Estimating the elasticity of substitution in the South African manufacturing industry during Apartheid, 1950 – 1985¶
Martine Mariotti (2010). Estimating the elasticity of substitution in the South African manufacturing industry during Apartheid, 1950 – 1985. Research School of Economics, ANU (draft, 'please do not cite'). Australian National University.
| Type | working paper |
| Scope | Economics and economic history of apartheid |
| Method | econometric |
| Period studied | 1950–1985 |
| Regions | South Africa |
| Themes | labour markets wages, manufacturing industrial policy, unions strikes, methods data sources, business and state, comparative theory |
| Citation key | mariotti2010 |
Argument¶
Mariotti tests the claim by Porter, Lundahl and Lowenberg that job reservation unravelled because rising white skills turned Africans and whites from substitutes into complements in manufacturing. Using a CES production function and the first-order conditions, she regresses the log African/white wage ratio on the log employment ratio in 3-digit manufacturing census data. She finds that African and white workers stayed substitutes throughout apartheid, especially in production jobs, with no clear decline in substitutability. She reads this alongside the historical literature on the 'floating' colour bar: Africans could already do semi-skilled tasks, and reclassifying occupations opened those jobs to them. Any complementarity probably lies between sectors or between very highly skilled whites and other workers, which the data cannot test.
Evidence and method¶
South African manufacturing censuses 1950-1985 (18 census years; 143 4-digit and 23 3-digit ISIC sectors) with employment by race and occupation group and wages by race only. Production and administrative wages are estimated with a Goldin-Sokoloff-type regression. Also uses secondary literature, a case study of the metal and engineering industrial council agreement, and a table of 28 job reservation determinations from Government Gazettes and SAIRR.
Place in the debates¶
- Did rising white skills turn African and white labour from substitutes into complements, ending job reservation (Porter/Lundahl/Lowenberg model)? (see The job colour bar, white labour and its erosion). Disputes it for manufacturing: the estimates show continued high substitutability, with no evidence of a decline for production workers.
- How rigid was job reservation in practice (liberal vs revisionist readings of the colour bar)? (see The job colour bar, white labour and its erosion). Sees it as flexible: unions and employers 'floated' the colour bar by recoding and downgrading occupations, so Africans crept up the job ladder, consistent with Lipton and Feinstein.
- Causes of the post-1973 jump in African real wages (see Racial wage gaps, wage convergence and wage determination). Suggests it may be a wage response to Africans' higher productivity in semi-skilled work, and notes that wages rose sharply after the 1973 Durban strikes.
Findings¶
-
African and white production workers in manufacturing remained substitutes throughout apartheid, with elasticity estimates far above 1. (p. 2)
African and white workers in the South African manufacturing industry remained substitutes throughout the Apartheid period with an elasticity of substitution between 7 and 14 %.
-
Production-worker elasticity estimates range from about 7 to 14 depending on fixed effects, and are higher than for all workers. (p. 19)
Estimates of the elasticity range from 7.1 to 14.4 depending on the inclusion or exclusion of year and industry fixed effects.
-
For all workers combined, there is some decline in elasticity from the mid-1970s, but it is constant until 1970. This is likely biased downward by pooling production and administrative jobs. (p. 18)
The elasticity decreases from 3.9 in 1950 to 2.1 in 1985, however, the elasticity appears to be constant from 1950 until 1970.
-
For production workers, only the 1985 interaction is significant, and no decline in substitutability is found. (p. 19)
Only one of the year – employment interaction terms is significant, that for 1985. These results suggest that Africans and whites in production remained substitutes throughout the Apartheid period.
-
Administrative workers show the highest, effectively infinite, substitutability. The author attributes this partly to very few Africans holding such jobs. (p. 20)
Elasticities range from 19.4 to 165.1, however all the estimates larger than 48 are insignificant suggesting that African and white administrative workers were infinitely substitutable.
-
African wages rose slowly until 1973 and then sharply after the 1973 strikes, while white wages became more volatile from the mid-1970s. (p. 15)
African wages increase slowly until 1973 at which point they increase sharply following the 1973 strike action in Durban, East London and Johannesburg.
-
The 1967 metal and engineering agreement let unions retain closed-shop skilled categories while opening telescoped, downgraded jobs to Africans. Employers then preferred cheaper Africans when vacancies arose. (p. 9)
This meant that even though no white person could be replaced or displaced by an African, when vacancies opened up employers always preferred to hire cheaper Africans.
-
Formal job reservation was used sparingly: 28 determinations, many amended, affecting few workers directly. (p. 7)
All in all, the government made 28 job reservation determinations as shown in Table 1. Table 1 also shows that many of these were amended and that few workers were ultimately directly affected by the determinations.
-
The erosion of job reservation probably reflects white semi-skilled workers losing control as their share of the industry fell, and Africans possibly already did the same tasks under different titles. (p. 21)
It is further possible that even prior to the decline of job reservation African workers executed the same tasks as white workers did but under different job titles
-
Omitting fixed effects biases the all-worker elasticity estimate; including year and industry dummies lowers it from 4.5 to 2.2. (p. 18)
The inclusion of both year and industry dummies reduces the elasticity to 2.2 in column V from 4.5 in column I.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| South African Manufacturing Census 1950-1961, 1963, 1965, 1967, 1970, 1972, 1976, 1979, 1982, 1985 (143 4-digit and 23 3-digit ISIC sectors) | industrial mining census | 1950-1985 (18 census years, excluding 1954 and 1955) | Supplies employment by race, gender and occupational category and wages by race, which are the basis for estimating the African-white elasticity of substitution. |
| Government Gazette 1961-1979 (Job Reservation Determinations) | legislation gazette | 1961-1979 | Source for Table 1 listing the 28 job reservation determinations, with jobs affected, exemptions and cancellations. |
| South African Institute of Race Relations publications (1958-1979) | secondary literature compilation | 1958-1979 | Second source for the Table 1 summary of job reservation determinations and the number of jobs lost. |
| Industrial Conciliation Act 1924 and 1956 (incl. Section 77 Industrial Tribunal) | legislation gazette | 1924, 1956 | Establishes the institutional mechanism of industry agreements and job reservation determinations that restricted African employment. |
| Group Areas Act (1950), pass laws, Factories, Machinery and Building Work Act (1941), Apprenticeship Act (1944) | legislation gazette | 1941-1950 | Describes the legal restrictions on African residence, mobility and training that constrained employment in manufacturing. |
| National Industrial Council for the Iron, Steel, Engineering and Metallurgical industries agreement, 1967 | employer association records | 1967 | Case study showing how occupation recoding and wage categories opened semi-skilled jobs to Africans while formal job reservation was retained. |
| Riekert and Wiehahn Commissions (1979) | commission or official report | 1979 | Cited, via Feinstein, as having recommended abolition of job reservation, which the government accepted. |
| Mariotti (2008) PhD dissertation, UCLA | secondary literature compilation | Apartheid era | Relied on for the finding that the industries subject to Job Reservation Determinations faced labour scarcity. |
| Horrell (1969), South Africa's Workers (SAIRR) | secondary literature compilation | 1969 | Source for white worker complaints to the Minister of Labour about being replaced by Africans in metal and engineering. |
| Feinstein (2005) data on white employment shares by sector | secondary literature compilation | twentieth century | Justifies the choice of manufacturing: its share of white blue-collar jobs, versus mining and agriculture. |
New data¶
Panel built from the manufacturing censuses: 23 3-digit ISIC industries over 18 census years 1950-1985, with employment by race and occupation and wages by race, plus estimated production and administrative wages by race for each 3-digit sector (373 observations; 370 for administrative workers). Estimated using the Goldin and Sokoloff method.
Inference¶
Quantification: Econometric (causal design).
Structural CES production function. The first-order condition gives a log wage-ratio on log employment-ratio regression across 3-digit industries and years, with year and industry fixed effects and year interactions. Production and administrative wages are not published, so they are estimated via an OLS decomposition of the wage bill (Goldin-Sokoloff style) and standard errors are bootstrapped. Institutional narrative on job reservation is drawn from secondary literature and a single industry case study.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| African and white production workers in manufacturing remained substitutes throughout Apartheid, with elasticity of substitution of roughly 7 to 14. (p. 19) | OLS on the census-based 3-digit by year panel (Table 3), using estimated production wages and bootstrapped standard errors. | Partial data |
| There is no evidence of a decline in substitutability for production workers; only the 1985 interaction is significant. (p. 19) | Specification VI interaction terms in Table 3, with mostly insignificant coefficients. | Partial data |
| For all workers combined, elasticity is lower (about 2 to 5) and declines from 3.9 in 1950 to 2.1 in 1985, but this is likely biased by pooling production and administrative jobs. (p. 18) | Table 2 regressions on total employment and average wages from the census. | Partial data |
| African and white administrative workers appear highly or infinitely substitutable, possibly because very few Africans held administrative jobs. (p. 20) | Table 4 regressions using estimated administrative wages; large elasticities are mostly insignificant. | Partial data |
| Job reservation was flexible: employers and white unions recoded occupations (the 'floating' Colour Bar), opening semi-skilled jobs to Africans, as shown in the metal and engineering industry. (p. 9) | Case study of the 1967 industrial council agreement, Table 1 determinations and exemptions, and secondary literature (Horrell, Lewis, Crankshaw). | Illustrative or anecdotal |
| Only 28 job reservation determinations were made, many were amended, and few workers were directly affected. (p. 7) | Table 1 compiled from the Government Gazette and SAIRR. | Systematic data |
| African wages rose sharply after 1973 following strike action, while white wage growth slowed after job reservation was removed. (p. 15) | Estimated wage series by race and sector (Figures 1-8) derived from the manufacturing census. | Partial data |
The author's caveats¶
Production and administrative wages are not published and have to be estimated from total wage bills, so the dependent variable is generated and standard errors are bootstrapped. Fixed-effect estimates at the 4-digit level are inconsistent, so analysis is aggregated to 23 3-digit sectors. Census data for later years are not available by race, and the census does not allow analysis of substitution between different skill levels, so complementarity between highly skilled whites and semi-skilled Africans cannot be tested.
Gaps in the evidence¶
The argument relies on estimated rather than observed wages by occupation and race, and it assumes workers are paid their marginal product with no wage discrimination. That assumption is questionable under an administered wage system with industrial council agreements. Establishment-level or firm payroll records, industrial council agreement wage schedules, or occupational data by skill and race would allow direct tests of skill complementarity; some of these exist in archives and in SAIRR and Department of Labour publications. The institutional account of job floating rests on one industry case and secondary literature, with little systematic evidence on how many tasks were reclassified.
Replicability: Data described but not public.