Discrimination and favouritism among South African workers: Ethnic identity and union membership

Highlights

• We use a lab-in-the-field experiment to study ethnic and labour discrimination between coal mineworkers in South Africa.

• In repeated dictator games, we distinguish discrimination and favouritism compared to an uninformed baseline behaviour.

• We find that trade union identity encourages broad altruism towards workers, while ethnic affiliation does the opposite.

• The ethnic majority (Zulu) discriminates against minority groups and even against co-ethnics (although to a smaller extent).

• Subcontracted workers from the ethnic majority, with less stable jobs and lower wages, drive this discriminatory behaviour.

Abstract

This paper analyses the ways in which ethnic identity and labour institutions shape favouritism and discrimination among workers. We conduct a lab experiment in the field with South African coal miners from various ethnic groups and with different trade union membership status. Our analysis suggests that union identity and ethnic identity are two social constructs that operate in a distinct and opposite fashion. Unionization acts as a factor of workers solidarity beyond the confine of union membership. Conversely, ethnicity operates as the linchpin through which discrimination among workers is infused not only between ethnic majority and minorities, but also within the majority group itself. We find that the widespread practice of subcontracting in the mining sector exacerbates ethnic discrimination among workers both between and within ethnic groupings.

Introduction

The labour market is one of the main economic realms where discrimination manifests itself, and numerous studies suggest that workers around the world regularly experience unfair treatment based on personal traits such as gender, nationality, religion, age, and sexual orientation.¹ The economic literature has explored the roots of labour discrimination from many angles – ranging from preferences or tastes, to statistical discrimination, to models of firms’ demand for different types of workers – but mostly focusing on biases originating from the employers. The literature on discrimination in developing countries’ labour markets is no exception (Ito, 2009, Galarza and Yamada, 2014, Mobius et al., 2016). To date, there is still a limited understanding of whether a segmented society can also influence attitudes among workers, which affects their behaviour towards their peers.

This issue is particularly relevant for highly fragmented economies. In developing countries, one of the most carefully documented source of discrimination is ethnic diversity. This problem has received much attention in the development literature because of the link between ethnic fractionalization and poverty (Montalvo and Reynal-Querol, 2005, Churchill and Smyth, 2017). The most recent studies, however, go beyond the boundaries of ethnic groups alone, and emphasize the interplay between ethnicity and other forms of economic and political status-see for example Grossman & Honig (2017) on ethnicity and class and Lazareva & Mironova (2018) on ethnic and political minorities. Labour status is linked to various forms of groups and hierarchies, since workers differ in wages, contractual benefits, and degrees of job security. For this reason, the analysis of discrimination in the labour market of fragmented societies cannot ignore the interplay between ethnicity and workers’ position.

In our study, we examine the role of ethnic and labour hierarchies in determining group favouritism and discrimination in a highly polarised context, South Africa’s mining industry. This country is a paradigmatic example of a divided society. The legacy of colonialism and apartheid in South Africa has shaped labour and ethnic relations in the mining sector, and has created a segmented labour market that often opposes unionised and protected workers to non-unionised, temporary and subcontract workers. Even in the most polarized contexts, however, measuring discrimination and favouritism is notoriously difficult, and relationships among workers are particularly challenging to assess. Social and labour status are not the only factors influencing how workers treat their colleagues, and in the actual job place it is virtually impossible to disentangle the effect of group biases from unobserved interpersonal dynamics specific to each individual. Hence, in order to study the presence of potential biases among workers, we use a lab experiment in the field. In particular, we rely on a dictator game to compare workers’ behaviour towards their peers in response to random economic shocks (lotteries), depending on the ethnic characteristics and union status of their colleagues. We find ample evidence of biases along existing ethnic lines, mediated by union status and contractual position, which we interpret in light of economic theory and the latest research in social psychology.

In recent years, there has been a burgeoning interest in the economics of social identity and discrimination (Chen & Mengel, 2016), and in particular experimental economists have increasingly adopted lab experiments to gain insight into discriminatory behaviours (Lane, 2016). Two broad mechanisms emerge from this experimental literature: altruism towards people with similar characteristics or status (in-group favouritism) and negative biases against outsiders (out-group discrimination) (Charness et al., 2007, Heap and Zizzo, 2009, Chen and Li, 2009, Eckel and Grossman, 2005). In our study we concentrate on the interplay of personal characteristics, such as ethnicity, with factors directly linked to one’s economic position, such as job status and trade union membership. The literature on ethnic preferences in the lab so far found heterogeneous results, ranging from evidence of discrimination towards minority groups (Fershtman & Gneezy, 2001), to no coethnic bias (Berge et al., 2015). Some lab experiments look specifically at labour markets’ discriminatory dynamics (Gachter & Fehr, 2002), but we depart from these studies by focusing on altruism among workers, rather than on the interactions between employers and employees.

Lab experiment in behavioural economics have mainly focused on favouritism or discrimination towards generic out-groups, and yet a wealth of research from social psychology has highlighted that not all out-groups are alike (Everett, Faber, & Crockett, 2015). In particular, groups can have different size and social status, so that the relative hierarchy of the groups matters for behaviour, not just the insider/outsider divide (Leonardelli and Brewer, 2001, Lücken and Simon, 2005). In sociology, the theory of group position (or group threat theory) has long emphasized how the alienation of minority groups determines in- and out-group behaviour (Blumer, 1958) and how group size in particular can drive the perception of potential threat from the out-groups, for example in the case of immigrants (Schlnter & Scheepers, 2010). Sociological studies have also emphasized that group hierarchies can be particularly relevant for workplace discrimination, since intra-workers harassment can enforce formal and informal status hierarchies (Lopez, Hodson, & Roscigno, 2009).

Few economic studies have looked at hierarchical group dynamics, and have typically done so in trust games. For example, Tsutsui & Zizzo (2014) show that low status induces deference towards high status and majority subjects. Their study, however, relies on artificially assigned group status. In South Asia, Gupta, Mahmud, Maitra, Mitra, and Neelim (2018) find that religious minorities show positive in-group bias, while majorities positive out-group bias. Hence they show that behavior is not driven by religious identity per se (Hindu or Muslim), but rather by the relative status it generates within the population. In Bosnia, Lazareva & Mironova (2018) find that minority status reduces trust in out-groups. Closer to our study, Haile, Sadrieh, and Verbon (2008) find evidence in South Africa of cross-racial envy stemming from poorer players: low-income subjects, both blacks and whites, invest significantly less when paired with high income subjects of the opposite racial group. All these economic studies were framed as trust games. One interesting dictator game that looks at relative status of dictators and recipients in the US is by Fong & Luttmer (2011), who find that, when choosing how much to give to different charities, non-black dictators perceive those who help mostly black people significantly less worthy of donations.

Our work contributes to the experimental literature on labour and ethnic discrimination, going beyond the standard in-group versus out-group dichotomy (Habyarimana, Humphreys, Posner, & Weinstein, 2009). Conceptually, our approach builds on the distinction between different types of discrimination and favouritism: Feld, Salamanca, and Hamermesh (2016) classify behaviours in endophilia/endophobia – like or dislike for similar people of the same group, and exophilia/exophobia – like or dislike for people with different backgrounds or from different groups. We enrich this mapping by adding a vertical dimension: a subject’s in- and out-group behaviour towards others may vary depending on the perceived relative status of the groups (e.g., a majority or privileged group vs. a minority or disadvantaged group). In our analysis, group interactions can be asymmetric depending on who has a higher or lower status (as already noted in the social psychology literature, for example in Fiske, Cuddy, Glick, & Xu (2011)). This new conceptualization has important policy implications, because instead of considering discrimination as a unidirectional issue, it allows for new forms of mutual support or injustice arising within the workforce.

Our laboratory experiment consisted of a series of dictator games in which coal miners shared the outcomes of a lottery, knowing only some information about the opposite player’s mother tongue and union affiliation. These characteristics reflect the most salient features that could spur discrimination in the South African mining sector, due to the high degree of social segmentation (Barr & Oduro, 2002). There is anecdotal evidence that unions can create a wedge among mineworkers by successfully driving an effective agenda for the defense of the interests of its members. Non-unionised workers, who tend to have precarious positions, lower pay and are often required to perform more dangerous tasks, could resent the preferential treatment that union workers enjoy. At the same time, union members may distrust their non-unionised counterparts, perceiving them as a threat because they are prepared to accept lower standards to secure employment. Arguably, the presence of a strong labour union such as the National Union of Mineworkers (NUM) has entrenched a segmented, dual labour market in the mining industry. In addition, South African mines have witnessed a re-emergence of old ethnic divisions that were not fully overcome by class consciousness and union solidarity. The ethnic divide has become more salient as the unions are perceived as an instrument of patronage at the disposal of self-serving union officials. Against this backdrop, we find that the trade union generates in-group favouritism, acting as an ‘elite’ club. However, in our study, union solidarity extends also to non-union members. Discrimination is instead driven by ethnic affiliation. We find that members of the majority ethnic group discriminate against ethnic minorities, but also against coethnic workers. These surprising results operate through the contractual employment status of workers, in that subcontract workers from the majority group tend to be considerably less generous.

These findings reflect some key principles highlighted in the economics of identity, whereby individuals derive utility from being part of a social group (Akerlof & Kranton, 2000). This literature, however, does not fully account for all the different nuances of discrimination that can arise, depending on how privileged an individual is. Our findings indicate that group behaviour does reflect vertical hierarchies of status: workers belonging to a more privileged group like a trade union or to an ethnic majority treat their opponents in the dictator game more or less favourably. Instead, people belonging to minor or less privileged groups are less prone to display consistent group biases, apart from some rare evidence of deference. We discuss our results in relation with the economic theories of preferences (or taste)-based discrimination and the evidence from the psychology research on inter-group relations (Jost et al., 2004, Ashburn-Nardo and Johnson, 2008, Tajfel and Turner, 1979, Brown, 2000, Balliet et al., 2014).

The rest of the article proceeds as follows: Section 2 provides some background on the South African mining labour market; Section 3 presents a simple conceptual framework and our broad hypotheses on the effect of status on altruism; Section 4 discusses the experimental design, the data, and the empirical strategy; Section 5 comments on the key results; Section 6 discusses the main interpretation of these results, their implications and some caveats; and finally Section 7 concludes.