Black South African Unions: Relative Wage Effects in International Perspective

Literature › Works

P. G. Moll (1993). Black South African Unions: Relative Wage Effects in International Perspective. Industrial and Labor Relations Review, Vol. 46, No. 2. Cornell University, School of Industrial & Labor Relations.

Type	journal article
Scope	Economics and economic history of apartheid
Method	econometric
Period studied	1924–1985
Regions	South Africa, Johannesburg, Pretoria, Durban, Cape Town, Port Elizabeth, Vaal Triangle
Themes	unions strikes, labour markets wages, inequality poverty unemployment, comparative theory, methods data sources, sanctions divestment
Citation key	moll1993

Argument

Using 1985 household microdata, Moll estimates union-nonunion wage differentials for black South African blue-collar workers. He finds a self-selection-corrected differential of about 24%, in the high 'American' range rather than the lower 'European' range, despite disenfranchisement, late legalisation of African unions and repression. A second finding is that unions compressed wages across skill levels: gains were concentrated among unskilled workers and negligible for skilled and semi-skilled ones. This is attributed to the emergent unions' plant-level organising around a 'living wage'. A small white sample (13% differential) is used to argue the black skilled result is not an artefact of industrial-council 'ergo omnes' extensions.

Evidence and method

Bureau of Market Research 1985 household surveys of urban black households (about 4,963 full-time blue-collar workers, 1,155 union and 3,808 nonunion) and the 1985 Central Statistical Service household survey for whites (72 union and 235-237 nonunion men, three cities only). Also industrial council data (TUCSA 1985; Williams 1990), strike statistics from the Department of Manpower and ILO, Friedman (1987) for union history, and international union-wage studies compiled in Table 6.

Place in the debates

• Size of union wage effects under apartheid / strength of black unions (see Black trade unions: growth, wage effects, Wiehahn and strategy). Argues black unions achieved wage gains comparable to those in developed democracies by 1985, rejecting the expectation that the racial oligarchy's power would keep them small.
• Why union wage premia differ internationally (centralised vs decentralised wage-setting) (see Black trade unions: growth, wage effects, Wiehahn and strategy). Partly supports Blanchflower and Freeman's centralisation typology (large South African effects reflect decentralised plant-level bargaining) but notes the typology is inadequate, since the UK and Switzerland do not fit.
• Wage compression and union equity goals (see Black trade unions: growth, wage effects, Wiehahn and strategy). Finds black unions compressed skill differentials, in line with standard US and Canadian findings, and attributes this to the living-wage strategy of the emergent unions.
• Role of foreign ownership and disinvestment pressure (see Black trade unions: growth, wage effects, Wiehahn and strategy). Suggests multinational pressure to recognise unions and raise wages may have contributed to the union effects, but cannot test this with the data.

Findings

• The self-selection-corrected union wage differential for black blue-collar workers was 24% overall (19% men, 31% women). (p. 12)

  According to this measure, the black union-nonunion differential was 19% for men, 31% for women, and 24% overall.

• Black South African union effects fall in the high American range, not the lower European range. (p. 13)

  The results for black South African unions thus fall in the higher "American" range rather than in the lower "European" range.

• Emergent unions' gains came mainly for unskilled workers, with large differentials for unskilled men and women. (p. 13)

  Taking all sectors together, the average union differential for unskilled Africans was 19% for men and 35% for women.

• Skilled African workers gained almost nothing from union membership, consistent with wage compression. (p. 13)

  skilled African workers benefited hardly at all from their union membership. The coefficients are small and insignificant.

• Union membership exploded after legalisation, with a large share of new members choosing to stay outside the official industrial council system. (p. 4)

  By 1985, the number of workers in registered unions had increased by 680,000 to 1.4 million, while the number of workers in unregistered unions had increased by 390,000 to 450,000.

• Strike activity by black workers rose steadily after legalisation to levels comparable with Britain in the 1950s-60s. (p. 6)

  following legalization in 1980, the number of days lost increased steadily, reaching 187 per 1,000 by 1985, which is comparable to the rate of days lost in Britain during the 1950s and 1960s.

• White skilled and semi-skilled workers had a substantial union premium, whereas black counterparts had none, so the black result is probably not due to ergo omnes extensions. (p. 16)

  The differential for white skilled and semi-skilled workers was about 13%.

• Returns to schooling and experience for blacks were about a quarter of those for whites, attributed to low-quality schooling and discrimination. (p. 15)

  Black schooling returns were low due to low-quality schooling, discrimination in education spending, and labor market discrimination

• Moll concludes that apartheid failed to prevent black unions from winning wage gains similar to those in democracies. (p. 17)

  even in 1985, five years before political negotiation became an acceptable concept for the minority white regime, African union workers had won wage concessions of a size similar to those in democratic developed countries.

Evidence ledger

Sources used

Source	Kind	Period	How used
Bureau of Market Research (BMR) 1985 household surveys of urban black households (incl. Income and Expenditure Patterns of Urban Black Multiple Households in Johannesburg, 1985, Research Report 130.9)	household or labour survey microdata	1985	Main microdata for probit of union membership and for union/nonunion wage regressions of black blue-collar workers (N=4,963), and for black union density by sector.
Central Statistical Service (CSS) 1985 household survey (white households)	household or labour survey microdata	1985	Microdata for estimating the union wage differential among white male skilled and semi-skilled workers (72 union, 235 nonunion observations) as a comparison with black results.
Trade Union Council of South Africa (1985) Official Trade Union Directory and Industrial Relations Handbook; Williams (1990) Industrial Council Wage Rates in South Africa (1978-1988)	secondary literature compilation	1984	Source for the number of industrial councils, those extendable to non-parties, and workers covered, in Table 1.
Central Statistical Service (1989) economically active population by sector	official statistical series	1985	Denominator for the share of workers covered by industrial councils (Table 1).
Central Statistical Service (1990) South African Labour Statistics, Table 2.2.1.1 (nonagricultural employment by race)	official statistical series	1970-1986	Employment denominators for strike person-days lost per 1,000 employed by race (Table 2).
Department of Manpower Annual Reports (person-days lost in strikes)	official statistical series	1970-1986	Days lost in strikes by race, showing black strike activity rising after 1980 as an indicator of union strength.
U.S. Department of Labor, Handbook of Labor Statistics 1989; ILO Yearbook of Labour Statistics	international trade or finance data	1950s-1980s	Comparison strike rates (person-days lost per 1,000 employed) for three other countries in Table 2.
Union membership figures (claimed membership 1979 and 1985; registered vs unregistered unions)	secondary literature compilation	1979-1985	Establishes growth in union membership from 760,000 to 1.8 million, with split between registered and unregistered unions.
Friedman (1987) Building Tomorrow Today: African Workers in Trade Unions, 1970-1984	secondary literature compilation	1970-1984	Source of facts on union recognition, FOSATU factory bargaining rights, the 1973 strikes, living-wage demands, and state/employer violence against unions.
Blanchflower and Freeman (1990) NBER Working Paper 3342 and other international union wage studies (Lewis 1986, Freeman and Medoff 1984, Duncan and Leigh 1980, Lee 1978, Simpson 1985, etc.)	secondary literature compilation	1967-1987	Union wage-effect estimates compiled in Table 6 as the international benchmark for the South African 24% effect.
Industrial Conciliation Act 1924 and Industrial Conciliation Act of October 1979 (with amendments); Wiehahn Commission (1977)	legislation gazette	1924-1979	Describes the legal framework of industrial councils, racial exclusion of African unions, and the legalization of African unions after 1979.

Inference

Quantification: Econometric (causal design).

Cross-sectional Mincerian log-earnings regressions on 1985 microdata with a union dummy, separate union and nonunion equations (Oaxaca-style differential), and Lee/Maddala selectivity-corrected switching regression with jackknife standard errors. Skill- and sector-specific subsample regressions assess compression, a white-worker comparison tests the ergo omnes objection, and the results are compared with a table of published international union differentials.

Key claims

Claim	Basis	Strength
The selectivity-corrected union wage effect for black blue-collar workers in 1985 was about 24% overall (19% men, 31% women), in the U.S. range rather than the European range. (p. 255)	BMR 1985 microdata regressions (Tables 4 and 5), compared with Table 6 literature estimates.	Systematic data
Black unions compressed wages by skill level: skilled African workers gained almost nothing from union membership while unskilled gained most (19% men, 35% women). (p. 256)	Skill- and sector-specific regressions with union dummy on BMR data for African workers (Table 7).	Partial data
The absence of a union effect for black skilled and semi-skilled workers is probably not due to ergo omnes extension of industrial council rulings, since white skilled and semi-skilled workers showed a ~13% union differential. (p. 259)	CSS 1985 white household survey (small, nonrandom sample) vs. BMR black male skilled/semi-skilled subsample (Tables 8 and 9).	Partial data
Black union membership is determined by industry, experience, tenure, firm size proxy, and region, not by education or marital status, as elsewhere; workers are attracted by the prospect of wage gains. (p. 254)	Reduced-form and structural probits on BMR data (Table 3).	Systematic data
Black strike activity was minimal in 1975-79 and rose steadily after legalization in 1980, reaching a rate comparable to Britain in the 1950s-60s, suggesting growing union strength. (p. 249)	Department of Manpower strike days and CSS employment (Table 2), compared with U.S./U.K./Japan figures.	Systematic data
Union membership grew from about 760,000 in 1979 to 1.8 million in 1985, with a dual system of industrial councils and plant-level recognition agreements; industrial councils covered about one-fifth of non-primary workers. (p. 247)	Claimed union membership figures; TUCSA and Williams IC data, CSS economically active population; BMR-based union density (Table 1).	Partial data
The source of the large black union effect cannot be pinned down, but may reflect emergent unions' combativeness and foreign ownership/disinvestment pressure. (p. 260)	Strike rates, Friedman's accounts of foreign-firm recognition; no direct test.	Illustrative or anecdotal

The author's caveats

BMR data are household not worker samples; hours worked and hourly wages are unavailable; earnings are annualized gross wage; BMR data have no industrial council coverage variable, so IC and plant-agreement effects cannot be separated; union membership figures are claimed and probably exaggerated; the CSS white sample is small, covers only three cities, only household heads, is nonrandom, lacks tenure; gender differences must be treated with skepticism given the data set; cross-country comparisons differ in samples and methods.

Gaps in the evidence

The argument lacks firm-level or establishment-linked data to separate industrial council extension effects from plant-level recognition agreements, and hourly wage data to remove hours bias. Panel or repeated cross-section data (before and after legalization) would give a better causal estimate of union effects rather than a single 1985 cross-section with selection correction. Matched employer-employee data and direct industrial council wage schedules (partly existing, e.g. Williams 1990) could test the compression and ergo omnes arguments more convincingly; the causal explanations (strikes, foreign ownership) are untested.

Replicability: Data described but not public.