Corps de l’article

This is a study of horizontal and vertical linkages within a national labour movement, based on a survey of members of affiliated unions of the Congress of South African Trade Unions (COSATU). The South African labour movement has been seen as one of the most effective worldwide and a cornerstone of this strength has been its internal solidarity. However, this solidarity may, more recently, be under pressure from changes in the wider economy: from new managerial strategies externally and from the need to reconcile the competing demands of different interest groupings within the movement. Both the relative strength of the South African labour movement, and the challenges it faces, hold important lessons for unions operating in other contexts: the South African experience underscores the importance of organizing and solidarity, and the problems and possibilities of outreach.

Outside South Africa, unions have had to contend with twenty years of unprecedented external and internal pressures resulting in falling membership, as Table 1 shows.

Table 1

Union Density in Selected Countries

Union Density in Selected Countries

ILO data, cited in Bamber and Lansbury (1998: 359).

-> Voir la liste des tableaux

Externally, unions have faced a range of forces generally associated with a process of “globalization”: the integration of financial markets and shifts in consumer taste, rapid technological advance, increased international competition, and reductions in the willingness and ability of national governments to intervene in ameliorating market forces (Cerny 2003: 85–89). Meanwhile internal pressures include increasingly diverse workforces and less secure employment contracts, particularly in those Anglo-Saxon countries that have been removing restrictions on employer liability labour market policies in the name of “flexibility.”

Union decline has long been debated, in terms of causes, whether such decline is long-term or cyclical, and the appropriateness of various approaches that might constitute an appropriate renewal strategy. The problem is principally due to there being no obvious single cause of union decline. Thus, while the impact of overtly anti-union governments may well have contributed to union decline in, say, Britain (Brown, Deakin and Ryan 1997; MacInnes 1987; Martin et al. 1995), it would not explain decline elsewhere (France, Australia). In the wider context of economic policy, union decline has also been attributed solely to the economic restructuring of the economy (financial and employment deregulation, and privatization), within an increasingly adverse global environment (Wood and Brewster 2002). However, whilst there is little doubt that the intensification of global competition, overproduction, and changes in consumer demand and taste have placed firms under considerable pressure—encouraging the adoption of labour repressive policies—the effects of “globalization” have not been uniform (Block 2003: 440–451). For example, union membership trends remain uneven within continental Europe (Hyman 2001a), have remained relatively resilient in Canada, and have witnessed some growth in a number of developing nations—despite, in the latter two cases, the deregulatory drive being very much in the Anglo-Saxon mould. In short, while a range of external forces have put unions under great pressure, labour movements in some national contexts seem to have been more successful in mitigating these forces than in others. In particular, labour unions remain strong and internally vibrant in a number of semi-industrialized/industrializing countries (Jose 2002). The latter would include Brazil, South Korea, the Philippines, and South Africa. The question, here, is what holds unions together in such circumstances.

Frost (2000) argues that the ability of unions to cope with changes in the external environment and the resultant managerial strategies are contingent on union capabilities. In turn, union capability is contingent upon horizontal and vertical linkages (Frost 2000: 562). In this sense, the labour “movement” could be considered to be a network of relations between actors, with its fortunes contingent on the strength and nature of internal contacts and ties (Walsh 1994; Frost 2000). Vertical linkages concern the ability of unions not only to educate and mobilize the rank-and-file, and the responsiveness of leadership to the concerns of members, but also the state of ties between local and national union structures, and the union and the relevant national and international federations; this allows for more effective mobilization, joint action, and information sharing. Horizontal links are those between different unions, social coalitions and community groupings; the nature and strength of these ties partially determines the capacity for collective action, good ties making for more effective joint campaigning and resource distribution (Frost 2000: 561–562; Walsh 1994: 5–8).

Locke (1992), from a study of restructuring in the Italian auto industry, argues that the uneven fortunes of unions between and within nations reflects external pressures, local conditions and the manner in which unions balance national and local priorities. The adoption of new managerial strategies and forms of work organization at enterprise level has forced local unions to experiment with new forms of bargaining and accommodation, leading to a proliferation of practices that may challenge national union priorities. Conversely, the need to reach accommodations with political parties has the potential to conflict with workplace level needs and rank and file priorities (James 2004) and national unions may discount the concerns of local groupings of workers under such circumstances (Locke 1992: 245).

In the case of South Africa, as Table 2 illustrates, unions have retained—indeed have expanded – their penetration in the labour market, during the period of decline elsewhere.

Table 2

Union Density in South Africa

Union Density in South Africa
Source: NALEDI (1999).

-> Voir la liste des tableaux

In part, COSATU has owed its strength to a legislative dispensation that has been relatively labour-friendly, and to the legitimacy accrued from playing a central role in opposing the apartheid regime during its closing years. However, COSATU’s strength has also been due to its ability to accommodate and meet the needs of a relatively diverse membership, straddling gender and racial divides. South Africa’s economy shares attributes in common with both the developed and the developing world. In few other contexts have the pressures facing organized labour been so great: the need to accommodate the intensification of global competition, the challenges of dealing with new managerial strategies at plant level, the legacies of entrenched racism both within the workplace and wider society, and the complexities of dealing with a political partner increasingly under the thrall of neo-liberalism whilst battling to accommodate the needs of different constituencies. At the same time, unions need to overcome the challenges posed by regional socio-economic variations, differing political traditions, and the need to be relevant to the concerns of the growing numbers of workers in peripheral categories of labour. The extent to which the South African labour movement has been able to cope, and prosper, in such an adverse environment holds important lessons for organized labour worldwide.

Horizontal and Vertical Linkages in South African Labour

During the apartheid era, the political exclusion of Africans was mirrored at the workplace by “racial fordism,” characterized by a rigid racial division of labour (Joffee at al. 1993; Webster 1987). Prior to the 1970s, attempts to organize African workers through periodic upsurges in service organizations generally proved short-lived. However, by 1985, through a gradual process of mergers, the bulk of the South African labour movement coalesced into industrial unions under the umbrella of COSATU. In the twilight of apartheid, COSATU played a central role in the internal mass-based opposition, fulfilling the role of a social movement. The ending of apartheid saw COSATU entering a formal alliance with the recently “unbanned” African National Congress (ANC), and its junior partner, the South African Communist Party (SACP).

Horizontal Linkages: Dealing with Diversity

A range of union responses to the “hard times” of the early 2000s have been proposed and debated over recent years. This has ranged from paying greater attention to the “servicing” of individual members’ needs; to promoting mergers in order to benefit from economies of scale; to rediscovering the benefits of active recruitment and representation of under-organized groupings—the “organizing” approach; to the promotion of new “partnerships” with management; and to the promotion of a broader role in promoting social upliftment (Hyman 2001b; Claydon 1998; Fairbrother 2000; Kochan and Osterman 1994; McIlroy 2000).

Apart from “partnership,” COSATU unions have been relatively active in all of these initiatives. COSATU has generally been successful in its policy of one-industry, one-union, and has also experienced some expansion into the less organized sectors (Wood 1997). Finally, COSATU has successfully reconstituted its role as a social movement and remains committed to the principles of militant organization (Wood and Psoulis 2001: 310).

However, the question emerges as to the extent to which such activities are mutually reinforcing or contradictory. To Hyman (2001b), it is often the latter: the pursuit of a social partner role can often erode broader worker solidarity. Conversely, Kochan et al. (2001) argue that a mix of strategies may be complementary, tailored to attract, organize and retain different groupings of workers, whilst building coalitions with community groupings and other labour friendly groupings. Such co-operation is, of course, easier if workers in different unions and sectors have similar aspirations and concerns, and similar attitudes towards organizational activities, regardless of whether the union is emerging or mature. This would highlight the need to test sectoral variables, above all between sectors where partnerships with management are advanced, and those where union organization is relatively recent and associated with a high degree of militancy.

Successfully maintaining internal cohesion while remaining externally effective is a key issue. To Kochan et al. (2001), union effectiveness in today’s world is about being able to appeal to a diverse membership and potential membership categories, straddling different sectors, demographic categories, and skill levels. In other words, it is about effective horizontal linkages. All of this is only achievable if unions fulfil the role of a popular movement, able to make a meaningful impact in the battle for ideas.

The internal challenges facing unions include that of being relevant to increasingly diverse constituencies and dealing with changes in work organization, employment contracts, and societal and generational changes (Hyman 2001b; Kelly 1998; Moody 1997). The potential cleavages identified by the wide range of authors mentioned, as likely to pose specific challenges to unions are, then, diversities of age/generation, gender, ethnicity, sector, occupation, and socio-cultural group.

Contingent on it being able to make a worthwhile case in the “battle for ideas,” the actual effectiveness of a union may be gauged by membership coverage, capacity to mobilize for collective action and the extent of participation in the internal life of the union (Hyman 2001b; Kelly 1998; Heery et al. 2000).

Vertical Linkages: The Nature and Extent of a Democratic Rupture

Political alliances are always a fraught business for unions. Invariably, unions have to trade off influence for a loss of independence (Regini 1984). No alliances can mean marginalization, particularly in a multi-party democracy. As noted earlier, COSATU is in a formal alliance with South Africa’s ruling ANC. On the one hand, this has facilitated in the promulgation of a body of legislation that is some of the most labour-friendly in the world. On the other hand, COSATU has proved unable to check the ANC’s gradual drift rightwards, marked by the adoption of the neo-liberal GEAR (Growth, Empowerment and Redistribution) macro-economic policy framework. Critics of the federation’s current stance of “strategic engagement” are based either on the view that COSATU is acting as a “selfish elite” or that it has sold its members short by being incorporated into an anti union/anti working class agenda.

The “selfish elite” critique of COSATU is based around the familiar conservative view that the federation represents a privileged grouping of labour market insiders, pursuing their interests regardless of the long term employment consequences (Olson 1982). In South Africa, this critique is taken up by some who brand the alleged over-use of the strike weapon as being detrimental to COSATU members and to the most marginalized members of society (Baskin 1991; Habib and Taylor 1999). Further criticisms also focus on the inefficient running of COSATU, with its members representing a radical “labour aristocracy” who are barely held in check by moderate leaders (Adam and Moodley 1993; Etherington 1992).

The “sell-out” critique is equal-but-opposite to that of the selfish elite critique. While trade union identity, in most national contexts, has centred on an ideology of class struggle, in the case of COSATU, issues of class contestation have been diluted by the exigencies of national liberation (Regini 1984: 134). This led to protracted debates at COSATU Congresses as to the relative importance of developing an explicitly socialist policy manifesto, and the nature and extent of relations with the ANC (Baskin 1991: 280; Ramaphosa 1985: 45). The formal ending of apartheid with the election of the ANC government in 1994 brought this debate further to the fore. Thus, opponents of COSATU’s alliance with the ANC argue that COSATU’s influence on the ANC has progressively waned (Habib and Taylor 1999).

It is further argued that union leadership has become tied to both the state and sections of capital through a proliferation of complex deals covering issues ranging from productivity and profit sharing to social policy (Murray and Wood 1997: 167). This relationship between senior union figures and the ANC resulted in discontentment amongst the rank-and-file and an increasing sense of being marginalized. These tensions have been exacerbated by the desire of national union leaders to reach strategic accommodations with employer associations, and the differing needs of grassroots constituencies, mobilizing in response to specific local managerial strategies and actions (Locke 1992; Frost 2000). However, Barchiesi (1999) and Rachleff (2001) argue that instead of union members being dissatisfied with the COSATU leadership, discontent has centred on a more conspicuous target: the alliance with the increasingly neo-liberal ANC: “The unions, tied to COSATU, which in turn has been tied to the ANC […], have been little able to articulate an alternative economic agenda, despite their once outspoken advocacy of a grassroots activism” (Rachleff 2001: 162–163).

For different reasons, then, critiques of COSATU’s current direction state that its leadership is pursuing policies against the wider interests of its constituent membership. In order to test the extent of a vertical rupture between rank-and-file and leadership, we assess the opinions of different categories of member towards the Alliance with the ANC. An analysis of COSATU union members’ views on specific issues relating to horizontal and vertical ruptures was conducted in order to identify the basis of any emerging cleavages. To this end, an analysis of responses to questions about members’ propensity to strike was used to identify the extent to which the COSATU leadership may be acting as a “selfish elite.” Also, an analysis of members’ stated intention to vote for the ANC was used to identify the extent to which support for the tripartite alliance was waning, thus vindicating the “sell-out” critique.


This study is based on a nation-wide survey conducted in 1998 amongst rank-and-file members of COSATU affiliated unions, replicating similar studies conducted in 1994 and 1992 (Ginsburg et al. 1995). As with its predecessors, the 1998 survey explored the nature of worker organization and the attitudes and expectations of members towards both parliamentary and workplace democracy.

The study was conducted with the support of COSATU and the co-operation of employers; this and the fact that the interviews were conducted on a face-to-face basis allowed for an extremely high response rate, with very few potential respondents declining to be interviewed. The sample used was multi-layered and area based and provided 646 respondents. The final sample size and format of the questionnaire were based on detailed pilot studies in each area.[1] As an accurate list of trade union members was not available, final sampling was done systematically at individual workplaces, the number of workers selected being in proportion to the firm’s overall workforce size. This precluded the over-representation of workers from smaller firms. The advantage of such a method is that the sample is compiled during the interviewing process (Bailey 1982).

In order to more formally test for unification/ruptures within COSATU, the propensity to undertake strike action was used as an indicator of horizontal solidarity, and support for the ANC was used to indicate vertical solidarity. In both cases, behavioural differences across regions, occupations, gender, and racial group are explored. Thus once these factors are controlled for, hypotheses can be tested, with any significant cross-union differences in the likelihood of industrial action and the level of ANC support indicating the presence of internal and external ruptures respectively.

The empirical analysis presents the data both in summary tabular form and produces binary probit models so that the effect of specific factors (noted below) can be compared to each other. The models created estimate the probability of a specific outcome as a function of the variables deemed to influence that outcome, and are of the form:

where z reflects the probability of a particular outcome, x is the vector of explanatory variables and b their estimated coefficients, with y being the dependent variable. In this analysis, the dependent variable takes the value 1 if strike action has taken place and 0 otherwise in the first instance, then 1 if the respondent intends voting for the ANC and 0 otherwise for the second model.

For simplicity, the largest group in each category is used as a dummy variable. Together they make up the base model, which is a male, semi-skilled, Isizulu speaking, STD6 educated (the benchmark level of education at age 14), National Union of Metalworkers (NUMSA) member who resides in Gauteng. A general to specific methodology is adopted where, for clarity, insignificant variables are dropped from the analysis.

For model 1 the likely determinants of strike action are: the trade union organization the worker is a member of, region, age, education, occupation, gender and whether a shop steward is present in the workplace and if so, whether the steward is elected by fellow union members. The hypothesis is that no differences in militancy exist within COSATU, with any deviation from this indicating a rupture. Any significant coefficients on the trade union dummy variables will indicate such a rupture, with the greater frequency of significant terms in the model indicating a greater problem. The shop steward variable is expected to be positive, with the elected shop steward variable also expected to be positive since his/her position is determined by fellow workers.[2] Trade union members of certain occupational groups may be more/less likely to undertake strike action in accordance with their role within the firm (e.g., supervisor and unskilled worker).

Finally, a probit model is used for identifying the allegiance of COSATU members to the ANC, with any significant explanatory variables again providing evidence of divergence within COSATU. The model uses the same explanatory variables as the previous one. Therefore, it seeks to explain voting behaviour as a function of region, age, education, occupation, gender, union and its presence within the workplace. This analysis will allow the diversity and unity within COSATU to be tested across different trade union workers whether by fixed “indices” (e.g., gender), variable indices (e.g., membership of a specific union), or region of residency.[3]


The findings present the results of the probit models described above. Overall the indicators of vertical or horizontal rupture (defined by support for the Alliance and propensity to strike, respectively) demonstrate high levels of solidarity among union members. Thus, 420 respondents—some 68% of those interviewed—had taken part in strike action within the previous four years and support for the Alliance is overwhelming, as Table 3 shows.

Table 3

Support for the Alliance

Support for the Alliance

-> Voir la liste des tableaux

However, within this picture of solidarity, the probit models indicate some diversity. The first model tests vertical solidarity by comparing factors (discussed in the methodology section, above) affecting the likelihood of trade union members having undertaken strike action. The dependent variable takes the value of 1 if an individual has been involved in strike action and 0 if they have not. The results for this model are presented in Table 4.

Table 4

Likelihood of Trade Union Members taking Industrial Action, 1994–1998 (Probit)

Likelihood of Trade Union Members taking Industrial Action, 1994–1998 (Probit)

Mean dependent variable: 0.6619. Observations: 630. Chi-squared: 154.5. Log likelihood (LL): -325.79. Restricted LL: -403.05.

-> Voir la liste des tableaux

Here, then, membership of a particular union has some significance as to the likelihood of strike action taking place, but region, language, occupation or education do not. Since the probit model is not a linear function, the coefficients are not the effect of a unit change in that explanatory variable, so for clarity the marginal effects are also reported. These marginal effects are the derivative of the probability function which estimates the slope of that function at each point and can be interpreted as the average effect of a unit change in that variable (Greene 2000). So, for example, Table 4 indicates that those living in the Eastern Cape are 11% less likely to have taken industrial action than comparable workers in the base group of Gauteng and that members of NEHAWU are 50% less likely than members of the reference category NUMSA. In terms of region, and education, only those workers from the Eastern Cape and/or employed as a supervisor and/or with no formal education are less likely to undertake strike action, these being 11%, 16% and 37% respectively. In terms of language, members who speak Tsonga are less likely (31%) to strike than Isizulu speaking union workers.

The presence of a shop steward in the place of work significantly increases the probability of industrial action by 48%. The key findings, though, are that membership of particular unions affects the likelihood of striking. There are 15 unions included in the sample. When controlled for age, education, region, etc., seven are significantly different from the base group in terms of the likelihood of strike action, and seven are not. Of those with a significant difference, most appear more averse to strike action, indicated by the negative coefficients, than the base group NUMSA. The exception to this is the Chemical Workers Industrial Union (CWIU) who is 20% more likely. This would probably reflect the introduction of centralized bargaining in that industry, which has resulted in periodic localized protests by sectional interests discontented with national-level deals reached. In addition, the chemical industry is one of the few sectors where a COSATU affiliate is not the dominant union. In this area, the ideologically Africanist, and rather smaller, National Council of Trade Unions’ (NACTU) and the South African Chemical Workers Union (SACWU) enjoy a substantial following.

Workers are far more likely to strike if there are functional shop steward structures present, which would reflect a high mobilizing capacity on the part of the union (Batstone 1988; Batsone, Boraston and Frenkel 1978; Crouch 1982).

Differences in Political Voting Behaviour: The General Case

The results for the likelihood of voting for the African National Congress in the 1999 general elections are reported in Table 5.

Overall 75% of those surveyed expressed an intention to vote for the ANC. Once again, any significant variation from the dummy variables suggests that, once the model has been controlled for the other explanatory factors, membership of a specific union is likely to cause the probability of voting ANC to differ from the sample mean, indicating a potential rupture. Union members in the Western Cape are 9% more likely to vote for the ANC than are those in the base group, Gauteng. The education dummies were insignificant with regard to affecting the voting behaviour of union members, and were dropped. Those union workers whose first language is either English or Afrikaans are less likely to vote for the ANC than union members whose “mother-tongue” is an African language (the base group), indicated by the negative coefficients observed, both being in the region of 51%. Within COSATU these union members would most likely be coloured (persons of mixed racial origin). This finding is consistent with actual election results in 1999 that found that the majority of “coloured” workers voted for the New National Party, the former party of apartheid. Few opted for more radical alternatives to the ANC in 1999 such as the Socialist party of Azania or the Azanian Peoples Organization.

Table 5

Likelihood of Trade Union Members voting for the ANC in 1999 (Probit)

Likelihood of Trade Union Members voting for the ANC in 1999 (Probit)

Mean dependent variable: 0.7456. Observations: 625. Chi-squared: 164.86. Log likelihood (LL): -272.02. Restricted LL: -354.45.

-> Voir la liste des tableaux

All trade union dummies were included in the regression model, with only the significant variables shown here. The election of a shop steward positively affects the likelihood of union members voting for the ANC (33%), reflecting the new democratic process in South Africa since 1994. The interesting finding here is that a rupture appears to exist between NUMSA and the South African Combined Catering and Allied Workers Union (SACCAWU) members regarding their allegiance to the ANC and the alliance. Whilst NUMSA members are 12% more likely to vote ANC than the base group, CWIU, SACCAWU members are 21% less likely to vote ANC than the CWIU. This may indicate a horizontal rupture within COSATU across skills groups regarding support for the ANC party, and represents a potential source for a vertical rupture for the tripartite alliance if SACCAWU members gain support. These findings are surprising. NUMSA is recognized as a left-wing union who would perhaps be less likely to support the ANC than other unions within COSATU (Habib and Taylor 1999). A possible reason for this finding is a lack of alternative left-wing political parties for NUMSA members to vote for in the 1999 elections. With regard to the findings for SACCAWU, evidence from Baskin (1991) indicates that there is evidence of long-term division within this union and all its precursors. In addition, SACCAWU incorporates significant numbers of non-African members. Other union dummies were included in the model but whilst they were all positive, the coefficients were not significant at the 10% level.

Finally, the 1998 questionnaire asks a series of questions about improvements in life since 1994, and these were added into the model to see if they had any impact on trade union members’ propensity to vote for the ANC. Table 5 illustrates that improved housing and improved access to healthcare all positively affect the likelihood of voting for the ANC in 1999, being 22% and 11% respectively.

Gender Differences in Political Voting Behaviour

Having undertaken the basic analysis for both male and female union members, gender specific analysis was next undertaken to see if differences across gender exist. The results in Table 6 are for female union workers only, indicating that 62% intended voting for the ANC. Those whose first language is English or Afrikaans are, again, less likely to vote for the ANC, 34 and 32% respectively, than union members whose “mother-tongue” is Isixhosa, a finding consistent with a horizontal rupture.

Table 6

Likelihood of Female Trade Union Members voting for the ANC in 1999 (Probit)

Likelihood of Female Trade Union Members voting for the ANC in 1999 (Probit)

Mean dependent variable: 0.6196. Observations: 184. Chi-squared: 50.8. Log likelihood (LL): –96.8. Restricted LL: –122.2.

-> Voir la liste des tableaux

Of particular interest is the importance of the questions regarding public services and access to them since 1994. As with the all-sample analysis, the access to better housing since 1994 is more likely to result in voting for the ANC in 1999 (24%), with this also being the case for access to better healthcare (22%). The presence of an elected shop steward in the workplace positively affects the likelihood of COSATU members voting for the ANC, indicating their importance to COSATU and the tripartite alliance in disseminating information. Perhaps the most interesting finding is the position of SACCAWU members with regard to affiliation to the ANC. As in Table 5, SACCAWU members are less likely to vote for the ANC than CWIU members, potentially indicating a horizontal rupture across both gender and skills, again with potential implications for vertical rupture. However, when compared with the male results in Table 7, it appears that the division between male and female is within SACCAWU itself (Baskin 1991).

Table 7

Likelihood of Male Trade Union Members voting for the ANC in 1999 (Probit)

Likelihood of Male Trade Union Members voting for the ANC in 1999 (Probit)

Mean dependent variable: 0.7982. Observations: 441. Chi-squared: 120.2. Log likelihood (LL): -161.7. Restricted LL: -121.8.

-> Voir la liste des tableaux

With regard to male union members, a significantly higher number, 80%, intended voting ANC. The similarity across language with female union members indicates a possible racial rupture, except to a greater extent. This is indicated not just by Afrikaans and English speakers being 51 and 54% less likely to support the ANC than those who speak Isixhosa, but also those who speak Isizulu being 15% less likely. Those union members that are employed as supervisors are also less likely to vote for the ANC than unskilled workers (22%); this would reflect the partial persistence of a racial division of labour. The election of a shop steward in the workplace again positively affects the probability of union members voting for the ANC, increasing it by 25%. As with females, better access to housing increases the likelihood of voting ANC, with access to more land also positively affecting voting for the ANC. Again, an apparent division is found between different trade unions within COSATU and their apparent alliance with the ANC party. Male NUMSA members are more likely to vote for the ANC than CWIU members (the base group), whilst NEHAWU members are less likely to vote ANC. This is perhaps surprising given that the results from Table 4 indicated a possible divide between NUMSA and SACCAWU. Again, the same argument of NUMSA members not appearing to have a legitimate alternative to the ANC can explain some of this. Male members of NEHAWU, however, are clearly disillusioned with the ANC and alliance, and this is not completely unsurprising. Male workers from this union represent unskilled, blue-collared, largely public sector, workers (e.g., cleaners) as opposed to a large number of female workers who are mostly nurses/professionals. The male union members may be particularly disillusioned with the drift to neoliberalism, as this has translated into both a commitment to privatization and a greater “contracting-out” of unskilled support jobs. Again, there have been a number of highly conspicuous industrial disputes within the public sector (mostly involving blue-collar workers), which have done little to promote good relations between the ANC and the public sector rank-and-file.


This paper aimed to examine the extent to which there was a lack of trade union unity within COSATU. This was against the background of current debates regarding the presence of vertical and horizontal cleavages. In particular, these debates relate to the existence of a diverse membership and to criticisms, from both left and right, of COSATU’s current stance of “strategic engagement.”

The survey revealed that, in general, there were high levels of both vertical and horizontal solidarity. Workers seem generally supportive of the strategic choices the leadership has made in choosing to retain the alliance with the ANC. There appears little divergence too in the likelihood of strike action amongst COSATU members. Exceptions exist though across gender and racial divides and—perhaps more worryingly—between unions within COSATU.

Some of the vertical and horizontal divisions that are visible seem to reflect the persistence of racial tensions, with more conservative non-African members being critical of the Alliance. Another concern within COSATU relates to members of SACCAWU. Our findings indicate that members of this trade union, and in particular female members, were less likely to vote for the ANC in the 1999 elections. This represents a horizontal rupture between sectors, and the possibility of a vertical rupture between COSATU and the ANC if disillusionment with the ANC spreads within the union federation. At the regional level, a degree of ethnic division and distinct political traditions was found, with the possibility again of a horizontal rupture within COSATU regarding SACCAWU members in Gauteng. Such a rupture was also found in the degree of militancy towards strike action, with members of CWIU more likely than any other union to take industrial action. As noted earlier, this is possibly a reflection of disenchantment with relatively recently introduced centralized bargaining structures.

However, whilst some horizontal ruptures within COSATU appear to exist, the results indicate a large degree of unity, since many explanatory variables in the voting models drop out of the analysis. Of those that have been dropped due to insignificance, the majority indicate support for the ANC and tripartite alliance. This is confirmed in the findings that reveal that the bulk of labour friendly legislation and—albeit focused and partial—improvements in public services (e.g. better housing and wages since 1994) have ensured that both leadership and rank-and-file remain largely behind the Alliance.

Whilst the survey revealed that there was not a significant divide between the “struggle” and the “post-struggle” generations in terms of political allegiances (Wood 2002), there is little doubt that the continued economic domination of the white minority and recent memories of the struggle against apartheid still provide a powerful “glue” uniting black workers. The increasingly close ties—and periodic joint campaigns—between COSATU and historically conservative white dominated unions would seem to indicate that future class solidarity could gradually erode racial differences. In short, despite internal pressures, COSATU has been able to maintain and develop existing horizontal and vertical linkages and this represents a cornerstone of union strength (Walsh 1994).

These ties can be ascribed to the persistence of high levels of mass mobilization and the persistence of an embedded culture of internal democracy, making any existing networks more meaningful and effective. In turn, the high degree of mobilization represents a likely product of real material conditions: the persistence of high levels of social inequality, despite limited progress in some areas, limited opportunities for upward mobility, and the dominance of management by members from the ethnic white minority, whilst dramatic political changes would make a similar transformation of the economic arena appear more attainable. In other words, high levels of relative deprivation persist (Kelly 1998).

Finally, whilst the South African labour movement has been relatively successful both in maintaining membership in traditional areas of union activity in the private sector, and in taking advantage of the political reforms of the early 1990s to expand activities in the public sector (Wood and Psoulis 2001), it has failed to make significant inroads amongst workers in temporary and insecure occupations, or to make any headway in the informal sector; as is the case with many other industrializing economies, the latter provides an increasing proportion of the overall number of jobs. The South African labour movement has traditionally had good relations with community based political structures; however, the relationships with the rapidly proliferating new grassroots organizations representing “socially excluded” groups remain informal and, often, uneasy (Desai 2002). Making headway in these areas is contingent on the forging of new solidarities, founded on the development of clear policy alternatives that are meaningful to both existing and new constituencies. Whilst an essential prerequisite for union effectiveness, the deepening of existing horizontal and vertical ties, and the use of tried-and-tested organizing strategies in traditional areas of activity are unlikely to resolve the new challenge of outreach beyond the formal workplace; the fortunes of organized labour are contingent not only on the strengthening of existing networks, but also on their expansion into previously-neglected areas.

Conclusion: The Limitations of Existing Solidarities

What this article highlights is that whilst there are both centrifugal and centripetal tendencies within COSATU, the overwhelming pressures are towards unity. Moreover, the nature and extent of existing internal divisions seem due to a complex range of causes, including changes in the levels of bargaining, distinct regional political traditions, long-standing debates within and between unions regarding the advisability of ties with political parties and community organizations, and strains imposed through union mergers.

Whilst there is little doubt that unity is a foundation of strength, the relationship should be two-way. High levels of internal solidarity make for organizational effectiveness at both national and plant level, which, in turn, would make both membership and active participation more rewarding, even amongst members of historically marginalized groupings. COSATU’s success over the past 16 years has resulted in relatively high degrees of confidence in the manner in which the federation is run, and the alliances it has entered into. In turn, this has led to continued union growth, even in areas such as the public sector, where, up until the late 1980s, independent unions were relatively weak. The challenge facing the federation now is to consolidate these gains, and actively draw in those categories of members that are still marginalized in key areas. This, however, not only depends on focused organizational interventions, but also on the articulation of viable policy alternatives which have broad appeal. The South African experience provides an inspiring example to unions elsewhere, but also highlights the persistence of distinct national institutional configurations and the impact of specific historical traditions, and underscores the importance of strategic choice and the ability to respond rapidly to changing external conditions.

The South African experience holds important lessons for organized labour worldwide: it underscores the importance of vertical and horizontal linkages as a basis for greater union effectiveness. The greatest strengths of the South African labour movement include a vibrant shopfloor democracy, an ability to hold elected representatives to account, close linkages between individual unions, and the capacity to mobilize support for collective action both in the workplace and in the community. Its weaknesses include persistent divisions on gender and racial lines, and within some unions and regions; again, there is the ongoing challenge of deepening ties between leadership and rank and file in a rapidly changing political-economic context. Whilst an emphasis on recruitment and mobilization may help organized labour retain its presence in key firms and industries, it is unlikely to make union membership more attractive to those employed in the informal sector and in other highly peripheral occupations: the latter are of increasing importance in industrializing countries, yet there is little evidence that unions have developed new policies more closely suited to the needs of workers in these areas. The challenges facing unions in both the mature economies and the developing world include not only the ability to mobilize and organize, and reconstitute and deepen vertical and horizontal solidarity, but the development of clear policy alternatives that reflect the concerns both of existing members and constituencies, and the growing bodies of workers condemned to more marginal areas of the economy.