The Institutional Work of Exploitation: Employers’ Work to Create and Perpetuate Inequality

Social inequality is underpinned by exploitative labour institutions, yet the agency of employers in establishing and maintaining such institutions remains underexplored. We thus adopt the lens of institutional work in analysing South African mining employers&#8217; purposive efforts to ensure reliable access to cheap labour from the 1860s through until the infamous Marikana Massacre in 2012. We find that while labour is scarce, employers engage in forcing: creating exploitative institutional devices through conscripting and controlling. But as labour becomes abundant (and political winds shift), employers engage in freeing: liberalizing institutional controls to give workers &#8216;choice&#8217;, while simultaneously outsourcing responsibilities and costs associated with the unjust employment relationship to others, including workers themselves. We thus explain how employers purposefully create and perpetuate their advantage in interaction with labour market dynamics, contributing to our understanding of inequality and the role of actors&#8217; intentions in impacting social systems.


INTRODUCTION
What is the role of employers in the unsettling and growing gap between the life chances and living conditions of the poor and those of the rich, and its intractability in the face of diverse efforts to address it? Take South Africa, where, for more than a century, colonial and Apartheid policies entrenched racial oppression and inequality through formal institutional means to the benefit of mining employers. With the transition to democracy in 1994, expectations were high that inequality would decline; however, instead, it has continued to rise (Bundy, 2016;Leibbrandt et al., 2012;Piketty, 2014). Indeed, the failure to reverse the trend of inequality became painfully evident in the infamous Marikana Massacre in 2012, when police killed 34 platinum mine workers who were striking for higher wages and better living conditions .
Building on Marx and Weber, sociologists point to the appropriation of labour effort as underlying persistent inequality, whereby employers rely on 'active institutional devices' (Wright, 2000(Wright, , p. 1565 and 'effort-inducing mechanisms' in the workplace (Burawoy and Wright, 1990). Yet the agency of employers has remained underexplored in the processes of establishing and maintaining such institutional devices, especially in the context of a political transition premised on undoing exploitation (Adler and Webster, 1995). Thus, anchored in the lens of institutional work (Lawrence and Suddaby, 2006), we focus on the agency and motivations of employers, asking: How do employers leverage and even create institutional structures that foster inequality, and how do they perpetuate inequality despite the appearance of reform in the context of political change?
We address these questions through an historical analysis of archival and interview data examining the institutional work of South African mining employers during the 150 years from the growth of the diamond and gold mining industries to the Marikana Massacre. We find that employers engage in institutional work to create and maintain access to cheap labour that shifts based on labour availability. When labour is relatively scarce, employers engage in forcing, undertaking conscripting work to secure a supply of low-cost workers and controlling work to keep them in their employ. In turn, when labour is abundant, employers shift to freeing: liberalizing institutional controls to give workers 'choice' and to distance themselves from the most visible features of oppression, while simultaneously outsourcing responsibilities and costs associated with the unjust employment relationship to others, including workers themselves.
We contribute to conversations about how organizations affect inequality (Beal and Astakhova, 2017;Cobb, 2016;Mair et al., 2016) by demonstrating how inequality is created and perpetuated by employers' purposive efforts, and how these efforts interact with social systems (Stern and Barley, 1996), especially labour market dynamics (North and Thomas, 1973). We build on prior studies of exploitative labour relations by challenging the assumption that for employers, 'manipulation is primary and coercion is held in reserve' (Braverman, 1974, p. 150), especially when labour is scarce. We extend prior research on how inequality is maintained across political transitions Robinson, 2008, 2012) by identifying shifts in employers' institutional work that mitigate the most visible expressions of exploitation, but nevertheless perpetuate inequality because of their links to changing labour market dynamics. Reconnecting organizational studies to an earlier interest in how organizations affect social systems (Parsons, 1956;Selznick, 1949;Stern and Barley, 1996;Walsh et al., 2003), we highlight how employers' intentions give rise to these effects through concealment (Mair et al., 2016) and the 'imperious immediacy of interest' (Merton, 1936, p. 901). gullible enough to believe him' (Rousseau, 2010(Rousseau, [1754, p. 172, own translation). Rousseau thus identified private property and legitimacy as key components of a theory of social inequality, heralding a tradition that emphasises social organization in contradistinction to neo-classical economists' emphasis on the market (Berger, 2004;Cobb, 2016;Sørensen, 2000). Marx (1968Marx ( [1867) echoed this focus on ownership in what has become one of the most influential theories of social inequality: that the capitalist class owns the means of production and is thus able to exploit wage-earning workers. However, even most Marxist scholars have dismissed the labour theory of value (Sørensen, 2000;Wright, 2000), with some responding by broadening capitalists' means of exploitation to focus on authority (Dahrendorf, 1959) or 'control of organizational assets' (Wright, 1985, p. 79). Sørensen, 1996Sørensen, , 2000 offers a theory of exploitation that emphasises property ownership but builds on neo-classical economics by focusing on the ownership of rentproducing assets. Wright (1997Wright ( , 2000 responds by returning to a focus on labour: 'inequalities occur, in part at least, through the ways in which exploiters, by virtue of their exclusionary rights and powers over resources, are able to appropriate [the] labour effort of the exploited' (Wright, 2000(Wright, , p. 1563. He illustrates this by means of historical example, comparing interactions between European settlers and Indigenous Peoples in North America and South Africa. In so doing, Wright (1997) sets out three conditions of exploitation. The first is the 'inverse interdependent welfare principle': European settlers' material welfare increased as a consequence of the material disadvantages of Indigenous People. The second is the 'exclusion principle': Europeans' material advantages were based on the exclusion of Indigenous People from productive resources, particularly land. These first two conditions applied equally in North America and South Africa. These regions differed, however, with regard to the third condition, the 'appropriation principle'. In North America, Indigenous People's labour was generally not appropriated, which meant that settlers 'could adopt a strategy of genocide in response to the conflicts generated by the exclusion of [I]ndigenous [P]eople from the land' (Wright, 2000(Wright, , p. 1565. In contrast, in South Africa, 'the settler population appropriated the fruits of labour of the [I]ndigenous population, first as agricultural labour working the land and later as mine workers' (Wright, 2000(Wright, , p. 1564). An implication of this is that the exploiters become dependent on the exploited, and this leaves open the possibility of resistance: The extraction of effort within exploitative relations is thus always to a greater or lesser extent problematic and precarious, requiring active institutional devices for its reproduction. Such devices can become costly to exploiters in the form of the costs of supervision, surveillance, sanctions, and so forth (Wright, 2000(Wright, , p. 1565).
However, missing from Wright's (2000) theory is an explanation of how exploiters go about achieving and maintaining the appropriation of labour effort, or in other words, what the 'active institutional devices' look like. In the absence of feudalism or slavery, the question arises how elites ensure that workers participate in a 'formally voluntary' way, but are actually forced 'to comply with the authority of others in order to obtain any return at all for the utilities they can offer on the market' (Weber, 1978(Weber, [1922, p. 110).
Following Braverman (1974), labour process analysts have addressed this question, focusing on how employers shape the workplace 'to give [them] the most effective control and profitability' (Thompson and McHugh, 2002, p. 367), making them so 'successful in extracting effort from workers' (Burawoy, 1996). Even though Braverman's arguments were revised or problematized by subsequent writers (for overviews, see, e.g., Adler, 2007;Adler et al., 2007;Burawoy, 1996; for an extension in the South African context, see Webster, 1985), one of his key claims remains axiomatic: in designing and controlling the labour process, 'manipulation is primary and coercion is held in reserve. . . The apparent acclimatization of the worker. . . grows out of the destruction of all other ways of living [and] the striking of wage bargains' (such as Ford's '$5 day' introduced to allay high labour turnover in 1914) (Braverman, 1974, pp. 150-1). This manipulation not only responds to workers' resistance but actively avoids it by 'manufacturing consent' (Burawoy, 1979) to create a hegemonic system, 'in which the conflicting interests of workers and capitalists are coordinated in such a way that workers spontaneously consent to their own exploitation' (Burawoy and Wright, 1990, p. 256). Burawoy and Wright (1990, pp. 252-3, emphasis added) thus develop a typology of 'mechanisms for eliciting labour effort', with an emphasis either on 'domination, [when] labour effort is performed because of the continual presence of various kinds of threats by bosses. . . [or] asymmetrical reciprocity, [when] labour effort is based on consent, on the positive agreement by each of the parties concerned over the mutual, if still unequal, benefits of the exertion of such effort'. Workers' responses to these mechanisms are premised on their 'strategic rationality', that is, their calculation of the costs and benefits associated with their actions, such as the costs of being fired (domination) or the career benefits associated with the firms' success (asymmetrical reciprocity). Such rational calculations are supported by workers' norms regarding their own or others' behaviour.
These effort-inducing mechanisms may co-exist in work settings, but especially in more developed capitalist settings, hegemonic mechanisms are more prevalent because of the interdependent and skilful nature of work (which makes surveillance of effort difficult), the existence of welfare systems (which reduce the costs to workers of being fired), and firms' multiple layers of managers, whose career prospects are linked to workers' cooperation. Conversely, conditions conducive to domination mechanisms -low interdependence of workers, low skill levels, absent welfare systems -'were common in certain sectors in the early phases of the industrial revolution, and they remain common in industrial settings in the Third World today' (Burawoy and Wright, 1990, p. 259). For instance, Burawoy (1976Burawoy ( , pp. 1054Burawoy ( , 1061 highlights the prevalence of domination mechanisms in the South African migrant labour system, with the state playing a prominent role to ensure that black workers were 'cajoled into selling their labour power by expropriation of land, imposition of taxation, and similar nonmarket inducements', and then ensuring their submission by treating the migrant 'as an alien without rights of citizenship'. Accordingly, if the organization of the state changes, as in Zambia in the 1950s, the migrant labour system collapses. Yet it remains unclear in these accounts whether or how such a political transition and corresponding shifts in the organisation of work create significant changes in the institutional underpinnings of exploitation and inequality. This concern has been given attention, meanwhile, by institutional economists comparing economic development in different nations or regions over time Robinson, 2008, 2012;North, 1990;North and Thomas, 1973;Olson, 1982). Equitable economic growth is constrained by the degree to which elites can shape 'labourrepressive' or 'extractive' Robinson, 2008, 2012) institutions to further their interests. Significantly, elites may maintain control over economic institutions even in the wake of changed political institutions, because they can offset a decline in their de jure political power with an increase in de facto power. This is illustrated in the case of the US South, where former slaves remained poor and exploited even after the abolition of slavery: 'The persistence of labour repression in the US South is consistent with changes in political institutions because they were offset by the exercise of de facto power; slavery was replaced by monopsonistic arrangements, policies designed to impede labour mobility, political disenfranchisement, intimidation, violence and lynching' (Acemoglu and Robinson, 2008, p. 269). Such persistence occurs also when the elite itself changes due to political transition, because the new elite has incentives to maintain key aspects of the established exploitative institutions (Acemoglu and Robinson, 2012). Other analyses combine these possibilities, in that the old elite establishes an 'elite pact' with the new elite, and this has been applied also in the South African case (Marais, 2001).
Sociologists and economists thus provide important arguments for why and how employers create and perpetuate exploitative (Burawoy, 1976;Wright, 2000) or 'extractive'' (Acemoglu and Robinson, 2012) institutions to appropriate workers' labour effort, and how they may maintain such exploitation even in the wake of political transition that ostensibly enhances the prospects of the exploited. Yet many of these analyses focus on the macro-level, leaving open the question of how employers go about enrolling workers into an exploitative system, how they ensure their ongoing compliance, and how they react when political circumstances shift. Labour process analysts provide more fine-grained studies focused on the workplace, but despite considering some broader aspects of the 'reproduction' of labour (Burawoy, 1976) and describing 'effort-inducing mechanisms' (Burawoy and Wright, 1990), there is a need to better understand employers' agency in influencing change over time, especially across political transitions, going beyond a focus on 'immanent' dynamics or 'external forces' (Burawoy, 1976(Burawoy, , p. 1078. More specifically, while labour process analysts have explained how employers manufacture consent in the workplace (Burawoy, 1979), we know less about why and how employers may support or even initiate institutional changes that elicit support from workers, but nevertheless entrench exploitative relations. To address these concerns, we turn to the micro-level lens of institutional work for assistance.

The Agency of Exploitation -Applying the Lens of Institutional Work
To better understand the creation and maintenance of the 'institutional devices' of exploitation we must better understand the role of employers in appropriating labour. We propose that the lens of institutional work, which draws attention to the 'purposive action of individuals and organizations aimed at creating, maintaining and disrupting institutions' (Lawrence and Suddaby, 2006, p. 215), may provide fruitful insights into these processes of social construction. To date, though, studies of institutional work (with notable exceptions) have often remained silent on issues of power, domination, oppression and resistance (Lawrence et al., 2011), ignoring the outcomes of these efforts, 'however unjust or dysfunctional' (Dover and Lawrence, 2010, p. 313). Perhaps, as Willmott suggests, it is because the focus for institutional theorists has been 'on processes of institutionalization per se, and not on these processes as a medium and outcome of historically specific forms of domination, oppression, and resistance' (Willmott, 2015, p. 109).
Nevertheless, prior theorizing does offer some indications of the work that may be undertaken by employers as they attempt to establish and maintain the institutional conditions that facilitate their appropriation of labour effort. For instance, Lawrence and Suddaby single out defining 'the construction of rule systems that confer status or identity, define boundaries of membership or create status hierarchies ' (2006, p. 222) as having great potential for the creation of new institutions. Furthermore, from Mart ı's and Fern andez's (2013) harrowing account of the institutional work of oppression in the Holocaust, we learn that categorising work -what we describe as the purposeful creation of devices to identify and signal a sub-class of people (for instance on the grounds of race, age, sex, education or qualifications) -may be an important precursor to exploitative arrangements by creating a set of potential workers 'deserving' of unjust or prejudicial treatment. Categorising often includes the repeated discursive delineation of a group by giving it a specific (often derogatory) name. Whether it be in colonial or post-colonial settings (Nkomo, 2011;Prasad, 2003;Said, 1978;Young, 1990), antebellum slavery (Cooke, 2003), modern slavery (Crane, 2013), or the Holocaust (Mart ı and Fern andez, 2013), categorisation work was and is necessary to create 'the other', often by defining rules and practices that 'create status hierarchies' (Lawrence and Suddaby, 2006, p. 222). This process is well reflected in Mart ı's and Fern andez's (2013) study where they describe how an elaborate system of categorisation was created to define the Jewish race as a class of sub-humans and then demonstrate how this social distancing was further reinforced through authorization and routinization, complemented by the physical segregation of those excluded from the rosters of 'ordinary' Germans (segregation that was ultimately taken to the extreme in the form of extermination). This provides indications that the work of categorisation and segregation may also be important elements of the institutional work of establishing exploitative labour arrangements.
Prior analyses also provide indications about why inequality might persist and the work that employers might undertake to maintain it. For instance, incumbent elites are likely to engage in deterring and policing to ensure compliance with rules (Lawrence and Suddaby, 2006). We learn that ruling elites in Mugabe's regime in Zimbabwe engaged in both deterrence and policing by 'using the state's monopoly on violence to counter efforts within and outside the state to challenge elite power' (Goodstein and Velamuri, 2009, p. 492).
Yet, inequality may persist not only because elites maintain it, but also because efforts to dismantle the underlying institutions are too superficial. Dover and Lawrence (2010, p. 311) urge us to 'account not only for the institutions on which the work focuses but also the system of institutions within which that work occurs'. Hence, '[a]ctors may overthrow institutional structures (such as organizational forms), rejecting the status quo of how to do things, but. . . underlying patterns of privilege may remain untouched, or even be strengthened -reinforcing the status quo of who benefits' (Greenwood and Suddaby, 2006, p. 43; see also Seo and Creed, 2002, p. 239). Some efforts to address inequality may also result in unintended consequences that 'stem from the often indirect nature of institutional agency' (Lawrence, 2008, p. 185) or arise from the complexity of organisational and societal change driven by an 'imperious immediacy of interest' (Merton, 1936, p. 901) that fails to take into account the variegated and complex effects of an intervention (MacKay and Chia, 2013).
Thus, we are left with a picture of actors (perhaps even employers) attempting to construct inequality to serve their own interests by defining and reinforcing the notion of 'others' through categorisation and segregation. We learn that while workers in a state of disadvantage may work to resist their placement in this hierarchy, their agency is embedded in a system of rules and sanctions that is likely to constrain them, especially when the ruling elite continue to engage in work to maintain their own advantage. We are also left with the sense that much still remains to be learned about the institutional work of employers to create and perpetuate exploitation and the impact of that work on social inequality.

Research Approach
We respond to the call for 'historical case studies' of organizations as 'the primary site of the production and allocation of inequality in modern societies' (Stainback et al., 2010, pp. 226, 241). Our research setting is the South African mining industry from the late 1860s until the Marikana Massacre in 2012. Our first author has been collecting data on the South African mining industry since 2001, with a particular focus on platinum mining companies in the area around the town of Rustenburg and their relationships with workers, local communities, and the state. When, in 2012, a wildcat strike at the Lonmin mine in the Marikana area, close to Rustenburg, turned into a violent confrontation in which 34 mineworkers were shot dead by the police, it launched a renewed national discussion about inequality in the sector (Alexander, 2013;Chinguno, 2013). The Marikana Massacre, as it came to be known, galvanised us to re-interpret our existing data, and to collect new data, with a focus on inequality and employment relations.
These analyses and interviews led us to focus on the issue of mineworkers' housing and specifically the decision by mines in the 1990s to offer a 'living out allowance' (LOA) so that workers could leave the massive housing compounds and live where they wanted, reducing their -and the companies' -reliance on what was a crucial and much-hated aspect of the migrant labour system. This was a change demanded by workers, but it had profoundly negative, unintended consequences, as workers swelled growing squatter camps around the mines. Deleterious living conditions in these settlements have become a source of suffering, resentment and frustration among workers and other community members. We thus saw this as a profound example of a situation in which efforts ostensibly intended to reduce inequality only seem to exacerbate it, and we wondered what role employers played in identifying and implementing this shift.
We also became increasingly attentive to the institutional work being undertaken by employers in the setting, and this led to another set of more focused interviews. However, we soon realised that to understand the motives and effects of this decision to institutionalise the LOA, and to develop a broader, more generalizable theory, we also needed to investigate how the institutional structures of the migrant labour system had been established and maintained during the previous 120 years. Echoing the approach taken by Mart ı and Fern andez (2013), we undertook an historical analysis (Maclean et al., 2016) to examine the institutional work of mining company employers between the discovery of diamonds (in 1867) and the Marikana Massacre (2012).

Data Collection
Our data consisted of diverse kinds of documents and interviews as outlined in Table I. Our analysis was grounded in archival material on the topic of inequality in South Africa and, more particularly, labour relations in the mining industry. These data include historical accounts, news articles, extensive transcripts of the Marikana Commission of Inquiry (set up to investigate the Marikana Massacre), and reports from companies, government agencies and non-governmental organizations, including archival documents from the library of the National Union of Mineworkers (NUM). Attempts were also made to obtain archival records from mining companies and the Chamber of Mines, but these were unsuccessful. Our analysis of the institutional work undertaken to establish the migrant labour system relied on these secondary data.
For our focus on the 1990s, we supplemented these secondary sources with data from three waves of interviews conducted by the first author between 2001 and 2015. The first set consisted of over 70 interviews, together with observations from repeated visits to the study area, conducted between 2001 and 2004 with platinum mining companies (including the three largest and most dominant companies in the industry: Anglo Platinum, Impala, and Lonmin), local governments, traditional authorities, labour unions, and non-governmental organizations. A recurring topic in many of these interviews was mining companies' historical reliance on the migrant labour system and single-sex compounds, and the relatively recent changes to companies' approaches to labour procurement and housing. At that point in time, the LOA had been implemented only a few years previously, and its negative effects on local settlement dynamics started to become apparent to interviewees. These interviews thus allowed us 'to obtain both retrospective and real-time accounts by those people experiencing the phenomenon of theoretical interest' (Gioia et al., 2013, p. 19).
The second set of interviews was conducted in 2014 in the wake of the Marikana Massacre and consisted of a dozen interviews with platinum companies' managers and representatives of labour unions, NGOs, and the government (including some who had also been interviewed in the previous round, as indicated in Table I). Again, mineworkers' housing and the LOA were prominent themes in these interviews. Discussing the emerging findings, we decided to focus on employers' institutional work in this setting. To collect more targeted data on this new focus, a third set of interviews was conducted in 2015. These interviews concentrated more specifically on the rationale for, the implementation of, and the outcomes of the LOA. The 14 interviewees in this set Phase 3 (2015) Mining companies (M) 3 (1) Semi-structured interviews, lasting between 45 and 60 minutes (approx. 300 pages of text -verbatim transcriptions from audiotape). Focus was on the rationale for, the implementation of, and the outcomes of the LOA.
Business associations (A) 2 Consultants/academics (C) 5 (1) These visits involved face-to-face interviews, informal discussions; observation in some meetings; site visits to mines and surrounding settlements. They advanced insights into employer-worker interactions and workers' living conditions.

Source Description Use in analysis
Historical accounts of the mining industry 9 books and 11 articles, collectively spanning the entire period under analysis Insights into the establishment and maintenance of the migrant labour system Archival documents from the library of the Approx. 500 pages of text, focusing mostly on the period 1987 to 1997 Insights into labour relations in the mining industry and worker housing, especially the LOA were identified with special regard for their experience and insight into the circumstances and discussions giving rise to the LOA, and thus their roles or positions in the period 1994 to 2000 were primary motivations for their inclusion. They included officials in the dominant labour union at the time (the National Union of Mineworkers, or NUM); managers in mining companies in the area, including one who was an executive in the period of interest; consultants or analysts, two of whom were working in labour relations in the study area at the time; academic researchers; and a representative of the Chamber of Mines, the dominant employers' association. Questions focused on respondents' views of the economic, political, and social context and how this motivated the adoption of the LOA; how the LOA came to be identified as an option; what arguments were made in support or opposition to the LOA (if any); and whether the consequences of this policy shift were considered or anticipated. In total, we analysed data from 121 interviews, as outlined in Table I. All interviews were semi-structured and recorded and transcribed with the permission of the participants. When referring to interviews in the text, we use three-digit labels consisting of a number for the interview phase, a letter representing the sector (as indicated in Table I), and a final unique identifying number. In the text, interviews are labelled by the interview phase, the interviewees' background, and a unique identifying number within this set. E.g., 3M1 represents the first mining company representative interviewed in phase 3. b Number in brackets indicates interviewees who were interviewed also in previous phases.
This combination of sources situated in a named study area responds to organisation scholars' expectations of replicability as well as historiographers' expectation of verifiability (Rowlinson et al., 2014, p. 258). It also goes some way to address the risks of teleological bias in retrospective case histories (Rowlinson et al., 2014;Van de Ven, 1992). Moreover, our data are 'longitudinal, rich, and varied', as recommended by Langley et al. (2013) for an analysis of temporal phenomena.

Data Analysis
We adopted an inductive approach to data analysis, iterating between efforts to write (and make sense of) an overall narrative compiled from many stories found in the primary and secondary data, coding work in which we worked from specific instances of the data to identify conceptual groupings, and reviewing prior research for connections to our own developing theory (Glaser, 1998(Glaser, , 2005Locke, 2011). Focusing on the institutional work of mining employers in different time periods, we attended to recommendations for historical analysis in organization studies (Maclean et al., 2016;Rowlinson et al., 2014), as well as guidelines for the theorization of temporal phenomena (Langley, 1999;Langley et al., 2013). The first author's immersion in the study area and the themes occupying the actors was offset constructively by the second author's lack of firsthand knowledge of the study area and the interviewees, 'combining intimacy with local settings and the potential for distancing' (Langley et al., 2013, p. 6).
We constructed an 'analytically structured history' (Rowlinson et al., 2014, p. 263) of the institutional work undertaken to create, maintain, and -to some extent -disrupt the migrant labour system, connecting events, actions, and actors' legitimating accounts to identify the most important forms of institutional work and their outcomes. Based on numerous discussions between the authors and others, as well as anonymous and friendly reviews, we defined the associated forms of work and theorized their relationship to the political and labour market contexts. We bracketed the timeframe under consideration into two key periods (Langley et al., 2013, p. 7), reflecting different political and labour market conditions and yielding different forms of institutional work. Forcing: 1867Forcing: -1987 Our findings are structured around the institutional work of South African mining employers over two key temporal periods. The first commences in 1867 with the discovery of diamonds and the subsequent formation of joint stock mining companies, and it ends in 1987. Soon after the discovery of diamonds came the discovery of large gold deposits in the 1880s. Partly because employers wanted to mine as much of these deposits as possible, including the lower-grade ores, they needed a reliable supply of cheap labour (Van Der Merwe, 1989). Employers achieved this by engaging in two essential forms of work: conscripting and controlling.

FINDINGS
Conscripting entailed identifying a target group and enrolling people from this group in a labour bond. As outlined in our review above, categorisation work helps to delineate a target group suitable for conscription. In our case, similar to other colonial contexts, the South African colonizers drew upon pseudo-scientific and Social Darwinist theories to encourage race-based, hierarchical distinctions between Black Africans and themselves that would allow them to justify their own right to appropriate land, resources, and labour (Marx, 1997;Posel, 2001). In fact, mining employers were not merely bystanders and passive beneficiaries of their government's institutional work of categorising: they played an active and decisive role in it in order to suit their interests, categorising Black Africans as a distinct group of 'Kaffir labour' (a term used at the time to refer to a Black person -now considered an offensive ethnic slur). Once successful, these employers were often referred to as 'Kaffir magnates' (see Jones, 1995, p. 42).
Once the target group was identified, employers needed to enrol Black Africans into a labour relationship without relying on high wages as an inducement. They established an association, the Chamber of Mines, which essentially created a monopsonistic cartel to avoid employers competing among each other for workers, and to ensure workers' wages were kept low. The Chamber also played an active role in lobbying the government forcefully in 1890 for 'an adequate supply of Kaffirs' through 'direct "encouragement" of mine labour by government officials in the rural areas' (quoted in Jeeves, 1975, p. 10). This included payments to chiefs who would use their traditional authority to compel members of their villages to enlist for work in the mines. As insidiously, employers lobbied for the imposition of special taxes on Black Africans. Whereas previously most Black Africans relied on a pastoral economy, the imposition of such taxes forced many of them into wage labour, which effectively meant that many ablebodied males were compelled to work on the mines -the main source for such wages. To be sure, such taxes were widespread in many colonial contexts, partly to raise revenue for colonial authorities. But mining employers played an active role in making this tax a more prominent imposition. This is strikingly expressed in statements by an employers' representative to a government commission in 1897 where the Commissioner asked, 'Suppose the kaffirs retire back to their kraals [traditional rural villages]. . . Would you be in favour of asking the government to enforce the labour?' The employers' representative responded: 'Yes, I would make it compulsory and without using force. A tax could be levied' (quoted in Jones, 1995, pp. 16, 17).
The mining employers were able to significantly influence the government not only by appealing to their shared interests in a growing mining industry, but also through their own involvement in government. The most prominent example is Cecil John Rhodes, whose company, De Beers, monopolised diamond mining from the mid-1880s and who was also Prime Minister of the Cape Colony from 1890. One of his key legislative accomplishments was to introduce, in 1895, a labour tax to force Africans into cheap wage labour. In answer to liberal criticism of these measures, Rhodes declared facetiously that the tax was 'not slavery but a gentle stimulus' (Van der Merwe, 1989, p. 353). Responding to advocacy by the Chamber of Mines, a similar tax law was enacted by the Transvaal parliament, also in 1895. [1] The combined effect of these rules would be to compel Africans to work as cheap labour for the mines (Jeeves, 1975;Yudelman, 1984). This conscripting was necessary as 'an active institutional device' (Wright, 2000(Wright, , p. 1565) until the 1913 Natives Land Act, which confined Black people to a small proportion of the arable land and helped to create a steadier stream of Africans entering companies' recruitment centres in rural areas. [2] Once they had been conscripted, black workers were channelled into an elaborate and highly institutionalised system of control that determined their movements and actions to a remarkable extent. As early as 1872, diamond mines imposed 'pass laws' constricting Black workers' movements, rationalised in part as a measure to reduce diamond theft. Later on, the Chamber of Mines provided drafts of such pass laws that were then enacted by the Transvaal government. They required Black Africans to carry a numbered, metal arm badge, which showed that they had a labour contract with a 'master', and they punished desertion. Such rules and practices restricting the movement of Indigenous populations were prevalent in many colonies, including most parts of the British Empire, but they were especially prominent and coercive in South African mining areas.
In 1902, the Chamber of Mines set up a dedicated migrant labour organization with local recruitment offices that contracted Black workers from far-flung rural areas around southern Africa, for periods ranging from six to eighteen months, and transported them to the mines and back. Once contracted, a worker's movements were constricted to his work in the mine during shifts and to a single-sex housing compound (also referred to as hostel) outside shifts. The compounds were 'closed', ensuring 'the total isolation of African workers' (Turrell, 1984, p. 61). They were generally large buildings housing workers in bunk beds in very crowded conditions -their scale and density kept costs low. In addition, they were designed for surveillance and control based on force, with a prominent role for security forces provided by the company, as well as the government, which was increasingly concerned about 'native' insurrection. On the diamond mines, they were also organised to facilitate highly invasive 'stringent searching against diamond theft' (Marks, 2012, p. 126). Companies abrogated unto themselves powers of coercion and force, giving rise to the 'company-state' (Bezuidenhout and Buhlungu, 2011, p. 241;Burawoy, 1985). Though such coercive control measures became less severe in the 1980s, they were still a defining feature of life in the compounds, as noted by a mineworker and trade unionist whom we interviewed: The use of compounds was a control system. . . The workers would stay there, they would eat there, and so on. There will be a security at the gate. When you go out, when you come in, there was a wristband you had to wear (interview 3U1). Similar appraisals were offered by managers. For instance, an interviewee who was a human resources manager in one of the compounds in the early 1990s noted: Peoples' lives were being controlled in many ways. Their jobs were chosen, and the bus they would get on [in their village], and all of that stuff in the compounds. . . the compounds were cut off from society. . . you have this total institution, like a prison. . . and so you had the company maintaining control (interview 3C1).
Writing in 1991, Crush and James (1991, p. 301) estimated that over 97 per cent of mineworkers 'currently live in these single-sex, regimented barracks. . . the central institution in the migratory labor system'. Over and above facilitating control of workers, these measures 'constituted an attack on the dignity of workers and their households, intervening in the most private spheres of human existence, from cooking food to making love' (Bezuidenhout and Buhlungu, 2011, p. 246).
With the industry's rules and practices established and implemented within the institutions developed and upheld by the colonial and -after 1948 -Apartheid governments, this migrant labour system retained its essential features for much of the 20 th Century (Bezuidenhout and Buhlungu, 2011;Crush and James, 1991;Moodie and Ndatshe, 1994). It provided cheap labour to the mines: no increase in real wages took place between 1897 and 1969 (Bundy, 2016;, and black workers' wages in mining were lower than those in other sectors: 'The mining industry would have to double the wages if they hoped to compete with the manufacturing sector for labour' (Wilson, 1972, p. 153;see Burawoy, 1976).
Yet, from the outset, controlling met with workers' resistance. In the mines and in the compounds, black mineworkers employed a range of informal means to support each other, negotiate improvements with white managers or overseers, and prevent the worst forms of cruel treatment by white 'masters' (Bezuidenhout and Buhlungu, 2011;Moodie and Ndatshe, 1994). Attempts at collective action were brutally opposed by the company-state's security apparatus (Crush, 1989), and the National Union of Mineworkers (NUM) was only legalised in 1982. In the mid-1970s, growing protests, resistance, and violence by workers in the compounds, as well as broader social resistance to Apartheid, including the Soweto uprising in 1976, demonstrated to employers that they could not respond only with repression. Employers thus acceded to some workers' demands and this entailed making small changes that left key features of the migrant labour system intact.
A telling example was the response by Anglo American, by far the largest mining company in the country at the time, to growing labour unrest in the compounds. In 1975, the company commissioned sociologist Dunbar Moodie to probe black miners' views and experiences. The disturbing findings of the Moodie report captured media headlines even beyond South Africa. Its basic message was that living conditions in the compounds were so deleterious that revolt was to be expected. A fundamental revision of the migrant labour system was called for, focusing on the conversion of the compounds to family accommodation. In response, Anglo American calculated the costs of establishing family compounds, but they were too high for significant changes. More minor changes were agreed to, including a decrease in densities and some concessions to workers' self-determination in the compounds. In 1979, Anglo American's CEO stated, 'Migrant labour is here to stay' (quoted in Flynn, 1992, p. 222).

Freeing: 1987 to 2012
Our second period commences in 1987, amidst an unprecedentedly large and longlasting mineworkers' strike and deepening political conflict. From this point onwards, employers entertain -and then, a few years later, implement -significant changes to the migrant labour system, including the payment of a living out allowance (LOA) that allows workers to opt out of staying in the controlled environment of the single-sex compounds. Most analyses interpret these shifts primarily as a response to the growing power and militancy of the NUM, following its establishment in 1982, and the broader political transition that culminated in the first democratic elections in 1994 (Bezuidenhout and Buhlungu, 2011;Crush and James, 1991;. However, we highlight that employers' shift away from the compounds entailed more intrinsic economic and strategic objectives, and these must be seen in the context of a fundamentally changed labour market. The South African labour market, as a whole, experienced a fundamental shift to large-scale unemployment and a relative oversupply of labour in the 1970s and 1980s (Bundy, 2016). Crush (1989, p. 19) explained: In the last ten years [i.e., the 1980s] the mines have moved from a position of perennial labour shortage to one of perpetual oversupply. Recession, retrenchment, and burgeoning unemployment in South and southern Africa since the mid-1970s have produced a highly visible reserve army of labour for the mines. This decline was due to broader labour market shifts, but it was acute in mining due to the declining gold price from the late 1980s and employers' increasing efforts to replace workers with mechanisation, especially in connection with punitive dismissals in the wake of the 1987 strike (Moodie, 2009). Thus, gold mining employment was at an alltime high in 1987 of 480,000 workers, but by 1995 it had reached lows of under 200,000 (Harington et al., 2004). Total mining employment was around 700,000 in 1987 (Harington et al., 2004), but decreased to around 540,000 by 1996 (Simkins, 2012). Unemployed mine workers had little chance to find work in other sectors, so large numbers flocked to the platinum mines, which were ramping up production in the context of an increasing price for platinum group metals.
It is within this broader labour context, coupled with a changing political context in which workers were calling for better housing options, that the institutional work of employers shifted. Employers began to abandon their controlling work, instead engaging in liberalizing -dismantling institutional structures of control and thus offering workers more 'choice'. The NUM saw the compounds as a manifestation and symbol of Apartheid, and so the union mobilized not only for the improvement of conditions in the compounds but also for their eventual abolition. A former mineworker and current trade unionist explained: 'In the compound, you don't enjoy a normal life. People want to stay out of that compound. . . So, in '87, '88, we said, "away with the compound system!"' (interview 3U1; see also Crush, 1989). The LOA emerged as the most prominent mechanism for this, which both workers and employers could agree upon, but for diverse and different reasons.
The LOA was something employers initially offered to white, skilled employees who elected not to live in the segregated and relatively luxurious villages provided for them. In the late 1980s, it was identified by the NUM as a privilege that it wanted to achieve for its members. The LOA was implemented in some gold mines as early as 1990 (National Union of Mineworkers, 1990), but in our study area, it became a standard feature among the platinum companies in 1996 (as explained by interviewees including company managers and union representatives: 3M1, 3M2, 3M3, 3A1, and 3U1). As noted by an interviewee who was an executive at the time, 'all the mining companies. . . took the decision [to adopt the LOA] about the same time' (interview 3M1).
One way to understand employers' decision to implement the LOA for all mineworkers would be to cast it as accession to demands for change by workers. But, in contrast to the changes made to compounds in the 1970s, the LOA was not just a concession that was easy enough to make; it was a much more significant change to the migrant labour system, driven by intrinsic economic and strategic motives. An industrial relations consultant noted, 'the problem that the living out allowance is trying to address was, if you like, the ugly face of the migrant labour system. . . [the] highly stratified, militarist, top-down industry was starting to change' (interview 3C2). Implementing the LOA and allowing workers to leave the hostels was a way for employers to distance themselves from the most tangible elements of the institutional structures of exploitation. Overtly, liberalizing work was meant to 'free' workers from the overbearing control enacted upon them in the migrant labour system. Yet, for employers, the primary incentive was to 'free' themselves -not only of the politically delegitimised control structures, but of the associated costs and responsibilities. This required an additional form of work.
Whereas liberalizing gave the appearance of dismantling the institutional structures of exploitation, a second form of employers' work served to preserve them by passing on the responsibility to maintain exploitative institutional structures to others: outsourcing. While the LOA gave workers more 'choice' in deciding where to live, it also shifted the responsibility for housing workers to workers themselves and others, including the government. The connection between liberalizing and outsourcing is powerfully represented by the following quote from a human resources executive in one of the large platinum companies in the 1990s: We had come from a mindset which was totally paternalistic, driven by a colonial mindset. . . to a mindset which was almost liberating and said, actually, we have been paternalistic; it's not our role, it's the government's role to make sure that people are provided with housing. . . actually, we're doing the right thing by distancing ourselves from what is none of our business. . . So there was a glorious coming together. . . at the time this was an obvious and a business-wise decision to make, and it met all the expectations of the stakeholders (interview 3M1).
Employers thus felt a palpable relief in distancing themselves from having to provide housing in compounds and this must be seen in the context of a broader effort to modernise and focus on 'core competencies' (Segal, 2000). The same executive noted: 'I can remember at one particular executive meeting where our then chief executive said, "We've got to get it into our heads that we're a mining company, not a housing company"' (interview 3M1). This was particularly salient because the compounds -as far as they were still relied upon -needed to be improved in response to the NUM's and the new government's expectations, 'to provide one room per worker, or family accommodation in compounds' (interview 3U3). In this context, the executive argued, 'Do you want to be spending capital on building compounds, or would you rather spend capital on building new mines? Of course, the preference would always be for the latter' (interview 3M1).
These financial incentives were powerful, of course, as long as the size of the LOA could be kept sufficiently low as part of the wage negotiations with unions. A human resources manager in the company that first implemented the LOA explained: 'the living out allowance was pegged at 300 Rand [per month] at the time and 300 Rand was a lot of money [for workers]' (interview 3M3), representing about a quarter of total income. He continued, '[Workers] could use the living out allowance, they could use anything that they could find in order to augment [their income] because workers were earning next to nothing' (interview 3M3). [3] For the employers the LOA was an economically beneficial course of action. A senior executive noted, 'Certainly, the fewer people that we have to provide accommodation for, the less capital intensive was going to be that exercise' (interview 3M1).
Recognising the LOA as a consequence of employers' increasing focus on core competencies places it in conjunction with the simultaneous move by employers to shift from the centralized recruitment and contracting of workers to the use of 'labour brokers' that would supply workers on a short-term basis as needs arose, but who were largely unregulated by the state. The practice of sub-contracting became significantly more prominent in the mid-1990s in the study area, meaning that in addition to housing, companies reduced their responsibilities associated with a contract of employmentincluding minimum pay and benefits (Hamann and Bezuidenhout, 2007). Importantly, these liberalizing and outsourcing efforts further exacerbated the labour supply glut. With workers being recruited and contracted by brokers, employers no longer needed to make a commitment to house them for long periods, and instead could now hire and fire workers in much shorter periods, focused on more specific labour needs in delineated projects. Employers' liberalizing and outsourcing work thus created a more efficient labour market and this further facilitated their appropriation of labour effort.
We found that the business benefits of employers' liberalizing and outsourcing efforts were explicit in the motivations and rationalizations of senior managers among themselves. As noted above, the CEO of one of the companies emphasized among colleagues and investors the need to focus 'on the business of mining' and not housing, and a senior executive emphasized to us that 'this was a business-wise decision to make. . . [and] less capital intensive' (interview 3M1). But in external communications -and also in many interviews -managers prioritized a narrative in which the LOA was implemented because of the union's demands, and because of intrinsic motivations to enhance human rights in the industry and to 'free' workers from the previously controlling and paternalistic approach of mining companies. Some interviewees were explicit about this distinction. For instance, a human resources manager in one of the companies argued: The living out allowance. . . was an opportunity to kill two birds with one stone. The first opportunity is to resolve the problem [of the union's demands. . .] the unions demanded this. . . The second opportunity, which is rather hidden, is to get rid of the compounds (interview 3M3, emphasis added).
As a consequence of liberalizing and outsourcing work, employers benefited from lower costs and responsibilities and increased labour market efficiencies, but mineworkers suffered increasing competition for jobs, greater precariousness in their work, worse employment conditions (including lower wages in many instances), and unexpected negative consequences of the LOA in terms of their living conditions. In the decade following implementation of the LOA, there was a massive growth of slum settlements that was clearly linked temporally and/or causally to the LOA by a range of interviewees (including managers and government and NGO representatives: e.g., interviews 1G1, 3C1, 3M1, 3M3, 3N1) and is also documented in municipal reports and scholarly studies (Bezuidenhout and Buhlungu, 2015). A human resources manager noted, 'When we started issuing. . . the living out allowance we created the squatting camps around the mines' (interview 3M3). Workers did not use the incremental income of the LOA to rent formal accommodation because there was very little affordable rental housing available around the mines, and furthermore staying in a shack allowed them to send more money home to their families in labour sending areas, and/or to establish and maintain a second family in the mining area. Given a lack of state support in terms of settlement planning and provision of infrastructure, workers became dependent on land owners, 'backyard' lessors, informal shack rental markets, 'slum lords', and loan sharks, increasingly living in abhorrent slum settlements around the mines.
In the area around Rustenburg, these settlements grew during the late 1990s by about 25 per cent per year, accommodating 34,000 in the Greater Rustenburg area alone (Plan Associates, 2001, pp. 1, 2). Company internal reports showed that large numbers of mineworkers were living in these slums from the outset (as explained by a company manager in 2001: interview 1M7). Living conditions in these settlements are terrible (Conflict Resolution Consortium, 2001), so much so that they have been credited with contributing to the severe labour strikes in 2012-14 and the Marikana Massacre . One manager exclaimed, 'The conditions that our employees are living in, in the informal settlements -ethically, it isn't right (3M2).
Most interviewees argued that both management and unions were surprised by the extent to which workers chose -or were forced by circumstances -to live in these informal settlements. An executive noted, 'We didn't sufficiently make those connections at the time. . . there was a singular failure on the part of everybody concerned. . . to fully appreciate all of the implications of that decision-making process' (interview 3M1). One union leader is quoted (by managers and unionists: 3M2, 3U2) as saying, 'We couldn't expect people would live like this!' This is despite prescient warnings from within the union as early as 1990, before the LOA was implemented in the study area: The living out allowance is offered by mines to excuse themselves from housing their black employees. . . [They say:] All workers must have 'freedom of choice'. But mine wages are too low to allow workers a real choice -to the mines the most easy solution is to pay a 'live-out allowance'. But accepting the live-out allowance means agreeing that the mines do not have a responsibility to house workers. . . and [to] make it the responsibility of the individual worker. . . Workers will accept the extra money and then find themselves unable to find accommodation, [living. . .] in shacks (National Union of Mineworkers, 1990, pp. 1, 8).
To summarise, employers' liberalizing and outsourcing work contributed to significantly reduced responsibilities and costs for employers associated with the recruitment, contracting, and housing of workers. In a fundamentally changed labour market context, employers no longer needed to recruit and house workers in a rigorously controlled and highly institutionalised way, because the labour oversupply and proliferation of labour brokers ensured that there were always sufficient job-seekers on mines' doorstep. Instead, they could liberalize and undo these previously predominant structures of control, and simultaneously outsource responsibilities for maintaining exploitative labour relations to others, with adverse consequences for workers.

DISCUSSION
We set out to explain employers' role in the creation and perpetuation of inequality. Our analysis yields insights about employers' institutional work of exploitation and how this manifests in interaction with the underlying social system (Stern and Barley, 1996). In so doing, we build on North and Thomas' (1973) explanation of how institutions are established in interaction with prevailing political and market conditions, highlighting how labour market conditions not only motivate elites to establish certain institutions, but also determine the social outcomes of these institutions. We explore these themes in more detail below, focusing first on employers' role in creating and perpetuating inequality, and then on the role of actors' intentionality.

Employers' Institutional Work of Exploitation
Our primary contributions are to our understanding of the creation and perpetuation of inequality and specifically the role of employers (Beal and Astakhova, 2017;Cobb, 2016;Mair et al., 2016;Stainback et al., 2010), by building upon prior work of sociologists and labour process theorists. This prior work has pointed to the need for 'active institutional devices' (Wright, 2000(Wright, , p. 1565 to maintain the exploitative appropriation of labour, and to 'effort-inducing mechanisms' in the workplace involving domination or asymmetric reciprocity (Burawoy and Wright, 1990). Yet the agency of employers has remained underexplored in the processes of establishing and maintaining such institutional devices, especially in the context of a profound political transition focused on undoing exploitation (Adler and Webster, 1995). We argue that employers engage in institutional work to create and maintain access to cheap labour that is highly influenced by labour availability. When labour availability is low, employers engage in forcing to secure a supply of low cost workers. In turn, when labour is abundant, employers shift to freeing, limiting their direct involvement in upholding institutional structures of exploitation and distancing themselves from the most visible features of oppression.
Forcing sees employers take an active role in shaping and enforcing the institutional structures that render exploitation possible. Two forms of institutional work -conscripting and controlling -are crucial to this process. Conscripting entails a proactive effort to identify and integrate targeted workers into exploitative institutional structures, while controlling ensures their on-going compliance. While prior work has signalled that defining (Lawrence and Suddaby, 2006) and categorising (Mart ı and Fern andez, 2013) are likely to be important elements of the institutional work of establishing exploitative labour arrangements, it has not addressed how exploiters go about establishing the appropriation of labour effort. Here, we reveal that in situations of labour scarcity, employers conscript workers through advocating (Lawrence and Suddaby, 2006) for the categorisation of a target group suitable for conscription, but also by advocating for institutional structures of enrolment, such as taxation. Once workers have been conscripted, employers maintain their hold on workers through a bundle of work we label controlling. As suggested by prior theory, they segregate workers (Mart ı and Fern andez, 2013) and coerce compliance through policing and deterring (Lawrence and Suddaby, 2006). The migrant labour system and especially the single-sex compounds, in which South African mining employers housed and controlled workers for most of the 20 th Century, are a striking illustration of this.
These findings develop our understanding of exploitative labour relations by challenging the assumption that for employers, 'manipulation is primary and coercion is held in reserve' (Braverman, 1974, p. 150), especially when suitable labour is scarce. After all, it was Ford's challenge of attracting workers that made him institute the '$5 day'. In our context, labour was in demand, yet employers opted not to pay higher wages but rather established an elaborate system of domination mechanisms. They were able to enrol the state in this process, in combination with prevailing racial ideologies.
We thus build on sociologists' and historians' accounts of the crucial role of the state in establishing exploitative labour relations -especially those premised on domination mechanisms (Burawoy and Wright, 1990). Yet we point to a more proactive and direct role for employers in influencing the content and process of the state's institutional work than is commonly assumed. For instance, Burawoy (1976Burawoy ( , p. 1054) notes how black workers in South Africa were 'cajoled into selling their labour power by expropriation of land [and] imposition of taxation' -institutional devices that clearly required 'government intervention' (Burawoy, 1976(Burawoy, , p. 1067, but the prominent role of employers in motivating and guiding such devices has remained underexposed. While the employers in our case relied and built upon racial discrimination and exploitation driven by the 'internal colonial' South African state (Burawoy 1976;Mamdani, 1996), we argue that such forcing work is not confined to this setting. Recent studies on the migrant labour system in China suggest that employers and the state also collaborate to create institutional structures that keep wages artificially low: The 'systemic provision of dormitories for internal migrant labour within or around factories facilitates continuous access to fresh labour reserves from the countryside and depresses wage demands' (Ngai and Smith, 2007, p. 29). We find elements of this work also in Canada where controversy rises over abuse in the Temporary Foreign Worker Program (Auditor General of Canada, 2017). First introduced in 1973 at the urging of employers, the program aimed to fill a gap in the labour market for jobs Canadians could not or would not fill. In 2002, again at the urging of employers, the government added a new category of 'low-skilled workers'. The result has been a rise in employers' reliance on conscripting migrant workers to fill low-wage, undesirable jobs. Criticism has mounted that the system leaves these workers vulnerable to excessive control and exploitation due to their dependence on employers, since workers risk both their livelihoods and their residency in Canada if they complain about working conditions (Farraday, 2016). Concerns have been raised also over other consequences such as general wage suppression (Auditor General of Canada, 2017).
Sociologists emphasising the role of the state in domination mechanisms imply that changes in the organisation of the state would undermine the viability of such mechanisms (Burawoy, 1976). Institutional economists, meanwhile, argue that elites can maintain exploitative relations even in the wake of political change by relying less on de jure influence and more on implicit, de facto means of power (Acemoglu and Robinson, 2008). Our analysis confirms expectations that a profound political transition will involve shifts in employers' institutional work, focused on mitigating the most visible and symbolically charged expressions of exploitation -in our case, doing away with controlling measures and giving workers choice about where to live. But we argue these shifts were only partly a response to changing political winds. Employers' motives were linked to more intrinsic, strategic drivers associated with changing labour market dynamics, with profound implications for the perpetuation of exploitative relations.
When the labour market shifts and labour supply exceeds demand, the character of employers' work may shift fundamentally to focus on freeing. New labour market conditions allow employers to simultaneously liberalize and outsource the key institutional structures of exploitation. Employers engage in liberalizing, 'freeing up' workers by dismantling the most contested institutional features of exploitation (such as, in our setting, single sex compounds), but also 'freeing' employers from the stigma of their association with these key features. Similarly, through outsourcing, employers 'free' themselves from the exclusive responsibility to maintain institutional structures of exploitation by outsourcing these responsibilities to the broader market, the state, and the workers themselves.
We argue that this coupling of liberalizing and outsourcing work raises particular concerns because it conceals the transfer of exploitative structures, making them less likely to be recognised and opposed. And, by imbricating a range of other actors into structuring unequal social systems, it becomes more difficult to make sense of, or even affect, the underlying inequality. For instance, in this setting, we observed that as the institutional structures of exploitation became more broadly entrenched in the marketplace, mining companies were better able to exploit their monopsonistic position in the labour market, much like the white estate owners in the US South after the abolition of slavery (Acemoglu and Robinson, 2008). In the wake of freeing work, structures of inequality disadvantaging workers evolve into an increasingly complex web of actors with their own interests in and contributions to these structures, and this web of actors becomes increasingly difficult to regulate or influence.
Again, we suggest that freeing is not just performed by employers in the South African mining sector. By describing employers' liberalizing and outsourcing work, and their interrelationship, we explicate the agentic dimensions of trends described and discussed by scholars in diverse disciplines and perspectives, and in diverse contexts characterised by labour surplus (e.g., Barker and Christensen, 1998;Barrientos, 2013;Coe and Jordhus-Lier, 2011;Davis-Blake and Broschak, 2009;Kalleberg et al., 2000;Peck and Theodore, 2006;Phillips, 2013;Polivka, 1996;Preibisch, 2010). These trends have also attracted attention by some organisation scholars (Barley and Kunda, 2006;Barley et al., 2017;Cobb, 2016;Davis, 2017;Pfeffer and Baron, 1988). As Barley and Kunda (2006) explain: 'Although firms have long employed temporal workers for seasonal and short-term needs. . . during the late 1980s corporations began to view temporal labour as an extension of the broader strategy of outsourcing'. Such contingent work has been associated with lower wages and fewer employment benefits, and its prominence and its negative impacts have been particularly high among women and minority groups. Cobb (2016, p. 335) argues that 'outsourcing. . . should be associated with greater income inequality at the societal level'.
The 'free agent perspective', on the other hand, recommends that workers 'use the demise of traditional employment as an opportunity to set oneself free. . . and to develop and market their own skills to the highest bidder' (Barley and Kunda, 2006, p. 47). Of course, as Barley and Kunda note, 'low-and high-skilled occupations substantially differ' (Barley and Kunda, 2006, p. 47), and so injunctions 'to set oneself free' are problematic in a context such as ours, where the limited skills that miners might have are difficult to transpose to other industries. To this we add, based on our analysis, that relying on the market to offer 'skills to the highest bidder' is especially problematic in the context of a labour glut and employers' monopsony. Given such obstacles, the fact that Barley and Kunda's (2006) recommendations for technical contractors have some relevance for mineworkers -e.g., 'removing [labour brokers'] legal advantage' (Barley and Kunda, 2006, p. 61) -provides only limited consolation.
We also see broader implications. The shift to contingent work brought about by employers' liberalizing and outsourcing work is particularly prominent in the 'gig economy': 'Like all contingent workers, those in the gig economy participate in spot labour markets except that "gig workers" typically land their jobs through online platforms and may never meet their "employer"' (Barley et al., 2017, p. 111). In this vein, Davis (2017, p. 736) bemoans the process of 'Uberization", by which labour is hired by the task rather than by the job. . . with dire implications for career mobility'. We see this as linked to the institutional work of liberalizing and outsourcing, as was evident in Uber's combative stance in over 70 lawsuits in 2016 alone, most of which surrounded the company's claim that Uber drivers are independent contractors despite it maintaining the power to hire and fire them and change conditions of employment (Lowe, 2017). In South Africa, meanwhile, Uber has attracted desperate job-seekers into its ranks as Uber drivers, though many of them are increasingly concerned about their dependence on a quasi-monopsonistic operator, especially given Uber's unilateral imposition of changes, such as fare reductions (Dreyer et al., 2017).

Surfacing Concealment and the Imperious Immediacy of Interest
By responding to Dover and Lawrence's (2010) advice to account for the system of institutions within which institutional work occurs, we reconnect to earlier interests in how organizations affect social systems (Parsons, 1956;Selznick, 1949;Stern and Barley, 1996;Walsh et al., 2003), and how actors' intentions interact with their social systemic context to give rise to the entrenchment of inequality. Following North and Thomas (1973), we attend particularly to labour market dynamics as both a motivation and consequence of employers' work. Given that intentions play an important role in defining institutional work (Lawrence and Suddaby, 2006), we discuss here employers' intentionality in creating and perpetuating inequality, focusing on two aspects: concealment and the 'imperious immediacy of interest' (Merton, 1936, p. 901).
First, our analysis complements and counterpoises Mair and colleagues' recent (2016) suggestion that organizations can transform local social systems by intentionally concealing the underlying goal of social transformation and instead focusing attention on a less contested objective. According to Selznick (1949, p. 255), 'The very necessity to keep your eye on the ball -which demands the construction of a rational system explicitly relating means and ends -will restrain the actor from taking account of those consequences which indirectly shape the means and ends of policy'. Mair andcolleagues (2016, p. 2038) thus argue that in their case, Gram Vikas was able 'to proactively manage villages to keep their eyes on the water-and-sanitation "ball" [. . .] and to hide the emergence of a new social order that [reduced inequality, but] was neither desired nor anticipated by some social groups'.
In contrast, our case focuses on employers' efforts to enhance and maintain their own advantaged position in an unequal social system. The underlying intentions are thus diametrically opposed to those of Gram Vikas, but the mechanism of concealing is similarly at play in the combination of employers' liberalizing and outsourcing work. Employers framed their efforts as responding to worker demands and increasing workers' 'choices', aligning themselves with a new political dispensation focused on human rights and restoring worker freedom. In so doing, employers concealed the intended outcomes of their outsourcing work -to reduce their own responsibilities and costs, and to maintain ready access to cheap labour in a more efficient labour market. By concealing their underlying interests, employers helped focus attention of all those involved on addressing a shared problem, the single-sex compounds. But this deflected attention from the underlying social system (Stern and Barley, 1996), in which employees were becoming increasingly dependent on diminishing employment opportunities. The result has been perpetuated inequality, with workers becoming immiserated in overcrowded and desperate slum settlements. By focusing attention on a particular problem, such as the single-sex compounds, and by replacing the institutional work of controlling with liberalizing, without taking into account the inter-relationship between this work and labour market dynamics, actors 'overthrow institutional structures. . . rejecting the status quo of how to do things, but. . . reinforcing the status quo of who benefits' (Greenwood and Suddaby, 2006, p. 43, emphasis in original).
Second, while we argue that employers refashioned modes of exploitation by concealing their underlying motivations for liberalizing and outsourcing, we recognize that employers did not intend for workers to experience deteriorating conditions in slums. Their intention was primarily to advance their own interests by reducing their responsibilities and costs, focusing on core competencies, and enhancing the efficiency of the labour market. Managers, whom we interviewed, conceded that with hindsight, it should have been recognised that the changes implemented by companies disadvantaged workers and led to their immiseration in slums, but these consequences were not considered at the time.
Managers were thus blinded by their overbearing focus on their own 'end-in-view'their intrinsic strategic and economic interests associated with liberalizing and outsourcing. They were profoundly struck by one of the factors leading to unanticipated consequences: the 'imperious immediacy of interest', when 'the actor's paramount concern with the foreseen immediate consequences excludes the consideration of further or other consequences of the same act' (Merton, 1936, p. 901). The burgeoning slums around the mines became such a 'further consequence', with particularly egregious implications for workers. But they also became a liability for the employers, as workers' productivity suffered due to crime and disease, as workers' grievances grew, and as the slums became increasingly potent symbols of inequality and the negative social impacts of mining. Managers' imperious immediacy of interest thus becomes analytically and practically important not only for its strategic implications (MacKay and Chia, 2013), but perhaps especially so for organizations' impacts on social systems.
As outlined in our analysis, employers were not the only ones afflicted by this imperious immediacy of interests. The workers and the unions representing them were also preoccupied with their immediate interests associated with doing away with the single-sex compounds and adding an increment to salaries. More specifically, workers focused attention on reacting to the offending work of controlling by replacing this with existing 'solutions' already in play for advantaged groups. They were thus disposed towards a form of mimicry displayed by disadvantaged actors in postcolonial contexts, what Srinivas calls (2013) 'inhabiting the habitus'. Apart from some remarkably prescient, but largely ignored internal critiques warning of unintended consequences, little heed was given by union leaders to the longer-term and cumulative effects of the LOA. Further research should unearth how this imperious immediacy of interest constrains the agency of workers or other disadvantaged groups, and how it might be corralled or mitigated. As far as our analysis is concerned, however, we maintain that employers bore a particular responsibility to withstand the imperious immediacy of their self-interests, given their position of power and advantage in an unjust system of their making.

Limitations
We are conscious that in our quest for parsimony, we glossed over nuances in South Africa's complex history. We did not attend to the role of droughts and plagues in the late 19 th century, or the importation of Chinese labour in the years following the South African War, which impacted labour market dynamics; however, while these impacts were significant at the timescale of some years, we maintain that they do not erode the overall characterisation of relative labour scarcity prior to the 1980s and abundance thereafter. Similarly, we gave no explicit attention to the vital role played by organised White mine workers in putting pressure on employers and the state to avoid Black workers competing for better-paid positions (see, e.g., Burawoy, 1976). This further increased pressure on employers to keep Black workers' wages low, so it did not materially change their institutional work to achieve this. By focusing on employers in the platinum subsector in the second period of our analysis, we did not attend to variations across subsectors (see, e.g., Crush and James, 1991), but these do not contradict our arguments. Also, while the apartheid state's decision to ease restrictions on Blacks' movements in the 1980s (Crush and James, 1991) contributed to the 'glorious coming together' (interview 3M1) of circumstances and motivations, this does not undermine our emphasis on the role of labour market dynamics.
It may be argued that liberalizing and outsourcing were dominant forms of institutional work in the 1990s, but that employers began internalising some aspects of the employment relationship again more recently, in response to government demands as well as their own concerns about workers' health, safety, and productivity due to the deleterious living conditions in the slums. One of our company interviewees argued: 'It's come full circle, and I think events like Marikana. . . have led the industry to fundamentally re-look at how it positions itself in society. What was previously seen as paternalistic, it's now being re-looked at' (interview 3M1). Such 're-internalising' may be expected as 'we have come full circle. . . [as a focus on the market] also facilitates the responsibilisation of market entities' (Shamir, 2008, p. 10). However, we maintain that liberalizing and outsourcing have continued to be the predominant forms of employers' work and, in comparison, companies' more recent efforts to contribute to workers' housing and welfare have been of limited consequence. The Marikana Massacre is a tragic testament to this. That said, we suggest that researchers should attend to the possibility of companies in some circumstances 're-internalising' their responsibilities for employees or even other stakeholders to address social systemic problems that impinge on their own strategies and operations, such as the burgeoning slums around South African mines.

CONCLUSION
We set out to better understand employers' role in the creation and perpetuation of inequality. Based on our historical analysis of employers' institutional work in the South African mining sector, we argue that the character of employers' institutional work shifts in response to labour market dynamics. Employers engage in forcing when labour is scarce, and they engage in freeing when it is abundant, laying claim to the ostensibly shared objective of giving 'choice' to workers and concealing their own economic and strategic objectives. Hence, we build a foundation for exploring whether and how employers' institutional work of exploitation is instantiated in other contexts, and how these contexts might affect the consequences of such work. Over and above contributing to our understanding of employers' role in creating and perpetuating inequality, we join others in seeking to revitalize organizational scholarship's interest in analysing -and indeed improving -organizations' impacts on social systems. Our analysis suggests that employers, workers, activists, regulators, and scholars should pay attention to both forcing and freeing as forms of institutional work that may insidiously advantage employers at the expense of workers, and should consider the possibility of concealed objectives and unanticipated consequences.