Discrimination against colored people in South Africa¶
P.G. Moll (1991). Discrimination against colored people in South Africa. Elsevier Science Publishers B.V.
| Type | journal article |
| Scope | Economics and economic history of apartheid |
| Method | econometric |
| Period studied | 1970–1980 |
| Regions | South Africa |
| Themes | labour markets wages, inequality poverty unemployment, education human capital, methods data sources, unions strikes, business and state |
| Citation key | moll1991 |
Argument¶
Moll measures wage and job discrimination against coloureds relative to whites in 1970 and 1980. He separates younger (1-10 years' experience) from older (11-20 years) cohorts. He uses a Blinder-Oaxaca-type residual-difference decomposition extended with a multinomial logit model of occupational attainment, so the unexplained gap splits into 'job' and 'wage' discrimination. He argues that discrimination fell over the decade, that the decline was larger for younger cohorts, and that the decline is consistent with the erosion of job reservation, rising demand for skilled labour, multinational and political pressure, and the 1970s strikes and union legalisation. The paper states that job discrimination is small relative to wage discrimination. This is partly because of the broad ten-category occupational classification.
Evidence and method¶
Public-use samples of the 1970 (10 percent) and 1980 (5 percent) South African population censuses. The sample is non-agricultural private-sector employees only, white and coloured, divided into young and older experience cohorts, giving eight samples. The 1960 census, Central Statistical Services earnings series and the secondary literature (Knight and McGrath, Lipton, Schlemmer and Boulanger, Bluen, Human and Greenacre) supply context. Africans are excluded because the census collected no earnings data for them in 1960 and 1970.
Place in the debates¶
- Sources of the racial wage gap: productivity differences vs discrimination (see Racial wage gaps, wage convergence and wage determination). Moll attributes a large unexplained residual to discrimination, after controlling for schooling, experience, diploma/degree, marital status and city size. He acknowledges that omitted variables may overstate it at a point in time.
- What eroded the colour bar in the 1970s: market forces, pressure and reform vs state design (see The job colour bar, white labour and its erosion). He attributes the erosion to rising demand for skilled labour, multinational and political pressure, the Wiehahn reforms and the union wave. He does not test any specific theory of discrimination.
- Job (occupational exclusion) vs wage discrimination (see Racial wage gaps, wage convergence and wage determination). Occupational discrimination is a smaller component than within-occupation wage discrimination, in line with Banerjee and Knight (India) and a US study of the gender gap. He notes that the coarse occupational grouping may understate job discrimination.
Findings¶
-
Coloured men moved substantially up the occupational structure between 1960 and 1980, though large gaps with whites remained. (p. 3)
Coloreds made substantial progress during the period, as managers and professionals increased from 0.1 percent to 6 percent, and laborers fell from 66 percent to 33 percent of the workforce.
-
Coloureds had trade union rights while Africans were denied them until 1980, which is why coloureds are an informative comparison group. (p. 3)
Coloreds were permitted trade union rights, while Africans were denied them until 1980.
-
Real earnings by race diverged sharply between the 1960s and 1970s, with white earnings stagnating while African earnings rose fastest. (p. 4)
By contrast, during the seventies, white earnings rose a mere 2 percent, while Asian, colored, and African earnings increased bj 40 percent, 17 percent and 73 percent, respectively.
-
Moll ties the shift in relative earnings to the Durban strikes, the rise in African mining wages and the growth and legalisation of African unions. (p. 4)
the massive wave of strikes by African workers that started in Durban in 1973, the subsequent quadrupling of African mining wages, and then, in the late seventies, an expansion of African trade unions
-
Job reservation and workplace segregation were dismantled across the 1970s, which leads Moll to expect lower measured discrimination by 1980. (p. 6)
In the steel and engineering sector, practically all categories of work, including the artisan grades, were opened to non-whites by 1978
-
Statutory job reservation was formally abolished following the Wiehahn Commission. (p. 6)
Following the recommendation of the Wiehahn Commission of 1979, virtually all forms of statutory job reservation in all industries were phased out by 1984.
-
The earlier Knight and McGrath (1977) work located the main source of discrimination in exclusion from jobs and crowding into others. (p. 6)
They concluded that the major source of discrimination was the exclusion from certain jobs of non-whites whose economic characteristics were appro- priate
-
Knight and McGrath (1987), using the same method for 1976-85 and all non-white groups, found much larger declines in discrimination than Moll. (p. 7)
they concluded that discrimination against Africans, coloreds, Asians and women had decreased by 50 percent or more.
-
Job discrimination is defined as the part of the gap in occupational distribution not explained by personal characteristics. (p. 9)
JD is the part of the gross differential attributable to occupational differences not explained on the basis of personal characteristics, and may be termed job discrimi;lation.
-
Measuring change rather than level rests on the maintained assumption that omitted-variable biases were stable between 1970 and 1980. (p. 10)
The analysis of this paper proceeds on the maintamed assumption that these misspecification biases did not change between 1970 and 19S0.
-
Job discrimination is found to be a small share of the wage gap, similar to findings for Indian castes and the US gender gap. (p. 18)
This finding is similar to those o Banerjee and Knight (1985) who found that job discrimination
-
Moll's explanation of the 1970s change combines market demand for skills with foreign pressure on multinationals and political unrest. (p. 18)
Foreign pressure upon multinationals combined with growing political unrest led to attitudinal change on the part of business leaders who came to accept the necessity for advancement of non-white workers.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| 1970 population census public use sample (10 percent sample) | household or labour survey microdata | 1970 | Main microdata for estimating occupational-attainment logit and occupation-specific Mincerian earnings regressions for whites and coloreds in 1970. |
| 1980 population census public use sample (5 percent sample) | household or labour survey microdata | 1980 | Main microdata for the 1980 decompositions of white-colored earnings gap into explained, wage discrimination and job discrimination. |
| 1960, 1970 and 1980 population censuses (occupational distributions) | population census | 1960-1980 | Table 1 shows occupational distribution of white and colored males, showing coloreds' shift from laborers to managerial/professional jobs. |
| Census-based schooling attainment figures (author's computations from censuses) | population census | 1960-1980 | Documents convergence in white and colored years of schooling and tertiary qualifications. |
| CSS (various years) real earnings series by race, manufacturing, mining and all non-agricultural sectors | official statistical series | 1960-1988 | Table 2 documents real earnings by race, showing convergence in the 1970s (white earnings +2 percent vs African +73 percent). |
| Schlemmer and Boulanger (1978) interviews with managers in Durban, 1972 | secondary literature compilation | 1972 | Used to show employer-reported factors inhibiting upward mobility of non-whites (white employee resistance, separate facilities costs, customers, unions, job reservation). |
| Knight and McGrath (1977) analysis of racial wage discrimination | secondary literature compilation | 1970s | Prior evidence of job segregation and unequal pay for similar job titles within firms. |
| Bluen (1984) simulated recruitment exercise with 140 personnel practitioners | secondary literature compilation | 1984 | Evidence that whites would receive higher proposed salaries than non-whites for bogus applicants. |
| Human and Greenacre (1987) regression study using 1980 census | secondary literature compilation | 1980 | Background finding that race and gender were pivotal in income determination. |
| Knight and McGrath (1987) residual difference study 1976-1985 | secondary literature compilation | 1976-1985 | Comparison benchmark: discrimination fell 50 percent or more, against which the author's smaller decline is judged consistent. |
| Wiehahn Commission report (1979) | commission or official report | 1979 | Cited as basis for phasing out statutory job reservation by 1984. |
| Apartheid legislation: job reservation laws and Factories Act provisions on separate facilities | legislation gazette | 1950s-1984 | Establishes legislative origins of occupational and workplace discrimination and subsequent loosening. |
| Lipton (1985) and van der Horst (1976) on desegregation of employment | secondary literature compilation | 1970s | Supports claims on Vorster's 1973 announcement, government use of colored labor, and opening of steel/engineering grades. |
New data¶
Author constructed eight samples from the 1970 (10%) and 1980 (5%) census public use samples: white and colored non-agricultural, non-government employees, split into young (1-10 years) and older (11-20 years) experience cohorts, with author-computed occupation logit and 40 occupation-specific regressions.
Inference¶
Quantification: Econometric (causal design).
Oaxaca/Blinder-style residual difference decomposition (Banerjee-Knight form) of white-colored log earnings gaps into explained and discrimination components, extended with multinomial logit occupational attainment and Lee/Maddala selectivity correction, compared across 1970 and 1980 and across experience cohorts; contextual legal/anecdotal evidence supports interpretation.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Coloreds made substantial occupational progress 1960-1980 but large gaps with whites remained. (p. 3) | Table 1 from 1960, 1970 and 1980 censuses. | Systematic data |
| Real earnings converged sharply in the 1970s, as white earnings stagnated while non-white earnings rose. (p. 4) | CSS real earnings series (Table 2). | Systematic data |
| Legal and institutional barriers to colored advancement eroded in the 1970s, so discrimination was expected to fall. (p. 6) | Legislative and commission evidence, Lipton, van der Horst, Schlemmer and Boulanger interviews. | Illustrative or anecdotal |
| Measured total discrimination against coloreds declined between 1970 and 1980, especially for younger experience cohorts. (p. 18) | Census public use sample decompositions (Tables 6-7) for four cohort-year samples. | Partial data |
| Wage discrimination was larger than job discrimination in all cases. (p. 18) | Decomposition results across the four samples, with a coarse ten-category occupation classification. | Partial data |
| Omitted-variable and ability-distribution biases do not undermine the finding of declining discrimination over time. (p. 10) | Argument based on maintained assumption that unmeasured factors were stable; footnote check on education quality and business cycle. | Asserted |
The author's caveats¶
Author notes census has only earnings, not wage rates and hours; occupation categories too coarse to separate crafts from operators; independent variables limited for computation; Africans excluded since 1960/1970 earnings not collected; agricultural and government/parastatal workers omitted; unmeasured variables assumed stable over time.
Gaps in the evidence¶
The study lacks establishment-level or matched employer-employee data to verify within-job pay differences and test the job-reservation mechanism directly. Hours worked, job tenure, school quality and finer occupational detail would improve the decomposition, and only two cross-sections were used, so no panel or intermediate years. Some such data (later October Household Surveys, firm payroll records) exist, but not for 1970-80 census-level comparisons.
Replicability: Data described but not public.