Causes and Consequences of Schooling Outcomes in South Africa: Evidence from Survey Data¶
Kermyt Anderson; Anne Case; David Lam (2001). Causes and Consequences of Schooling Outcomes in South Africa: Evidence from Survey Data. PSC Research Report No. 01-490. Population Studies Center, Institute for Social Research, University of Michigan.
| Type | working paper |
| Scope | Wider apartheid literature |
| Method | synthesis review |
| Period studied | c.1930 (oldest cohorts observed in 1995)–1996 |
| Regions | South Africa (national), Guguletu, Cape Town |
| Themes | education human capital, inequality poverty unemployment, labour markets wages, methods data sources |
| Citation key | anderson2001 |
Argument¶
Using household survey data (1993 SALSS, 1995 OHS, 1996 census) the paper reviews the apartheid-era legacy in schooling. Racial gaps in attainment persist but have narrowed steadily across cohorts, and there is essentially no gender gap. The authors argue the African-white gap comes mainly from high grade repetition, not from low enrollment or early dropout. School quality, measured by pupil-teacher ratios, and family structure shape attainment and later earnings. Returns to schooling are high above about grade 5-7, and higher for Africans than whites. These returns, combined with wide dispersion in schooling, help explain extreme income inequality. The paper is a synthesis built largely on the authors' own earlier work, and it ends by identifying data gaps for future surveys.
Evidence and method¶
1995 October Household Survey (about 130,000 individuals in 30,000 households); 1993 SALDRU/World Bank South African Living Standards Survey (about 9,000 households, 40,000 individuals); the Education Atlas and School Register of Needs; the 1996 census 10% sample (via Case and Yogo); and the 1998 Guguletu High School Study (603 students with retrospective schooling histories).
Place in the debates¶
- Source of the racial schooling gap: non-enrollment/dropout versus grade repetition (see Education, Bantu Education and skills). Argues that repetition, not non-enrollment, drives the gap, citing enrollment of 97% or more for Africans to age 14 and about 0.8 grades per year of age.
- Does school quality matter for outcomes (education production function debate)? (see Education, Bantu Education and skills). Sides with Case and Deaton and Case and Yogo that pupil-teacher ratios strongly affect attainment, returns to education and employment, treating South African families as having had little control over school quality.
- Returns to schooling in South Africa: low primary returns versus high returns at higher levels (see Education, Bantu Education and skills). Finds low returns through about grade 5, then high returns of 15-20% per year above grade 7. These are higher for Africans than whites, though the authors say they should be discounted for repetition. The paper notes this is higher than Moll's earlier estimates for primary schooling.
Findings¶
-
The white-African attainment gap has narrowed sharply across cohorts, from about 8 years among the oldest to about 3 among the youngest. (p. 6)
For the oldest age groups the gap between whites and Africans is almost 8 years, while for younger cohorts it is around 3 years.
-
Enrollment among African children is near universal through early adolescence, so non-enrollment cannot explain the gap. (p. 7)
the enrollment rates of Africans are 97% or above between ages 10 and 14.
-
The main driver of the gap is grade repetition. (p. 7)
It is grade repetition rather than non-enrollment that is primarily responsible for the lower schooling attainment of African children.
-
In the Guguletu township school, most students had repeated grades by the end of secondary school. (p. 9)
roughly 2/3 of the students have repeated at least one grade, and more than 1/3 have repeated at least two grades
-
Lower pupil-teacher ratios raise the return to education for African men. (p. 12)
a decrease in the pupil/teacher ratio of 5 students is on average associated with an increase in the return to education of roughly 1 percent.
-
School quality also raises years of completed schooling. (p. 13)
reducing the pupil/teacher ratio by 10 students would, all else equal, increase completed schooling by 0.6 years.
-
A very large racial earnings gap remains at equal schooling, partly reflecting African joblessness. (p. 14)
whites have four times higher mean earnings than Africans.
-
Wage regressions imply high returns to each year of schooling above grade 7. (p. 14)
implying returns of 15% to 20% per completed year of schooling at most schooling levels above grade 7 (Lam 1999).
-
Prime-age African male joblessness is very high at every schooling level through grade 8. (p. 15)
Over 25% of African men age 30-39 report zero earnings, with levels over 30% at every level of schooling up through 8th grade.
-
Extreme inequality is linked to dispersed schooling combined with high returns to schooling. (p. 15)
South Africa's extreme income inequality can be attributed to the combination of a wide dispersion in schooling and high returns to schooling.
-
The state old-age pension is an important household resource that affects many African children. (p. 11)
One third of African children live with a grandparent or an older adult receiving a state old-age pension.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| October Household Survey 1995 (OHS), Statistics South Africa | household or labour survey microdata | 1995 | Main data for schooling by age, race and gender, grade progress, enrollment, work, family background, school expenditure and earnings by schooling (Figures 1-4, Tables 1, 3, 4). |
| South African Living Standards Survey (SALSS) 1993 | household or labour survey microdata | August-December 1993 | Provides educational attainment, enrollment, earnings and family structure (e.g. co-residence with parents) and underlies Case and Deaton's school quality results. |
| Guguletu High School Study (1998) | author survey or fieldwork | 1998 | Retrospective schooling histories used to document grade repetition and delay by grade (Table 2) and the link between age and matric scores and between parental schooling and child schooling. |
| Education Atlas of South Africa | official statistical series | Not specified (data matched to 1993 SALSS) | District-level pupil-teacher ratios used by Case and Deaton as a measure of school quality linked to SALSS respondents. |
| School Register of Needs | official statistical series | Not specified | Mentioned as school-based data on school quality, drawn on via Case and Yogo and Case and Deaton. |
| 1996 South African Population Census (10 percent public release sample) | population census | 1996 | Used by Case and Yogo to estimate the effect of 1991 district school quality (pupil/teacher ratios) on returns to education and employment of 24-34 year olds. |
| 1991 school quality measures by magisterial district | secondary literature compilation | 1991 | Assigned to young adults as the school quality of the district in which they were raised. |
| School records from the Ministry of Education | official statistical series | Not specified | Mentioned only to note that they provide little direct information on grade repetition. |
New data¶
Guguletu High School Study (1998): survey with retrospective schooling histories and family information for 603 students at one African township high school in Cape Town, collected by the authors and colleagues.
Inference¶
Quantification: Descriptive tables.
Review of survey evidence: descriptive cross-tabulations of the 1995 OHS by age, race and gender (using age profiles as a proxy for historical cohort trends), indirect inference about grade repetition from grades-per-year and enrollment, a small-sample school survey, and summaries of regression results from the authors' and others' prior work (Anderson 2000, Lam 1999, Case and Deaton 1999, Case and Yogo 1999).
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Racial gaps in schooling persist but have narrowed across cohorts; the white-African gap falls from almost 8 years for the oldest to about 3 years for younger cohorts. (p. 6) | 1995 OHS mean years of schooling by age and race (Figure 1), subject to assumptions about migration and differential mortality. | Systematic data |
| There is virtually no gender gap in schooling in any racial group. (p. 6) | 1995 OHS schooling measures by sex (Figure 2) and prior studies. | Systematic data |
| Grade repetition, not non-enrollment or dropout, mainly explains the African-white gap in grade attainment. (p. 7) | Indirect inference from OHS 1995 grades per year and enrollment rates (Table 1), supported by the Guguletu retrospective histories (Table 2). | Partial data |
| Children behind in school have less spent on their education, and family structure (living with neither parent) is associated with worse schooling outcomes. (p. 9) | 1995 OHS expenditure data for African children (Figure 3), and family-type comparisons with multiple-comparison tests (Table 3), from Anderson 2000; cross-sectional only. | Partial data |
| School quality (pupil-teacher ratios) strongly affects schooling, returns to education and employment for Africans. (p. 12) | Case and Deaton's SALSS matched to Education Atlas data, and Case and Yogo's use of the 1996 census 10% sample and 1991 district school quality; reported as summaries of the other authors' regressions. | Systematic data |
| Large earnings gaps persist between Africans and whites at the same schooling level, and African men have very high rates of zero earnings. (p. 14) | 1995 OHS, males aged 30-49, mean earnings and zero-earnings shares by schooling (Table 4). | Systematic data |
| Returns to schooling are high above grade 5-7 (15-25% per year) and appear higher for Africans than whites, though they should be discounted for grade repetition. (p. 14) | Table 4 earnings ratios, Lam 1999 wage regressions, and comparison with Moll, Case and Deaton, and Mwabu and Schultz. | Partial data |
The author's caveats¶
The authors note that the OHS lacks direct information on grade repetition, school quality, attendance and achievement, and on non-resident parents. Age-based inference assumes no differential migration or mortality by education. The cross-sectional data cannot separate causal channels. The Guguletu sample covers only one low-quality township school and omits dropouts and those at better schools. School quality is measured at district level, not school level.
Gaps in the evidence¶
The argument would be better tested with longitudinal data or retrospective histories (age at entry, repetition, non-enrollment) in national surveys, and with school-level quality data linked directly to individuals. Administrative school records on repetition and matric pass rates by race would help. Linked data on non-resident parents are also lacking; the authors say none existed in 2001, and later panel surveys such as NIDS partly fill this gap. Historical evidence on apartheid-era school provision (per-pupil spending, teacher qualifications by race) is only cited indirectly via Fedderke et al.
Replicability: Data described but not public.