Long-run Effects of Forced Removal under Apartheid on Social Capital¶
Martin Abel (2014). Long-run Effects of Forced Removal under Apartheid on Social Capital. Harvard University (working paper, 19 October 2014).
| Type | working paper |
| Scope | Wider apartheid literature |
| Method | econometric |
| Period studied | 1948–2011 |
| Regions | Former homelands (Ciskei, KwaZulu, Gazankulu, Venda, Lebowa, Bophuthatswana), South Africa |
| Themes | bantustans removals influx, politics ideology elections, demography urbanisation, inequality poverty unemployment, methods data sources, comparative theory, resistance social movements |
| Citation key | abel2014 |
Argument¶
Abel treats the forced removal of roughly four million Africans to homeland resettlement camps (1960-1980) as a quasi-natural experiment. Within the same former homeland, people living near former camps report higher trust in relatives, neighbours, people in general and other ethnic groups, and report less crime, than people living farther away. The effect on inter-ethnic trust is concentrated among those born after 1975. He argues the mechanisms are greater ethnic diversity in camp areas (positively related to inter-ethnic trust only there) and a shared identity as displaced people. He supports this with placebo-camp, border/city/growth-point controls, clustering and balance tests. He admits that the identification assumptions and mechanisms remain inconclusive.
Evidence and method¶
Surplus People Project (1983) geocoded data on 45 camps (66 used) and numbers relocated; Afrobarometer waves 2000-2010, geocoded to town level (N about 700-1,700 per outcome); South African census data (1996-2011) at ward level; Statistical Yearbook 1974 and 1982; secondary literature (Platzky and Walker 1985, Rogers 1980, Posel 2011, Christopher 2001); newspaper and quoted primary snippets.
Place in the debates¶
- Ethnic diversity and trust/conflict (see Long-run legacy and persistence of apartheid). 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.
- Did Bantustan policy succeed in dividing Africans along ethnic lines? (see Long-run legacy and persistence of apartheid). Finds little evidence of success: camp areas are more ethnically mixed and trust other ethnic groups more, consistent with solidarity among the displaced.
- Purists vs pragmatists within apartheid (segregation vs economic integration of African labour) (see Origins, coherence and ideology of apartheid and Afrikaner nationalism). Follows Posel: the Bantustan system tried to serve both aims by exploiting labour migration while keeping Africans separated and politically weak.
- Persistence of historical shocks on culture and trust (see Long-run legacy and persistence of apartheid). 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.
Findings¶
-
Living near a former resettlement camp is associated with higher trust in relatives, neighbours and people in general (significant for relatives and general trust). (p. 14)
Estimates suggest that people living in these areas are more trusting towards their relatives (1), neighbors (3) and people in general (5).
-
Camp-area residents trust other ethnic groups more, and this is driven by cohorts born after the height of removals. (p. 17)
people living in former resettlement areas trust members from other ethnic groups about 20% more than those living in other homeland areas (3).
-
Camp areas are more ethnically fragmented than other parts of homelands, despite the regime's aim of homogeneous homelands. (p. 23)
The average fragmentation index for areas near resettlement camps is 0.256 compared to 0.232 of areas further away from the camps
-
Placebo camps placed on demographically matched wards produce mostly insignificant coefficients, supporting a removal effect over community characteristics. (p. 19)
-
Population growth did not differ between camp and comparison areas, which weakens the selective-migration concern. (p. 13)
resettlement and comparison homeland areas have comparable levels of population growth (annual rate of 1.66% vs. 1.72%, p-value=0.82, see Table 1).
-
Removal scale and conditions: at least 3.5 million removed to Bantustans; the population of Qwaqwa rose from 21,000 to 200,000 in 1971-78 amid overcrowding and violence. (p. 5)
most studies estimate that a total of at least 3.5 million people were removed to Bantustans over the next decades (SPP 1983, Unterhalter 1987).
-
The author concedes that the identification tests remain inconclusive, and the data cannot separate relocated people from prior residents. (p. 14)
it is important to emphasize that tests (e.g. of selective migration) remain inconclusive with the historic data available.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| Surplus People Project (SPP) report, Forced Removals in South Africa, Volumes 1-5 (1983): geocoded resettlement camp locations and numbers removed | commission or official report | Fieldwork 1980-1983; published 1983 | Provides the location and size of resettlement camps, used to build the treatment variable (proximity to camp, radius scaled by number relocated). |
| South African population census 1996-2011, ward-level data (4,278 wards) | population census | 1996, 2001, 2011 | Used for population growth comparison (migration check), Mahalanobis matching for placebo camps, and 2011 ethnic fractionalization index. |
| South African Statistical Yearbook 1982 (growth point locations) | official statistical series | 1982 | Locates growth points to control for distance to industrial areas in the robustness test. |
| National Income Dynamics Study (NIDS) 2008 wave | household or labour survey microdata | 2008 | Shows trust levels are nearly identical in urban and rural areas, justifying that urbanicity is not driving the result. |
| Geographic data: homeland borders (Moerdijk 1981 map), distances to cities and borders computed in ArcGIS | maps spatial data | Contemporary and 1980s | Used to geocode respondents and camps and to compute distance to camp, homeland border, nearest city (200,000+) and growth point. |
| Tomlinson Commission report (1954) | commission or official report | 1954 | Cited for estimate that homelands could sustain 1.8 million people and for its view that homeland fragmentation produced incoherence. |
| Contemporary newspapers: The Voice (1978, 1979), The Star (1976), Daily Dispatch (1969), Financial Mail (1968) | newspapers periodicals | 1968-1979 | Anecdotal evidence of conditions, violence, and minister statements in resettlement areas. |
| Platzky and Walker (1985) Surplus People, including fieldwork quotations from Sada camp | secondary literature compilation | 1980s fieldwork | Qualitative evidence of the mechanisms (reliance on informal networks, camp conditions) and the Gazankulu Tsonga share. |
| Government statements and policy documents (1955 government report, 1959 White Paper, Minister of Bantu Administration speech 1973) | commission or official report | 1955-1973 | Quoted to document the stated rationale for ethnic homeland policy and removals. |
| Apartheid legislation (Group Areas Act 1950, Population Registration Act 1950, Bantu Authorities Act 1951, Bantu Resettlement Act 1954, Bantu Self-Government Act 1959, Homeland Constitution Act 1971, Land Acts 1913/1936) | legislation gazette | 1913-1971 | Describes the legal basis of the homeland system and the removals. |
| Secondary estimates of total number removed (SPP 1983, Unterhalter 1987); Christopher 2001 atlas; Rogers 1980 population data for Qwaqwa | secondary literature compilation | 1960-1983 | Provides aggregate figures on people removed and homeland populations that frame the treatment scale. |
New data¶
Newly digitized and geocoded dataset of 66 resettlement camps (45 with exact SPP locations) with estimated numbers relocated, linked to geocoded Afrobarometer respondents, ward-level census covariates and a ward-level ethnic fractionalization index; analysis samples of roughly 700-1,700 respondents.
Inference¶
Quantification: Econometric (causal design).
Quasi-natural-experiment: OLS comparison within former homelands of respondents near SPP resettlement camps (radius scaled by camp size) versus those within 60km, with homeland, province, wave and ethnicity controls, clustered SEs, age-cohort interactions for persistence, robustness checks (distance controls, clustering, placebo camps via Mahalanobis matching on census wards) and a within-camp diversity-trust slope comparison.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| People living near former resettlement camps report higher trust in relatives, neighbours and people in general than others in the same homeland. (p. 14) | Afrobarometer regressions with SPP camp proximity (Table 2); robust to distance controls and placebo test, though neighbour trust is insignificant in the binary specification. | Partial data |
| Residents near camps experience less theft and physical violence than others in homelands. (p. 15-16) | Afrobarometer crime items regressed on camp proximity (Table 3); significant mainly for physical violence in Panel A and in Panel B continuous measure. | Partial data |
| Inter-ethnic trust is higher near camps, with the effect concentrated among cohorts born after 1975, suggesting persistence. (p. 17) | Afrobarometer waves 3-4 trust-in-ethnic-groups items with cohort interaction (Table 4); small samples (N≈680). | Partial data |
| Areas near resettlement camps are more ethnically fragmented than other parts of homelands, despite the regime's goal of homogeneous homelands. (p. 23) | 2011 census ward-level fractionalization index (0.256 vs 0.232) plus 1974 Statistical Yearbook and SPP Gazankulu Tsonga share. | Systematic data |
| Diversity is positively associated with inter-ethnic trust only in camp areas (negative elsewhere), pointing to contact rather than adversity as a mechanism. (p. 25) | Fitted slopes of fragmentation vs inter-ethnic trust (Figure 4) comparing camp and non-camp areas; slope difference p=0.07; author calls it inconclusive. | Partial data |
| Results are not explained by selective post-apartheid migration, location of camps near borders/cities/growth points, or demographic differences in camp areas. (p. 13) | Census 1996-2011 ward growth rates (1.66% vs 1.72%), controls in Table 5, and Mahalanobis-matched placebo camps (Table 6). | Partial data |
| Forced removal under apartheid created economic hardship and overcrowding in resettlement camps, with both conflict and solidarity. (p. 6) | Qualitative accounts: newspapers, Rogers 1980, Platzky and Walker 1985, Tomlinson Commission figures; no systematic data. | Illustrative or anecdotal |
The author's caveats¶
Author notes the SPP data's accuracy and representativeness are unclear, no data exist on who was actually relocated so effects are community averages, the pre-removal social capital assumption cannot be tested, selective migration tests are 'inconclusive with the historic data available', census data begin only in 1996, KwaZulu camp sizes are missing, and the 1974 Yearbook's reliability is doubtful.
Gaps in the evidence¶
The argument lacks pre-removal or panel data on trust, individual-level information on who was relocated, and data on within-camp ethnic mixing and network formation, so persistence versus selection and the diversity/adversity channels are not identified. Treatment is inferred from SPP camp locations at town level with imputed radii and a density assumption, and the Transkei and KwaZulu are missing. Archival removal records, linked census or NIDS microdata with migration histories, and village-level survey modules on networks could test this better; some of these exist now but not for the pre-1994 period.
Replicability: Data described but not public.