The politics of labour relations in global production networks: Collective action, industrial parks, and local conﬂict in the Ethiopian apparel sector

In this paper we examine the emerging politics of labour agency as new manufacturing locations are incorporated into existing global production networks, using the example of the Ethiopian apparel industry. The Ethiopian state has employed an active space-based industrial policy to attract leading apparel manufacturers into a series of new industrial parks in the country. Both investors and the Ethiopian government expected to ﬁnd a large and pliant labour force willing to work for low wages. However, the new sector has already seen a wave of collective and individual resistance from workers. We ask which factors contribute to drive and constrain labour agency and shape the speciﬁc forms it takes in ﬁrms tied into leading global production networks. Drawing on a large-N quantitative survey of factory workers and in-depth qualitative interviews with managers, workers, trade union representatives and government ofﬁcials, we show how the quality of industrial relations depends not just on state action and the business strategies of lead ﬁrms in production networks, but also on variegated forms of labour agency used both by organised and unorganised Ethiopian workers. We ﬁnd that many industrial conﬂicts result from the collision of the productivity imperatives of manufacturing ﬁrms tied into demanding, but low value-added, segments of global production networks with the expectations of workers with limited prior experience in industrial jobs, but are compounded by the contradictory actions of different state agencies, a lack of formal unionisation, and the contingent interactions of factory based grievances with local political conﬂicts. Industrial parks emerge as spaces of particular contestation. Our ﬁndings highlight the need to adopt an understanding of labour regimes grounded in local political realities. These ﬁndings have implications for the design of industrial policies and labour market institutions aiming to support ﬁrms and workers in emerging manufacturing clusters.


Introduction
Global manufacturing is increasingly organised via relationships between suppliers and transnational lead firms who organise where and under what conditions production takes.These relationships, variously described as global value chains (GVCs) or global production networks (GPNs), create opportunities and challenges for firms and workers in low-income countries (Hess, 2009).Firms might benefit from access to overseas markets, while workers might benefit from increased employment opportunities.
However, these networks are characterised by asymmetric power relations between lead firms and suppliers which constrain their developmental potential (Gereffi, Humphrey, & Sturgeon, 2005;Yeung & Coe, 2015).Supplier firms in many low-and middleincome countries are seeing their mark-ups squeezed as lead firms capture the benefits resulting from lower production costs (Selwyn, 2019;World Bank, 2020).Outcomes for workers therefore depend on the conditions of their inclusion into global production networks, and the agency of workers inside and outside the workplace plays an important role in shaping conditions in supplier firms (Coe, 2013).In the workplace such agency comprises both compliance with and resistance to management strategies (Taylor et al., 2015), while in the wider political sphere the balance of power between workers and employers influences the institutional norms that govern interactions between the state, labour and capital (Selwyn, 2013).
In this paper we ask what factors drive labour conflict and worker resistance in new export manufacturing sectors and what implications such conflicts have for working conditions.To address these research questions we illustrate the dynamics of encounters between international firms and local workers and the contingent nature of outcomes for workers with an analysis of the apparel sector in Ethiopia, which offers a case study into the establishment of a new export-oriented manufacturing industry.Ethiopia has embarked on an ambitious industrialisation strategy centred around an active industrial policy aiming to attract foreign direct investment (FDI).For Ethiopian policy elites attracting such FDI is a political necessity.It can help overcome constraints on industrialisation in a context where domestic capitalists have limited access to capital and capabilities, which hampers their ability to compete in international markets.At the same time, inflows of export-oriented FDI can help address balance of payment constraints (Cramer et al., 2020), while new manufacturing jobs provide employment for a rapidly growing labour force.
To streamline service provision and attract a new industrial labour force, the apparel industry in Ethiopia is concentrated in a series of industrial parks (Whitfield, Staritz, & Morris, 2020;Whitfield & Staritz, 2020).Establishing such a labour force is a necessary condition of sustained access to the global production networks that structure the sector, but building an industrial workforce is an uneven, conflictive and slow process (Thompson, 1967).The Ethiopian state is engaged in constructing a national labour regime that can support insertion into demanding global production networks by regulating the labour market, employment relations, and the institutions of collective representation.This process is, however, co-constituted in important ways by workers, who gain access to new avenues for exercising their agency (Selwyn, 2012).
Understanding labour relations at the factory level requires a combination of different layers of analysis, from the macropolitical economy of growth in the country, through the sector and global production network level, to the micro-level of capitallabour encounters in the workplace.To capture these factors and their interactions, we combine a quantitative large-N survey of factory workers with in-depth qualitative interviews with managers, workers, trade union representatives and government officials.We show how labour relations are not simply determined by the strategies of lead firms and their suppliers, but also depend on state-level regulations and responses by organised and unorganised labour.We find that labour conflict, particularly in the form of wildcat strikes, was more common in industrial parks, which host mostly foreign firms and have low levels of unionisation.The incidence of conflict varied significantly among firms within parks, with internationally recognised companies being more affected than smaller and less visible firms.This variation is driven in part by localised political conflicts.The acceleration of strikes and labour mobilisation coincided with a tumultuous period of political tensions and protests that began in 2014.The targeting of highprofile factories appears to boost the visibility of regional and local opposition groups.However, despite these entanglements, genuine labour grievances were at the heart of labour conflicts.Local political mobilisation provided an opportunity to make these strikes more effective.Our analysis further suggests that, although strikes are common, 'exit' is a far more widespread form of resistance.Exit manifested in high rates of absenteeism and turnover, particularly for workers in their first three months into the job.
This paper aims to contribute to the literature in three ways.First, we provide new empirical evidence on labour outcomes and worker agency in a country which only recently entered apparel GPNs controlled by global buyers.With the partial exception of Bangladesh (see Anner, 2020;Saxena, 2019;Kabeer et al., 2020), empirical evidence on labour dynamics within GPNs is still thin, and especially so for 'late-late comers' like Ethiopia.Second, while research on labour agency in GPNs tends to focus on formal organisations such as trade unions (Cumbers et al., 2008), we offer evidence on forms of resistance not mediated by unions, such as wildcat strikes and high turnover, which are important features of the Ethiopian context (Hardy & Hauge, 2019).Third, we demonstrate the importance of considering local politics and political conflicts for both the timing and targeting of labour resistance.Dominant narratives about low-income countries stress how plentiful supplies of labour, a supposedly docile labour force comprised of women with limited employment experience, and the stability that industrial jobs offer should dampen industrial strife.However, in new centres of production workplace encounters are expected to eventually produce new waves of resistance (Qi & Pringle, 2019;Silver, 2003).Ethiopia is of particular interest in this regard as conflict arose in the very early stages of global capital penetration, albeit with significant variation across firms.Understanding this variation requires taking local political dynamics into account.
The remainder of this article is organised as follows.Section 2 proposes a theoretical framework to analyse labour outcomes within distinct labour regimes shaped by global, national and local forces.This section also briefly lays out how we collected our data.Section 3 provides the necessary background to understand capital, the state and labour in the emerging apparel industry in Ethiopia.Section 4 contains the primary empirical material on the dynamics of labour conflict and resistance with regard to strikes and constraints on collective action, the role played by localised political conflict, and struggles over effort and mobility.Finally, Section 5 concludes.

Analysing labour conflict and variegated labour regimes: Theory and data
Labour relations in the Ethiopian apparel export sector have undergone rapid transformations with the arrival of new foreign investments linked into global production networks organised by global buyers.To analyse these dynamics, we draw on labour process theory, mobilisation theory and GPN analysis.We use the concept of labour regimes as a heuristic tool to explore how interconnections between factors inside and outside the factory help explain differences across workplaces.This concept stems from work on the politics of production (Burawoy, 1985;Lee, 1999) which connects workplace labour relations with the macro-politics of capital-labour relations in a national or global context.Labour regimes build on labour process theory to frame the interrelations of labour market segmentation and workforce mobilisation, conditions of employment, and forms of enterprise authority and control for the appropriation of surplus value (Bernstein, 2007;Thompson, 1989).Employers and managers are under permanent tension between manufacturing consent, i.e. creating internal labour markets that provide some stability to their workforce, and exercising coercion to ensure work effort matches productivity and quality imperatives (Burawoy, 1979;Collinson & Ackroyd, 2006).Beyond the workplace, labour regimes incorporate the institutions of social reproduction which ensure that workers can be mobilised, motivated, utilised in production, and reproduced (Lerche et al., 2017;Taylor & Rioux, 2018).In this sense, labour regime analysis is ''necessarily multi-scalar", incorporating the global, national, and local (Selwyn, 2016).The associated concept of labour control regimes emphasizes the role of local/national labour institutions and the different sources of disciplining labour, whether governments, labour market conditions or employers (Jonas, 1996;Anner, 2015;Pattenden, 2016;Smith et al., 2018;Coe, 2015).At all levels labour regimes are coconstituted by the actions of workers and their collective organisations.The extent to which workers engage in struggles, and the strategies they employ, influence and are influenced by the balance of power between capital and labour (Selwyn, 2012) as well as broader social relations such as ethnicity and gender (Barnes, Krishna, & Pratap, 2016;Mezzadri, 2016), all of which impact the organisation of GPNs.
To understand our central research problem, i.e. labour conflict and resistance in the emerging apparel sector, we conceptualise labour regimes in the dialectic of structure and agency (Van der Linden, 2008;Wright, 2000;Atzeni, 2009).Particularly relevant to this line of inquiry are Ohlin-Wright's concepts of structural and associational power (Ohlin-Wright, 2000), which have influenced work on labour relations in a wide range of contexts (see for instance Rizzo & Atzeni, 2020).Associational power flows from the organisation of workers into collective bodies.Structural power can be subdivided into workplace and marketplace power (Silver, 2003).The former describes power resulting from workers' strategic place in particular production processes and their ability to disrupt activities further along the chain, while the latter stems from tight labour markets which impede the ability of firms to replace workers.There are different forms of labour mobilisation and resistance, and their occurrence relates to different combinations of workplace and marketplace structural power, as well as perceptions of 'injustice' and 'opportunity', following Kelly's work on mobilisation theory (Kelly, 1998).These concepts help explain what triggers workers' action, whether such mobilisations reflect labour-processgenerated solidarity (Atzeni, 2009), and how contingent opportunity factors enhance workers' structural power. 1 In cases where there is no open conflict, such as strikes, workers can still exert pressure that may impact aspects of the dominant labour regime.Smith (2006) has suggested that work contracts contain two types of labour indeterminacy and stressed the importance of workers' mobility or 'exit' and 'mobility-effort struggles' in shaping labour relations.By labour indeterminacy Smith (2006, 390) means that labour power, i.e. the capacity to work over a given period of time or 'what the employer hires and the worker exchanges', is indeterminate, first, because the amount of effort cannot be fixed ex-ante and is subject to bargaining at the shop floor level; second, because of labour mobility, i.e. workers' ability to leave the job.We argue that these manifestations of workers' resistance can be as important as strikes.
As many of the new apparel factories in Ethiopia are part of GPNs with a strong manufacturing base in Asia, the characteristics of labour regimes can be linked to practices that 'travel' across the geography of GPNs (Anner, 2020).The GPN framework has generally focused more on the processes of value creation and capture, via structures and dynamics of power (corporate, collective, etc.), by considering the social, network and territorial embeddedness of actors, markets and processes (Coe & Yeung, 2015).Although labour remains a significant gap in the GPN literature, there have been important theoretical and empirical contributions on social upgrading (Barrientos et al., 2011;Selwyn, 2013;Rossi, 2013), the role of labour agency in producing geographies of labour (Herod, 2012), workplace micro-dynamics, gender and migration as key elements of labour dynamics within GPNs (Baglioni, 2018), and the incorporation of the aforementioned concept of labour regime (Coe & Yeung, 2019;Taylor & Rioux, 2018).
Based on these insights, we adopt a framework that combines three sets of categories of analysis at different scales.First is the analysis of micro-level workplace dynamics and 'raw' encounters between employers and workers over wages, productivity, safety, effort, labour time, and mobility considering structural and associational power and drivers of mobilisation at factory level.Second, we consider the characteristics of a particular sector or set of global production networks, which cut across national boundaries and generate specific imperatives of labour control.The apparel sector has become increasingly concentrated since the phasing out of the multi-fibre arrangement (Gereffi and Frederick, 2010).Highly concentrated global buyers have been able to reduce the unit prices they pay their suppliers.At the same time, due to the dominance of 'fast fashion', suppliers are expected to help lead firms reduce the 'time to market'.Suppliers operating in these GPNs are often subject to a double profit and sourcing squeeze, and face pressures to tighten mechanisms to manage work effort (Anner, 2020).They may therefore impose methods of labour control that are harsher and more intrusive than those in firms not subject to these squeezes, which has implications for the incidence and nature of labour conflict at workplace level.Thus, governments competing to attract GPN-linked manufacturers may be compelled to create suitable conditions for these labour control regimes (Selwyn, 2019).However, within this type of 'global' labour regime, variation remains possible and is shaped by 'varieties of capital' (Lee, 2017), particular buyer-supplier relations (Amengual et al., 2020), government regulations and their enforcement (Anner, 2015), specific firm-level corporate ethos and management discourses (Goger, 2013) and, overall, variation in labour's social reproduction costs alongside the specific balance of power between capital and labour (Selwyn, 2019).
Finally, we incorporate key aspects of the state-society relations which shape labour supply dynamics and the arenas of labour struggles, whether over the extent of commodification, the limits to labour reproduction, or claims over representation.This 'national political economy' scale encompasses the institutions that underpin these relations, including labour legislation and its enforcement, trade unions and their power of representation, employer organisations, and social protection systems, all critical elements to understanding workers' structural and associational power and their effects on labour outcomes (Anner, 2015;Lee, 2017;Silver, 2003).
To capture these combinations of context, mechanisms and outcomes, we draw on a mixture of quantitative and qualitative data collected in Ethiopia in 2017 and 2018.Our quantitative data comes from survey of 531 workers across 25 apparel companies in Ethiopia, 17 of which were foreign-owned and eight were Ethiopian-owned.We purposively selected the most successful enterprises in terms of employment creation, production volume and export performance.Our sample includes leading firms both inside and outside of the new industrial parks.Within each company we constructed a random sample of production workers, stratified by skill group.Following our main survey in 2017 we conducted a phone survey with 120 workers in 2018 to capture developments in employment status and take-home pay.We also collected firm-level data from each sampled enterprise.In parallel, we conducted 150 in-depth qualitative interviews with managers, workers, trade union representatives, government officials, NGO officials, and members of international organisations. 2 .
1 Opportunity factors may include conditions that alter the labour-capital balance of power, the costs of repressions, and the ongoing contestation of state power.Admasie (2019) provides a very useful analysis of the dialectics of opportunity and collective action in Ethiopia.
2 Throughout this paper we will refer to interview sources with a simple coding system.The first two letters of each code indicate the profile of the respondent and the numbers show the order of interviews (e.g.CM01).All interviews were conducted in Ethiopia between 2017 and 2018.The letter codes have the following interpretations: CM = company manager, usually factory GM or deputy manager; GO = government official in different departments and agencies, mainly middle-to-high level officials; TU = trade union representative, either at federal level or from factorylevel basic unions; WI = worker interview, specifically qualitative work/life histories conducted with a sub-sample of workers extracted from the initial quantitative survey; Oth = Other such as NGOs, or other key informants (e.g.researchers).This system is designed to preserve the anonymity of respondents, as promised in our consent form.Our sampling and data collection methodology is discussed in detail in Schaefer and Oya (2019).
C. Oya and F. Schaefer World Development 146 (2021) 105564 3. The state, capital and labour in Ethiopia's emerging apparel sector

Unpacking the state
The Ethiopian state has played a central role in managing foreign direct investment and the subsequent interactions of foreign firms and domestic workers.State officials, aiming to generate foreign exchange, create jobs, and accelerate structural transformation, have helped attract international companies involved in leading apparel global production networks.The state has adopted a strategy of industrial agglomeration by concentrating incoming FDI in a series of new industrial parks. 3The most important parks in terms of employment are the government-owned Bole Lemi and Hawassa Industrial Parks (HIP), operational since 2015 and 2016, respectively, and the privately-owned Eastern Industrial Park (EIP, previously Eastern Industrial Zone), which opened in 2012.Smaller government-owned parks opened in Mekelle and Adama in 2017 and 2018. 4In late 2018 total employment in the industrial parks stood at around 65,000 (see Table 1), a figure which had risen to 82,000 by December 2019 (EIC, unpublished).
To understand how policy impacts the politics of production in Ethiopia we must unpack the state institutions dealing with FDI management, industrial development, and labour control.As part of Ethiopia's industrial policy, a network of government institutions has taken the lead in managing FDI and industrial parks.The Ethiopian Investment Commission (EIC), the state-owned Industrial Parks Development Corporation, the Ministry of Trade and Industry, and the team advising the prime minister on industrial policy form the apex of this structure, with the EIC acting as a coordinating hub.Interventions feed a continuous learning process, leading to differences in institutional design across the different 'generations' of industrial parks (Oqubay & Kefale, 2020).Labour management is, however, much less coherent.The Ministry of Labour and Social Affairs (MoLSA), which regulates the labour market, plays a contradictory role: on the one hand officials seek to avoid potential deterrents to new investors, such as the promotion of trade unions, but on the other hand officials do sometimes act to protect workers in cases of disputes.5Regional and municipal authorities, including the regional Bureaus of Labour and Social Affairs (BOLSAs) play an important role in labour mobilisation: directly, by supporting recruitment efforts in industrial parks and, indirectly, by managing local services.These bureaus answer to regional governments and are not directly accountable to federal authorities.Therefore, the coordination across these levels of government and the contradictions that arise from differing political priorities are major challenges for the industrial policy eco-system.This coordination is made more difficult by the fragmented nature of the Ethiopian polity, which is characterised by political conflict increasingly fought along ethnic lines.

Varieties of capital
Government industrial policy has been successful in attracting new capital from abroad.Between 2012 and 2019 nominal FDI stocks in Ethiopia expanded from USD 5.1 billion to USD 24.9 bil-lion (UNCTADSTAT, 2020) and by 2017 foreign firms had created close to 100,000 jobs (Cheru & Oqubay, 2019).A significant proportion of these investments originate in Asia, and Chinese investors constitute the largest group.These investments take place within the context of shifts in the production locations and modalities of the global apparel sector.Large buyers in the US and EU demand low costs and flexibility from manufacturers, leading them to focus on a smaller number of large-scale suppliers.In turn, the largest suppliers have themselves become transnational corporations capable of offering manufacturing-related services such as design, logistics, and inventory management (Kumar, 2020).In search of new spaces of low-cost production these firms are moving beyond their production bases in Asia.Concentration among buyers and fierce competition among manufacturers results in highly asymmetrical power relations among firms in the value chain, even if some of the large transnational suppliers have been able to capture higher-value activities (Staritz, 2011).
Foreign apparel investors came into Ethiopia in three waves, which form different varieties of private industrial capital (Staritz & Whitfield, 2017).The first wave comprised mostly Turkish investors, who arrived before 2010 when industrial parks were not yet a core government strategy.Most of these investors were aiming for the domestic market.The second wave consisted of Chinese investments clustered in the EIP and again included many firms producing for the domestic market. 6The most recent wave, which began in 2014, is dominated by suppliers integrated into well-known apparel GPNs moving to government industrial parks, such as Bole Lemi, Mekelle and especially Hawassa.These parks were designed to attract foreign investors and host almost no domestic companies (Whitfield, Staritz, & Morris, 2020).Despite their particularities, these three waves are connected because some of the first largescale export-oriented firms to settle acted as catalysts of subsequent waves of investors (Lin & Xu, 2019).The domestic firms in our sample were on average much older than the foreign firms and not sited in industrial parks.

Labour force characteristics
The apparel workforce in Ethiopia is characterised by four basic features: most workers are young, female, have migrated internally, and have relatively high levels of education for the jobs they fill.In our sample of 531 workers the mean age was 25 years and 75% were women.While most workers in so-called low-skilled positions are women, especially in foreign firms, men dominate in semi-skilled jobs.The labour force in sampled factories, especially in foreign ones, is substantially more educated than the national average.Among so-called low-skilled workers 52.8% had completed at least grade 10 and another 17.4% had additionally completed a technical and vocational training (TVET) course.For comparison, at national level, the (primary) grade-8 completion rate was 54% in 2015/16 and gross enrolment in secondary education barely reached 30% (Rekiso, 2019).
The other key characteristic is mobility.Persistent un-and underemployment make labour migration a common feature in Ethiopia.Most workers in our sample are internal migrants, and workers in industrial parks were more likely to have migrated than workers outside of parks.There are also differences across industrial parks.For example, in the case of Hawassa, workers' mobility was partly induced by a government policy of far-flung catchment areas, while in the Eastern Industrial Park most migrants came from surrounding areas.Across the whole sample of workers, the majority (58%) was in their first factory job, and 46% of workers had no formal work experience.Among workers in industrial parks 65% had no previous experience in manufacturing and 47% had no work experience outside of the household. 7

The politics of labour relations in a fast-moving scenario
According to the narrative put forward by the government Ethiopia has a large pool of workers in need of stable jobs and ready to work for 'competitive' wages (EIC, 2017). 8Large numbers of eager workers were expected to flock to the gates of industrial parks.Such an excess supply of labour would reduce workers' structural marketplace power and limit conflict, at least in the early stages of manufacturing growth.As in some Asian settings, managerial expectations were further conditioned by racist and sexist discourses concerning the supposed docility of Ethiopians in general and of young women in particular.
However, capital-labour relations in Ethiopia's apparel sector have proven much more fractious than anticipated.Workers in new manufacturing firms, particularly the ones supplying global buyers, have used both direct and indirect forms of labour resistance.Direct forms of resistance entail a confrontation with the employer and include collective mobilisation and work stoppages.These can be highly disruptive for manufacturers selling to global markets, potentially giving workers substantial structural workplace power.By contrast, indirect forms of resistance avoid collective confrontation with employers.These include 'exit' strategies, i.e. workers leaving the factories or not being available for recruitment.Alternatively, workers who remain may employ 'everyday' resistance strategies such as absenteeism, resistance to overtime, and foot dragging, which can have a detrimental effect on productivity.These actions affect the amount of work effort over which employers and workers bargain (Smith, 2006).
We found that these forms of collective and individual resistance were common, but frequently unanticipated by foreign managers.Not all firms or parks suffer in equal intensity, though.An examination of the conditions increasing the likelihood of such struggles reveals a complex interplay of mismatched expectations between workers and employers and the role of local political conflicts.This section therefore explores the dynamics of labour relations in relation to three key themes.First are the instances of direct labour conflict, in the form of strikes and collective action.Here our focus is the limited role played by formal trade unions.Second, we argue that labour relations are not simply shaped by GPN dynamics and the import of prevailing labour regimes.The observed patterns of labour conflict, and especially strikes, cannot be understood without considering the interaction with political conflict between central and regional political actors.Third, we look at less open forms of worker resistance, in particular around turnover, which became a serious short-term problem for many investors.

Conflict in action: Strikes and constraints on collective mobilisation
Collective mobilisation in the form of strikes is a common form of labour resistance and, according to labour process theory, is particularly effective where it builds upon a combination of strong workers' structural and associational power (Ohlin- Wright, 2000;Silver, 2003;Collinson & Ackroyd, 2006).However, the increasing fragmentation of manufacturing production across developing countries appears to have driven a 'global quiescence', manifested in lower union densities and less collective action (Piazza, 2005;Taylor & Rioux, 2018).
The recent experience of Ethiopia does not conform to this pattern.Despite the scarcity of stable jobs for the vast number of people entering the labour market every year, collective resistance by workers is common in Ethiopian manufacturing.After a period of 25 years of relative tranquillity (Admasie, 2019), and a rapid growth in manufacturing employment since 2015 (Cheru & Oqubay, 2019), the incidence of strikes in the apparel industry has increased.An ILO-led survey of 43 firms in the Ethiopian apparel sector reported that 20% of firms had experienced a strike in the previous 12 months ( ILO, 2016).According to our survey data, 38% of a sample of 531 workers reported having witnessed a strike in their company in the 12 months prior.Strikes in apparel manufacturing coincided with strikes in other sectors, especially between September 2017 and June 2018, which saw the 'greatest wave of labor unrest in Ethiopia since the mid-1970s' (Admasie, 2018, 432).During the same time political tensions led to the destruction of a number of foreign-owned factories and plantations by protestors.
Strikes in the apparel sector were concentrated in industrial parks and in certain politically volatile locations, such as the periurban rim of Addis Ababa and the city of Hawassa.According to our data, at companies located in industrial parks nearly 50% of workers had witnessed a strike, while outside of the parks only 25% of workers had.There is substantial variation across industrial parks, with the highest incidence, 65% of workers, reported among workers in Bole Lemi and the lowest, 39% of workers, in the Hawassa park (see Fig. 1).9These two parks are the ones with the highest concentration of factories producing for GPNs.After 2018, however, the frequency of strikes declined in Bole Lemi, with most firms finding ways of minimising unrest (CM52, Admasie, 2020).Meanwhile, work stoppages increased in the Hawassa park, especially when a major strike erupted in March 2019 (IndustryALL, 2019).Variation was also remarkable within industrial parks.For example, in the EIP, while a few firms, including the largest and most 'visible', were hit by waves of two or more strikes in 2017-18, most other firms were spared.In one of the largest apparel firms operating outside industrial parks, company managers complained that strikes happened regularly to push for annual wage revisions (CM25).Most of the reported labour conflict was in the form of wildcat strikes and generally short-lived, not lasting more than a few days.
Overall, we find that low wages and wage setting systems lie at the heart of conflict in many firms, a conclusion that confirms earlier research by the ILO (2016).However, as with strike incidence, we observe a significant difference between factories in industrial parks and those outside.While outside of industrial parks 95% of strikes were about wages, inside the parks, which host almost only foreign companies, concerns such as excessive working time, lowquality canteen food and ill-treatment of workers by supervisors accounted for between 26% and 67% of strikes, depending on the park. 10 In the case of the Hawassa park, grievances included obstacles to union representation, and issues of safety and harassment outside the industrial park, with young women being particularly vulnerable (IndustryALL, 2019; Admasie, 2020, TU03, WI01, Oth01).Wage and non-wage grievances are of course linked, as low wages often led to unsafe accommodation and dangerous commutes, fuelling requests for housing near the park (GO28, TU03).According to company managers in the Bole Lemi park, grievances around living costs in Addis Ababa and the lack of affordable housing were particularly acute and drove some firms to consider the in-house provision of food, accommodation and other amenities (CM52).
While monthly wages varied substantially across companies, low wages are a common complaint by workers and trade unionists, and both employers and government officials frequently conceded that wages levels were too low.It is hard to compare monthly wages to 'average' levels in the economy, given a lack of reliable data and substantial heterogeneity across sectors and firms.However, we found that, in purchasing power parity (PPP) terms, no workers in our sample earned less than the $3.20 international poverty line and all workers earned more than twice as much as $1.90 per day, which is the reference for poorer countries.Low-skilled workers in our sample earned between $130 and $155 per month in PPP terms.Reported wages were roughly in line, accounting for inflation, with wages in flower farms, estimated at between ETB 800 and ETB 1100 in 2014 (Melese, 2015), but lower than wages for low-skilled workers in the construction sector (Schaefer & Oya, 2019). 11Several company managers stressed that base wages were always 'over' the national poverty line (CM35, CM40).
However, international poverty lines have been criticised for being too low to sustain dignified livelihoods (Selwyn, 2012).In our sample, workers systematically complained that they did not receive living wages, given rising living costs and their need to support families. 12Overall, just 27% of low-skilled workers and 41% of semi-skilled workers stated that their earnings were sufficient to cover monthly expenditures.Workers made up the shortfall by relying on income from other household members and going into debt.While wage levels were reported as an issue by workers in all types of companies, they were a particular issue in the international com-panies sited in industrial parks. 13Our longitudinal evidence also shows that, with the exception of Bole Lemi, wage growth inside the parks was no faster than outside of parks between 2017 and the end of 2018 (see Table 2).The significant rise in nominal wages (40%) experienced in Bole Lemi appears to be related to the frequent strikes and a shared perception among workers and employers that living costs were rising fast in and around Addis Ababa, which led to concessions from some of the employers (Admasie, 2020).
Wage levels reflect managerial calculations of profitability in companies that operate within GPNs with thin profit margins.However, interviews with both trade unions representatives and company managers suggested that the government also played a role by influencing wage-setting systems.At the time of the survey, a few months after most factories in the Hawassa park had started operations, wages there were especially low.Some of the differences may be explained by higher living costs in and around Addis Ababa.However, the firms in the Hawassa park and the Ethiopian government also operated a unique 'cartel' system that was established to keep starting wages in Hawassa sufficiently low for firms to commit.Factories inside the park agreed on a common base salary but discretion could be used for performance pay, thereby allowing for limited variation in wages across factories.In the early stages of park operations many workers reported wages that were close to their low base salaries, leading to significant differences with respect to other locations.In addition, government park managers appear to have contributed to initial wage repression by recommending out-of-date base wage standards to newly arrived companies.Low base benchmarks were repeatedly reported by managers in Hawassa, where companies assumed acceptable wages levels to be substantially lower than around Addis Ababa, and 'above national poverty lines' (CM35).Government sources suggested that officials set low benchmarks in the expectation that additional performance-related pay would soon lead to much higher wages (IO02, GO28). 14However, the combination of poorly understood benchmarks, limited knowledge of context on the part of newly arrived foreign managers, and the government's fear of discouraging investors in their quest to compete with other countries in attracting FDI, led to wage setting systems which drove many of the strikes and part of the high labour turnover analysed in the section below.In the eyes of some senior managers at the park low starter wages were justified by low productivity.As one manager put it: 'there is no point in rewarding work if not good enough for the demands of the industry' (CM35).This manager, like many others, expected rapid increases in productivity over a three-to five-year period, from 10 to 15% to more than 60% of the global norms, leading to higher wages via performance-based pay (CM40, CM41). 15By contrast, low benchmarks were not an issue in the first industrial park, the Eastern Industrial Park, where the government did not play an active role in organising the arrival of companies and managing the park.In this park, operated by Chinese companies, average wages for same jobs were significantly higher than in Bole Lemi and the Hawassa park (see Table 1). 16These contrasts show that struggles 10 Food provision and the presence of foreign managers and supervisors was far more common in foreign firms in industrial parks than in firms outside the parks, increasing the potential for conflict over a wider range of issues. 11The construction sector has grown rapidly in the past two decades on the back of a vast infrastructure construction drive and accelerated real estate development.
Several key informants reported that firms in industrial parks had to compete with a strong labour demand for (especially male) workers in the urban construction sector as well as in other non-tradable services. 12Melese (2015) estimated required living wages for a family at ETB3,367, significantly higher than reported wages in the flower sector.This calculation assumes a family of five with 1.6 workers per household. 13See also Selwyn (2019) on living vs poverty wages in 'global poverty chains'. 14They also argued that firms should pay higher wages if they could afford them and that benchmarks for base salaries, such as the national poverty line, were only indicative of a minimum wage floor. 15These factory managers suggested that low levels of time efficiency were normal in early stages.For reference, they quoted levels of 85% in the best performing factories in China and over 75% in Vietnam (CM41, CM36, CM52).These values are determined by calculating operator's efficiency in relation to a benchmark of 'standard allowed minutes" (or standard minute value) per operation. 16These wage comparisons are based on the workers surveys in the sampled firms in each park.Firms in the EIP sample were primarily export-oriented and GPN-linked with the exception of one textile company.Therefore, they are comparable to other GPN-linked firms in Bole Lemi and Hawassa.The main difference, apart from nationality of ownership, was the years of operation in the country, i.e. longer in the case of EIP firms.
C. Oya and F. Schaefer World Development 146 (2021) 105564 over wages and workers' effort need to consider the contingent role of actors external to the factory floor.
Collective action in the contemporary Ethiopian apparel sector is largely organised outside of formal union structures.Strikes in the sector were generally wildcat strikes.Officials from the Confederation of Ethiopian Trade Unions, the main trade union body in the country, acknowledged that most strikes were 'unorganised' in that they happened in factories without union representation.There were some reports of wildcat strikes even in firms with a basic union in place (TU03).This is not a new phenomenon.The history of labour conflict in Ethiopia since the end of World War II is marked by recurrent wildcat strikes, even when unions existed, partly because the labour law makes legal strikes difficult (Admasie, 2019).
Contemporary wildcat strikes are driven both by legal impediments to union-organised strikes and by the low trade union density in Ethiopian manufacturing in general, and in industrial parks in particular.An ILO study found gross trade union density in 2010, i.e. before the establishment of the main industrial parks, was around 15% and less than 1% of wage workers were covered by a collective bargaining agreement (ILO, 2016). 17Trade union data show more than 60% of firms in the sector did not have any union presence and most workers were not union members (TU03, CCOO Industria, 2018;Admasie, 2020).
As in many other countries, larger firms are more likely to have a basic union.Among the larger firms in the sector, foreign firms were much less likely to allow trade unions to form than Ethiopian companies, at least initially.The aforementioned ILO (2016) report found that foreign investors seemed especially resistant to union presence in their factories. 18Our survey confirmed that most of the foreign firms in industrial parks have no formal trade union presence (see Figure 1 above).Trade unions have de facto (though not de jure) been discouraged in the three largest parks.Several union representatives reported that there was active discouragement of unions by government institutions, a fact corroborated by other reports (Admasie, 2020;ILO, 2017).This is again not new in Ethiopia, where historically the relationship between organized labour and the state has often been fractious, and labour has been particularly militant in apparel companies (Admasie, 2018(Admasie, , 2020)). 19According to our interviews with trade unionists, the EIC opted for a handsoff approach to labour matters, whereas the MOLSA and BOLSAs seemed reluctant to take the part of workers in cases of conflict unless management misbehaviour was clear-cut.
Different degrees of government discouragement across parks partly explain variation in trade union presence.The perceived discouragement was particularly strong in the government-owned Hawassa industrial park (TU04).Here the collective preference of company managers was to avoid formal trade union representation and instead operate with more artificial and undemocratic organi-  and 2018 the proliferation of many small and 'poor' enterprises means that trade union density and especially collective bargaining generally remain low in this sector (TU03).
18 However, trade union representatives at Ethiopian-owned firms also consistently reported that despite the presence of a trade union collective bargaining agreements were not adhered to. 19Admasie (2019) in dialogue with Kelly's framework, suggests that labour conflict in Ethiopia has followed a series of cycles since the 1960s.These 'proletarian insurgencies' in the late 1960s, early 70s, then early 1990s, were followed by periods of calm, partly driven by repression and partly by consent, depending on the configuration of organised labour and the relationship between trade unions and the state.The recent wave of strikes documented in this article can be viewed as part of this long history of labour mobilisation in the country.
sational forms such as workers' councils.Such joint workermanagement committees are common in the apparel sector, but tend to have a limited mandate, often focused on health and safety issues (ILO, 2016). 20Moreover, according to interviews with factorylevel union representatives such joint committees, along with some of the basic union structures, were dominated by management personnel from human resources departments, particularly in foreign owned firms (TU06, WI02).Workers' councils organised by senior management without involving trade unions reflect a combination of two of the four dominant patterns of work practices in developed countries according to the classification proposed by Katz (2006), i.e. a 'low wage' managerial archetype and the HRM model.Anti-union animus prevails in the former while the latter is characterised by union substitution of the kind observed in the Hawassa park.
The preference for union substitution in the Hawassa park reflected negative perceptions of trade unions shared by foreign investors and the government agencies managing the park.Managerial bargaining leverage is often enhanced when alternative production sites are perceived to pose a threat to the scaling up or even the continuation of an investment.In Ethiopia, the government seemed particularly sensitive to the mobility of capital and eager to demonstrate that conflict could be managed, as shown in interviews with MOLSA.In this regard, the government's stance vis-à-vis trade unions has contributed to a predominance of 'low wage' workplace patterns with hierarchical work relations and significant managerial discretion over working conditions.These, in turn, fuel grievances and perceptions of injustice that can result in unorganised strikes, as suggested by labour mobilisation theory.
Finally, how strikes were resolved also tells us much about the shifting politics of labour relations.In the initial stages, a common managerial tactic, especially in non-unionised firms, was to fire organisers, a practice that is against the labour law.This is a common method to counterbalance collective action and demobilise workers (Kelly, 1998).The narrative was two-fold.Some managers argued that workers only took action due to the intervention of a minority of 'trouble-makers', implying that grievances were not widespread (CM27).This is clearly not true considering the ample evidence of complaints.The second justification was that the 'instigators' were in fact political mobilisers using factory workers to further their political agendas (CM63).We will come back to this issue below.
Less belligerent firms attempted to appease workers by reviewing their management procedures to better respond to grievances.For example, some managers in Chinese firms argued that welltrained managerial staff was key to successfully managing a restive labour force.They also admitted problems in transmitting the corporate culture of the firm to workers, and an excessive tendency to attribute problems to a 'lack of work ethic' (CM51, CM63).Better trained staff would be more able to communicate productivity imperatives and organise a labour force with different levels of efficiency.Managing such differences was seen as critical to improving workers' attitude to production targets and avoiding widespread discontent.
At the same time there is a growing divergence of attitudes towards unions among foreign firms.For instance, in Bole Lemi and the Eastern Industrial Park firms were initially openly opposed to trade unions, as managers associated unions with more labour conflict.Over time, managers at a small number of firms realized that having adequate worker's representation may help them respond to labour conflicts, a point often repeated by union leaders (Admasie, 2020).One of the pioneering firms in accepting formal union representation in Bole Lemi indeed suffered no labour con-flict thereafter and other factories in the same park took note (CM44).Similarly, the CEO of one of the largest factories in the sector reported a succession of wildcat strikes and decried the difficulty in finding consistent interlocutors on the side of workers, as each strike came with different representatives (CM27, CM51).This factory eventually called in the sectoral trade union to establish a basic union, but spontaneous mobilisations continued to flare up, especially in connection to local protests.In short, despite initial animosity, some factory managers were able to shift practices and view unions as part of the solution to labour grievances.Their changes in attitude were not directly linked to being part of a particular GPN, but rather to the micro-level dynamics of conflict management and the variation in the incidence of such conflicts.

Local political conflicts and their impact on labour relations
It is tempting to read all instances of labour conflict as evidence of class conflict between capital and labour.However, the antagonistic capital-labour relation does not necessarily generate conflictual behaviour (Kelly 1998).In Ethiopia, the interpretation of drivers of labour conflict is complicated by the fact that the recent acceleration of strikes coincided with a tumultuous period of political tensions and protests that began in 2014 and escalated further in 2016 (see Fisher & Gebrewahd, 2019 for an overview). 21We must stress that our own research, along with other studies, finds a host of genuine labour grievances and, in many instances, these are clear drivers of conflict.However, it is difficult to disentangle the role played by opposition politics in some parts of the Oromia and SNNP regions of Ethiopia, where three of the sampled industrial parks are located.These areas saw widespread protests against the federal government, with protesters demanding greater political rights and the administrative recognition of regional and subregional identities.
Two clear examples of this articulation between labour grievances and local political conflicts are the wildcat strikes that occurred in Bole Lemi Industrial Park in 2018 and in the Hawassa Industrial Park in 2019.Workers in Bole Lemi went on strike for five days in May 2018.While these strikes were nominally about wages levels and other labour-related issues they occurred in the context of sustained uprisings against the federal government driven by grievances about the political representation of the Oromo ethnic group.The unrest provided additional motivation to workers and made striking easier, as widespread demonstrations had weakened the ability of state security forces to intervene (Admasie, 2018).At least some of the workers saw the strike as directly political.One worker reported that factory gates were blocked by workers from Oromia who proclaimed that ''we cannot continue working in the situation that our brothers and sisters are being killed by armed forces" (WI03).In 2019, workers in the Hawassa Industrial park went on strike twice, in March and again in July.As in Bole Lemi these strikes were fuelled by both labour grievances and wider political struggles, in this case around the issue of statehood for the Sidama area which surrounds the city of Hawassa.Beginning in the 1970s, residents of Sidama have led a long campaign seeking to have their area recognised as its own region within the Ethiopian federal state.Their campaign was frequently met with violent suppression by the central state which helps explain the ferocity of the more recent clashes.The conflict led to violent demonstrations in 2018 and again in July 2019 after activists threatened to unilaterally declare an independent regional state (Tronvoll et al., 2020).While the strike in March 2019 was triggered by working conditions (IndustryALL, 2018), the strike in 20 These are also well-known labour control mechanisms as suggested by the industrial relations literature (Collinson & Ackroyd, 2006). 21There are historical precedents of links between industrial contention and political strife such as the period 1974-76 (Admasie, 2019, 96).
C. Oya and F. Schaefer World Development 146 (2021) 105564 July was preceded by large number of workers in the Hawassa Park wearing t-shirts displaying a symbol of Sidama statehood (Astatike, 2019).Large foreign flagship firms, which have a high public profile, were particularly targeted by strikes. 22Wages and working conditions were not generally worse in these firms.For instance, a particular factory that suffered a number of strikes in recent years offered the highest wages in its industrial park (CM27 and own survey results).Two factors are behind this targeting of high-visibility enterprises.On the one hand, workers in such firms have a greater incentive to strike, as they can expect that such companies will be keen to avoid reputational damage.On the other hand, this targeting reflects the interests of local political actors.Some company managers in parks in the affected areas reported that their informants linked several wildcat strikes to local opposition groups pushing workers to strike as part of the political mobilisation against the federal government (CM51) and suggested that high-profile factories were seen as strategic targets to boost the visibility of protests.The claim was that while grievances had previously been resolved on an individual basis or with small groups of workers, once political protests erupted factories witnessed an escalation of strikes (CM50).Some of the strikes in 2018-19 spared medium-sized factories and instead focused on the largest factories inside the parks, especially in the Eastern Industrial Park (CM51).The presence of flagship firms and their importance to the Ethiopian government, which is eager to use them as catalysts of more FDI, are a contingent opportunity factor in Kelly's (1998) framework and, given the potential vulnerability of these firms, enhance workers' structural power.
Many companies failed to anticipate these political developments.Ethiopia had been perceived, perhaps naively, as a haven of comparative political stability.Suddenly, foreign capitalists found themselves in a political climate at odds with their image of a strong state friendly to overseas investors.Over time, a stark contrast emerged between the views of some company managers, who claimed that state institutions, especially at local level, usually took the side of workers, and the perceptions of trade unionists, who reported facing an alliance between the central state and foreign investors, especially those situated in industrial parks.Indeed, during some strikes and local protests spilling into industrial parks, armed federal police and even military units were deployed to protect factories (CM25, CM50).These divergent perceptions reflect the complexity of centre-periphery relations within the Ethiopian state (Clapham, 2018;Markakis, 2013), where local-level actions may not be aligned with federal institutions.For example, according to several foreign company managers (CM53), the regional Oromia BOLSA became much less investor-friendly.This change in attitude was reflected in greater support for workers' grievances, and fewer (or slower) interventions in strikes and protests affecting factories.Many factory managers also complained about the 'bias' in local courts, which apparently routinely ruled in favour of workers. 23This dynamic reflected the perception that industrial parks were 'federal government projects' that had been imposed on regional and local governments without sufficient consultation (GO30).
Several foreign investors seemed confused as to whether the state was 'on their side' or not.An important implication is that employers, especially newly arrived foreign investors, are not generally sufficiently tuned into the complexities of national and local politics.Understanding and anticipating these developments and the possible effects on their operations might have helped reduce labour conflict.The troubles facing many of the most affected foreign firms reflect missing organizational capabilities, in the sense of contextual knowledge and the relationships needed to organise collective activities in unfamiliar social and political contexts (Khan, 2019).
The complex interaction of labour grievances, low union density, national-and local-level politics, and fragmented statecapital and state-labour relations has contributed to uneven manifestations of open labour conflict, despite shared grievances.This evidence underscores the need to account for multiple dimensions of the politics of production, and the variety of factors that underpin labour mobilisation, with the opportunity factor of localised political conflict being a key contributor in this recent wave of wildcat strikes.

A plentiful pool of disciplined labour? Struggles over effort and mobility
The history of industrial capitalism is marked by repeated struggles over the content of the labour process, work effort, and the extent of managerial control over workers.The imposition of strict factory discipline has frequently generated conflicts between workers and managers, and the regulation of work effort has been a central concern of labour movements since their inception (Kocka, 2014).In industrialised countries despotic labour regimes slowly gave way to manufacturing consent (Burawoy, 1985).However, struggles over work intensity and work time remain unresolved in most high-income economies (McGovern, 2020), and in contemporary low-income countries effective labour standards face additional structural barriers, including the unwillingness of many lead firms to pay for improved working conditions (Frenkel, 2018).
Ethiopia is no exception to these struggles over work effort.Our survey data shows that the labour process in all factories was controlled using daily production targets.About 75% of workers were set individual targets.Production targets were demanding and around 30% of workers report regularly failing to meet targets.Work effort is ensured through close supervision by line managers.Abusive behaviour by managers and line supervisors was a common complaint among workers in foreign firms.According to our survey data 53% of workers in foreign firms have witnessed superiors verbally abusing workers, while 38% report having been subjected to such abuse themselves.A frequently expressed sentiment in qualitative interviews was that workers did not feel treated with dignity.
While the labour process literature has rightly devoted much attention to formal conflict and bargaining, forms of nonorganised resistance by individual workers can have at least as great an impact on labour relations as strikes.Such resistance can take a wide range of forms (Collinson & Ackroyd, 2006).Sabotage, theft, work limitation such as refusing to accept overtime, absenteeism, time-wasting and discursive opposition to 'management speak' are all everyday forms of resistance at the workplace (Smith, 2006).Although the incidence of strikes in Ethiopia is significant, 'exit' is a far more widespread form of resistance.In our sample of workers, the three most common forms of resistance were turnover (i.e.workers leaving the factory), absenteeism, and withholding of work effort, including reluctance to do overtime.Overtime is common in the garment industries of Asia (Alamgir & Banerjee, 2019;Anner, 2020;Oka, 2016), but in our sample 28% of workers report never having done overtime work.
A common complaint among foreign managers was that these forms of resistance were far more disruptive (and therefore effective from the workers' point of view) than strikes.These forms of resistance all affect firms' competitiveness and even engender the risk of being cut out of global production networks if orders are consistently missed.High turnover affected both the domestic and the foreign firms in our sample but was a much more serious 22 These included high-profile companies located outside industrial parks which were severely hit by waves of strikes first and then local riots linked to political mobilisations. 23They, however, also complained about the rigidity of the labour law, implicitly acknowledging that courts were simply applying the law.
C. Oya and F. Schaefer World Development 146 (2021) 105564 concern in foreign firms, as evidenced by interviews with managers. 24Some factors driving turnover, such as low wages and unmet expectations on the part of workers, are shared by foreign and domestic firms, but the latter have had more time to develop methods for stabilising the labour force and have greater freedom to do so as they face less stringent pressures from their buyers.According to union officials and workers, high rates of absenteeism and turnover are direct responses to low wages.Despite a shared view among managers and government officials that turnover rates represented a significant obstacle to industrial development in Ethiopia, data are contradictory and it is hard to obtain consistent estimates of monthly turnover rates.One company in the Hawassa park reported a 6% monthly turnover rate compared to only 1% annual turnover in their Chinese factories.Monthly turnover rates in 2018 could be as high as 15% in many factories in other parks (CCOO Industria, 2018).We conducted a longitudinal analysis of employment trajectories for a sample of 120 apparel workers with an interval between surveys of 12 to 16 months.We found that 41% of workers were no longer in their original jobs. 25 While all industrial parks experience similar levels of turnover, there is substantial variation at the level of the firm, suggesting that turnover is related to the issues in each particular factory, and not simply linked to a particular GPN.Whether turnover persists, according to several managers, is mostly linked to the company's failure to examine the causes and implement corrective measures.Our longitudinal evidence suggests that the primary cause of workers changing jobs was the search for higher wages. 26Another common source of frustration, according to interviews with workers and union representatives, was the lack of understanding of wage setting systems on the part of workers, which led to perceptions of unfairness.This was especially the case in newly established industrial parks like Hawassa and for workers in their first factory job.The complexity of bonus systems, deductions, and other variations in take-home pay left many workers disgruntled because their expectation of a stable wage was not met.Wages were not just low and variable, but also insufficient to fund the lifestyles workers expected.A salary high enough to save money and support their families seemed distant for many of the new factory workers (Schaefer & Oya, 2019).In the words of a senior manager in the Hawassa park 'workers' expectations were unrealistic partly because of a lack of previous experience, and partly because of misleading word of mouth suggesting salaries would be higher' (CM34).Moreover, turnover and absenteeism were compounded by difficulties in sourcing enough workers, according to a report on the Hawassa park.The gap between factory requests for workers and candidates channelled through the park's labour sourcing system widened between 2017 and 2018, leading to unexpected labour shortages for some factories (Butler, 2018).Unmet expectations thus contributed to the 'exit' (or no entry) option for poorly organised and inexperienced employees, who did not regard the exercise of 'voice' as a realistic option.
However, high turnover rates were not uncommon in early stages of similar investments in other countries (Oya, 2019). 27Some managers reported having faced comparable rates in their operations in Southeast Asia, illustrating the frictions that accompany processes of transition to industrial work (CM41).The most experienced suppliers in GPNs seemed the least concerned by the evidence of high turnover (CM41, CM34).Their expectation was that, after one or two years, worker turnover would stabilise and eventually decline as the firm was able to retain the best workers. 28To a certain extent, as suggested by the literature on low-wage classic labour regimes (Smith, 2006), high turnover may be tolerable from a managerial perspective if it leads to the most productive workers remaining in the job.For some firms, high turnover in occupations requiring basic skills may even be a strategy to reduce workers' potential for collective action.This may explain why many globally integrated firms in Ethiopia confronted this problem without making significant concessions to workers (Hardy & Hauge, 2019).
This does not mean that there were no gains for workers.As reported above, changes in attitudes towards unions in some companies led to increased union representation.A number of Chinese firms also provided better meals and improved hygiene standards in response to strikes (CM27, CM60, CM63).Yet, there is only patchy evidence on gains in relation to wages.In some flagship companies, periodic strikes eventually led to wage gains, but this was not a widespread phenomenon, hence the variation in wage increases across parks and even among firms within the same park (Table 2, CM25, CM52).In 2019, the Ethiopian government issued a new labour proclamation which included provisions for a commission to decide upon minimum wages, but so far no minimum wage has been set (ILO, 2019). 29 There seems to be more consensus on the need to tackle the challenges of housing and commuting costs.A number of firms in our sample attempted to reduce living costs through subsidising transport to the factories and were considering options for providing housing.Bole Lemi and especially the Hawassa Park were also the object of feasibility studies and government-business discussions on housing solutions (CM52, CM35, CM34, GO29).At the time of research, in 2018, there was growing, but not uniform, acceptance of the idea of investing in workers' accommodation to reduce commuting and subsidise housing costs for workers (CM52, CM35, CM57). 30Since then, according to high-level officials, land adjacent to industrial parks has been made available for firms to build workers' accommodation. 31In addition, the government and state financial institutions promoted housing development by property owners, to offer workers affordable rental accommodation (GO29, GO30). 32Thus, the approach is to induce private-led solutions rather than to directly intervene in housing provision, primarily due to the government's limited financial capacity to undertake such endeavours. 33 The effectiveness of these measures is yet to be determined as there has been very limited construction until now (GO30).However, the fact that a significant debate on labour grievances has ensued and 24 While high turnover has been identified as concern in domestic firms before (Hardy & Hauge, 2019), our survey data show that workers in domestic firms on average have much longer tenure and receive higher wages than in foreign firms (Schaefer & Oya, 2019). 25This includes workers who moved to other factories in the same sector and those who left the sector altogether. 26As noted before, in areas surrounding the parks, male workers can find better paid jobs in construction and transport services and some women may find better paid jobs in services.However, for young women a key reason is that factory jobs were a steppingstone towards far more lucrative jobs as domestic workers in the Middle East (Schewel & Fransen 2018). 27See remarkably similar stories and challenges in the cases of Cambodia (Derks, 2008), Bangladesh (Kibria, 1998) and Malaysia (Ong, 1987). 28Anecdotal evidence from recent conversations with factory managers in Hawassa suggest that by 2020 the turnover problem was significantly less acute than in the early stages of operation. 29The new labour legislation has also introduced some gains for employers, notably more flexibility in overtime use, extended probation periods and simplification of rules for redundancy. 30Few companies were reluctant to associate their brand to workers' accommodation given potential reputational damage (CM35).Thus, they had a strong preference for the local or federal government to own these initiatives. 31Projects are overseen by the Industrial Park Development Corporation to ensure they confirm to ILO standards and are affordable for workers.Several government officials are adamant to avoid 'factory dormitories' of the kind found in parts of China to house migrant workers.See Pun and Smith (2007) on the dormitory labour regime in China. 32Some of these initiatives in Hawassa have been further developed through the Hawassa Sustainable Urban Development Partners' Platform (ILO, 2020). 33An EIC document (2020) recognises that the incorporation of workers' housing was a planned part of the industrial park programme but limited finances left this aspect out of the final development of the parks.
C. Oya and F. Schaefer World Development 146 (2021) 105564 that both companies and the government have articulated some responses suggests that labour mobilization and resistance may eventually pay off even in these early stages of industrial development.

Conclusion
This paper has analysed the dynamics of industrial labour relations during the process of setting up new manufacturing locations serving global production networks in a low-income country, using the example of the Ethiopian apparel sector.The worldwide restructuring of production networks by lead firms in the sector and a proactive space-based industrial policy programme in Ethiopia have combined to entice suppliers of leading apparel GPNs to set up factories in a series of new industrial parks.Both investors and the Ethiopian government expected to find a pliant labour force willing to work for low wages.However, they underestimated the extent to which new labour regimes, which are necessary for the profitable insertion into demanding GPNs, are coconstituted and contested by workers.The new sector has already seen a wave of collective and individual resistance from workers, driven in part by grievances over wages and non-wage working conditions.Part of the problem lies in crude characterisations of labour regimes and what they miss: the historical militancy of workers in the sector, the role of opportunity factors, and the resulting variety of drivers of labour conflict (and hence contingency of actions).We have argued that the best way to understand how certain forms of labour agency are produced, and the impacts they have on workers and firms, is via a combination of analytical categories from labour process theory, mobilisation theory and GPN analysis, in order to capture developments at the level of the country, the sector and the firm.National governments frequently play a contradictory role in regulating labour markets and working conditions as they seek to maintain inflows of FDI while not excessively undermining their own political legitimacy (Smith et al., 2018).Firms, in turn, differ in their target markets, business experience in different contexts, and conditions of integration into global production networks.
Our analysis uncovered widespread 'informal' resistance to low wages and strenuous working conditions on the part of workers.This activism comprises a variety of forms, from large-scale collective action, in particular wildcat strikes, to individual struggles over work effort and mobility, resulting in absenteeism and high turnover.Industrial parks emerged as spaces of particularly intense contestation.Many conflicts result from the collision of productivity imperatives on the part of firms tied into demanding, but badly remunerated, segments of global production networks with the expectations of workers with limited prior experience in industry, who assumed international firms would pay better.Conflicts were sharpened by the suppression of trade union activity in many, but not all, industrial parks and firms, which drove workers to organise outside of union structures and hampered industrial dialogue.The prevailing opinion among government officials and foreign managers, that democratically organised unions could be substituted for by less democratic organisational forms with more limited mandates such as workers' councils, is no longer shared by all firms and some have allowed the formation of union branches in their premises.Individual resistance by workers unused to the intense discipline and submission to the demands of the production processes expected of factory workers frequently took the form of exit.Disgruntled workers would leave the sector altogether, while others would seek to move employer in search of better wages and working conditions.The government reacted by putting in place programmes to help recruitment by assisting with the sourcing and screening of workers.However, in some cases even this labour sourcing was insufficient, resulting in challenges for some newly establish factories to meet their increasing labour requirements.The gains resulting from the various manifestations of labour resistance have been so far limited and not systematic, but there is evidence that workers' 'voice' and 'exit' can contribute to improvements in employment outcomes.
Crucially, our approach has highlighted the importance of taking into account local political conflicts when analysing drivers of labour resistance.In Ethiopia, industrial strife resulted from a complex interplay of factory-based grievances and political mobilisation against the national government.Local political actors sought to capitalise on existing grievances of workers by targeting prominent foreign factories; a political strategy made feasible by the close identification of the national government with FDIdriven industrialisation and the construction of the new industrial parks.The result are fleeting, unstable, and highly localised politics of production that are structured, but not determined, by the dynamics of global production networks and the strategic aims of national political actors.

Table 1
Overview of Ethiopian industrial parks, August 2018.
Schaefer and Oya (2019)2019)for more statistical details on labour force characteristics. 8 Te rural population grew by 23 million between 2001 and 2014, of whom 13.1 million are of working age (EEA, 2016).Every year, thousands of new labour market entrants join the growing labour force.

Table 2
Development of estimated mean take-home wages by industrial park, 2017-2018.
Note: 2018 USD wages were converted to purchasing power parity (PPP) using the 2017 World Bank's conversion factor for Ethiopian private consumption.Source: Own surveys, 2017 and 2018.17According to representatives of the main textile and garment trade union in 2017