The collapse of primary schooling returns in South Africa 1960-90

Literature › Works

Peter G. Moll (1996). The collapse of primary schooling returns in South Africa 1960-90. Oxford Bulletin of Economics and Statistics, 58(1). Blackwell Publishers.

Type	journal article
Scope	Economics and economic history of apartheid
Method	econometric
Period studied	1960–1990
Regions	South Africa, 13 urban areas (Johannesburg, Pretoria, Vaal, Durban, Cape-region and other cities)
Themes	education human capital, labour markets wages, inequality poverty unemployment, mining migrant labour, unions strikes, methods data sources, transition post apartheid
Citation key	moll1996a

Argument

Moll shows that the wage return to primary schooling for urban Africans was already low by 1965 and fell further to about 2-3 percent a year by 1975-90. Returns to secondary schooling stayed high for all races. He attributes the low level to poor African primary school quality (very high pupil-teacher ratios, unqualified teachers, low spending), which worsened after the 1953 Bantu Education Act. The fall is attributed to a larger supply of primary-educated urban workers, the 1970s mine wage rises that put a floor under wages for the uneducated, and 1980s wage compression from unions, Industrial Councils and Wage Boards. He concludes that wage regulation cut primary returns, probably added to African unemployment, and calls for reallocating spending towards primary education.

Evidence and method

Mincerian log-earnings regressions on microdata for urban male private non-agricultural employees: the 1960, 1970 and 1980 censuses, BMR household surveys of 1975 and 1985, the 1980 Current Population Survey and the 1990 CSS/HSRC survey. Grouped BMR data for 1965-75, an international comparison of returns to schooling, and series on pupil-teacher ratios, teacher qualifications, per-pupil spending, enrolments, workforce education distributions and Industrial Council minimum wage rates (Williams and Homer).

Place in the debates

• Labour market discrimination vs school quality as the cause of low African returns (see Education, Bantu Education and skills). Rejects discrimination as the explanation for low primary returns, since African secondary returns were high. Points to school quality (pupil-teacher ratios, teacher qualifications, spending).
• Effects of wage regulation, unions and minimum wages on unemployment and skill differentials (see Education, Bantu Education and skills). Argues that Industrial Council and Wage Board minimum wages and union 'living wage' policies compressed skill differentials, lowering primary returns and probably raising African unemployment.
• Quantity vs quality in education policy for post-apartheid South Africa (see Education, Bantu Education and skills). Favours improving primary quality and access, funded by reallocating from secondary and tertiary spending and raising fees at those levels, on equity grounds.
• Effect of 1970s mine wage rises on the wider labour market (see Education, Bantu Education and skills). Sees the mine wage increases and local recruitment as raising wages of the uneducated and so compressing the primary-schooling differential.

Findings

• Government spending per pupil was massively skewed by race. (p. 2)

  The government spent approximately ten times as much on schooling for each white child as for each African child, with coloured and Asian children receiving intermediate amounts.

• The annual return to African primary schooling was very low from 1975. (p. 4)

  The annual effect of primary schooling on the logarithm of earnings of Africans was a low 0.025 in 1975 (BMR).

• Grouped-data estimates show the African primary return was already low in 1965 and then fell further. (p. 10)

  It is clear that the return to African primary schooling was already low by 1965, and that it fell to extremely low levels during the late 1960's and early 1970's.

• African primary classes were too large to sustain instruction quality. (p. 12)

  The 60:1 ratio for African primary schools was likely to be of a size that impaired the quality of instruction;

• African primary pupils received far less funding than coloured pupils. (p. 14)

  African primary pupils enjoyed only 36 percent of the financing that their coloured peers did, and African secondary pupils some 51 percent.

• After the Bantu Education Act enrolments boomed but real spending barely grew. (p. 15)

  Whereas enrolments grew by 109 percent from 938,000 in 1953 to 2.0 million in 1965, real spending grew by only 21 percent over the period.

• Large 1970s mine wage rises placed a floor under wages of the uneducated. (p. 17)

  By 1975, the real wage of African mine workers was 2.4 times higher than the wage in 1972, and by 1980, 3.2 times higher.

• Minimum-wage regulation covered most urban African employees outside the primary sectors. (p. 18)

  Summing the Industrial Council and Wage Board coverage, about 82 percent of African employees in the non-primary private sector enjoy minimum wage protection.

• The overall decline of the African primary return, 1965 to 1980 and 1990. (p. 20)

  The effect of a year's primary schooling upon wages of Africans in urban South Africa fell from 6 percent in 1965 to 2 or 3 percent in 1980 and 1990.

• Defending the real unskilled wage probably contributed to African unemployment. (p. 21)

  The defence of the real unskilled wage, at a time when wages in a wide variety of other occupations were falling, no doubt contributed to the high unemployment rate among African workers.

• Earnings inequality among urban African men rose sharply as the labour market opened up. (p. 23)

  The Gini coefficient of earnings of urban African men increased from 0.25 in 1980 to 0.38 in 1993

Evidence ledger

Sources used

Source	Kind	Period	How used
Population Censuses 1960, 1970 and 1980 (microdata)	population census	1960, 1970, 1980	Microdata used to estimate Mincerian log-earnings regressions and the annual effects of primary and secondary schooling by race, and to compute workforce educational distributions and mean schooling by occupation.
Bureau of Market Research (BMR) household surveys 1975 and 1985	household or labour survey microdata	1975, 1985	Microdata for earnings regressions yielding the African primary effect of 0.025 in 1975 and 0.027 in 1985 and the comparisons across races.
Current Population Survey (CPS) of May 1980 (1979-82 survey)	household or labour survey microdata	1979-82 (May 1980 round used)	Provides a second 1980 estimate of the African primary schooling effect (0.034) for cross-checking against the census.
CSS/HSRC household survey, November 1990	household or labour survey microdata	November 1990	Microdata for the 1990 earnings regressions showing the African primary effect of 0.020 and secondary effect of 0.11.
Published grouped BMR data 1965-1975 (BMR Research Reports 1965-1978)	household or labour survey microdata	1965-1975	Grouped data used in weighted least squares spline regressions to extend the series back to 1965 and show that African primary returns were already low.
Bezuidenhout (1989), Comparative Education Statistics Part 2: Black Education 1960-1983 (HSRC)	secondary literature compilation	1960-1989	Source of African pupil-teacher ratios, unqualified teacher percentages and enrolment figures by level.
Department of Statistics, Education: Coloureds and Asians reports (08-04-01, 21-03-04, 21-03-09) and Education: Asians 1979 (21-03-13)	official statistical series	1957-1979	Provides coloured and Asian pupil-teacher ratios and teacher qualification data for comparison with African schooling.
Department of Education and Training (1983) Annual Report; van Rensburg (1975) HSRC Report WS-10	commission or official report	1930-1965	Source for the long-run African pupil-teacher ratios from 1930 to 1965.
Per-pupil government expenditure series (Jones 1970; Horrell 1964, 1970; SAIRR Race Relations Survey; Trotter and Shave 1988)	secondary literature compilation	1930-1985/6	Establishes the per-pupil spending gap by race, the fall in African spending after 1950, and the primary/secondary spending breakdown for 1985/6.
Central Statistical Service (1991), Population Census 1991, Economic Characteristics of the Population (Report 03-01-23)	population census	1991	Provides the 1991 educational distribution of the economically active population used to show the rising relative supply of primary-educated workers.
Williams (1989) Wage Determinations in South Africa; Williams and Homer (1990) Industrial Council Wage Rates in South Africa 1978-1988 (SALDRU)	secondary literature compilation	1975-1988	Minimum wage rates for Industrial Council occupations and Wage Board coverage, used to show wage compression and the 82 percent coverage estimate.
Project for Statistics on Living Standards and Development (PSLSD) 1993 microdata	household or labour survey microdata	1993	Used for mean family income by child's schooling and the 1993 Gini and white/black earnings ratio in the policy discussion.
International returns-to-schooling studies (Knight and Sabot 1990; Psacharopoulos and Ng 1992; Herz and Khandker 1991; Kugler and Psacharopoulos 1989) and World Bank World Development Reports	secondary literature compilation	1980-1990	Benchmarks South African primary and secondary schooling effects against other developing countries' urban estimates; WDR gives the pupil-teacher ratio comparison.
Trotter and Shave (1988) and Weekly Mail (3 April 1992) on private school enrolments	newspapers periodicals	1985, 1992	Supplies private school pupil numbers for the discussion of privatization.

Inference

Quantification: Econometric (causal design).

Semi-parametric Mincerian log-earnings regressions on several microdata sets (1960-90) give annual primary and secondary schooling effects by race. A weighted least squares spline regression on grouped 1965-75 BMR data extends the series back. The comparison is cross-race, cross-time and cross-country. Causes are inferred by matching the timing and pattern of descriptive indicators (pupil-teacher ratios, teacher qualifications, per-pupil spending, workforce education shares, mine and minimum wage changes) to the observed returns, without a formal causal test.

Key claims

Claim	Basis	Strength
The annual effect of primary schooling on African earnings was very low (about 0.02-0.034) from 1975 to 1990, far below the secondary effect and below the international range. (p. 4)	Earnings regressions on the BMR 1975/85, 1980 census, CPS 1980 and CSS/HSRC 1990 microdata (Tables 1, 2), compared with studies in Table 4.	Systematic data
African primary returns were already low in 1965 and fell during the late 1960s and early 1970s, while coloured and Asian returns also declined. (p. 10)	Grouped BMR data 1965-75 with weighted least squares splines (Table 3), adjusted by one third for grouping bias using the 1975 overlap with the microdata.	Partial data
Low African primary returns reflect poor school quality: very high pupil-teacher ratios, more unqualified teachers and lower per-pupil spending than other groups. (p. 12)	Descriptive tables 5, 6 and 7 compiled from Bezuidenhout, the Department of Statistics, Horrell and Trotter and Shave, matched with the regression results.	Partial data
The fall in primary returns was partly due to a decline in schooling quality from 1955 to 1965, when pupil-teacher ratios rose and real per capita spending fell as enrolments expanded under Bantu Education. (p. 15)	Time series in Tables 5 and 7 on ratios and spending, and enrolment figures from Bezuidenhout.	Partial data
Rising relative supply of primary-educated urban workers reduced the primary wage premium. (p. 16)	Table 8 educational distributions from 1970 and 1980 census microdata and the 1991 census, giving primary-to-no-education ratios.	Partial data
The 1970s mine wage increases and increased local recruitment put a floor under the wages of uneducated workers, depressing the primary differential. (p. 17)	Cited mine wage ratios and census microdata on the education of mining workers, without a cited source for the mine wage series and with no direct test.	Illustrative or anecdotal
Industrial Council and Wage Board minimum wage regulation compressed skill differentials in 1975-88 while holding up the unskilled minimum wage, reducing returns to later primary years. (p. 20)	Williams and Homer (1990) minimum wage rates for selected occupations in four industries, matched to 1980 census mean schooling (Table 9), plus coverage estimates.	Partial data

The author's caveats

The author notes that none of the data was collected to examine returns to education and that the surveys give earnings, not wage rates. The samples are restricted to urban male private non-agricultural employees. No microdata including Africans exist before 1975, and grouped data overestimate primary returns by about one third. The 1960 and 1970 censuses omit African earnings, and categories are lumped in some surveys, requiring interpolation. The author adds that the mining effect had no satisfactory unemployment measure and that the demand-side factors tested had no explanatory power.

Gaps in the evidence

The explanations rest on timing coincidence and descriptive indicators rather than identification. The paper has no direct school-quality measures linked to individual workers (such as test scores or cohort-specific schooling) and no cohort or panel data separating quality, supply and wage-regulation effects. Rural, female and self-employed workers are excluded, so the findings do not cover most Africans. Longitudinal or retrospective data, which the author says are needed, and archival Industrial Council and Chamber of Mines wage records could test the mechanisms better; some of these existed at the time, and the later PSLSD and census microdata offer partial tests today.

Replicability: Data described but not public.