Living Standards in South Africa's Former Homelands¶
Martine Mariotti (2012). Living Standards in South Africa's Former Homelands. ANU Working Papers in Economics and Econometrics. Research School of Economics, Australian National University.
| Type | working paper |
| Scope | Economics and economic history of apartheid |
| Method | econometric |
| Period studied | 1974–2008 |
| Regions | Transkei, Ciskei, Bophuthatswana, KwaZulu, Eastern Cape, KwaZulu-Natal, South African gold mines, Malawi, Mozambique |
| Themes | mining migrant labour, bantustans removals influx, inequality poverty unemployment, labour markets wages, agriculture land, demography urbanisation, education human capital, methods data sources |
| Citation key | mariotti2012 |
Argument¶
Mariotti treats the mid-1970s loss of foreign (Malawian and Mozambican) gold-mine labour as a quasi-natural experiment. The Chamber of Mines recruited domestically, mainly in four homelands (Transkei, Ciskei, Bophuthatswana, KwaZulu), which sharply raised household incomes there. Using NIDS 2008 adult data and a difference-in-differences design, she compares men and women born in treated districts in 1976-77 with other homeland-born Africans. Men born in the treated regions at the shock are about 6 cm taller and have better long-run health, which supports the foetal origins hypothesis and implies pre-shock malnutrition and stunting in the homelands. No effects appear for education, other health measures, children born before or after the shock, or women. The paper also documents poor homeland conditions to show that a small mining wage was a large income gain.
Evidence and method¶
National Income Dynamics Study Wave 1 (2008), with 10,433 African respondents (4,098 men, 6,335 women) matched to birth district councils. Mining employment by origin from SAIRR annual surveys and Crush et al. (1991). Homeland conditions from the Tomlinson Commission (1955), Horrell (1969, 1973), Wilson (1972), Butler et al. (1978) and Feinstein (2005). Mining wage series from SAIRR.
Place in the debates¶
- Foetal origins hypothesis (in utero nutrition and adult outcomes) (see Living standards, health and social policy). Supports it: the effect is concentrated in men born at the shock, with no effect on those born slightly before or after.
- Whether household income gains are passed to children (intra-household allocation) (see Living standards, health and social policy). Finds no effect on existing children, which is consistent with Duflo (2003) that male income is not necessarily passed on to children.
- Cheap migrant labour thesis and living standards in the reserves/homelands (see Living standards, health and social policy). Accepts that the mining industry relied on cheap foreign labour and that homeland poverty was severe. Interprets stunting as the physical consequence of that poverty.
Findings¶
-
Foreign labour on the gold mines fell sharply after 1973, while domestic recruitment more than doubled. (p. 4)
The number of foreign workers in the mining industry fell from 297, 000 in December 1973 to 189,000 at the end of 1977
-
The shock was large relative to homeland incomes: about a quarter of households in affected areas gained at least R14.22 a month. (p. 4)
For approximately 24 % of households in the affected areas this signified an increase in family incomes of at least R14.22 per month
-
Foreign workers dominated the gold mining workforce before the shock, and domestic Africans had been largely excluded from mining. (p. 8)
By 1973, foreign workers made up 75 % of the gold mining workforce in South Africa.
-
Homeland populations far exceeded what the land could carry, according to the Tomlinson Commission. (p. 11)
the land could only sustain 36 - 43 % of the current population and that amount only if the remaining 57 - 64 % moved off the land
-
Mining wages rose rapidly after the 1973 labour disputes. (p. 13)
Following labour disputes throughout 1973, the wage increased to R46 in 1974, R79 in 1975, R92 in 1976, R102 in 1977, R123 in 1978 and R141 in 1979
-
The preferred DiD specification shows a large positive height effect for men born in treated regions in 1976-77. (p. 22)
The results suggest that male heights of those born during the years of the shock increased by between 5.9 and 6.3 centimetres.
-
Women born in treated regions at the shock appear shorter, not taller, which is an unexplained puzzle. (p. 23)
All specifications yield height decreases of between 1.8 and 3.7 centimetres for women born in the treated regions at the time of the shock.
-
The height effect is robust to controls for age and language. (p. 23)
In all three specifications the effect on the treatment group is significant and the increase in heights is between 5.9 centimetres and 6 centimetres.
-
No effects on men born before or toddlers, and men born after the shock in treated regions are shorter, so the effect operates through in utero exposure. (p. 24)
males born one to five years after the shock in the treated regions are 3.12 centimetres shorter than males born before the shock
-
The shock had no significant effect on educational attainment, and a significant effect only on long-run health among health measures. (p. 26)
I do find a positive and significant effect of treatment on long run health outcomes, males from the treatment group are less likely to be suffering from any long term health complaints.
-
African men were substantially shorter than Whites, indicating poor apartheid-era living standards. (p. 6)
on average African men living both within and outside of a homeland were approximately 8 centimetres shorter than Whites
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| National Income Dynamics Study (NIDS) Wave 1, 2008 (adult sample, derived household file) | household or labour survey microdata | 2008 survey; respondents born c.1941-1990 | Supplies measured adult height, education, health measures and birth year/location for the difference-in-difference analysis. |
| South African Institute of Race Relations Annual Reports/Surveys 1969-1979 (foreign and domestic mine employment, Table 2) | official statistical series | 1969-1978 | Documents the fall in foreign mine labour and rise in domestic recruits that defines the labour demand shock. |
| SAIRR Annual Surveys (mining wage series 1972-1979) | official statistical series | 1972-1979 | Shows nominal mining wages rising from R21 to R141 per month, used to size the income shock. |
| Tomlinson Commission Report 1955 (Summary, UG 61/1955) | commission or official report | c.1951-1955 (data around 1953) | Provides baseline farm incomes, livestock holdings, land carrying capacity and erosion to show pre-shock poverty. |
| Crush, Jeeves and Yudelman (1991) South Africa's Labour Empire and Crush (1986) - mine employment by homeland | secondary literature compilation | 1970-1978 | Timing and regional distribution of increased domestic recruitment; employment numbers used to construct the mining-proportion variable. |
| Butler, Rotberg and Adams (1978) The Black Homelands of South Africa - homeland population and income estimates | secondary literature compilation | 1973 estimates | Population denominators for the proportion-employed-in-mining variable and 1973 income range. |
| Horrell (1969, 1973) SAIRR studies of African reserves/homelands | secondary literature compilation | 1969-1973 | Population density, land quality and remittance share figures describing homeland conditions. |
| Wilson (1972) Labour in South African Gold Mines | secondary literature compilation | 1960s-1972 | Remittance rate (18% via official channels), overcrowding and wage-income comparisons. |
| Mauder (1982) Atlas of Transkei - population estimate | secondary literature compilation | 1982 | Population of 2.3 million used in the ad hoc estimate of the share of households affected. |
| 1970 South African population census | population census | 1970 | Discussed to show the lack of reliable apartheid-era data on Africans, motivating use of NIDS. |
| NIDS survey documentation and weighting guidelines (SALDRU 2009) | commission or official report | 2009 | Sample weighting specification followed (without stratification) in regressions. |
| Municipal Demarcation Board district council boundary data and Perry-Castañeda map collection (homeland maps) | maps spatial data | 2011 boundaries; apartheid-era homeland map | Matching district councils to former homeland boundaries to assign treatment status. |
| Abasebenzi workers' newsletter, 1973(3) | newspapers periodicals | 1973 | Cited for a Poverty Datum Line of R20 a week for family survival. |
| Leys (1975) 'South African Gold Mining in 1974' | secondary literature compilation | 1974-1975 | Contemporary account used to argue the shock was unanticipated and homeland allocation to industries was independent of conditions. |
| Personal conversation with Francis Wilson, October 2011 | interviews oral history | 2011 | Supports the claim that workers brought home cash beyond formal remittances. |
New data¶
Author constructed an analysis sample from NIDS Wave 1: 10,433 African respondents (4,098 men, 6,335 women) born in districts containing homelands, with treatment status assigned by matching birth district councils to affected former homelands (19 of 28 homeland-overlapping district councils out of 53) and birth year 1976-77; plus a 460-observation regional mining-employment proportion variable.
Inference¶
Quantification: Econometric (causal design).
Difference-in-differences using OLS and ordered probit regressions comparing adult outcomes of Africans born in affected homeland district councils during the 1976-77 shock with other cohorts and regions, with birth-year, province, age, language and father's occupation controls; supported by an ad hoc calculation of income shock size and descriptive tables.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Men born in affected homeland regions in 1976-77 are about 6 cm taller than other African men, indicating pre-shock malnutrition/stunting. (p. 6) | NIDS 2008 height measurements; DiD regressions (Tables 4 and 6) with treatment assigned by birth district council. | Partial data |
| The shock did not affect heights of men born before or after, consistent with a foetal-origins in utero effect. (p. 24) | NIDS regressions on cohorts born 1-5 years before/after and teenagers (Table 7). | Partial data |
| The mining recruitment raised household incomes substantially relative to very low pre-shock homeland incomes, with about 24% of Transkei births affected. (p. 13) | Tomlinson Commission income data, SAIRR mine employment and wage series, Wilson/Horrell remittance rates, and an ad hoc population calculation. | Illustrative or anecdotal |
| No effect of the shock on education or most health measures, but some improvement in long-run health for men. (p. 26) | NIDS ordered probit and OLS regressions (Tables 8 and 9). | Partial data |
| No positive effect on women born during the shock; treated women appear shorter. (p. 23) | NIDS regressions for women (Table 5, Tables 10-12). | Partial data |
| The mining labour demand shock was unexpected and homeland selection was independent of outcomes, with little in-migration due to pass laws, supporting identification. (p. 5) | Leys (1975), Wilson (1972), Crush et al., and qualitative reasoning about pass laws. | Asserted |
| Homelands suffered extreme poverty, overcrowding and poor land before the shock. (p. 11) | Tomlinson Commission 1955, Horrell 1969/1973, Wilson 1972, Butler et al. 1978. | Partial data |
The author's caveats¶
Acknowledges that data collection in the homelands was unreliable and inadequate, income estimates rest on unverifiable assumptions, treatment group samples are small (82 men, 137 women) lowering power, not all treated individuals were actually exposed, birth district councils only approximately match homeland boundaries (Transkei and Ciskei indistinguishable), father's occupation is too aggregated to identify mining, no parental height/income data, Tomlinson data date from 1953, and population estimates for the mining proportion ignore growth.
Gaps in the evidence¶
The argument lacks direct data on household exposure to mine employment, remittances and pre-shock child nutrition, relying on an ad hoc calculation and 1953 income figures; recruitment records by magisterial district (TEBA/NRC archives) or district-level births would permit finer treatment assignment. Placebo tests on other birth cohorts, and checks for selective survival, migration and measurement of birthplace are limited, and effects of the economy-wide wage rises in the same years are not separated. Larger or administrative datasets (e.g., later censuses with birthplace, conscript/army height records) could test the large 6 cm effect; such data exist now only partially.
Replicability: Data described but not public.