Labor Markets During Apartheid in South Africa¶

Literature › Works

Martine Mariotti (2009). Labor Markets During Apartheid in South Africa. Working Papers in Economics & Econometrics, Working Paper No. 503. School of Economics, College of Business and Economics, Australian National University.


Type	working paper
Scope	Economics and economic history of apartheid
Method	mixed
Period studied	1948–1985
Regions	South Africa (national), manufacturing sector
Themes	labour markets wages, education human capital, unions strikes, manufacturing industrial policy, business and state, politics ideology elections, origins segregation, methods data sources
Citation key	`mariotti2009a`

Argument¶

Mariotti argues that the desegregation of semi-skilled jobs in the late 1960s and early 1970s was driven by white workers' economic interests, not by capitalist pressure, state policy or rising domestic and international opposition. The Industrial Conciliation Act of 1956 let white unions and employers negotiate which jobs were reserved. As whites, especially Afrikaans speakers, gained education and moved into skilled jobs, fewer whites were left to fill semi-skilled posts. Skilled whites then opened those posts to Africans to protect the skill premium. A two-sector CES model is tested with 1970 and 1980 census micro-data, Manpower Surveys and manufacturing censuses. The tests suggest desegregation was unrelated to industry growth or African education. The paper concludes that the opening was limited to semi-skilled work, that it reflected no change in white racial ideology, and that it may have had unintended consequences such as the 1973 unrest.

Evidence and method¶

1970 and 1980 South African population censuses (individual-level micro-data, with a separate and thinner African census in 1970); Manpower Surveys 1969-1985 (271 occupations by race); manufacturing censuses 1950-1985 (23 industries, race-specific employment and wages); the Ganzeboom-De Graaf-Treiman socio-economic occupation index; Horrell's accounts of industry labour agreements; SAIRR surveys and Rand Daily Mail 1973.

Place in the debates¶

What caused the erosion of job reservation and the eventual end of apartheid (political pressure vs economic forces) (see The job colour bar, white labour and its erosion). Rejects the conventional wisdom that international and domestic political pressure from the mid-1970s ended apartheid. Argues that labour-market desegregation began earlier, at the height of apartheid, for different reasons.
Liberal vs revisionist views on capital, the state and apartheid (see Liberal vs revisionist: was apartheid functional or dysfunctional for capitalist growth?). Denies that capitalist pressure or state intervention drove desegregation. Says capital held little political power, because the return to capital was lower under segregation and capital would otherwise have forced change earlier. Gives white workers the agency.
Becker vs Heckman on whether discrimination is competed away (see The job colour bar, white labour and its erosion). Supports Heckman's point that collective worker action can sustain segregation without state backing. Adds that white workers' preference for higher incomes eventually outweighed their preference for segregation, even with explicit government support.

Findings¶

The paper's headline claim: rising white education reduced white resistance to opening semi-skilled jobs to Africans. (p. 2)

Increased educational attainment among whites reduced resistance to opening semi-skilled jobs to Africans.
Job reservation under the IC Act was arbitrary and depended on which jobs whites found attractive, not on skill. (p. 6)

Evidence from the negotiation process indicates that decisions on the skill requirement were fairly arbitrary and depended more on which whites found those jobs to be attractive than on skills per se.
The white-African education gap was very large by 1970. (p. 14)

whereas 85% of Africans had never been enrolled in high school (see Table 1).
White occupational rank rose between 1970 and 1980, consistent with upward white occupational mobility. (p. 15)

I find that the weighted average rank for white males was 46 in 1970 and 49 in 1980, demonstrating an increase in the skill level of employment (see Table 4).
Few occupations were proportionally integrated in 1969. (p. 15)

Column 1 of table 5 shows that in 1969 only 12.5 % of the occupations listed in the industry-wide Manpower survey employed Africans and whites in a proportion equivalent to that in the population.
Integration rose but remained limited by 1985. (p. 15)

This proportion grew to 27.7 % by 1985; still suggesting significant segregation in the labor market.
African access to higher-skilled occupations increased. (p. 16)

access to higher skilled occupations as demonstrated by the increase in the weighted average African SEO skill rank from 27.22 in 1969 to 34.07 in 1985.
The white skill premium stayed roughly constant between 1970 and 1980, which the author reads as skilled whites protecting their premium through desegregation. (p. 17)

therefore remained constant in the ten year period, consistent with the claim that skilled whites enabled desegregation in order to maintain the skill premium.
The 1970 time dummy in the manufacturing regression shows a jump in the growth of African production-worker share before the 1973 unrest, and industry conditions have only tiny effects. (p. 18)

In contrast, the coefficient on the dummy variable for 1970 suggests an increase in the growth rate of 6 % from 1970.
African education explains only a minor part of African occupational advancement. (p. 19)

I find that education is responsible for only 22% of the change in African occupation rank using the 1970 coefficients.
An Oaxaca-Blinder decomposition attributes a small and falling share of the occupational gap to labour-market discrimination, with most of the gap due to pre-labour-market education differences. (p. 19)

discrimination accounted for only 15.8 % of the gap in 1980, down from 26.5 % in 1970.
Desegregation reflected economic interest rather than changed white racial attitudes. (p. 20)

Changes in white characteristics do not imply a change in racial ideology; rather, white workers began to be affected by the economic constraints imposed by apartheid.

Evidence ledger¶

Sources used¶

Source	Kind	Period	How used
South African Population Census 1970 (individual-level micro data; separate, shorter census for Africans)	population census	1970	Documents white and African educational attainment, occupations, incomes and region, and supports the income and occupation-rank regressions.
South African Population Census 1980 (individual-level micro data)	population census	1980	Provides the second wave for comparing white and African occupational rank and wages, and for the stacked wage regression (Table 6) and Table 8.
South African Manpower Surveys (biennial 1969-1985 except 1979), 271 occupations by race and gender	official statistical series	1969-1985	Used to measure the share of occupations with African employment near the population share and the weighted African and white occupational skill rank (Table 5).
South African Manufacturing Census 1950-1961, 1963, 1965, 1967, 1970, 1972, 1976, 1979, 1982, 1985	industrial mining census	1950-1985	Industry-level panel (race-specific employment, wage bills, capital stock, output) for the regression of African production-worker share growth on economic conditions (Table 7).
Ganzeboom, De Graaf and Treiman (1992) Socio-Economic (SEO) Index of occupational status	secondary literature compilation	cross-national, 16 countries	Assigns skill ranks to census and manpower survey occupations, used as the occupational skill measure throughout.
Horrell (1969) / Horrell (1978) Laws Affecting Race Relations in South Africa; labour agreements (e.g. 1967 metal and engineering industry agreement)	secondary literature compilation	1967-1976	Documents the renegotiation of industrial agreements reclassifying semi-skilled jobs for non-whites in the late 1960s.
Industrial Conciliation Act of 1956 (including Section 77 job reservation)	legislation gazette	1956-	Institutional basis for white control over racial job allocation and the timing dummies used in the manufacturing regression.
Population composition figures (Ballinger 1969; Statistics South Africa Mid-year Population Figures 1980)	official statistical series	1969-1980	Gives the African share of the population against which occupational representation is benchmarked.
Rand Daily Mail (January-May 1973), South African Institute of Race Relations Annual Surveys 1967-1973, Abasebenzi 1973-1975	newspapers periodicals	1967-1975	Listed in the references, apparently as background for the January 1973 labour unrest and labour agreements; use not shown in text.
Central Statistical Service, South African Statistics (biennial 1970-1990) and South African Reserve Bank Quarterly Bulletin	official statistical series	1970-1990	Listed as macro series (likely for deflators/economic conditions); specific use not described in text.

Inference¶

Quantification: Statistical analysis.

Theory-driven argument: a stylised two-sector CES model of white skill premium motivates predicted patterns. These are checked through descriptive tabulations of education, occupational rank and African job shares, OLS regressions on stacked 1970/1980 census data and an industry panel, and an Oaxaca-Blinder decomposition. Rival explanations (growth, African education, politics) are rejected by the absence of correlation, not by a causal identification strategy.

Key claims¶

Claim	Basis	Strength
White educational attainment, especially among Afrikaans speakers, rose sharply across birth cohorts and converged toward English speakers by 1970, while African attainment stayed low. (p. 13)	1970 census tabulations by birth cohort (Tables 1-3).	Systematic data
Whites, particularly younger Afrikaans speakers, moved into higher-ranked occupations between 1970 and 1980. (p. 14)	SEO-weighted occupational ranks from the 1970 and 1980 censuses (Table 4).	Systematic data
African access to occupations, including higher-skilled ones, rose from 1969 to 1985, but segregation remained substantial. (p. 14)	Manpower Survey occupation counts weighted by SEO rank (Table 5).	Systematic data
The white skill premium was maintained between 1970 and 1980, consistent with whites opening semi-skilled jobs to protect their incomes. (p. 15)	Stacked census wage regressions (Table 6); the decline in the white-collar premium is significant only for manufacturing without education controls.	Partial data
Growth in the African share of manufacturing production workers was unrelated to output, capital intensity or wage conditions but rose after 1970. (p. 17)	OLS industry panel of manufacturing census data, 327 observations (Table 7).	Partial data
Improvements in African education explain only a small part (about a quarter) of the rise in African occupational rank in 1970-1980. (p. 18)	Census-based occupation rank regressions (Table 8) and counterfactual substitution of education differences.	Partial data
Desegregation of semi-skilled jobs was driven by white economic interests, not by capital, state or political pressure, and preceded mid-1970s unrest. (p. 19)	Inferred from the pattern of the model and regressions, plus the government's stated commitment to apartheid; no direct archival evidence of union or labour negotiations.	Asserted

The author's caveats¶

The author notes that the 1970 African census had far fewer questions and no income data, that 1970 and 1980 occupation and industry classifications had to be harmonised, and that the 1970 electronic African education coding was incorrect and had to be recoded. Income is available only in categories (midpoints used). The manufacturing regression excludes 1958 because of a data discontinuity. The African and white SEO rank changes are not directly comparable, and individual labour contracts are not analysed.

Gaps in the evidence¶

The central causal claim (white workers initiated and benefited from opening semi-skilled jobs) is not tested with direct evidence on union and employer negotiations; the industrial agreements, Department of Labour files and union records exist in archives and could be coded, as the author suggests. Evidence that whites actually gained is limited to a weak premium regression with no controls for selection or composition, and the null results on economic conditions do not rule out capital or state influence. Occupation-level, firm-level or wage data by race for semi-skilled categories before 1970, which exist partly in manufacturing and mining records, would test the timing better.

Replicability: Data described but not public.