Desegregating Labor Markets in the Apartheid Regime¶
Martine Mariotti (2008). Desegregating Labor Markets in the Apartheid Regime. Job Market Paper, UCLA Department of Economics (January 2008). UCLA Department of Economics.
| Type | working paper |
| Scope | Economics and economic history of apartheid |
| Method | mixed |
| Period studied | 1948–1985 |
| Regions | South Africa |
| Themes | labour markets wages, origins segregation, education human capital, manufacturing industrial policy, politics ideology elections, business and state, unions strikes, methods data sources, comparative theory |
| Citation key | mariotti2008 |
Argument¶
Mariotti argues that job reservation in manufacturing began unravelling from the late 1960s and early 1970s, before the 1973 strikes, the intensified sanctions pressure of the late 1970s, or improvements in black education. The cause was the government's own success in raising the education and occupational status of poor Afrikaans-speaking whites. In her two-sector model, once the pivotal white worker became skilled, skilled whites gained from opening complementary semi-skilled jobs to blacks and no longer needed protection. Because job reservation was determined industry by industry, desegregation proceeded industry by industry, led by high-skill industries. Census and Manpower Survey regressions, Oaxaca-Blinder decompositions and manufacturing-census wage ratios are used to show that black gains were not driven by black human capital. She concludes that the change reflected white preferences and not a softening of apartheid ideology.
Evidence and method¶
Manpower Surveys 1969-1985 (271 occupations, ranked with the Ganzeboom-De Graaf-Treiman SEO index); manufacturing censuses 1950-1985 (23 industries, production wages estimated by a Goldin-Sokoloff procedure); 1970 and 1980 population census micro data (excluding blacks in homelands); SAIRR annual surveys; Hansard quotations; Rand Daily Mail; Horrell's legal compendium; secondary histories (Lipton, Feinstein, Thompson, Terreblanche).
Place in the debates¶
- What caused the relaxation of job reservation (labour demand vs civil unrest vs international pressure vs black education vs capital lobbying) (see The job colour bar, white labour and its erosion). Rejects each as the main cause. Labour demand is necessary but insufficient. She argues the decisive factor was rising white, especially Afrikaner, education and occupational status, which shifted political power within industries.
- Liberal vs revisionist views on apartheid and capitalism (see Liberal vs revisionist: was apartheid functional or dysfunctional for capitalist growth?). Capital is not treated as the driving political force. Because the return to capital was lower under segregation, yet reform came late, she infers that political power lay with white workers, not capitalists.
- Why apartheid ended: external pressure and unrest vs internal dynamics (see Why apartheid ended: reform, transition and elite bargaining). Argues that internal labour-market change preceded and was independent of the 1973 unrest and 1970s sanctions. These later events may still have contributed to apartheid's ultimate collapse.
- Fear of revolution as elite motive for reform (Acemoglu and Robinson) (see Why apartheid ended: reform, transition and elite bargaining). Finds no evidence that fear of revolution motivated the relaxation of job reservation, though it may have mattered for the later collapse.
- Competitive markets erode discrimination (Becker) vs workers' ability to enforce it politically (see The job colour bar, white labour and its erosion). Sides with those who stress worker-enforced segregation: discrimination persists while the politically dominant group benefits from it, and ends when its interests change.
Findings¶
-
Headline claim: labour market change in the early 1970s substantially restructured manufacturing labour and narrowed the racial wage gap, before the 1980s events usually credited. (p. 1)
Increased educational attainment among whites reduced resistance to opening semi-skilled jobs to blacks, the result was a massive restructuring of the manufacturing labor force that narrowed the black-white wage gap.
-
Black occupational skill rank in the Manpower Surveys rose between 1969 and 1985, indicating movement from unskilled to semi-skilled blue-collar work. (p. 6)
the black rank increased from 27.22 to 34.07 suggesting that black employment moved from predominantly unskilled labor to slightly higher skilled blue collar jobs
-
Black production wages rose relative to white wages from the late 1960s, across industries. (p. 7)
Black real wages in production began to increase relative to white by the late
-
Afrikaner educational attainment rose sharply across cohorts, underpinning the shift in white interests. (p. 22)
by 1970 around 60% of the sample attained at least the third year of high school and over 20% received a high school completion certificate.
-
Black education remained very poor, so it cannot explain black access to semi-skilled jobs. (p. 22)
whereas 85% of blacks never even got to high school
-
The 1973 strikes are rejected as the cause of relaxed job reservation on grounds of timing. (p. 13)
However, the strike action took place after the job reservation changes that started taking place in the late
-
White unions shifted from violent opposition to accepting the relaxation of job reservation, focusing on wage rates and union access. (p. 11)
In contrast to the violent response of white unions to the employment of blacks in the first half of the century, they now accepted the relaxation of job reservation
-
The government did not soften its apartheid stance during the relaxation of job reservation. (p. 11)
Yet there is no evidence of a softening of the government
-
Capital did not hold the political power to end job reservation, since it would have forced reform earlier. (p. 18)
If capital held any political power then it would have forced transformation onto the economy much earlier than actually happened.
-
Black education changes explain only about a quarter of the rise in black occupational rank. (p. 27)
I find that in 1970 education is responsible for only 24% of the change in black occupation rank.
-
The Oaxaca-Blinder decomposition shows the unexplained (discrimination) component of the black-white occupational gap shrank between 1970 and 1980. (p. 27)
there was a decline in the unexplained component of the difference between black and white occupational attainment between 1970 and 1980 from 18.3% to 10.5%
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| South African Manpower Surveys (Department of Labor/Manpower), biennial 1969-1985 except 1979 | official statistical series | 1969-1985 (biennial, no 1979) | Establishes rising share of occupations majority-black and rising black/white weighted occupational skill rank; also used in the equation 4 regression on black share changes. |
| South African Manufacturing Census, 1950-1985 | industrial mining census | 1950-1961, 1963, 1965, 1967, 1970, 1972, 1976, 1979, 1982, 1985 | Source of black and white employment and wage bill data used to build production wage ratios showing black wages rising relative to white from the late 1960s. |
| South African Population Censuses 1970 and 1980 (micro data) | population census | 1970 and 1980 | Micro data used to document white and black education and occupational attainment, run wage/occupation regressions and the Oaxaca-Blinder decomposition. |
| Socio-Economic Index of Occupational Status (SEO), Ganzeboom, De Graaf and Treiman 1992 | secondary literature compilation | 1992 index (16 countries, 73,901 men) | Provides occupational skill ranks assigned to South African Manpower Survey and census occupations, the core measure of skill used throughout. |
| South African Reserve Bank, Quarterly Bulletin (real GDP) | national accounts macro | 1950-1980 | Real GDP and growth used to show no obvious macroeconomic change around the period of labour market adjustment. |
| South African Institute of Race Relations (SAIRR) Annual Surveys/Reports 1967-1973 | secondary literature compilation | 1967-1973 | Cited for firms and workers negotiating job reservation concessions, continued new determinations, and ministerial statements. |
| Hansard (as quoted in SAIRR Annual Report 1972) | parliamentary debates | c.1972 | Minister of Finance statement used to show government remained committed to apartheid while allowing job reservation concessions. |
| Rand Daily Mail, January-May 1973 | newspapers periodicals | January-May 1973 | Used to show white unions accepted relaxation of job reservation but sought equal rates for the job. |
| Abasebenzi (Workers' Newspaper), 1973-1975 | newspapers periodicals | 1973-1975 | Cited for claim that 1973 strikes contributed to black real wage rises and the 1973 amendment to the Bantu Labor Regulation Act. |
| Horrell 1978, Laws Affecting Race Relations in South Africa (job reservation determinations) | secondary literature compilation | to end of 1976 | Source on Industrial Conciliation Act 1956 procedures and the pattern of job reservation determinations across manufacturing industries. |
| Legislation: Industrial Conciliation Act 1956 and other labour/apartheid statutes (Appendix III timeline) | legislation gazette | 1911-1970 | Timeline of pre-apartheid and apartheid legislation establishing the job reservation framework. |
| Statistics South Africa Mid-year Population Figures 1980; Central Statistical Service South African Statistics; special wage reports | official statistical series | 1970-1990 | Listed as general data references supporting population and wage background, not used distinctly in the text. |
| Fedderke, de Kadt and Luiz 2000 education data | secondary literature compilation | 1910-1993 | Used to show divergence between black and white education and lower black school quality. |
| Census questionnaire 1970 (used to recode black education data) | population census | 1970 | Used to correct errors in electronic coding of 1970 black education. |
New data¶
Author collected the South African manufacturing census data (1950-1985, 23 industries, employment and wage bills by race) using research grants; also harmonized 1970 and 1980 census micro data occupations, industries and education into comparable categories.
Inference¶
Quantification: Statistical analysis.
Theory-driven argument: a two-sector model generates predictions (decentralized, industry-by-industry desegregation; changing returns to skill), tested with descriptive statistics on census cohorts, OLS regressions, an Oaxaca-Blinder decomposition and a manpower-survey panel regression; alternative hypotheses are dismissed mainly by timing arguments from secondary literature.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Black access to semi-skilled jobs and the black occupational skill rank rose from 1969 to 1985, with the share of mostly-black occupations growing. (p. 6) | Manpower Surveys, SEO-ranked occupations (Tables 1-2). | Systematic data |
| Black production wages rose relative to white from the late 1960s across manufacturing industries, before the 1973 strikes. (p. 7) | Manufacturing census wage bills and estimated production wages (Figures 1-3, Appendix II). | Partial data |
| Afrikaans speaking whites caught up with English speakers in education and occupational rank between 1970 and 1980. (p. 24) | 1970 and 1980 census micro data, cohort tables and SEO weighted ranks (Tables 4-8, Figs 5-7). | Systematic data |
| Industries with higher proportions of skilled white workers transformed more (larger gains in average black rank), consistent with industry-by-industry job reservation relaxation. (p. 24) | Census 1970/1980 industry-level scatter plots (Figures 8-9) with ad hoc explanations of outliers. | Partial data |
| Changes in black education explain only about a quarter of the rise in black occupation rank; most is unexplained. (p. 27) | Census-based regression of occupation rank on education (Tables 10-11). | Partial data |
| The unexplained (discrimination) component of the black-white occupation rank gap fell from 18.3% to 10.5% between 1970 and 1980. (p. 27) | Oaxaca-Blinder decomposition on 1970 and 1980 census data (Table 12). | Partial data |
| Relaxation of job reservation was not caused by civil unrest, international pressure, black education, emigration or capital lobbying, on timing grounds. (p. 12) | Chronology drawn from secondary literature (Thompson, Hofmeyr, Terreblanche, Fedderke et al.) compared with the timing of the occupational shifts. | Illustrative or anecdotal |
| Return to occupation rank rose for skilled occupations (rank 66+) and fell for semi-skilled occupations between 1970 and 1980, as the model predicts. (p. 26) | OLS wage regression with interaction terms on white male census data (equation 1); coefficients not tabulated. | Asserted |
The author's caveats¶
Notes that census excludes blacks in homelands (claims this biases results toward the hypothesis), that black income is unavailable in 1970, income is in categories (midpoints used), 1970 black education coding was erroneous and recoded, occupation and industry classifications had to be harmonized, and the manufacturing census does not separate production wages (estimated). Also notes comparing rank gains between races is difficult.
Gaps in the evidence¶
The argument lacks direct evidence on the political process: Industrial Tribunal and Industrial Council determinations, union and party records, and Hansard debates were not systematically examined, so the white-preference mechanism is inferred from timing and theory. Only two census years and a short manpower-survey panel are used, so pre-1970 trends in white education and black occupations are not observed in micro data, and the regression results rely on ad hoc interpretation of industry outliers. Archival data on job reservation determinations by industry and year, and firm-level data, exist and could test the industry-by-industry hypothesis more directly.
Replicability: Data described but not public.