Labor Unions and the Distribution of Wages and Employment in South Africa¶
T. Paul Schultz; Germano Mwabu (1998). Labor Unions and the Distribution of Wages and Employment in South Africa. Industrial and Labor Relations Review, Vol. 51, No. 4. Cornell University, School of Industrial & Labor Relations.
| Type | journal article |
| Scope | Economics and economic history of apartheid |
| Method | econometric |
| Period studied | 1993–1993 |
| Regions | South Africa, 14 regions (4 provinces and 10 homelands) |
| Themes | unions strikes, labour markets wages, inequality poverty unemployment, transition post apartheid, methods data sources, education human capital |
| Citation key | schultz1998a |
Argument¶
Using a 1993 national household survey, the authors estimate how union membership relates to the wage distribution and to employment for Africans and whites. Quantile regressions show that African union members earn large wage premia, concentrated at the bottom of the residual distribution, while white union members in the upper 70% earn less than comparable nonunion workers. Unions therefore compress wages within the union sector and narrow racial gaps there, but union members are a minority of African workers. Cross-regional logit and multinomial logit estimates suggest that a larger union wage effect is associated with lower employment and labour-force participation among African youth. The authors conclude that halving the union premium might raise African youth employment modestly, which poses a dilemma for the post-apartheid government's union alliance.
Evidence and method¶
SALDRU/World Bank 1993 national household survey (43,974 individuals, 9,000 households, 360 clusters, Aug-Dec 1993), with wage-earner subsamples by race and sex; union shares from Bureau of Market Research surveys and COSATU estimates as cited from secondary sources.
Place in the debates¶
- Do unions widen or narrow wage inequality (union vs nonunion, racial gaps)? (see Black trade unions: growth, wage effects, Wiehahn and strategy). Unions compress wages among members and narrow the racial gap within the union sector, but widen the union-nonunion gap. Most African workers are nonunion and may lose relatively.
- Harris-Todaro-Calvo view that union wage premia raise unemployment (see Black trade unions: growth, wage effects, Wiehahn and strategy). Finds a larger union wage effect associated with lower employment and participation among African youth, but also with lower unemployment, contrary to the model's prediction. Concludes mainly that the distortion pushes people out of the labour force.
- Union wage premium as rent versus productivity or compensating differential (see Black trade unions: growth, wage effects, Wiehahn and strategy). Cannot separate these because union status is treated as exogenous. The premia may overstate or understate union rents, and the authors flag this as a limitation.
Findings¶
-
Union membership is associated with very large wage premia for low-wage African men and small ones for high-wage African men. (p. 1)
Among male African workers in the bottom decile of the wage distribution, union membership was associated with wages that were 145% higher than those of comparable nonunion workers
-
The mean OLS union effect is a 60% premium for African men and a small, insignificant penalty for white men. (p. 10)
The coeffi- cient on union status is .468 for African men, and for white men it is -.051, implying that an African union worker receives a wage that is 60% higher
-
The union premium falls steeply across the wage residual distribution for African men, which implies compression of wages within the union sector. (p. 11)
among African men, the log wage gain associated with having a union job is eight times larger for those in the lowest decile than for those in the highest decile.
-
Union workers earn lower returns to schooling than nonunion workers, which weakens private incentives to acquire education. (p. 13)
if a stu- dent thought it likely that he would be employed in a unionjob, he would have less private incentive to obtain additional schooling
-
Union wage profiles are flatter with respect to experience but peak later. (p. 14)
the wage profile peaks at a later age for union workers (50) than for nonunion workers (37)
-
Industry-level union shares are highest in mining and manufacturing and very low in agriculture and domestic service. (p. 8)
The share of union members is highest in mining (73%); substantial in manufacturing (48%)
-
Africans face much higher unemployment and lower labour-force participation than whites. (p. 8)
The rate of unemployment among those who report themselves in the labor force (employed or looking for ajob in the prior week) is four times larger for Africans and coloreds than for whites
-
Within an industry, the union premium for African men remains substantial despite centralized bargaining; manufacturing union workers earn far more than nonunion peers. (p. 15)
the union manufacturing worker earns a wage that is about 70% higher than that received by nonunion workers in manufacturing
-
Regional union wage effects are associated with reduced employment of young African men, with a small elasticity. (p. 20)
the elasticity of young male employment with respect to the union relative wage is -.06 and is statistically significant at the 10% level.
-
Only a minority of African workers are unionized, which qualifies the egalitarian reading of unions. (p. 22)
Seven-tenths of African wage earners are not in a union and stand to lose relatively and perhaps absolutely because of union relative wage effects.
-
Halving the union premium is estimated to raise African employment by about 2%, redistributing from union to nonunion and marginal workers. (p. 23)
reducing the union relative wage effect by half could increase Afri- can employment by about 2%, roughly equal to an expansion of one-eighth in African youth employment.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| SALDRU/World Bank 1993 national household survey (Project for Statistics on Living Standards and Development), as described in SALDRU 1994, South Africans Rich and Poor: Baseline Household Statistics | household or labour survey microdata | August-December 1993 | Sole microdata for the wage, union status, employment and time-allocation regressions: 43,974 individuals from 9,000 households. |
| Bureau of Market Research surveys of African urban union membership (as cited via Moll 1993b) | secondary literature compilation | 1975, 1980, 1985 | Provides the background trend in African urban unionization (2.5% in 1975 to 19% in 1985). |
| COSATU membership estimates as reported in Baskin (ed.) 1994, Unions in Transition: Cosatu at the Dawn of Democracy | union or movement records | 1985-1993 | Gives the union membership totals (400,000 in 1985 to 1,205,612 in 1993) and incomplete industry union shares used to context Table 1. |
| South African Industrial Conciliation Act of 1924 | legislation gazette | 1924 | Describes the legal framework for Industrial Councils whose agreements could be extended to nonunion workers and firms, motivating the spillover and understatement arguments. |
| Moll (1993a) estimates of union relative wage effects for urban manufacturing workers | secondary literature compilation | 1993 (study) | Prior work whose coverage is limited to urban manufacturing; cited to position the paper's national estimates. |
| Mwabu and Schultz (1995, 1996) earlier analyses of the same survey | secondary literature compilation | 1995-1996 | Source of the race wage gap decomposition (roughly half explained by education and location) and of the sample selection correction methodology. |
| Schultz and Mwabu 1997, Yale EGC Discussion Paper No. 776 (Appendix Tables A-1 to A-5) | author survey or fieldwork | 1993 survey data | Supplementary tabulations of employment type, education cross-tabs, regional predicted values and older-age results, referred to for details not printed here. |
Inference¶
Quantification: Econometric (causal design).
Cross-sectional OLS and quantile (10th/50th/90th) wage regressions with bootstrap standard errors on African and white men, with union-education/experience/industry interactions. Logit and multinomial logit models of youth time allocation on constructed regional union variables, followed by a counterfactual simulation of halving the union wage effect. The authors acknowledge union status is treated as exogenous, so the estimates are associational.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| African male union members earn substantially higher wages than comparable nonunion workers, with the premium largest at the bottom of the residual wage distribution (145% at 10th decile vs about 11% at 90th); white union men show no premium or a penalty. (p. 700) | Quantile and OLS regressions on 1993 SALDRU survey wage earners (2,364 African men, 653 white men), Tables 3 and 4. | Systematic data |
| Union membership compresses wage dispersion and lowers returns to schooling and experience among African union workers compared with nonunion workers. (p. 691) | Table 7 interaction models of education and experience with union status, estimated at deciles and at the mean. | Systematic data |
| Part of the union wage effect is associated with industry, and the within-industry union-nonunion differentials remain substantial for Africans despite centralized bargaining. (p. 694) | Tables 5, 6 and 8 with one-digit industry dummies and union-industry interactions on the survey data. | Partial data |
| In regions with larger union wage effects, youth employment is lower and labor force exit higher, so halving the union wage effect would raise African youth employment by roughly two percentage points. (p. 701) | Logit and multinomial logit regressions on 14 regions using constructed regional union variables (Tables 9 and 10), plus a counterfactual calculation. | Partial data |
| Unionized workers are a minority of African wage earners, so benefits accrue to a relatively privileged group while nonunion Africans may lose. (p. 700) | Union membership proportions from the 1993 survey (Tables 1 and 2). | Systematic data |
| Unemployment and nonparticipation are far higher among Africans than whites, and the African/white wage gap is a factor of five. (p. 686) | Descriptive statistics from the 1993 survey (Table 2) and the prior Mwabu-Schultz decomposition. | Systematic data |
The author's caveats¶
Single cross-section; union status is treated as exogenous with no instrument for selection into union jobs; no data on firm size, multinational status, working conditions or fringe benefits; industry is only one-digit (9 groups); the Industrial Council spillover to nonunion workers cannot be measured, so the union effects likely understate or misstate the true effect; the regional variables are predicted and the number of regions is small (14); no recent estimates of effective protection were available.
Gaps in the evidence¶
The argument would benefit from panel or longitudinal wage data to separate union effects from worker selection, plus firm-level data on size, ownership and Industrial Council coverage and enforcement, which were not collected in the 1993 survey. With only 14 regions and a cross-section, the employment and unemployment claims rest on thin regional variation and cannot establish causality. Later household and labour force surveys and employer-employee data now exist and could test these results.
Replicability: Data public and well described.