Racial wage gaps, wage convergence and wage determination¶
What determined black wages and the racial wage gap, and why did black wages converge after the 1970s?
Works engaging the debate: 20
Positions¶
- Gold price, foreign labour supply, unions and influx control (Siebert, Moll, Hofmeyr)
- Discrimination vs productivity or skills explanations of the gap
- Market-determined vs administered wages (Lucas)
- Wage Board and state regulation vs market forces
Evidence base¶
The empirical core of this debate rests on a small number of econometric studies built from two types of evidence: published aggregate series and, later, individual-level samples. The time-series side is anchored by Siebert (1987a) (duplicated as Siebert (1987b) and Siebert (1987c)). It estimates simultaneous supply and demand equations for black labour in gold mining and manufacturing on roughly 38 annual observations, 1947-84. The inputs are earnings and employment from the Bulletin of Statistics and Statistical Yearbooks, censuses of 1946-1980 for the economically active population, and capital-stock benchmarks. Constructed regressors include a foreign-mineworker share from Lipton and SAIRR, a pass-law prosecution index spliced from SAIRR (1963), the Bulletin of Statistics (1947-50) and later SAIRR Surveys, and a union-density series interpolated from Clack's thesis, SAIRR and the 1984 Department of Manpower report. On this basis it documents the rise in the black/white earnings ratio after 1971, from 5% to 19% in mining and from 16% to 26% in manufacturing. Lucas (1985) is the main comparator. It fits a 3SLS derived-demand model on 1946-80 annual series for the gold, diamond, coal and other-minerals sectors. Its supply-side exogenous variables cover the labour-sending countries (population, rainfall, wages), alongside TEBA novice shares for 1981 and interviews. However, it does not name its underlying mining series. Earlier descriptive work uses the same official statistics at lower resolution. Spandau (1972) relies on Department of Labour and Wage Board reports, the 1942-43 Census of Industrial Establishments, South African Statistics 1968 and a six-sector 1970 cross-section. Nattrass (1977) combines Statistical Year Book and South African Statistics wage and employment series, Manpower Surveys No. 8 (1969) and No. 11 (1975), homeland national accounts, Bureau of Market Research (BMR) Johannesburg and Durban household studies, and a 1973 survey of manufacturing firms with 282 respondents. From these it reconstructs racial income shares, with black real wage growth of 6.6% a year in 1970-75.
The cross-sectional and microdata side is more recent. Moll (1991a) decomposes white-coloured log earnings gaps using the 1970 (10%) and 1980 (5%) census public use samples, with a multinomial logit for occupation, and adds CSS real earnings series by race. Moll (1993b) estimates Mincerian industry-premium regressions by race and occupation on 1960, 1970 and 1980 census microdata. It adds BMR urban African surveys (1975, 1980, 1985), a 1980 homelands census subsample and the 1979-82 Current Population Survey, with N = 15,386 for the clerks regression. Hofmeyr (1994a) and Hofmeyr (1994b) run human-capital regressions and decompositions on the BMR urban African household samples of 1975 and 1985. They benchmark these against the CSS employer-survey averages in South African Labour Statistics 1987 and 1990. Mariotti (2010) builds a 3-digit industry-by-year panel from the Manufacturing Census (1950-61 and selected years to 1985), with estimated wages by race and occupation. It pairs this with a count of the 28 job-reservation determinations compiled from the Government Gazette and SAIRR. The remaining works supply theory or synthesis rather than evidence. Lundahl (1982a) and Lundahl (1982b) offer comparative-static models with no data. Rex (1975) and Nattrass (1991) are historiographical reviews. Kraak (1995) provides descriptive occupational tables from CSS and education statistics plus one 1985 NTB/HSRC employer survey. Bendix (1989) is a textbook drawing on Department of Manpower strike statistics. Houghton and Dagut (1973) is a documentary anthology.
Contested evidence¶
The sharpest disagreement concerns timing and mechanism after 1975, and it arises because the two sides measure different things.
Unions versus institutional reform. Siebert (1987a) attributes most of the 1980-84 rise in real manufacturing wages to black union density, using a post-1974 coefficient of 0.083 on an interpolated membership series. Hofmeyr (1994b) reads the national SALS wage series as rising before mass unionisation in 1984 and under slack labour demand. Its BMR decomposition attributes almost all average wage growth (1.28 of 1.39% a year) to changes in worker characteristics, while unskilled reference-category wages fell about 3.3% a year. That finding directly contradicts the rising occupational wages in Knight and McGrath's graded private-sector wage survey. Hofmeyr explains the difference by sampling. In effect, Siebert's sector averages from employer returns and Hofmeyr's household samples of urban Africans are being used to answer the same question with incompatible units. Mariotti (2010) adds a third timing, dating the African wage acceleration to the 1973 strikes using census-derived wage estimates.
Influx control. Both Siebert and Hofmeyr credit relaxed influx control, but neither measures enforcement directly. Siebert uses pass-law prosecution counts as a proxy. Hofmeyr infers it from the disappearance of regional, sectoral and migrant-accommodation dummies between 1975 and 1985.
Foreign labour and the mines. On the localisation of mine labour after 1974, (Siebert (1987a)) treats falling foreign supply as a supply shock that raised wages; its foreign-labour coefficient is -0.199 with t = 1.94. Lucas (1985) reads the same shift as a demand-side preference by the mines not to rely on foreign labour. Lucas also finds monopsony in gold before 1974 and an inelastic, upward-sloping supply curve, against the labour-reserve view. Wage determination is the underlying split. Lucas finds no systematic wage discrimination by the mines. Moll (1991a) finds that wage discrimination exceeded job discrimination for coloureds. Mariotti (2010) simply assumes marginal-product pay, a premise its own critics call questionable under industrial council wage-setting.
Substitutability. Estimates conflict across sectors. Lucas (1985) finds white and nonwhite labour to be complements in mining. Mariotti (2010) finds African and white production workers to be strong substitutes in manufacturing, with elasticities of roughly 7 to 14.
Administered versus market wages. Spandau (1972) uses 1959-70 sector averages to argue that the Wage Board mattered little, that market forces dominated, and that racial differentials widened in the 1960s, rejecting Pursell's 'new deal'. Moll (1993b) uses the strong cross-race correlation of industry premia (0.70 white-African) to argue that union rent-sharing cannot explain wage structure.
Headline magnitudes. These also differ without reconciliation: Nattrass (1977) reports 6.6% a year black real wage growth for 1970-75, while Bendix (1989) asserts 5.6% over 1970-79 without a source.
Underused and missing sources¶
The influx-control mechanism invoked by Siebert, Hofmeyr and Lundahl rests on national pass-law prosecution counts. No work uses regional or district enforcement data, labour bureau records, or Section 10 qualification counts.
Manpower Surveys, 1965-94. These give occupation-by-race employment, and wages where available, so they could replace the estimated occupational wages in Mariotti (2010). They could also test whether Hofmeyr's composition effect appears in employer data. Yet only (Nattrass (1977)) draws on them, and only two issues.
Population census microdata, 1960-2011. These were last used here by Moll (1991a) and Moll (1993b) for 1960-80. Later rounds would extend the discrimination and industry-premium decompositions across the reform years.
Manufacturing census / Statistics of Production, 1917-85. Mariotti (2010) uses it only from 1950. The pre-1950 years could test (Spandau (1972))'s claim that wartime cost-of-living allowances compressed racial differentials more than Wage Board determinations did.
SAIRR Survey of Race Relations extracts, 1961-76. These cover part of the period of Siebert's spliced union and pass-law series and could be used to check those series.
Other sources are absent altogether. Chamber of Mines, TEBA and Anglo American archives would be needed to test Lucas's narrative that the 1971-78 tripling of mine wages was a Chamber decision; Lucas uses TEBA data for 1981 only. Wage Board determination-level and compliance data, and industrial council agreements, would be needed to test administered-wage effects. Firm-level matched employer-employee wage and union-recognition records would be needed to separate the union effect from contemporaneous reforms, a gap noted in nearly every econometric evidence ledger.
Strength of key claims¶
| Basis | Share of key claims |
|---|---|
| Systematic data | 17% |
| Partial data | 55% |
| Illustrative or anecdotal | 12% |
| Asserted | 16% |
Comparative figures for all debates are on the Literature overview.
Best-evidenced works¶
Siebert (1987a), Lucas (1985), Moll (1993b), Mariotti (2010), Hofmeyr (1994b), Moll (1991a)
Most-used primary sources¶
Holdings are matched by machine against the archive's file manifest; a dash means no match was found, not that the project certainly lacks the source.
Works and their positions¶
| Work | Position |
|---|---|
| Bendix (1989) | Cites Swart's finding that differentials reflect skill and productivity, but argues these in turn derive from unequal educational and labour opportunity and from a political wage policy. |
| Hofmeyr (1994a) | Rejects labour scarcity and union pressure as explanations. Argues that dismantling apartheid labour-market restrictions unified segmented markets and allowed occupational upgrading. |
| Hofmeyr (1994a) | Finds segmentation by sector, region and gender in 1975 that largely weakened or disappeared by 1985. Racial segmentation is supported only indirectly, because the data exclude Whites. Agrees with Knight and McGrath (1987). |
| Hofmeyr (1994a) | Finds sharply falling real wages for unskilled workers in the reference category, unlike Knight and McGrath, who found occupational wage levels rising in real terms. He attributes the difference to sampling. |
| Hofmeyr (1994b) | Rejects general labour scarcity and trade-union pressure, since employment growth lagged population growth and mass unionisation came only after 1984. Attributes the rise to labour-market unification after reform and to upward occupational mobility. |
| Hofmeyr (1994b) | Finds significant sectoral, regional and gender segmentation in 1975 that weakened or disappeared by 1985. Sees this as consistent with the erosion of influx control and other labour laws. |
| Hofmeyr (1994b) | Finds unskilled real wages fell sharply and occupational wage rates for males mostly fell, whereas Knight and McGrath found rising real occupational wages. Hofmeyr attributes the difference mainly to sampling. |
| Houghton and Dagut (1973) | Presents Viljoen and Steenkamp (cautious, incremental) against Katzen (statutory minimum wage). |
| Kraak (1995) | Rejects the neoclassical view (e.g. Human and Greenacre) that discrimination is a market distortion caused by racist policy; sees segmentation as a structural feature of South African capitalism shaped by state, capital and labour strategies. |
| Kraak (1995) | Reform diluted racial occupational barriers and created a unionised African semi-skilled core, but discrimination was reproduced through informal and micro-level mechanisms and has had 'little effect' on the racial occupational structure at the top. |
| Lucas (1985) | Finds no systematic evidence of wage discrimination beyond job barriers. Mines acted as monopsonists, and discrimination operates through the job colour bar in the production function. |
| Lundahl (1982a) | Treats white wages as endogenous, set by collective bargaining within restrictions on black labour supply. Job reservation creates conflicts between skilled and unskilled whites and between agriculture and industry. |
| Lundahl (1982b) | Treats wages as endogenous, with unskilled wages tied to reserve productivity or to influx limits, and white wages determined within restrictions on the supply of black labour. This is closer to Knight than to Porter. |
| Mariotti (2010) | 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. |
| Moll (1991a) | 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. |
| Moll (1991a) | 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. |
| Moll (1993b) | Dissents. The industry wage structure cannot serve as proof of efficiency-wage or rent-sharing models, since South African African workers with no bargaining power show the same premia as whites. |
| Moll (1993b) | Argues Krueger and Summers' rejection of it was unconvincing, but finds that the South African natural experiment (non-union African workers, homelands) still shows large premia, so unions plus threats cannot be the main cause. |
| Moll (1993b) | Finds no systematic exclusion of Africans from high-wage industries; the 'white' dummy has a small effect on access to industry premia. |
| Moll (1993b) | Notes the South African data do not reject expense preference (Heywood) and cannot test it for lack of firm-size data; it regards expense preference and internal equity as the only surviving candidates. |
| Nattrass (1977) | Contradicts the Spandau/McGrath finding of near-constant White share, arguing that 1970-75 saw the first redistribution toward Blacks. |
| Nattrass (1991) | Reports the 1950s-70s division between those wanting minimum wages (Katzen, Burrows) and those saying wages tracked productivity (Spandau, Viljoen). She does not adjudicate. |
| Rex (1975) | Rex questions attributing wage changes to a reserve crisis, since the crisis was recorded from the late 1940s. He calls for analysis of what happened in the reserve economy after 1950. |
| Siebert (1987a) | Multi-causal. Gold prices and reduced foreign labour mattered in the 1970s, but unions explain most of the real manufacturing wage rise after 1980. The 'employer sensitivity' explanation is dismissed as lacking a measurable metric. |
| Siebert (1987a) | Treats this as an empirical question and estimates supply and demand. He finds an upward-sloping supply of labour to mining and downward-sloping demand, so market forces are relevant. |
| Siebert (1987a) | Stricter pass-law enforcement lowers black wages in both mining and manufacturing, so relaxation or abolition would raise them. The mechanism is that mobility restrictions cause misallocation. |
| Siebert (1987b) | Multi-causal. Gold prices and falling foreign labour mattered for mining in the 1970s, but unions explain most of the real manufacturing wage rise after 1980. Relaxed influx control also helped. Gold prices alone are not the full answer. |
| Siebert (1987b) | Treats this as an empirical question. Finds an elastic mining labour supply as Lucas does, but with a supply elasticity above one (1.32) where Lucas found it below one, and keeps demand in the wage system. |
| Siebert (1987c) | Multi-causal. The gold price and reduced foreign labour mattered in the 1970s, but unions were the main driver of manufacturing wages after 1980, and pass-law relaxation also contributed. |
| Siebert (1987c) | Finds an elastic supply to mining, as Lucas does, but with an elasticity above unity (1.32) where Lucas found it below. Treats the wage as market-determined rather than purely administered. |
| Siebert (1987c) | Influx control depressed wages in both sectors by reducing labour mobility, so its abolition would raise wages without adverse employment effects. |
| Spandau (1972) | Rejects it. Wage Board differentials are not indicators of government policy, the earnings gap widened, and the Board never investigated mining. |
| Spandau (1972) | Holds that market forces, sectoral 'payability' and White bargaining strength shaped differentials far more than Wage Board determinations. |
| Spandau (1972) | Sees purposeful wage restraint consistent with government anti-inflation and growth policy, not aimless drift. |