced Labour Convention (ILO 1930), which reads: government estimates there are between 10,000 and 13,000 victims of forced labor within its borders (UK Home Office 2014). Second, while statistically less prominent than other forms of exploitation, forced labor and related practices like slavery and trafficking are much worse forms of exploitation, and the persistence of severe human suffering in modern industry warrants investigation. Third, forced labor provides a particularly interesting context to extend work on issues of regulation and governance, including the literature on public and private interactions (Bartley 2014;Eberlein et al. 2014) and on the limits of private governance (Verbruggen 2013;Fransen & Conzelmann 2015). The historical abolition of slavery in developed countries would seemingly make forced labor a story of effective state governance. Yet its reemergence in the contemporary period elucidates a crucial governance gap in the industrialized world, and points to the failure of existing public and private governance initiatives. As our study highlights, although a wave of responsive regulation has sought to strengthen governance systems to combat forced labor over the past few years, regulatory gaps continue to surround forced labor in both global and domestic supply chains.
To investigate the supply chain dynamics of forced labor in the domestic context, we first introduce the preliminary GVC conceptual framework that guides our study and provide an overview of how it has been used to study forced labor. In Section 3, we introduce our research questions and methodology, which is inductive but is guided by the preliminary GVC conceptual framework. Section 4 describes our main findings, focusing on how our study of forced labor in domestic supply chains in the UK requires us to conceptually modify and adapt the GVC framework, which may serve useful to scholars and practitioners examining similar issues in other domestic settings in the future. In Section 5, we discuss the implications of our key findings about the regulatory gaps surrounding forced labor.

Theorizing forced labor through the global value chain (GVC) framework
Anchored in the recognition that globalization of production and trade has reorganized large portions of the global economy, the GVC framework pioneered by Gereffi et al. "examines the different ways in which global production and distribution systems are integrated, and the possibilities for firms in developing countries to enhance their position in global markets" (2005 p. 79). Because the aim of GVC research has been a holistic analysis of global industries, the framework has rarely been extended to understand domestic industries that produce for local rather than exportoriented consumption. Nevertheless, the framework offers several important insights that serve as a preliminary conceptual guide for our study.
The GVC framework emphasizes four key components that are needed to understand the design and functioning of a supply chain: the input-output structure, geographical configuration, governance structure, and institutional context (Gereffi 1994(Gereffi , 2014Gereffi & Fernandez-Stark 2011). The remainder of this Section will review research that has sought to understand forced labor through these components. In Section 4 we describe the modifications we have made to the framework as a result of our domestic-level analysis.

Structure
The starting point of GVC analysis is to identify the key activities, actors, and segments of the chain that "brings a product or service from initial conception to the consumer's hands" (Gereffi & Fernandez-Stark 2011, p. 5). Described as the "input-output structure" (Gereffi 1994), these stages of production are typically depicted as horizontal product supply chains, with an emphasis on the input-output of physical products and services that add value as raw materials are transformed into final products.
Several recent studies have linked the structure of GVCs to the presence of forced labor. The literature holds that a complex and multi-tiered input-output structure is a key factor that can give rise to forced labor, as well as facilitate labor exploitation more broadly. For instance, Phillips' in-depth comparative study of forced labor in India and Brazil found it to thrive along heavily outsourced portions of supply chains (Phillips 2013;Phillips et al. 2011). These findings are confirmed by several studies suggesting that forced labor in GVCs often takes place among lower tiered suppliers, particularly when informal or illegal subcontracting takes place (Taylor 2011;McGrath 2013;Verité 2014a,b).

Geography
The second component of GVC analysis identifies the chain's geographic configuration. Given that a product like a smartphone can involve production activities fragmented across dozens of national boundaries, GVC analysis is focused on the global flow of products, and tends to assume a skilled/unskilled binary between labor in the developed and developing world. As Gereffi and Fernandez-Stark describe it, "usually developing countries offer low labor costs and raw materials, while rich nations with highly educated talent are behind research and development and product design" (2011, p. 7).
This assumption has been influential in shaping GVC research on forced labor. Most of the academic and policy literature has focused on how buyers and lead firms located in the developed world impact upon workers located in the developing world, including the labor conditions in supply chains as well as the codes and standards lead firms introduce to address them (Locke et al. 2009;Barrientos et al. 2011;Anner et al. 2013;Lund-Thomsen & Coe 2013;Nadvi & Raj-Reichert 2015). A key finding has been that imbalances in power, wealth, and value capture between actors in the developed and developing world can create barriers to raising labor standards. For instance, several recent studies have emphasized that the uneven value captured by retail and brand companies located in the global North compared to developing world suppliers, smallholder farmers, and workers can create pressure toward exploitation, including forced labor (Barrientos & Visser 2012;Barrientos et al. 2013).

Governance
The third component of the GVC framework relates to a chain's governance structure, which Gereffi defines as "authority and power relationships that determine how financial, material and human resources are allocated and flow within a chain" (1994, p. 97). Early iterations of the framework simply differentiated between "buyer-driven" and "supplier-driven" governance (Gereffi 1994, p. 97), but more recent work by Gereffi et al. (2005, p. 83) proposes a five-fold typology of GVC governance structures, differentiating between markets, modular, relational, captive, and hierarchy-based governance.
The majority of the literature on raising labor standards in GVCswhich again, tends to focus on workers in developing country contexts (but is not limited to forced labor)has focused on market-based forms of supply chain governance, such as the role and effectiveness of codes of conduct, multi-stakeholder initiatives, and other forms of business governance beyond (or in addition to) government (Fransen 2011;Locke 2013;Bartley 2014;Bartley et al. 2015). For instance, scholars have debated the legitimacy, potential, and effectiveness of global governance initiatives to combat labor exploitation through company codes of conduct (Locke et al. 2009;Fransen 2011), ethical compliance auditing (Locke 2013), and other forms of inspection and monitoring (O'Rourke 2003). Another interlocking body of research has analyzed the capacity and effectiveness of global governance instruments designed to heighten labor standards and human rights within supply chains, such as the United Nations Global Compact (Marx & Wouters 2016).
When this literature has dealt with forced labor, it has tended to approach it as a global governance issue, and more specifically, as a challenge of engineering the right balance of public and private initiatives to address the emergence of forced labor within multi-tiered, complex, transnational supply chains. At the global level, the literature has focused primarily on private regulation. When attention has been paid to domestic governance, it has revolved primarily on public regulation and the public-private intersection (Balch 2012;Waite et al. 2015).

Institutional context
The final component of the GVC framework is an analysis of the broader institutional context in which the chain operates. Essentially, this element of the framework endeavors to understand the social, political, and economic contexts at play in the local, national, and global levels in which chains operate and are governed (Gereffi & Fernandez-Stark 2011, pp. 11-12).
The key underpinning assumption of much of the GVC literature on labor exploitation is the claim that national and global political and economic structures have failed to adapt to the challenges sparked by the reorganization of production into GVCs and that this has created what Barrientos et al. (2011, p. 310) have described as "decent work deficits" across the developing world. Forced labor can arise in this context as a result of both the "supply" of vulnerable workers generated by economic factors like poverty (Phillips 2013;Howard 2014), as well as the business "demand" for forced labor that can thrive in the face of very low levels of state-based labor inspection and enforcement (Andrees & Belser 2009;LeBaron 2014).
Recent research has also identified the presence of corruption as an important institutional factor that can impact upon labor conditions in supply chains. This can include corruption in political or market based supply chain governance (e.g. double bookkeeping and bribes), as well as audit corruption. For instance, Toffel et al.'s recent study of 44,383 social audits in 47 countries found "evidence suggesting that violations recorded in audits might indeed be influenced by financial conflicts of interest and by auditor competence" (2015, p.16).

Research questions and methodology
The GVC framework as applied to forced labor offers an important conceptual framework to explore forced labor in the domestic context. The dimensions of structure, geography, governance, and institutional context provided the contours around which we developed and responded to our research questions in our empirical investigation. That is, the core research question driving our empirical research was: In what ways is forced labor in domestic supply chains similar and different to what is predicted by the GVC literature? Specifically, we sought to explore these similarities and differences in relation to structure, geography, governance, and institutional context.
Our empirical research focused on forced labor in the UK. We chose the UK as our domestic case study for three reasons. First, the UK government was in the midst of introducing a Modern Slavery Act, which helps us understand the current governance context and challenges surrounding forced labor. Second, given the UK has prosecuted forced labor under the Coroners and Justice Act 2009, there is a reliable body of secondary data. Finally, we were able to identify comparable industrial sectors in the UK, covered by public and private governance initiatives. Throughout the study, we deployed a narrow conceptualization of forced labor in order to ensure that findings were based on solid, unambiguous evidence. This meant that we only included cases that clearly involved both involuntary work and coercion (see Section 1). The UK's status as a developed country had the added advantage of shedding light into forced labor in a developed country setting, broadening the typical geographical focus of the GVC literature.
Given how little is known about the supply chain dynamics of forced labor in this context we adopted a largely inductive approach guided by the broad conceptual GVC framework, which involved three key analytical steps. First, we surveyed existing data to select and analyze industrial sectors that would give us insights into forced labor in the UK. We selected three industriesfood, construction, and commercial cannabis cultivationwhere forced labor had been well documented in previous research as well as media reports.
Because industry structure is crucial to understanding forced labor, we selected industries with different types of industry structure. The food industry is dominated by a handful of global food retailers and their large suppliers, and provides insights into how forced labor manifests itself in a highly concentrated industry. The construction sector, in contrast, comprises a large number of small firms and a high number of subcontractors. Commercial cannabis cultivation is illegal in the UK, but according to the National Crime Agency (2016), is a rapidly growing industry worth around £1 billion per year with a large number of organized criminal gangs operating. The cannabis industry therefore sheds light into the supply chains of forced labor when the product being produced is illegal, and where the industry is structured around criminal networks. Table 1 presents an overview of each industry structure.
Second, we mapped and compared how forced labor was manifested in the supply chains of each industry by combining secondary industry data analysis with primary data gathered through semi-structured interviews with key informants with firsthand knowledge of the patterns of forced labor in each industry ( Table 2). These included police officers involved in dismantling commercial cannabis operations, social auditors with experience of detecting forced labor in the food and construction industries, barristers who had represented victims of forced labor, and representatives of the Gangmasters Licensing Authority. The majority of these informants had knowledge of more than one of our industries.
In mapping forced labor in supply chains, we also reviewed audit firm data pertaining to patterns of detection of forced labor in the food and construction industries over the last decade, but were asked not to publish this data within our study. We then used court cases to triangulate our findings. For the food industry, we analyzed the Gangmasters Licensing Authority's conviction and appeal cases (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013), as well as other court documents. For the construction and cannabis industries, we reviewed recent court documents (2009)(2010)(2011)(2012)(2013) containing information about forced labor. On the basis of this triangulated data, we sought to understand and compare how forced labor was manifested in different types of supply chains within the UK context.
In total, our desk-based study encompassed a total of 62 court documents and appeal cases, 35 newspaper articles, 42 academic studies, and 63 reports. Our primary data collection involved a total of 28 in-person interviews and five phone interviews with experts on forced labor, including: academics, barristers, company chief executive officers, employer representatives, ethical auditors, law enforcement agents, non-governmental organization representatives, researchers, and trade unionists. The interviews were transcribed and combined in a database with our secondary data. These data were then analyzed using three levels of coding: (i) initial coding, where abstract concepts and themes that emerged from the interviews were linked with those extracted from the secondary data in relation to the four broad categories of structure, geography, governance, and institutional context; (ii) categorical coding, where initial codes were refined and examined using NVivo10 in order to determine the specific manifestations of structure, geography, governance, and institutional context; and (iii) thematic coding, where emerging themes were identified from the refined data and comparisons with the extant GVC literature to develop deeper-level insights. For example, the following piece of data was (i) initially grouped under "institutional context," (ii) then coded in the category of "immigration law," and finally, (iii) interpreted as "institutional causes of worker vulnerability" in the domestic context: "You often see immigration laws used to create, like engineer vulnerability in workers, so sometimes what you'll even see is that the employer might purposely wait for the visa to expire because then the worker's much more vulnerable." 6 Following these rounds of analysis, we convened a round-table discussion, during which consultation took place with eight experts in the field. The insights obtained contributed to a refinement of our results. Our findings are specific to the UK and cannot be immediately extended to domestic supply chains in other countries, but the ongoing documentation of forced labor suggests the need to investigate domestic supply chains in both developed and developing countries in more detail.

Findings
Our analysis reveals some clear overlaps between what could be predicted by the GVC literature about forced labor in domestic supply chains, as well as some significant differences. Table 3 summarizes these findings with respect to structure, geography, governance, and institutional context. • Over 2 million construction workers in 2012 • Approx 400,000 bogus self-employed workers.
• 250,000 firms in 2012 • 90 percent employ fewer than 10 workers • Labor providers are fragmented, informal and unregulated by the GLA.
• Highly volatilefell faster than GDP during recession, rose faster in the upturn. • Industry has grown in recent years with less reliance on imports.

Structure of domestic supply chains in forced labor
There are both convergences and divergences in the structure of domestic versus global supply chains associated with forced labor. Within the UK, a major difference is the clear dominance of labor providers and other intermediaries in many cases of forced labor. Although intermediaries are known to be part of the structural GVC problem (Barrientos 2013), the main emphasis in mapping forced labor in GVCs is on how it manifests within the different stages of the product supply chain. Within domestic supply chains, however, our analysis suggests that the labor supply chain is equally if not more important. Furthermore, in the domestic context, forced labor often occurs within very simple product supply chains, in contrast to the complex and multi-tiered input-output structures associated with forced labor in GVCs (see also Section 5). A product supply chain describes the discrete stages that a product goes through to transform it from raw materials to a finished product. It is much less commonplace to talk of labor supply chains, but the concept of a supply chain is also applicable to labor where intermediaries are involved. Labor market intermediaries are individuals or organizations that "mediate between individual workers and the organizations that need work done, shaping how workers are matched to organizations, how tasks are performed, and how conflicts are resolved" (Bonet et al. 2013, p. 339). Labor intermediaries are not directly engaged in production, but provide labor and labor-related services to producers. Examples of intermediaries are temporary agencies, recruitment agencies, labor providers, gangmasters, and payroll companies (see also Osterman 1999;Kalleberg 2011). A labor supply chain consists of the sequence of employment relationships that a worker goes through in order to be deployed in a productive capacity. These labor supply chains might be short, for instance in a direct employment relationship between a producer and a worker, where the worker has found the position independently. However, some chains are extended because labor market intermediaries are present. Intermediaries may either facilitate a direct worker-producer relationship (e.g. by helping the worker find employment) or substitute for a direct relationship by employing or controlling the worker directly. As such, the relevant dynamics of interest are not just the input-output of physical products but also the input-output of labor services as a "product" that is essentially traded by intermediaries.
Although labor has rarely been considered within the input-output model of GVCs (for an exception, see Barrientos 2013), the role and importance of labor market intermediaries within the domestic economies of developed countries has been well documented by scholars of employment law (Peck & Theodore 2007;Autor 2009;Fudge & Strauss 2013;Weil 2014). In addition, this literature has analyzed the challenges that employment relationships involving intermediaries pose to the enforcement of domestic labor law (Weil 2011;Dickens 2012). While scholars of employment law do not generally use the terminology of supply chains, their analysis of the role of labor market intermediaries was influential to our thinking about the structure of domestic supply chains associated with forced labor.
Labor market intermediaries featured prominently in all three of the sectors covered by our study. The UK has the largest agency sector in the European Union (EU) (Gallagher & O'Leary 2007), as well as one of the most fragmented, "with a significant proportion of small, local operators… supplying mainly low-status workers" (Balch & Scott 2011, p. 144). The construction and food industries use a considerable amount of temporary, casual, and other forms of contingent labor, the supply of which is often outsourced to third party labor providers (Forde et al. 2008;Lloyd & James 2008; Department for Environment, Food & Rural Affairs [DEFRA] 2012). According to our analysis, these intermediaries may in turn also source some of the required labor from further intermediaries. This can stretch to several tiers in the supply chain, potentially even as many as five or six. Our research revealed that when forced labor arises in the context of intermediaries, the labor supply chain is likely to be relatively long and complex, and the forced labor component is likely to be several steps removed from the core labor force at the producer company.
Within the food industry, informants suggested that forced labor often enters into the labor supply chain in the face of multiple subcontracted labor agencies. This tends to occur when a time-sensitive crop needs to be harvested, and intermediaries experience a sudden need for more workers. Thus workers may only be present on the work site for a matter of days or weeks, making it difficult for producers to detect abuses within labor supply chains. One social auditor informant described how a first tier labor intermediary would initiate a labor supply chain: You call someone, you say I want 20 people and that person maybe has five people at their disposal so he phones somebody else and he says, have you got some people and they go, yeah I have got three but I can get you another eight, so he phones his mate, so it is very, very informal. 7 The critical insight here is that forced labor needs to be understood in the context of the intersection of product and labor supply chains. Figure 1 provides an illustrative example of these dynamics in the UK food industry. As this shows, constructing the product supply chain helps us locate forced labor within the farming stage in the chain, but it is not often farming organizations themselves that perpetrate forced labor. Rather, by mapping the labor supply chain we can identify the different levels of subcontracted labor provision where the exploitation might have taken place.
As this suggests, outsourcing of economic activities is central to the occurrence of forced labor in the UK just as it is in GVCs. Our findings suggest, however, that the location of forced labor in the UK is not restricted to outsourced production but also to outsourced labor. An informant from the Union of Construction, Allied Trades and Technicians explained the intersection of product and supply chains: The key thing to remember is [that] the vast majority of the major construction companies directly employ very few construction workers themselves… the work is subcontracted… So you get kind of second or third tier subcontractors before you get anyone directly employing any workers. Now what increasingly happens is once we've gone down that chain, rather than directly employing people themselves, there's a very strong use of agency workers. 8 Figure 1 Illustrative product and labor supply chains giving rise to forced labor in food industry Forced labor in domestic supply chains A. Crane et al. Figure 2 provides an example of such an arrangement. Such complexity provides the potential for exploitation and, in some cases, forced labor. For instance, an informant from a human resources service provider suggested that most instances of forced labor within the construction sector's supply chain occur among subcontractors in tiers 4 and 5. These tiers tend to be comprised of smaller producers and intermediaries, or "basic one-man shops," facing high pressure to cut labor costs because these can comprise high proportions of the costs of doing business.  As a result, "illegal labor contractors [are] often involved" because "smaller contractors need labor immediately to fill orders." 9 In the context of cannabis growing, most cases of forced labor appear to be located within vertically integrated criminal gangs that both traffic victims in from overseas and exploit them in grow-ops. However, outsourced labor (where gangs source from unconnected traffickers) also features. That is, in the UK, organized criminal groups dominate commercial cannabis cultivation (European Agency for Law Enforcement Cooperation [EUROPOL] 2012), and often source basic labor from children and young adults trafficked into the country through debt bondage (where the debt is a result of the fees paid to be smuggled in), and the bondage often results in compliance brought about by threats made against family members of the victim. Figure 3 provides an example of the supply chain dynamics in the cannabis industry, which although less complex than food and construction, still exemplifies the importance of conceptualizing the industry in terms of intersecting product and labor chains.

Geography of domestic supply chains in forced labor
The geography of domestic supply chains also offers some similarities as well as differences from their global counterparts. Rather than a global flow of products as anticipated by the GVC literature (Phillips 2013), in our research, we observed that forced labor in domestic supply chains tends to be associated with a domestic flow of products. Within the UK, the relevant sections of the food, construction, and cannabis industries where forced labor emerged tended to be where enterprises were engaged in domestic production for domestic sale with domestic sourcing. Althoughgiven the limited empirical data on the overall prevalence of forced labor within domestic versus export oriented productsit is not possible to draw more general conclusions, given the emphasis on global product flows in the GVC literature, this finding is worth exploring further.
For instance, in one recent case of forced labor described to us by informants with knowledge of the food industry, 30 workers in Kent were "subjected to slave-like conditions and controlled through the use of violence" by a licensed gangmaster who supplied workers to Nobel Foods, a large UK processor of chickens and eggs (Gangmasters Licensing Authority [GLA] 2012). The eggs produced by forced workers at Nobel Foods were sold to Tesco, McDonalds, Sainsbury's, and Marks & Spencer, which sold them to consumers across the UK. As one of our informants remarked: So back to Noble Foods, well how many steps down the supply chain was it? One. M&S, Sainsbury's Tesco's, Asda, McDonalds, and then Noble Foodsthe manufacturersand their gang masters, they were the ones caught with the slaves. 10 As this quote suggests, in this instance, forced labor emerged within a relatively simple supply chain associated with a domestic flow of products.  A global flow is, however, evident in the context of labor in that a large proportion of the victims of forced labor in the UK appear to be migrant workers. As the country has increasingly restricted foreign unskilled workers' access to its labor markets, many have attempted to migrate through irregular channels and have ended up vulnerable to trafficking and exploitation through forced labor. Others enter and work in the UK legally, but are either made to believe they are working illegally by their employer, or become vulnerable because of restrictions placed on their right to work. For example, one social auditor recounted the following case involving a distribution warehouse: A group of Romanian women in their 40s were intentionally recruited because they couldn't speak very good English… They were made to think that they were illegal. So, [they were] forced to register under a false name, [their] passports [were] taken away, and [they] had to pay various monies… A fellow worker acted as a middleman with one of the local agency representatives to ensure all the women's wages were paid into the bank account of the middleman and he then distributed minimal monies to them. He claimed many deductions from their wages so they never received enough money to cover their debts or save money. So in effect they were debt bonded to this middleman. 11 As mentioned, scholars of forced labor utilizing the GVC framework have emphasized the uneven power and value capture between brand companies in the developed world and suppliers and workers in the developing world as a key geographic dimension that can give rise to forced labor. By contrast, a key geographic determinant of forced labor in domestic supply chains is this division of legal status and protections offered to workers depending on their country of origin. The labor market power and rights of migrant workers in the UK is governed by their immigration status, which allows free movement of EU workers, while, for instance, "simultaneously promoting the destitution" of asylum seekers and others by removing their right to work (Dwyer et al. 2011, p. 9). In our study, those most frequently identified as vulnerable to forced labor were workers from recent EU accession countries, most notably Romania and Bulgaria. Between 2007 and 2014, such workers were free to enter the UK but were rendered highly vulnerable to exploitation by a range of labor market and welfare restrictions and exclusions. These differing geographic dynamics engendered new forms of regulatory gaps in relation to forced labor in domestic supply chains compared to those most frequently emphasized in relation to GVCs. No doubt regulatory gaps surrounding migrant workers also contribute to vulnerability to forced labor in many GVCs; however, these have been understudied and underemphasized. Thus, the key role and importance of protections for migrant workers in the supply chains encompassed in our study is worth emphasizing.

Governance of domestic supply chains in forced labor
In developing country contexts, the focus on remedying extreme exploitation in GVCs inevitably focuses on private regulation by lead firms, typically through market-based mechanisms. While GVC scholarship has explored state-based regulation, including hybrid and co-regulatory approaches in relation to labor standards protections more broadly, when it comes to the challenge of tackling forced labor in the developing world, both policy and scholarship have tended to focus on private governance. By comparison, domestic supply chains are understood to be governed through a patchwork of private and public governance, involving both market-and hierarchy-based mechanisms. As Table 4 illustrates, our findings show that in the UK, the governance system to combat forced labor is currently fragmented, relying on a combination of regulatory enforcement, licensing, and business self-regulation through social audit. It bears noting that our data was collected in 2013, prior to the passing and implementation of the Modern Slavery Act 2015 and the UK's 2016 ratification of the 2014 ILO Protocol on Forced Labour. Further research is needed to investigate whether these new initiatives will address the fragmentation of governance systems.
Regulatory enforcement and policing targeting the business of forced labor is limited in the UK. As one report collected as part of our secondary data concluded, "the evidence presented here suggests that there is patchy and inconsistent regulation and enforcement around forced labor in the UK" (Balch 2012, p. 34). The criminalization of forced labor with the enactment of the Coroners and Justice Act 2009 and the obligations which flow from the European Human Rights Convention that "no one shall be required to perform forced or compulsory labor" means that law enforcement now has a much wider remit to consider crime in the workplace. Yet such enforcement remains very difficult, as contemporary law enforcement is geared toward crime in the public rather than in the private sphere, which has its own regulatory framework (Scheingold 2011). In addition, efforts to combat forced labor through a criminal law framework can inflict "collateral damage" on victims and those vulnerable to forced labor (Dottridge 2007). The focus on locating individual criminals, not victims, can sometimes criminalize victims of forced labor. As one of our informants from the National Crime Agency described of the cannabis sector: We still get referrals made to us to say: 'I think we have a trafficked victim here.' …they've been encountered in a cannabis cultivation, they've been arrested as an offender, they've had access to a defense solicitor, prosecution, barristers have been involved as well, [a] judge has been involved, if they're a child the local authority has been involved and then they've gone to prison or a young offenders institute. Throughout that process clearly nobody has identified the fact that these may be potential victims. 12 Indeed, our research suggests that at or near the national minimum wage, policing around the workplace is focused on immigration status rather than forced labor. Although established as a standalone crime in 2009, forced labor is generally dealt with by law enforcement within the regulatory framework related to trafficking, which has the UK Border Agency as one of its lead organizations in terms of the formal identification of victims. Because it falls under the umbrella of trafficking, labor exploitation tends to take a back seat to sexual exploitation, which carries more resonance with the general public, is given more emphasis as a political priority, and is more conducive to policing.
Regulatory enforcement within the private sphere is usually left to businesses themselves, much as is typically found in global supply chains. However, the use of intermediaries, such as labor suppliers, has blurred the line of responsibility between employer and employee, making it difficult to establish accountability for exploitation. Where hierarchical regulation does exist in the area of business, it mainly transpires through licensing.
Licensing of labor intermediaries is an important way in which supply chains can be regulated to combat the business of forced labor (although there are several others, including joint liability approaches; see Gordon 2015). Such regulation of UK-based food supply chains commenced with the creation in 2005 of the Gangmasters Licensing Authority (GLA), which has played a critical role in improving the standards of labor providers. However, the restricted scope of licensing (at the time of our research, the GLA's remit was limited to agriculture, horticulture, forestry, fish processing, shellfish, dairy farming, food packaging, and processing), has meant it has had limited success in combating forced labor.
The restriction of licensing to the food sector has prompted certain intermediaries to diversify to multiple industries. As a GLA informant suggested, some labor agencies "divisionalize so that there is a division that deals with providing workers into agriculture. They then mitigate and reduce their risk exposure in case we revoke their license." 13 If the GLA revokes an agency's license to operate within the food industry, workers can be moved into another industry, such as construction.
Non-GLA industries, such as construction, encourage the self-regulation of agencies, but there is no formal enforcement to secure compliance. As such, the Union of Construction, Allied Trades and Technicians union has made clear that "many construction workers encounter daily exploitation from agencies and gangmasters." While the Agency Establishes some monitoring over product supply chains of branded firms.
Limited coverage of labor supply chain. Auditors obligated to report abuses to their clients, not externally.
GLA, Gangmasters Licensing Authority; NGO, non-governmental organization. Workers Regulations came into force on October 2011, "giving agency workers the entitlement to the same basic employment and working conditions" after 12 weeks on the job (UK Department for Business, Innovation & Skills 2011, p. 3), the lack of licensing authority means that there is little accountability or enforcement. As one union informant noted, "unless it is covered by the GLA […] there is no one really looking." 14 Finally, in terms of private regulation, business self-regulation and social auditing is commonplace in domestic supply chains, as it is globally. However, within the UK, limited detection occurs because the path of social audits is typically built around a product supply chain rather than a labor one, and many incidences of forced labor are episodic and emerge in subcontracted workplaces beyond the scope of current auditing practice. As one social auditor commented regarding the inability of audits to trace through the labor supply chain in the food industry: So you go to a farm and they're growing onions and you say, 'Can I see your books?' … and then you say, 'Do you use an agency?' and they say 'Yes'… But say they were using another agent you didn't know. You can't check that all those three agents are responsible for all the people who've been on the farm. So there could always be another group of people who aren't written down and this happens commonly. 15 Audits also tend to exclude farms that are outside major supermarket supply chains, as another social auditor informant described: Not all [of] the farms are involved in the supply chains of supermarkets and if you are going into the cash and carries, if you are going into the markets, if you are going to those places, I am not aware of anybody who is checking. 16 This has particularly serious implications for industries such as construction that are primarily structured around a business-to-business market, and illegal industries such as cannabis production that have no legitimate, downstream retail stage. These design flaws combine with the gaps in public regulation to create a patchwork approach that leaves critical gaps surrounding the governance of forced labor in domestic supply chains. In combination with the institutional dynamics described below, these governance dynamics produced both similar and different regulatory gaps in relation to forced labor in domestic supply chains than those predicted by the GVC framework, as we describe in Section 5.

Institutional context of domestic supply chains in forced labor
The institutional context of domestic supply chains of forced labor reflected a similar picture to that in the global context, albeit with some shifts in emphasis. That is, just as there has been a limited institutional response to preventing forced labor in the reorganization into global supply chains, there has been a similar lack of effective institutional response to the growing intermediation of labor supply in the domestic context.
In the UK, forced labor tends to emerge in "fissured workplaces" where, because of the high use of outsourcing and labor market intermediaries, producer companies are separated from the actual employment of workers, and therefore traditional methods for enforcing labor standards are inadequate (Weil 2014). As one report collected as part of our secondary data put it: The current concern with vulnerable employment has meant that new forms of state intervention are required to deal with what some consider to be a more fragmented and 'hidden' workforce. Inspection and regulation find a more decentralised terrain of work structured in many cases through a range of employment agencies, gangmasters and forms of subcontracting. (Elliot & Lucio 2011, p. 26) This lack of an institutional response to changing employment practices also needs to be viewed in the context of other more active institutional changes that increased the supply of workers vulnerable to forced labor. For example, as discussed above, institutional arrangements around the legal status of migrants created a cadre of precarious workers who entered the UK facing severe restrictions on their ability to work legally or claim benefits. Thus, rather than vulnerability to forced labor being primarily the result of deep-rooted inequalities, such as poverty and gender discrimination, that are highly resistant to change (as in the case of GVCs in developing countries), in the UK, vulnerability primarily arises from specific institutional characteristics and decisions that could, at least in principle, be overturned, but which are deliberately used to exploit workers, as this NGO informant suggests: You often see immigration laws used to create… engineer vulnerability in workers so sometimes what you'll even see is that the employer might purposely wait for the visa to expire because then the worker's much more vulnerable. 17 Corruption also plays a part in a developed country setting, but in some ways that are different to the forms of corruption commonly emphasized in relation to forced labor in developing countries. First, we found no evidence or even suggestion of "hard" corruption, such as bribes to public officials or law enforcement officers, that would have enabled forced labor to flourish unimpeded by the authorities. Even informants who were critical of the lack of prosecutions tended to see this as the result of an inability to act in the face of institutional constraints rather the result of any kind of explicit corruption.
Although not to the same extent as in GVCs in developing countries (O'Rourke 2003), evidence or accusations of explicit cheating in the social audits of supplier firms was present in the UK context. For example, one social auditor respondent recounted: It used to be the standing joke that if you went to Alton Park (which is one of our amusement parks) on any weekday, what you did find is loads of illegal workers who had been given the day off by their employers, and all the difficult ones who were likely to say things to the auditors. 18 In many ways, a form of softer corruption in auditing practice appears to have contributed most to the persistence of forced labor in the UK. That is, the specific protocols of social audits in the UK lead to the limited detection and reporting of forced labor.
Detection can be limited by soft forms of corruption because auditors may be implicitly or even explicitly encouraged to not detect incidents of forced labor. As one social auditor described it, many auditors are "not trying to find things out, they are trying to prove that something is not there." 19 Or in the words of an NGO informant, "You have an industry of ethical auditors out there now who will find nothing if you pay them to go and find nothing." 20 Soft corruption can also limit the reporting of potential cases of forced labor by social auditors. Our data suggests that in the UK, the audit regime has not been organized to encourage reporting of criminal activity, but rather auditors report to their clients. As one social auditor informant described the situation that arises if an audit reveals potential incidents of forced labor: We would have to work with the commissioning company on that to decide how they wanted to play it because… unless you have real evidence that the law is being broken, you really cannot start going to the authoritiesnot if you want to stay in business, anyway. 21 Given these limitations on detection and reporting, it is perhaps not surprising that there have been several incidents in recent years where forced labor has been found amid UK businesses that have successfully passed social audits. In one such case, a police raid in Lincolnshire found 60 migrants subjected to forced labor harvesting leeks by a firm called A14 Vehicle Hire, which supplied labor to Emmett UK. A14 had recently passed two successful audits by Emmett, as well as an audit by the GLA. All of these audits failed to detect the abuse (Shankleman 2008). While these softer forms of corruption may also be present in relation to auditing and forced labor in GVCs, there is limited empirical evidence to confirm this. Thus an empirical basis to assess whether the differences in "hard" and "soft" corruption in GVCs and domestic settings differ in type or degree is currently lacking.

Implications and conclusions: Governance gaps surrounding forced labor in domestic supply chains
The GVC literature has characterized the problem of forced labor as one rooted primarily in the challenge of governing labor standards in complex global product supply chains (such as multi-tiered, input-output structures crossing dozens of national boundaries, and including unauthorized subcontracting and highly informal production networks) (Anner et al. 2013;Phillips 2013;Gordon 2015). Informed by the key assumptions made about the structure, geography, governance, and institutional context of supply chains outlined in Section 2, the literature has tended to locate the solution to the problem of forced labor in supply chains in public and private governance initiatives to address forced labor in the developing world, with an emphasis on public "disclosure" legislation and corporate social responsibility in developed countries (Gereffi et al. 2005;Locke 2013). This understanding of the problem and solution has had an important impact on recent policymaking.
Over the last five years, a number of governments have passed responsive regulation to combat forced labor. This body of legislation has framed the challenges of tackling the business of forced labor as one that concerns global supply chains. For instance, California's 2010 Transparency in Supply Chains Act (SB-657) requires retailers and manufacturers with annual global profits exceeding US$100 million and conducting business in California to report on the measures they are taking to address forced labor, trafficking, and slavery in their supply chain. US federal legislation recently introduced to Congress (the Business Supply Chain Transparency on Trafficking and Slavery Act of 2015 [HR 3226]), as well as the transparency provisions included in the UK's Modern Slavery Act 2015, have been modeled after the California Act. As UK Prime Minister Teresa May described the impetus behind the UK Act: Modern slavery is international and requires an international response. So rather than chasing individual criminals in Britain as they are reported, we need a radically new, comprehensive approach to defeating this vile and systematic international business model at its source and in transit. (UK Prime Minister's Office 2016) [Correction added on 18 September 2017, after first online publication: 'US$1 million' has been corrected to 'US$100 million' in the above paragraph.] Our findings, as laid out in the previous sections, suggest that this conceptualization of the contemporary problem of forced labor is not wholly accurate, and that solutions premised upon it may be limited in their effectiveness. Our work suggests that the governance gaps surrounding forced labor in supply chains needs to be re-thought along two dimensions: in relation to domestic (as opposed to global) supply chains, and in terms of the differences that surround forced labor that occurs in developed countries (as opposed to developing countries).

Governance gaps surrounding domestic supply chains
As mentioned, there is a need for larger-scale comparative study to confirm the patterns of forced labor in domestic supply chains and how these might differ from that of GVCs. Within the three industries we examined in the UK, there were a number of similarities and differences between domestic supply chain dynamics and those anticipated by the GVC literature. In short, our research reveals the need to isolate product and labor supply chains, and to engineer governance initiatives that respond to the specific forms of complexity typical of domestic chains. This is an important insight given that business, NGOs, and government have consistently argued that the key challenge in governing supply chains is their complexity, 22 and this cult of complexity has prevented us from seeing some of the simplicity of domestic product supply chains. Only by isolating the respective product and labor supply chains can we identify the types of complexity that need to be targeted in domestic governance initiatives to combat forced labor.
This finding has implications for policy, and particularly for the recent wave of public governance initiatives to combat forced labor. Much attention to date has focused on transparency in relation to complex global supply chains rather than domestic chains (UK Prime Minister's Office 2016). Further, as this round of legislation is focused along product supply chains and does not contain specific provisions related to the recruitment and subcontracting of labor, there has been very little discussion of the need to address forced labor within labor supply chains that give rise to forced labor.
By anchoring our exploration of governance gaps surrounding domestic supply chains in a case study of the UK, we do not mean to suggest that governance gaps surrounding domestic supply chains are limited to developed countries. Indeed, there is a similar need to rethink the governance of forced labor in domestic supply chains in the context of developing countries. For instance, GVC literature that has explored labor exploitation in countries like China or India, has tended to focus on the global connection and developed country consumers. Simply put, forced labor in garments destined for consumption in developed country markets tends to receive much more attention than say, bricks destined for domestic consumption. If governance initiatives to address forced labor are to do more than make consumers in developed countries feel good about the products they buy, there is a need to focus closer attention on forced labor occurring in domestic supply chains, and especially those that are not led by branded multi-national retail companies.

Governance gaps within developed countries
In addition to rethinking predictions about governance gaps from the domestic perspective in GVC literature, our research also suggests that it is important to carefully calibrate our thinking about these issues in the context of developing countries. Although to date the GVC literature has focused on governance gaps that give rise to labor exploitation in developing country contexts (Locke et al. 2009;Barrientos et al. 2011), our study of the UK confirms that many of these governance gaps exist in forced labor in developed country contexts as well.
Just as in many developing country contexts, the rolling back of the state regarding labor rights, inspection, and standards enforcement has exacerbated the risks of workers in the UK being exploited in forced labor conditions. However, the persistence of albeit patchy and inconsistent regulation and enforcement, combined with a relatively effective but structurally limited system of licensing, means that the key challenges are to coordinate governance across a patchwork of initiatives, both market and hierarchy-based, that are designed to address labor abuses, as well as coordinate with broader regulation around immigration and other institutional conditions that promote vulnerability to exploitation.
In some respects, these challenges of coordination were acknowledged by the UK government with the passing of the Modern Slavery Act 2015, designed to create a more comprehensive and integrated response to the problem of forced labor among other forms of extreme exploitation. The Act was first introduced in a draft bill in December 2013 (after we had concluded our fieldwork), and aimed to "consolidate and simplify existing slavery and trafficking offences," to provide coordination across government with respect to enforcement, and also to connect with a "wider non-legislative approach" that included "work with businesses on a voluntary basis so they can ensure their workforces and supply chains are not exploited." 23 Yet the design and criminal justice focus of the Act suggests that it may be ineffective in addressing the coordination challenges surrounding forced labor documented through our research.
Although further research will be required once the Act has been implemented long enough to evaluate its effectiveness, two weaknesses are important to note here. In the first case, the Act does little to correct the lack of coordination between labor and immigration law enforcement uncovered in our research. Furthermore, policymakers have raised concerns that a separate new piece of legislationthe 2016 Immigration Billwill further increase the vulnerability of migrant workers and undermine any improvements that the Modern Slavery Act does make in terms of recognizing victims (cf. Blomfield 2016). Secondly, rather than bolstering enforcement around labor standards for low and minimum wage workers, the Modern Slavery Act merely reinforces existing private initiatives to enforce labor standards, for example, through auditing (which, as we note in Section 4.4, can be limited in their detection and reporting of forced labor). Thus, the Act appears to do little to address concerns our informants raised about audit scope and protocols undermining protection.
In short, while limited coordination between public and private governance initiatives is often considered to be a problem typical of GVCs (Gereffi et al. 2005), our research suggests that this issue is also (and perhaps even more) significant in terms of forced labor occurring in developed countries. However, in such settings, coordination gaps are further exacerbated by limited intra-governmental agency coordination, as well as institutional changes that have increased the vulnerabilities of migrant and low-wage workers. Our findings suggest that the effectiveness of governance in developed countries critically depends on understanding these interactions.
Finally, it is important to note that many of the governance gaps occurring within a developed country context may be a matter of politics rather than capability deficit. The GVC literature asserts that "decent work deficits" arising from changing patterns of production and labor market regulation are typically associated with developing world contexts (Barrientos et al. 2011McGrath 2013;Nadvi & Raj-Reichert 2015). Developed country settings, by contrast, are often considered to have robust capability to enforce labor market regulation and social protections for workers (Eberlein et al. 2014). Yet as our research on forced labor demonstrates, there is a need to acknowledge that such deficits can be engineered through institutional and governance dynamics within developed countries too. That said, there may be important variation surrounding the institutional context, as suggested by our analysis of the similarities and differences in forms and degrees of corruption surrounding forced labor in domestic and global supply chains. Large-scale and deeper study of the continuities and differences between governance gaps surrounding forced labor in developed and developing country settings are sorely needed, and would make a fruitful contribution to the burgeoning literature on how public and private governance interactions might complement or compete with one another at a global level (Fransen 2013;Eberlein et al. 2014;Gulbrandsen 2014). Ultimately, addressing the governance gaps that surround forced labor will require scholars and policymakers to carefully refine their thinking about how we might design operative governance that effectively engages with local variation.