The Rise in African Wages: 1975-1985¶
Hofmeyr, J. F. (1994). The Rise in African Wages: 1975-1985. South African Journal of Economics 62(3).
| Type | journal article |
| Scope | Economics and economic history of apartheid |
| Method | econometric |
| Period studied | 1975–1985 |
| Regions | South Africa (main urban areas), Johannesburg, Cape Town |
| Themes | labour markets wages, bantustans removals influx, education human capital, inequality poverty unemployment, methods data sources, mining migrant labour |
| Citation key | hofmeyr1994 |
Argument¶
Average African wages in the modern sector rose strongly in real terms after 1970. This is puzzling because employment growth was slow and mass unionisation came only in 1984. Hofmeyr argues that the explanation lies in reform of the institutional structure of the labour market, which unified a labour market previously segmented by sector, region and gender. He tests this with regressions in a human-capital framework on BMR household survey data for 1975 and 1985, then decomposes wage growth into changes in the structure of earnings and changes in workers' characteristics. The decomposition shows that average growth came almost entirely from Africans moving up the occupational and educational ladder. Real wages of unskilled workers in the reference category fell.
Evidence and method¶
Raw household survey data on urban Africans from the Bureau of Market Research (UNISA) for 1975 and 1985 (the small 1980 survey was omitted). Only employed people aged 16-65 were used. Comparison series come from South African Labour Statistics (Central Statistical Services). The study also draws on Knight and McGrath's (1987) private-sector wage survey and Hofmeyr (1990), which holds the full regression results.
Place in the debates¶
- Why African wages rose 1975-85 (labour scarcity vs union pressure vs institutional change) (see Racial wage gaps, wage convergence and wage determination). Rejects labour scarcity and union pressure as explanations. Argues that dismantling apartheid labour-market restrictions unified segmented markets and allowed occupational upgrading.
- Erosion of apartheid in the labour market / market segmentation (see Racial wage gaps, wage convergence and wage determination). 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).
- Real wage trends for unskilled Africans (see Racial wage gaps, wage convergence and wage determination). 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.
Findings¶
-
Real African wages grew rapidly in the non-primary sectors while White wages fell. (p. 1)
Between 1975 and 1990 real wages rose by 2,9 per cent per annum on average for the non-primary sectors, while those for Whites actually dropped
-
Real wages of unskilled male workers in the reference categories fell sharply. (p. 5)
was approximately 8,6 per cent per annum over the decade, representing a fall of no less than 3,3 per cent per annum in real terms
-
The wage penalty for women fell substantially, which is consistent with a decrease in labour-market segmentation. (p. 5)
However, this difference decreased dramatically over the decade, to only 29 per cent in 1985.
-
Regional wage segmentation in 1975 had largely disappeared around Johannesburg by 1985. (p. 5)
Turning to the results for 1985, there is no longer any significant difference between Johannesburg and its hinterland.
-
African mining wages rose relative to manufacturing wages over the period. (p. 9)
average wages for Africans in mining rose from 70 to 89 per cent of those in manufacturing between 1975 and 1984
-
Almost all growth in average wages came from changes in workers' characteristics, not from changes in the structure of earnings. (p. 6)
Overall, the change in the earnings functions contributed practically nothing to the increase in average wages, while changes in characteristics were responsible for almost all the growth observed.
-
Occupational and educational upgrading was the main driver of growth in average wages. (p. 7)
upward movement through the occupational and educational structure, which contributed 1,4 per cent per annum to overall
-
Narrowing of the gender wage gap accounted for most of the structural contribution to women's wage growth. (p. 7)
Improvements in these contributed 1,1 per cent per annum, or over 85 per cent of the overall change in average wages.
-
The study's results differ from Knight and McGrath's private-sector survey in the size of the real wage rise, but agree that occupational upgrading was the largest source of growth. (p. 8)
an overall increase in real wages of 3,95 per cent per annum for the period 1976-85, of which 39 per cent was caused by the change in earnings functions
-
The conclusion confirms the segmentation-then-unification hypothesis for urban African labour. (p. 8)
there was significant segmentation of the labour market for urban Africans along sectoral, regional and gender lines, and that this tended to lessen or disappear over the ensuing ten years
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| Bureau of Market Research (BMR, UNISA) household income and expenditure surveys of urban Africans, 1975 and 1985 (1980 omitted) | household or labour survey microdata | 1975, 1985 (1980 survey omitted) | Raw household survey microdata, restricted to employed Africans aged 16-65, underpin the semilog human-capital wage regressions and the decomposition of wage change. |
| South African Labour Statistics 1990 (SALS 1990), Table 4.1.1 | official statistical series | 1975-1990 | Supplies real wage growth for Africans and Whites (2,9% p.a. for Africans) and the mining-to-manufacturing wage ratio (70 to 89 per cent) that motivate and support the argument. |
| South African Labour Statistics 1987, Table 4.1.1.1 (Central Statistical Services employer sample surveys of remuneration) | official statistical series | 1975, 1984 | National average wages are used as a benchmark to check the BMR sample averages for plausibility. |
| Current Population Survey based on the 1985 census, Statistical News Release P27.3, 6 March 1987, Table 4.3 | official statistical series | August 1986 | Used to argue that part-time hours explain little of the male-female wage gap (under 5% of women worked fewer than 35 hours). |
| Van Niekerk (1983) data on average school-leaving ages of Blacks | secondary literature compilation | c.1983 | Provides school-leaving ages used to convert age into total working life (experience). |
| Knight and McGrath (1987) analysis of private-sector graded wage survey, 1976-85 | secondary literature compilation | 1976-1985 | Their decomposition results on real wage growth are compared with the author's BMR-based results as a cross-check on direction of change. |
| Knight and McGrath (1987:48,49) education expenditure and enrolment data | secondary literature compilation | 1970s | Supports the claim that African education expanded rapidly in the 1970s. |
| Hofmeyr (1990), Occasional Paper No. 22, Economic Research Unit, University of Natal | secondary literature compilation | 1975-1985 | Holds the full regression results, time-dummy regressions and decomposition derivations that are only summarised in this paper. |
Inference¶
Quantification: Econometric (causal design).
Semilogarithmic human-capital wage regressions (with female interactions) are run separately on the 1975 and 1985 BMR urban African household samples. The change in mean log wages is then decomposed (Oaxaca-type) into changes in the earnings structure and changes in worker characteristics, and the results are compared with the hypothesis of labour market unification. The results are cross-checked against national averages and Knight and McGrath's study. The inference is descriptive and associational, and institutional reform is not directly measured.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Women's wage disadvantage relative to men, after controls, fell from about 48 per cent in 1975 to 29 per cent in 1985, consistent with decreasing gender segmentation. (p. 7) | BMR 1975 and 1985 regression coefficients on the female dummy; the hours caveat is addressed with the Current Population Survey. | Partial data |
| Real wages of unskilled reference-category workers fell by over 3 per cent per annum from 1975 to 1985, and only professionals saw real increases among men. (p. 8) | The regression constant and occupational dummies in the BMR cash-wage regressions, deflated by the low-income inflation rate. | Partial data |
| Almost all growth in average cash wages (1,39% real p.a.) came from changes in worker characteristics, especially upward occupational and educational movement, rather than changes in the earnings structure. (p. 6) | Decomposition of mean log wage change between the BMR 1975 and 1985 samples (Table 3). | Partial data |
| The BMR sample averages agree closely with national average wages, though nominal growth in the BMR was slower (13,2-13,4% against 15,4% p.a.). (p. 2) | Comparison of BMR means with Central Statistical Services employer-survey averages in SALS 1987 (Table 2). | Partial data |
| Rapid African wage growth cannot be explained by labour scarcity or union pressure; the explanation lies in institutional change in the labour market (reform of apartheid labour controls). (p. 1) | Asserted from aggregate SALS wage and employment trends and background on reforms (Hofmeyr 1994, Knight and McGrath); the reforms are not directly measured in the data. | Asserted |
The author's caveats¶
The author notes that the BMR sample is not a random sample of the country because areas and household types are combined arbitrarily, and the areas covered differ between surveys. There is no information on hours worked, which is a particular problem for women. Only a one-digit sector classification is available, and the 1975 mining sample has only 44 observations. The data do not cover Whites, national figures are not strictly comparable, and the results differ from Knight and McGrath's, which the author attributes to sampling differences.
Gaps in the evidence¶
The argument that labour market reform and the erosion of influx control drove unification is inferred from coefficient patterns, with no direct measure of policy enforcement or of trade union or employer behaviour. Data on hours worked, establishment-level wages and a nationally representative African earnings survey would test it better. These were largely absent then, but later October Household Surveys and archival employer records could help today.
Replicability: Data described but not public.