Governing Labour Standards Through Free Trade Agreements: Limits of the European Union's Trade and Sustainable Development Chapters

The EU has established a new architecture of international labour standards governance within the Trade and Sustainable Development (TSD) chapters of its Free Trade Agreements (FTAs). To examine the operationalization of this framework, we draw upon 121 interviews undertaken with key informants in three FTAs signed with the Caribbean, South Korea and Moldova. We engage with wider debates over external governance and the projection of EU power by showing how operational failings, including a lack of legal and political prioritization of TSD chapters and shortcomings in the implementation of key provisions, have hindered the impact of the FTAs upon labour standards. We also identify significant limitations to the EU's &#8216;common formulation&#8217; approach when applied to different trading partner contexts, alongside ambiguities about the underlying purpose of the trade&#8211;labour linkage. Reflection about the function and purpose of labour standards provisions in EU trade policy is therefore required.


Introduction
The EU has long sought to address labour standards in its trade policy. This has been done through references to labour rights in its unilateral Generalized System of Preferences (GSP) and by failed attempts to bring a social clause into multilateral trade agreements (Wilkinson, 1999;Young, 2007). But it is in its free trade agreements (FTAs) where the most notable changes are now occurring. FTAs are the most economically significant aspect of EU trade policy. They could cover as much as two thirds of EU trade if all current negotiations are successfully concluded (EC, 2015a, p. 9). They are also the most important legally binding instruments that the EU can use in its external policy (Jurje and Lavenex, 2014). Labour provisions within EU FTAs have 'widened and deepened' over the past decade. This is linked in part to the 2007 Lisbon Treaty which accorded greater influence in trade policy-making to the European Parliament; an institution which has emphasized the labour and human rights dimensions of trade policy (Van den Putte and Orbie, 2015, p. 264).
A significant point of departure for labour provisions in EU FTAs was the 2008 CARIFORUM Economic Partnership Agreement (EPA). Unlike its predecessors, this agreement contained more references to social policy norms and core labour standards. It also allowed for disputes on social issues to be referred to independent experts, and institutionalized dialogue about the trade agreement within a civil society mechanism (CSM). Since the negotiation of the 2011 EU-Korea FTA such provisions have been packaged with rules around environmental protection in a Trade and Sustainable Development (TSD) chapter. TSD chapters have become an integral part of the EU's 'new generation' trade agreements (Bendini, 2015). Such chapters were present in finalized agreements with a further 18 countries as of July 2017, 1 with bold claims made about their efficacy. TSD chapters are meant to ensure that economic growth goes hand in hand with higher labour standards, making trade policy 'not just about interests but also about values' (EC, 2015a, p. 5).
It is against this backdrop that we analyze the EU's approach to protecting and promoting labour standards. We do so by drawing on research into the negotiation and implementation of labour provisions in three 'new generation' EU agreements, adding to debates about the projection of EU power in three respects. First we argue that to determine whether the EU's approach to the external governance of labour can be characterized as a form of (potentially significant) 'normative power', qualitative analysis of how FTA provisions are translated into practice is vital. Second, by empirically substantiating the weaknesses of the EU's common formulation approach across three cases, we provide a critique of the intended functions of TSD chapters and argue for greater recognition of the distinctive third country contexts that shape their operationalization. Finally, we argue that different ideas among policy-makers and interest group representatives about whom and what the trade-labour linkage is for, suggests the need for greater reflection upon the intended purpose(s) of labour provisions.

I. Existing Academic Debates on Labour Standards in EU FTAs
While there is some variation across the different agreements, the TSD chapters common to EU FTAs share three key types of provisions. First there are substantive standards. All agreements require that the parties commit to upholding the International Labour Organization's (ILO) core labour standards and promote its decent work agenda. Second, there are procedural commitments. These include dialogue and co-operation between the parties, transparency in introducing new labour standards measures, monitoring and review of the sustainability impacts of the agreement, and a commitment to upholding levels of domestic labour protection. Third, there are institutional mechanisms. All agreements since the EU-Korea FTA have a tripartite format. Committees of state/EU officials from the two parties are established to oversee the implementation of the TSD chapter. These are advised by a CSM that takes the form of a Domestic Advisory Group (DAG) including representatives of business, trade unions, non-governmental organizations (NGOs) and occasionally academia, with the DAGs of the two parties meeting together on an annual basis. 2 Finally, there is an expert panel that investigates complaints made by the parties and makes recommendations on them. The implication of the agreement text is that these institutions will interact to effectively implement the TSD chapter.
In contrast to much of the literature on the EU's trade-labour linkage, which concerns itself with internal policy formation and the question of why labour standards provisions have been addressed by the European Commission (see, for example, Bossuyt, 2009;Lechner, 2016), we focus on external dynamics. Two inter-related debates are central: (1) the form of external governance which the 'EU model' represents and seeks to project beyond its borders; and (2) the effects of that model in the EU's trading partners.
In the first debate, two positions can be identified. One sees the EU's labour provisions as a weak form of 'market power' (Damro, 2012). Such claims are based on the 'best endeavour' language of many of the provisions and the use of 'soft law' enforcement to back them up, delinked from economic sanctions (Adriaensen and González-Garibay, 2013;Bartels, 2015;Vogt, 2015;Young, 2015) Moreover, since the reference points are the internationally-agreed core labour standards of the ILO this is not seen as an externalization of EU regulation per se (see Damro, 2012;Young, 2015).
Another position takes issue with this characterization of the EU's approach, suggesting that it does not give sufficient weight to the communicative or 'networkbased' forms of governance through which standards are promoted and which give it a European dimension distinct from the supposed 'sanctions-based' approach of the US. From this perspective, the potential for improvements in labour standards is said to reside in the mechanisms of dialogue and co-operation institutionalized through EU trade agreements (Oehri, 2015;Postnikov and Bastiaens, 2014;Van den Putte and Orbie, 2015). Rather than evaluating the model's effectiveness by looking solely at the legal obligations placed on third countries towards immediate regulatory convergence, this position allows for the possibility that improvements in labour standards may emerge through longer-term change in political norms and processes resultant from learning and socialization. TSD chapters can in this sense be understood as an exemplar of 'normative power Europe' wherein 'persuasion, argumentation and the conferral of shame and prestige' rather than 'coercion or solely material motivations' effect more sustained change in third countries (Manners, 2009, p. 793).
Informed by this debate on how the EU governs labour standards is a second set of claims about what its effects have been. Postnikov and Bastiaens (2014, p. 931) offer a mixed methods study of EU bilateral trade agreements in force prior to 2010, concluding that those 'with labour provisions have a positive and statistically significant impact on workers' rights in signatory nations'. Regression analysis is used to demonstrate the impact, while interviews in Brussels and with one informant in Chile are used to buttress the claim that what causes this positive effect is the way state officials are educated about, and normalized into, upholding labour standards. Also based on interviews in Brussels, Garcia and Masselot (2015) argue that the EU's insistence on the importance of core labour standards during its FTA negotiations with Malaysia during 2010-12 created the opportunity for the country to ratify the ILO conventions on freedom of association and discrimination.
Other assessments cast doubt on such conclusions. Under the EU-Peru FTA Orbie and Van den Putte (2016, p. 19) found that in Peru there were 'serious shortcomings' in the implementation of core labour standards and that standards of labour protection on health and safety at work had actually been lowered since the agreement came into force. Likewise Marx et al. (2016, p. 606) found that, with regard to the TSD chapter in the EU-Colombia FTA, there was a 'shared notion among stakeholders that the EU's approach constitutes little more than "window dressing"'. Other studies have challenged the EU's purported mechanisms of influence. They find that labour standards provisions have not been accompanied by increases in development assistance for labour-related activities (Ebert, 2016), that the functioning of the CSM in the EU-South Korea FTA has been impaired ( Van den Putte, 2015), and that in negotiations with Vietnam the EU watered-down its requirement for the country to ratify the ILO convention on freedom of association (Sicurelli, 2015). Collectively these critical evaluations pose questions about the causal impact and lasting influence of the EU model of labour provisions. Indeed, in response to similar concerns raised by actors within the EUincluding Members of the European Parliament and trade unionsin July 2017 the European Commission acknowledged that some reform to the model might be necessary, initiating 'a thorough stocktaking of the EU TSD provisions' (EC, 2017c, p. 5).
These debates on labour provisions in EU FTAs resonate with the wider EU studies literature about external influence and the conditions under which 'rules travel' (Lavenex, 2014). Three insights are particularly useful. First is the functionalist argument by Lavenex (2014) that even when looking at 'coercive' rules affecting access to the EU market such as regulatory standards, close attention must still be paid to the transgovernmental and technocratic networks through which regulatory convergence is enacted. Second is the recognition by Lavenex and Schimmelfennig (2009, p. 804) that third-country actors are more likely to accept modes of external governance that resonate with their domestic institutional mechanisms and are seen as normal and legitimate. And third is the point by Bicchi (2006, p. 293) that this is unlikely to take place when the EU bases its external policy on routine behaviour and 'exports its own norms unreflexively, with a single model promoted to all its partners regardless of their context'an 'our size fits all' approach. Countering the more uniform and uni-directional 'diffusionist' approaches to external governance (see Börzel and Risse, 2012) these insights suggest the need to interrogate the ways in which the EU's TSD chapters articulate with different third country contexts, via the networks intended to enact the processes of dialogue and co-operation on labour standards (Campling et al., 2016). Our methodology for interrogating this phenomenon is outlined next.

II. Methodology
While existing research on the impact of EU labour provisions has relied on single case studies and/or interviews with Brussels-based actors to generate data on causal mechanisms of influence, our approach examines three cases, selected both because of their diversity and because they reflect the most likely cases for labour provisions to have an impact. These cases are the CARIFORUM EPA (2008), a trade agreement with a stated development dimension, partnering the EU with 15 small Caribbean states; the EU-South Korea FTA (2011), a bilateral trade agreement with a country that is geographically remote from the EU and at a comparable level of development on a per capita income basis; and the EU-Moldova 'deep and comprehensive trade area' agreement (DCFTA), part of an Association Agreement (2014) with a small eastern European country on the EU's border.
In terms of most likely cases, the CARIFORUM agreement is the oldest of the 'new generation' FTAs and thus gives the best chance for longer-term change to become evident. This is important since 'network-effects' can take time to emerge; a point often made by the European Commission regarding the perceived lack of impact of the TSD chapters to date (DG TRADE, 2017). The EU-Korea FTA has been the agreement in which the European Commission has invested most energy on TSD issues, and has been touted accordingly as the success story with respect to dialogue. Meetings between the two parties purportedly demonstrate that 'the provisions are having a positive impact to promote sustainable development' (EC, 2015b). Finally, Moldova is geographically proximate to the EU and is dependent on trade with the EU. In 2016 the EU accounted for 49 per cent of Moldova's imports and 66 per cent of its exports; more than any other 'new generation' FTA signatory (EC, 2017a). 3 This material influence, of the sort routinely cited in the literature on the EU as external power, makes the Moldova agreement the one in which the EU has the most political leverage. If the TSD chapters do not appear to be affecting labour standards in these most likely cases, then it is unlikely that that they will have significant influence in other countries either.
The diversity of cases, meanwhile, allows us to reflect on the way in which third country context matters to the operationalization of the EU's consistent approach. In particular, where similar conclusions are reached about the functioning of labour provisions despite such diversity, the generalizability of findings concerning any inherent limitations is strengthened. Put another way, lack of variation in the selection of case studies can 'inhibit the discovery and development of theories, models and concepts that are broadly applicable' (Vaughan, 1992, p. 174). By comparing accounts across the three cases, we can also begin to explore how and why the domestic structures of third countries matter to the EU's attempts to govern labour standards.
The following analysis has a primary focus on the institutionalized practices of dialogue and co-operation which are meant to flow from the implementation phase of the TSD chapter, and where there is most debate regarding the potential for impact on labour standards. That said, in line with our 'most likely' case selection, we also take into account the negotiating phase of FTAs. Research on the trade-labour linkage has identified that during the negotiating period pressure upon trading partner governments can lead to increased support for, and positive changes in, labour rights; a point reflected by a number of EU representatives in our interviews (see also ILO, 2013).
We conducted interviews with an extensive range of key informants: those involved in the negotiation and implementation of agreements in the case study countries as well as in the EU (such as civil servants, politicians and spokespeople of prominent interest groups); members of the institutional mechanisms set up through the agreements (the joint committees, expert panels and CSMs); and knowledgeable 'outsiders' from business, civil society and trade unions who ought to be affected by the respective provisions, thus testing the reach of the TSD chapter beyond its immediate constituents. In 2015 and 2016, 121 key informants were interviewed; around 30 in each case study and in Brussels. 4 To protect their identity, interviewees are referenced anonymously according to an alphanumeric system (K23, E12 for example). The letter represents the location of the interview -C for Caribbean, K for South Korea, M for Moldova, E for EUand the number a unique interviewee. In order to minimize the dangers of interviewee bias affecting key conclusions, interviews were triangulated across multiple interviewee types and/or by verifying with documentary evidence. Any significant differences in opinions between key informants are explicitly identified.
The methodology adopted here is important in that it seeks to isolate the impact of labour provisions within FTAs from other factors, thereby interrogating the causal relationship between particular policy instruments and changes in third countries. In many existing studies on human and labour rights provisions in trade agreements, including those contained in the EU GSP and in US trade agreements (see, for example, Hafner-Burton, 2005;Kim, 2012), it is commonplace to assess the change over time of the 'average' civil/labour rights situation of entire countries, typically via quantified interpretations of reported rights violations aggregated from a limited range of textual reports. Pinpointing and substantiating the precise mechanisms at work is beyond the scope of that form of analysis. Such methods also miss the fact that, insofar as they are operationalized at all, labour provisions tend to privilege certain actors by virtue of the issues prioritized and the groups that mobilize around them, for instance focusing on freedom of association more than poverty wages, and involving trade unions rather than informal worker networks. By speaking with informants cognizant of these differences and involved in the negotiation and implementation of the TSD chapters, our approach allows us to investigate their contrasting understandings of what the trade-labour linkage is for, and who it is intended to benefit.

Negotiation
In all three cases there was a consensus among state representatives from the EU's negotiating partners, as well as others involved in the negotiations, that the TSD chapters and labour provisions contained within them were proposed and driven through by EU negotiators (C9, C19, C31, K23, K26, M16). Labour provisions were not seen as a natural part of trade agreements by CARIFORUM, Korean and Moldovan negotiators (C9, C19, K25, M16). Caribbean negotiators sought to safeguard against the provisions having significant impacts by clarifying that they could not become the basis for sanctions (C30, C31), while Korean negotiators successfully demanded fewer references to international standards and the removal of any immediate obligation to ratify all fundamental ILO conventions (K25). Moldovan negotiators did accept the TSD chapter as it was suggested by the EU on the basis that a similar formulation had been used in the Korean FTA and because the substantive standards were already part of Moldovan law as similar provisions had been included in earlier GSP+ and Autonomous Trade Preference agreements with the EU (M2, Smith et al., 2017Smith et al., , 2018. On the EU side, it was clear that TSD chapters were included primarily as a result of pressure from the European Parliament, supported by NGOs and the European trade union movement (E7, E17). Negotiators from the European Commission saw their role as a relatively narrow one; to obtain agreement from trade partners to the TSD chapters, rather than to create any detailed understanding of them amongst states and civil society in trade partner countries (E9, E12).
At the same time, organized labour and their advocates in partner countries were not able to open up space for discussion of labour issues in a meaningful way during the negotiation period. These groups had a low level of engagement with the negotiating process for two primary reasons. Some trade union interviewees stressed that they were either opposed to the agreement as a whole or did not see it as a political priority in comparison to pressing domestic issues, and so did not meaningfully engage with the negotiations (C7, C9, K5, K23). Others argued that trade negotiators and government officials did not consult with trade unions or NGOs on the substance of the TSD chapters (M2, M13, K23, C20). While it is possible for organizations in third countries to participate in DG TRADE Civil Society Dialogue events and inform policy from the EU side, no organizations from the three cases joined the official register (see EC, 2017b). Some consultation with organized labour was undertaken by the private sector contractors which carried out the respective Sustainability Impact Assessments on behalf of the EU. But, at best, it is unclear how such input fed into the final recommendations; at worst, it was merely teleguided participation designed to pay lip service to non-commercial interests (on the CARIFORUM assessment see Gammage, 2010). We now move on to discuss implementation of the labour provisions by the tripartite institutional structure of the TSD chapter.

Committees of State/EU Officials
Perhaps unsurprisingly given the negotiating process, civil servants based in the three case studies thought that labour standards were not a legislative or procedural priority in terms of the operationalization of the agreement (C4, C5, M2, K24). Rather, the primary focus was on implementing the commercial provisions and setting up institutional mechanisms. It was clear therefore, that improvements to the legal protection and effective implementation of labour standards were not going to happen simply as a result of the trade agreements coming into force. Of critical importance then, are the workings of the institutional mechanisms that the TSD chapters create: committees of state/EU officials, CSMs and expert panels.
In terms of dialogue between state officials in the EU and its trade partners, labour standards had not been discussed in any detail by the Trade and Development Committee in the CARIFORUM EPA (E8; see also CARICOM Secretariat, 2012). In the Korea FTA, by contrast, EU interviewees argued that labour standards are increasingly on the agenda, with South Korea's failure to ratify four of the eight ILO fundamental conventions the most important issue raised (E10, K24). But there was consensus among Korean interviewees that the EU was unlikely to make any progress on those fundamental conventions through the provisions and institutions of the FTA (K5, K10, K11, K25, K27, K28, K31). Multiple interviewees in Korea pointed to the fact that President Park Geun-hye's administration (2013-17) had been actively pursuing a legislative programme which involved significant weakening of labour law (K5, K11, K18, K20, K22, K23, K31). 5 For some connected to the labour movement, the move by the Park regime indicated that any form of dialogue on increasing labour protection would be futile (K23). In addition, state and labour representatives pointed to the fact that politicians and civil servants vital to the process of labour law reform were not engaged in TSD chapter-related discussions with EU interlocutors (K1, K16, K25). Finally, in Moldova, while the first meeting of the TSD Committee did discuss child labour and the labour inspection system, the primary focus, driven by pressure from civil society, was on the need to resolve the deepening political and economic crisis that the country has been facing in the wake of a 2014 banking crisis, which saw $1 billion allegedly stolen from the banking system (M11, M12, M13, M14).
There was also little EU funding targeted towards TSD chapter-related activities, reinforcing the findings of existing research (Ebert, 2016;Orbie and Van den Putte, 2016). We found only one relevant EU-funded project which was operational by 2016; a €2 million project led by the ILO aimed at promoting regional social dialogue around the obligations of the CARIFORUM agreement. Thus far, this had not led to any labour standards issue being raised in the agreement's committees (C12, C20, E8) and various interviewees suggested that Caribbean trade unions are a long way from being able to identify labour issues stemming from the EPA and formulate appropriate policies to address them (C20, C21, C27). In 2016 two labour projects were being developed in relation to the EU-Korea FTA. The first focused on the implementation of ILO Convention 111 on discrimination, the second on approaches to corporate social responsibility (CSR) in the EU and East Asia. While EU representatives saw these as stepping stones towards dealing with more contested labour issues in Korea (E10), Korean labour representatives were less convinced about whether they were strategic priorities in the context of much more politicized issues, such as freedom of association (K23).
A compounding factor limiting progress was that state officials in the EU and partner countries did not see the TSD chapters as their primary responsibility, but rather saw CSMs as being the primary mechanism for providing impetus on labour-related issues (C5, E8, M7). As one Moldovan official said of labour provisions: 'They are not exactly commitments for government, they are more for civil society' (M3).

Civil Society Mechanisms
As noted above, at the heart of the operation of the TSD framework and its potential 'network-effects' are mechanisms for civil society dialogue. It is clear that there have been serious difficulties affecting the functioning of the CSMs in all three agreements. In the Caribbean, there was a six year delay from the commencement of the agreement to the operationalization of the CSM. Interviewees suggested that the long delay resulted from government officials' unfamiliarity with, or even hostility to, incorporation of civil society within the formal structures of a trade agreement. This was combined with the difficulty of constituting a manageable and representative body from 15 different Caribbean states incorporating historically distinct Anglophone, Hispanaphone and 5 The new administration of Moon Jae-in appears more open to engaging around freedom of association and forced labour issues. This happenstance does not undermine our analysis here and later in the paper which demonstrates the problems and limitations of the EU's approach when dealing with an intransigent administration Francophone communities (C20, C31). The first CSM meeting therefore only took place in 2014, with further meetings in 2016 and 2017. A number of members suggested that, particularly on the Caribbean side, the CSM had been insufficiently resourced to perform its role (C12, E3, E25; see also Van den Putte, 2015). There is only one trade union representative on the CSM, the low-profile Caribbean Congress of Labour. Other members of the CSM on the Caribbean side had no knowledge of the content of labour provisions and therefore seemed unlikely to engage with them (C20). At the time of writing, labour standards issues have not been raised in any Caribbean, European or joint civil society meetings (C20, E6, E8), nor were they mentioned in the response of the parties after the agreement's five year review (CARIFORUM-EU, 2015). Overall, while a few interviewees claimed that the CSM has a useful knowledge-sharing function, most felt that there was no real purpose to their discussions, beyond demonstrating that the meetings themselves have taken place (C12, C13, C20, E6). As one informant close to the implementation process cautioned: 'At the moment we risk becoming another one of these institutions where you meet twice a year, and you put out a statement "Committee met in Brussels and raised concerns about this and that" and that is the end of it' (C20).
The Korean CSM was seen by interviewees from the European Parliament, the Commission and trade unions as the most advanced and developed of all CSMs, reflecting the importance placed on this 'flagship' agreement by the EU (E10, E16, E26). Funding was less of an issue and meetings occurred more regularly and predictably. Some EU officials and business groups were positive about the fact that dialogue is occurring (E14, E20, E21), but questions remained from representatives of both sides about what the mechanism is achieving and whether the discussions are valuable (E19, E26, K16, K23). As in the Caribbean case, European and Korean trade unions said that the CSM was unnecessary for creating international links as there were existing networks (E30, E31, K5, K11, K21, K22, K23). A number of Korean interviewees involved in the CSM also raised concern about the lack of meaningful engagement from EU interlocutors in the civil society forum (K16, K17, 26). Additionally, as in the CARIFORUM EPA, meetings have been treated as an end in themselves (K17). On the European side there was on-going scepticism from trade union and civil society representatives, as well as from Members of the European Parliament, about 'public interest' membership of the Korean Domestic Advisory Group (academics and other professional researchers), and the extent to which it is sufficiently independent from government (E6, E17, E30), along with a more prosaic recognition that linguistic differences make communication and relationship building difficult (E6).
In Moldova, implementation of the agreement is at an earlier stage. But there were already serious funding and institutional capacity problems, with complaints about lack of support for even the most basic structures of the CSM on the Moldovan side (M11), noted also in the joint declaration following the second meeting between the EU and Moldovan DAGs in October 2016. Members of the CSM reported little attention being paid to labour standards (M8, M11). This was reflected in the joint declaration of the Moldova and EU Domestic Advisory Groups (2015), which identified very general labour issues to address (such as 'adoption and strengthening of the domestic enforcement of the ILO conventions') and placed these alongside other priorities such as forming a functional government and easing lending to businesses. Partly this was a result of the wider crisis affecting Moldova which dominated political discussions generally, and partly because members of the Moldovan CSM also met to discuss a range of issues in committees constituted by other parts of the wider Association Agreement. There are a limited number of non-state actors in Moldova with a capacity to engage, and so the same organizations play multiple roles. Consequently, there was considerable confusion as to the proper remit of this CSM as distinct from other civil society fora (M7, M11).

Interactions between the Committee, CSM and Panel of Experts
Compounding the limitations of the civil society dialogue process, civil society representatives and state officials were unsure about how the inter-governmental TSD Committees were supposed to respond to issues identified by the CSMs (C12, E8, M11). This raises a broader issue of the interactions between the different institutional mechanisms established by the agreements. A number of civil society and trade union representatives thought that the opinions of the CSMs were not taken seriously by the European Commission or inter-governmental committees, and that interactions between them were superficial (C20, E6, E23, E30, K6). The only examples we found of meaningful interaction on labour-related issues occurred in relation to the EU-Korea FTA. The first issue concerned the exclusion of the Korean Confederation of Trade Unions from the CSM, and was successfully resolved with EU DAG pressure. But a second issue, described below, exposed more fundamental concerns about the TSD chapters and their potential to protect and promote labour standards.
After European and Korean trade unions had raised issues in joint CSM meetings about the Korean government's failure to make progress on ratifying ILO fundamental conventions, the EU DAG sent a letter to the Commission requesting that formal consultations be initiateda precursor to convening the panel of expertson the basis that widespread violations of labour rights, particularly freedom of association, were allegedly taking place (Jenkins, 2014). But the Trade Commissioner at the time, Karel De Gucht, rejected the request and instead promised to pursue the matter through intergovernmental dialogue (Vogt, 2015). Trade union and NGO representatives in Europe, however, remained sceptical that intergovernmental dialogue, even against a background of CSM pressure, would produce tangible results (E15, E19, E26, E29). One trade union actor argued that the presence of trade unions within the CSM was utilized to legitimize the agreement, without leading to any substantial change for labour rights protection on the ground (E26).
European Commission officials stressed that the labour situation is difficult in South Korea and that the TSD chapter cannot undermine the overall objectives of the agreement (E10, K24). Providing empirical support for Orbie's (2009) contention that the social dimensions of trade are potentially compromised by commitments to market opening, one Commission official told us that: 'It is important to have a positive forward looking agenda. Confrontation would lead to a backlash on behalf of Korea. We want to add investment protection into the agreement. If we took action under this chapter, we might lose benefits elsewhere. So we do need to think about the bigger context' (E14). This reluctance to commence formal 'complaint' procedures in Korea is symptomatic of a broader sense that there is inadequate legal duty and political will to at least try and enforce labour standards provisions. There have been no complaints taken to the expert panel in any EU FTA, putting a question mark over the efficacy of this institutional mechanism. Perhaps confirming this suspicion, members of expert panels in trading partners did not appear to have been well-briefed about the role they had taken on. Interviews with panel appointees revealed that some did not have a basic understanding of the types of functions they might be expected to perform and others did not know they had even been appointed.
Overall then, labour standards provisions were being undermined by lack of prioritization among key actors and shortcomings in implementation. Government officials from trading partners did not appear to see the externally imposed TSD chapters as their responsibility, and despite general commitment to the sustainable development agenda by DG TRADE (see EC, 2015a) officials were not deeply engaged with labour rights issues in third countries, the motivation for which has its origins in the European Parliament. The result is that the weight of expectations has been loaded onto CSMs. But CSMs are ill-equipped to play a meaningful 'networking' role. Dialogue between civil society actors has not been fruitful, and it is unclear how the outputs of that dialogue would impact on state or business policy, given shortcomings in the engagement processes with governmental bodies and the dispute settlement process.

IV. Common Framework, Different Limits
Our argument thus far is that the EU has arrived at a common approach to the governance of labour standards in its FTAs that has failed to have substantive effects within three most likely cases. Given the diversity of the cases selected, we can also suppose that the context in which the EU model is deployed does not matter greatly to the outcome. This raises a further question: could the TSD chapters be effective if they were better operationalized, or are there inherent limits to what they can achieve? The evidence from our three cases sides firmly with the latter.
In the Caribbean context, the vast majority of interviewees did not consider ILO core labour standards, which are at the heart of the EU's model, to be the most pressing problem and pointed to the near universal ratification of the fundamental conventions in the region as evidence of this (C2, C3, C9, C10, C11, C21, C28, C33). Similarly, European trade unionists thought violations of core labour standards were much more systematic in neighbouring Central America, meaning that support for Caribbean labour movements was not a priority (E23, E25). This is not to say there are no rights violations or labour disputes involving organized labour, but what many academic and NGO interviewees were at pains to point out was that small-scale farmers, informal traders, the selfemployed and those working on daily contracts had been worst affected as a result of the region's changing trade relationships (C3, C17, C18, C26). For instance, it was suggested to us that 'hucksters'petty traders who are predominantly female and often operate outside the formal economywere being adversely affected by the tightening up of customs operations, partly as a result of the CARIFORUM agreement's trade facilitation provisions (C17). Consequently, for Caribbean civil society actors in the CSM, the main focus was to try to understand what impacts the EPA was having on employment and working conditions in the region, and it was for this reason that concerns were raised over inadequate monitoring (C20). All TSD chapters contain an obligation to monitor the social and environmental impacts of the agreements, but resources for the CSMs to carry this out are very limited (E14). Eight years after coming into force the first efforts at developing a methodological approach for monitoring the social impacts of the CARIFORUM agreement were only just beginning (E8, E25).
A very different situation is apparent in South Korea, which has only signed and ratified four out of the eight fundamental ILO conventions. During the first five years of the FTA, the Korean government stood accused of cracking down on trade unions in a way that severely undermined freedom of association, attested to by complaints submitted to the ILO in 2013 and 2015 by coalitions of domestic and international trade unions. This was accompanied by conflict between government, trade unions and businesses about proposed relaxation to existing labour laws. Insofar as they speak to prominent concerns in Korea then, the standards and procedures contained in the FTA provisions are apt. However, trade union and business actors in Korea have called into question the use of social dialogue to achieve these ends, not least because of the antagonistic state-society relations that characterize labour governance, which is in turn linked to the influential political role played by the family conglomerates known as chaebols (K3, K5, K10, K30, K31). These problems are exacerbated by the fact that, in the eyes of well-placed interviewees, the EU has insufficient political influence over the Korean government to induce policy change, further aggravated by the aforementioned reluctance of the European Commission to take a strong stance on labour rights (E29, K25).
In Moldova, the government has signed and ratified all the fundamental conventions. ILO reports and interviews indicate there are problems with regard to child labour, especially in the agriculture sector, and with the labour inspectorate, which is covered indirectly in the procedural commitments of the TSD chapter to uphold domestic labour law (M1; see also ILO, 2016, Smith et al., 2017. But it proved difficult to retain focus on these labour issues when the political landscape was dominated by the profound governance problems mentioned above (M7) and when the predominant labour issues in Moldova relate to poor wages and conditions of employmentaspects not captured adequately in the ILO core labour standards framework. It may therefore be more productive to concentrate on the broader range of labour provisions contained in the Association Agreement, since these include obligations to introduce legislation influencing the quality and quantity of employment as part of the convergence of Moldovan legislation to meet the requirements of the EU's acquis communautaire. During the agreement's negotiation, these obligations were considered far more demanding and important than labour provisions in the TSD chapter (M2, M3). Indeed, in 2015, Moldova passed laws to transpose European Directives on fixed-term work and on obligations to inform employees about their conditions of work. Others including those pertaining to occupational health and safety, remain outstanding (Moldovan government, n.d.). However, the institutions of the TSD chapter are not designed to engage with issues outside the trade chapter of the Association Agreement.
This examination of the ways in which labour provisions are utilized (or not) and the elucidation of the concerns of key constituencies in our three case studies reveals the differentiated limits of the common formulation adopted by the EU. What we find is that the appropriateness of the EU model is brought into question in terms of substance (for instance, should the focus always be on ILO core labour standards?), process (for instance, what are the possibilities and limitations of dialogue in the context of particular political situations and relationships?), and scope (such as, should there be a focus on labour-related obligations beyond those contained in the TSD chapter?). At root, we suggest that the failure to design labour provisions to better articulate with third-country contexts has limited the possibilities for EU FTAs to protect and promote labour standards. Coupled with the challenges facing the operationalization of TSD chapters as identified earlierpresented summarily in Table 1 we therefore conclude that a serious reconsideration of the EU's trade-labour linkage is required.

Conclusions
A superficial account of labour provisions within EU FTAs tells a positive story. Trade agreements have been negotiated with relatively extensive substantive standards and

Primary labour issues
Trade-related redundancy, informal workers not protected Highly politicized struggles over labour rights, concerns over freedom of association and forced labour Poverty pay, child labour in agricultural sector, erosion of capacity of State Labour Inspectorate as part of wider process of labour market liberalization procedural commitments. Representatives of the respective parties are meeting with their international counterparts and civil society meetings are occurring. But in terms of addressing substantive labour standards issues, this article has shown that TSD chapters have delivered little. Scholars have already argued that the EU has not sought to 'aggressively' export labour standards through its trade agreements (Young, 2015). We show that neither have state officials in trading partners readily imported them. Rather, they have reluctantly acceptedor even actively softenedminimalist obligations around core labour standards. The weight of expectation has been loaded instead onto processes of dialogue, particularly via CSMs, widely considered as the key institutions for making progress on labour issues. However, CSMs are seriously hampered by various operational deficiencies as well as political marginalization within the broader institutional mechanisms and processes of the FTA. Overall, we found no evidence that the existence of TSD chapters has led to improvements in labour standards governance in any of our case studies, nor did we find any evidence that the institutionalization of opportunities for learning and socialization between the parties was creating a significant prospect of longer-term change. These findings thus offer the most robust refutation to date of the hypothesis that labour provisions in EU FTAs are actively advancing workers' rights. And in contrast to more optimistic assessments, they also suggest that future normative influence is unlikely to be realized via TSD chapters in their current form.
Should the EU therefore seek to put more of its market power behind its labour governance strategy? Such an approach could in part be realized by the most common suggestion for reform from European interviewees involved in the labour movement, namely to increase the enforceability of the TSD chapter by giving the EU the ability to withdraw preferential access to its market if labour standards are violated (E17, E19, E20, E25, E26). This would make the labour provisions more coercive, and in the shadow of sanctions, perhaps persuade trading partners to engage in more earnest dialogue and responsive action. Yet some interviewees in case study countries averred from this approach, with unionists and allied researchers expressing concern about the dangers of labour standards being utilized as a form of disguised protectionism (C7, K10). We agree, then, with the comments made by one labour representative who noted that more must be done to assuage such concerns and explain how specific forms of conditionality could benefit labour struggles in trade partners (E33). This demands consideration of a range of complex design issues including how a dispute is initiated, who it targets in relation to what labour-related allegations, who investigates those allegations, who decides on the kinds of corrective action and penalties, and what form of sanctions or fines are available to do this. Appropriately nuanced, a crude social clause founded on economic nationalism could thus be avoided.
But tackling enforceability issues will not by itself attend to the deficiencies identified above. In contrast to an 'our size fits all' model, Bicchi has argued that the EU needs instead to 'critically analyse the expected consequences of norm promotion for all parties involved and adapt … accordingly' (2006, pp. 289-290). In the Caribbean context, this could have been achieved by recognizing and responding to the aspirations of civil society to monitor the impacts of the EPA on employment and working conditions. In Korea, by developing alternative strategies for engaging key government officials with responsibilities for labour rights. And in Moldova, by giving the TSD chapter the remit to address labour obligations contained in the wider Association Agreement, as this would begin to address issues like health and safety at work that go beyond the ILO's core labour standards. In the same way that the EU has addressed context-specific commercial issues in a bespoke manner (such as in the annex to deal with non-tariff barriers in automobile parts under the Korea FTA), it ought to extend the same treatment to fundamental labour rights by designing provisions that articulate with the most pressing labour issues in each trade partner.
Stepping back from the question of efficacy, our final reflection is more scholarly in orientation. It concerns the underlying purpose of labour provisions. In the academic literature on the trade-labour linkage it is assumed that such provisions are intended to raise 'the national average' of labour rights protection in third countries, however measured and regardless of whether those jobs are connected to tradable goods and services. Yet the analysis presented here reveals that the groups and individuals central to the operationalization of the EU's TSD chapters had very different understandings of what they were meant to achievefrom monitoring trade impacts to reforming state institutions like the labour inspectorate. Further diversifying that sense of purpose is the notion that TSD chapters are ways of regulating global supply chains, as exemplified by the promotion of East Asian CSR in the Korea case and elaborated in the European Union's (2015) proposed text for the Transatlantic Trade and Investment Partnership and in speeches by the EU Trade Commissioner (Malmström, 2015). 6 This explicitly reorients labour provisions toward jobs in transnationally-organized export industries and away from those within discreet national economies. Lastly, we note that at the 2017 civil society forum of the EU-Korea FTA some participants turned the discussion toward inadequate protection of workers within the EU, complicating the assumption that such provisions are essentially externally orientated. While such alternative trade-labour linkages could create new possibilities for improving labour standards, they also demand new methodologies to scrutinize them along the lines we have developed here, paying more attention to causal mechanisms and intended beneficiaries.