Long-run legacy and persistence of apartheid¶
What are the causal long-run effects of apartheid policies on human capital, trust and inequality, and how much inequality predates apartheid?
Works engaging the debate: 8
Positions¶
- Removals caused lasting human-capital losses
- Divide-and-rule ethnic engineering failed in the long run
- Inequality rooted before 1948, exacerbated by apartheid
Evidence base¶
The debate rests on two largely separate bodies of evidence. The first, causal and microeconometric, deals with forced removals and homelands. Carrillo et al. (2023b) and its appendix (Carrillo et al. (2023a)) draw on individual-level census microdata: the 1980 complete count (with a separate Bophuthatswana file), 1985 and 1991 full censuses, and the 1996 10 percent sample, whose previous-residence and year-of-arrival questions calibrate an average 6.8-year lag between a homeland's formal establishment and the arrival of migrants from White areas. Supporting material comes from the October Household Surveys of 1994-95 (fertility checks) and the South African Internal Migration Survey 1999-2000. Homeland establishment dates are taken from Dinkelman and the boundaries from a 1986 political map. Treatment is an intent-to-treat exposure simulated from birth cohort and establishment date, and outcomes are completed schooling (about 0.116 years lost per year of childhood exposure in 1980) and 1991 employment and income. Abel (2014a) and Abel (2014b) instead geocode respondents to the 2000-2010 Afrobarometer waves for South Africa. Their outcomes are self-reported trust, national identity, perceived crime and trust in chiefs. The qualitative colour in both sets of papers comes from SPP interviews, Platzky and Walker, Rogers, and newspapers such as The Voice, The Star and the Daily Dispatch.
The second body of evidence, on how far inequality predates 1948, is far thinner and more aggregate. Krugell (2014) is the only work with pre-apartheid quantitative data. It transcribes place and district populations and vehicle ownership from the 1911 Union census and matches them to 2011 census places aggregated to 1911 district boundaries, with a 1996 check. It estimates a Zipf exponent that rises from 0.783 to 1.356, and Spearman correlations of 0.377 and 0.465 in vehicles per person between 1911 and 2011. Wilson (2011) is a narrative synthesis. Its only systematic data are PSLSD 1993 schooling tables (61 percent of whites against 11 percent of blacks with 12 years of schooling). It supplements these with Pillay's per-pupil spending ratios, the Commonwealth Expert Group's 1989 occupational shares, and the author's earlier mine-wage ratios, which widened from 12:1 in 1936 to 21:1, together with the Carnegie and Transvaal Indigency commissions and the 1911-1913 statutes. Cooper (2007) reports Seekings and Nattrass's household income distributions for 1975 and 1991 without naming the underlying sources. World Bank (1994) rests on unnamed national accounts, racial income and spending series, and a roughly 300-equation macroeconometric model estimated on 1960-1992 data.
Contested evidence¶
The sharpest internal tension is in the trust evidence. Abel (2014a) and Abel (2014b) use the same Afrobarometer and SPP materials, yet they support the 'divide-and-rule failed' position in different ways. Abel (2014a) reports higher inter-ethnic trust near camps concentrated among cohorts born after 1975, implying intergenerational persistence. Abel (2014b) finds that the negative homeland effects on trust, national pride and crime are concentrated among those born before 1975 and are not transmitted to later cohorts. The two papers also measure the key mechanism, ethnic fragmentation around camps, from incompatible sources with very different magnitudes. One uses a 2011 census ward index (0.256 against 0.232); the other uses an Afrobarometer language-based index (0.294 within 15 km against 0.079). Neither paper has pre-removal trust data or individual records of who was removed. The camp results and the homeland results therefore cannot be separated from non-random siting.
The human-capital and social-capital positions do not contradict each other directly, but they rest on incommensurable treatments and units. Carrillo et al. (2023b) identifies removal from census-recorded birth in a White area, an individual-level proxy measured with error. Abel identifies it from residential proximity to an SPP camp, an area-level proxy measured decades later. The two camps could thus find lasting damage and unexpected solidarity in the same removed population without either test touching the other.
On pre-1948 roots, the evidence is read in opposite directions. Krugell (2014) treats the 1911 vehicle counts, largely horse-drawn transport, and their modest correlation with 2011 as showing that spatial wealth inequality predates apartheid. The same rising Zipf exponent is read as showing that apartheid failed to halt agglomeration, but with only two endpoint censuses neither the timing nor the policy effect is identified. Wilson (2011) and World Bank (1994) assert that apartheid intensified inequality beyond what development would predict, without time series or cross-country comparators. Wilson's own mine-wage ratios show the gap widening from 1936, before 1948. Cooper (2007) reports that inequality began to deracialise under apartheid, with gains from 1975 to 1991 accruing to the richest African quintile. That claim rests on African income data the reviewer concedes were weak.
Underused and missing sources¶
The census microdata series for 1960-2011, together with the published 1904-1960 population distribution and the 1960/70 tables, could fill the 1921-1991 gap that prevents (Krugell (2014)) from dating any break in the rank-size distribution or in spatial wealth persistence. They could also extend (Carrillo et al. (2023b))'s cohort design beyond 1980. Linking them to census districts would turn Carrillo's simulated exposure into measured exposure, and would test whether Abel's camp effects reflect actual relocatees.
No work in the debate uses any pre-removal measure of social capital, and no such source is identified.
Strength of key claims¶
| Basis | Share of key claims |
|---|---|
| Systematic data | 20% |
| Partial data | 55% |
| Illustrative or anecdotal | 9% |
| Asserted | 16% |
Comparative figures for all debates are on the Literature overview.
Best-evidenced works¶
Carrillo et al. (2023b), Carrillo et al. (2023a), Abel (2014a), Abel (2014b), Krugell (2014)
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 |
|---|---|
| Abel (2014a) | Challenges the finding that fragmentation lowers trust. Argues that forced diversity in camps raised inter-ethnic trust, reconciling micro-level contact benefits with aggregate-level negative correlations. |
| Abel (2014a) | Finds little evidence of success: camp areas are more ethnically mixed and trust other ethnic groups more, consistent with solidarity among the displaced. |
| Abel (2014a) | Supports the Nunn-Wantchekon/Tabellini view that shocks change values transmitted across generations, though he cannot rule out that conditions in the camp areas persisted. |
| Abel (2014b) | Acknowledges the cheap-labour-reserve function as the focus of the existing literature but deliberately studies other long-run effects on social capital and identity. |
| Abel (2014b) | Finds attitudes shaped by removals only among those who experienced them, so no transmission to later cohorts. Resettlement camp areas show higher trust, which cuts against a simple persistence-of-mistrust story. |
| Abel (2014b) | Finds no higher own-group trust and no lower trust in other groups in homelands generally, so apartheid's ethnic division was not effective in the long run. National identity is weaker among older homeland residents. |
| Carrillo et al. (2023a) | Treats relocation as a quasi-experiment showing causal negative effects on education, employment and income, and argues the results are robust to inference, selection and specification checks. |
| Carrillo et al. (2023a) | Finds reduced years of schooling account for roughly 25-37 percent of the income effect, with the rest attributed to unmeasured quality or other inputs. |
| Carrillo et al. (2023b) | Provides rigorous causal evidence that homeland relocation reduced human capital and earnings, supporting the view that apartheid's spatial policies have long-lasting inequality consequences. |
| Carrillo et al. (2023b) | Argues place effects (poorer destination environments) explain about 70 percent of the schooling decline; discrimination, fertility change and school interruption are judged less important, and the eviction-induced education-insurance effect found elsewhere (Becker et al. 2020) is not found here. |
| Cooper (2007) | Endorses the view that inequality has been deracialised since before 1994 and that unemployment is now its main expression. He accepts that the post-1994 high-wage strategy encouraged capital-intensive growth and worsened unemployment. |
| Krugell (2014) | Argues wealth was already highly unequally spread in 1911, so apartheid exacerbated but did not originate spatial inequality, aligning with Fourie and Von Fintel and Wilson. |
| Wilson (2011) | Holds that the democratic government has done little to shift inherited poverty, unemployment and inequality, and that human capital deficits (Bantu Education) make the racial gap slow to close. |
| World Bank (1994) | Holds that apartheid intensified inequality beyond what would be expected at South Africa's level of development, and that public spending favoured whites. |