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 | 1975–1993 |
| Regions | South Africa, four provinces and ten homelands (14 regional labour markets) |
| Themes | unions strikes, labour markets wages, inequality poverty unemployment, transition post apartheid, methods data sources |
| Citation key | schultz1998 |
Argument¶
Using a 1993 national household survey, the authors ask whether South African unions widen or narrow wage inequality and what they do to employment. Quantile regressions show that the union wage premium among African men is largest at the bottom of the residual wage distribution and almost vanishes at the top. Among white men it turns negative over the upper 70% of the distribution. Unions compress wages within the union sector and flatten returns to schooling and experience. Seven-tenths of African wage earners are non-union and may lose from the premium. Regional estimates suggest a smaller union premium would raise African youth employment modestly, though the unemployment result runs against the Harris-Todaro-Calvo prediction.
Evidence and method¶
SALDRU/World Bank national probability sample of 43,974 individuals in 9,000 households (360 clusters), interviewed August-December 1993; 2,364 African male and 653 white male wage earners in the wage regressions. Union density figures from Bureau of Market Research surveys (via Moll 1993b) and COSATU estimates (Baskin 1994). Secondary literature on Industrial Councils.
Place in the debates¶
- Are unions an egalitarian force or a source of insider-outsider inequality in South Africa? (see Black trade unions: growth, wage effects, Wiehahn and strategy). Mixed. Unions compress wages among members and narrow the racial gap within the union sector. Because most African workers are non-union, the authors say the premium may disadvantage them and the marginalised poor.
- Do union wage premia raise unemployment (Harris-Todaro-Calvo)? (see Black trade unions: growth, wage effects, Wiehahn and strategy). Regional estimates give small negative employment effects (elasticity of about -.06 for young men), but unemployment falls where the premium is larger, the opposite of the Harris-Todaro-Calvo prediction. Higher premia seem to push youth out of the labour force instead.
- Role of centralised bargaining and Industrial Councils in setting wages (see Black trade unions: growth, wage effects, Wiehahn and strategy). Industry controls absorb much of the union premium, which the authors take to reflect Industrial Councils or spillover of administered wages. Their data cannot measure firm size or council enforcement, so they suggest the union effect is probably understated.
Findings¶
-
South African wage inequality is among the world's highest, with a fivefold African-white wage gap. (p. 2)
Few countries have higher wage inequality than South Africa, where wages of African and white workers differ by a factor of five.
-
Union density among urban African workers rose rapidly after unions were legalised. (p. 2)
2.5% of African workers in urban areas of South Africa were unionized in 1975, 5.5% by 1980, and 19% by 1985
-
Union membership is highest in mining and substantial in manufacturing. (p. 8)
The share of union members is highest in mining (73%); substantial in manufacturing (48%)
-
The mean union wage premium for African men is about 60%, whereas white men show a small negative differential. (p. 10)
implying that an African union worker receives a wage that is 60% higher (exp (.468) - 1.0) than that of a nonunion worker with the same observed characteristics.
-
Among white men the union premium turns negative over most of the residual wage distribution. (p. 11)
White men in the top 70% of the distribution of wage residuals (not reported) receive a lower wage if they are in a unionjob
-
Unions flatten returns to schooling, which may weaken private incentives to acquire education. (p. 13)
he would have less private incentive to obtain additional schooling, insofar as it would increase his wage.
-
Wage dispersion is lower within the union sector of each industry. (p. 16)
Wage inequality is generally less within the union sector than within the nonunion sector of each industry.
-
The regional employment elasticity with respect to the union premium is small for young African men. (p. 20)
hence, the elasticity of young male employment with respect to the union relative wage is -.06 and is statistically significant at the 10% level.
-
Unemployment moves opposite to the Harris-Todaro-Calvo prediction. (p. 22)
This observed pattern is the opposite of that predicted by the Harris-Todaro-Calvo models of labor market sectoral distortions caused by unions or government wage regulations.
-
Most African workers are outside unions and union members are a small share of the African population. (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.
-
African union workers are a small minority of the labour force. (p. 22)
African union workers are only one-seventh of the African labor force, and only 6% of the African population age 16-65
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 source for all wage, union-status, and time-allocation regressions (OLS, quantile, logit, multinomial logit). |
| Bureau of Market Research surveys of African urban workers, as reported in Moll (1993b) | secondary literature compilation | 1975-1985 | Provides background on the growth in African unionization from 1975 to 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 national union membership totals and the share of workers unionized, and a comparison for industry union shares. |
| Industrial Conciliation Act of 1924 and the Industrial Council system | legislation gazette | 1924-1993 | Institutional background on how Industrial Councils set and extend wage floors to nonunion firms and workers, motivating the interpretation of union and nonunion wage gaps. |
| Mwabu and Schultz (1995) analysis of the same survey on wage premia by race, gender, education and location | secondary literature compilation | 1993 survey | Supports the claim that about half of the African-white wage gap is accounted for by education and location differences, and the earlier sample-selection methodology. |
| Moll (1993a) study of union wage effects in urban manufacturing | secondary literature compilation | Pre-1993 (period not stated in this text) | Prior estimates of union effects, restricted to urban manufacturing, against which the national sample is positioned. |
| Schultz and Mwabu (1997), Appendix Tables A-1 to A-5, Economic Growth Center Discussion Paper No. 776 | author survey or fieldwork | 1993 survey | Supplementary cross-tabulations of union shares by employment type and education, regional predicted variables, and older-age and youth sample statistics not reported in the article. |
Inference¶
Quantification: Econometric (causal design).
Cross-sectional OLS and quantile (10th, 50th, 90th percentile) Mincerian wage regressions with bootstrap standard errors, comparing union and nonunion workers with interaction terms. Logit and multinomial logit models of time allocation among African youth use region-level predicted union variables with Huber standard errors. The authors call the approach descriptive, treat union status as exogenous, and have no identifying instrument or panel.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Unions raise African male wages substantially conditional on observables, by about 60% at the mean and far more at the bottom of the residual distribution. For white men the mean effect is slightly negative and insignificant. (p. 688) | Table 3 and Table 4 OLS and quantile regressions on the 1993 SALDRU survey. | Partial data |
| Union wage gains are inversely related to position in the residual wage distribution, so unions compress wage inequality among union workers. (p. 689) | Decile coefficients in Tables 3-6, with bootstrap t-ratios (20 repetitions). | Partial data |
| Returns to education and experience are lower for union than nonunion African workers, so unions flatten the wage structure with respect to skill. (p. 690) | Table 7 interactions of union status with education spline and experience. | Partial data |
| Union-nonunion differentials remain substantial within industries for Africans but are smaller for whites, possibly reflecting differences in firm size and capital intensity or Industrial Council enforcement. (p. 694) | Table 8 industry interactions; the interpretation about firm size is conjectural because firm size is not in the data. | Illustrative or anecdotal |
| Larger regional union wage effects are associated with lower employment of young Africans and with lower unemployment and labor force exit, contrary to Harris-Todaro-Calvo predictions. (p. 698) | Tables 9 and 10: logit and multinomial logit on 14 regions with three constructed regional variables, so effective identification rests on few regional observations. | Partial data |
| Halving the union wage effect would raise employment of African youth by a couple of percentage points, with a redistribution from union workers to nonunion and nonparticipating poor. (p. 701) | Counterfactual extrapolation from Table 9 logit coefficients applied to mean sample values. | Illustrative or anecdotal |
| Only a minority of African workers are unionized, so most nonunion Africans may lose relatively from union wage effects. (p. 700) | Union membership shares in Tables 1 and 2 from the 1993 survey. | Partial data |
The author's caveats¶
Single cross-section; union status cannot be endogenized and no exclusion restriction is available. The survey lacks firm size, multinational status and fringe benefit/working condition data. Industry is aggregated to nine groups. Spillovers to nonunion workers from Industrial Councils cannot be measured, so union effects are probably understated. Wage-earner selection correction did not matter much. Sample selection of hours cannot be handled. Some regions have no union workers in the youth samples. Regional variables are predicted, and bootstrap uses only 20 repetitions.
Gaps in the evidence¶
The argument needs panel or matched employer-employee data with firm size, ownership, and Industrial Council coverage to separate union effects from selection and administered-wage spillovers; such firm-level data were not in this survey but later labour force surveys and firm surveys might supply them. The labor-supply claims rest on 14 regional aggregates built from the same survey, so they could be tested better with repeated cross-sections or panel data over the 1980s-90s as unions grew. Archival Industrial Council agreements and enforcement records existed at the time and could document wage floor coverage.
Replicability: Data described but not public.