South African Apartheid Economy Data Archive¶
This archive contains digitized historical datasets covering the South African economy during apartheid (1948-1994). It is designed to support quantitative research into the economic mechanisms of apartheid's rise, transformation, and collapse.
What Research Can This Data Support?¶
The datasets here enable analysis of several interconnected questions about apartheid South Africa:
- Labor market shocks and mechanization: How did the 1974 mine labor crisis propagate through the economy, tripling Black mine wages and driving a national wage revolution in agriculture? Combines rose 30% and tractors 13% nationally between 1972 and 1980 while regular Black employment fell 10% (Agricultural census + TEBA recruiting panel)
- Electoral politics of labor market change: How did differential exposure to Black labor market competition fracture the white electoral coalition? The Eiselen Line — the boundary of the Coloured Labour Preference Area — divides constituencies that competed directly with mines for Black labor (east) from those that did not (west). Exact-line event study shows east-of-Eiselen constituencies shifted +14.2pp toward right-wing parties by 1987 (p<0.001, N=426), while the Cape moved toward reform (Elections + district-level economic data)
- Industrial action and regime change: Did strikes in the 1980s shift investor expectations about apartheid's durability? Across 222 matched strikes, the cumulative announcement-day Rand value lost was R 606 million (~$259 million USD), with the 1990 transition year alone accounting for 45% of the total. Struck and non-struck firms had identical Afrikaans board representation in 1983 (18.9% vs 18.8%, p = 0.94), suggesting strikes operated as a diffuse pressure across the formal economy rather than as targeted action against regime allies (Strike panel + share prices)
- Sanctions and the dual exchange-rate regime: Why did studies examining the commercial rand find near-zero sanctions effects while studies examining the financial rand and credit spreads found large effects? The September 1985 reimposition of the financial rand insulated the trade-linked commercial rate from capital-account sanctions — the bite was channelled through the capital account, not through trade (SARB monthly financial-rand series)
- Structural transformation: How did the economy shift from primary production to manufacturing, and how did racial employment barriers evolve? (Manpower survey + manufacturing census)
Datasets at a Glance¶
District-Level Panels (joinable on magisterial district)¶
These datasets share a common geographic unit — the 279 magisterial districts — and can be merged into a single panel.
| Dataset | Years | Level | Records | Key Variables |
|---|---|---|---|---|
| Agricultural Census | 1949-1990 | District | ~3,000 district-years | Employment by race, wages, mechanization |
| TEBA Mine Recruiting | 1953-1975 | District | ~800 district-years | Mine recruits by origin district |
| Population | 1904-1996 | District | 279 per year | Population by race, urban/rural |
| Forced Removals | 1960s-1980s | District | 631 events | Removal counts by category |
| Elections | 1938-1992 | Constituency | 2,000+ results | Votes by party, turnout |
National and Sectoral Panels¶
| Dataset | Years | Level | Records | Key Variables |
|---|---|---|---|---|
| Strike Panel | 1978-1990 | Incident | 4,100+ strikes | Company, union, workers, outcome |
| Manpower Survey | 1965-1994 | Sector-occupation | 165,583 | Employment by race, gender, sector |
| Mining Census | 1964-1987 | Mine | 67+ files | Gold production, costs, employment |
| Manufacturing Census | 1948-1985 | Region | 23+ files | Output, employment by industry |
| Industrial Wages | 1973-1988 | Industry-occupation | 825 files | Gazetted wage rates |
Financial and Macro¶
| Dataset | Years | Records | Key Variables |
|---|---|---|---|
| Sanctions & Macro | 1968-2015 | ~40,000 daily obs | Gold price, exchange rates, bond yields, credit spreads |
| Share Prices | 1973-2015 | 1,000+ equities | JSE market values, price-to-book |
See the full Data Catalog for all 1,900+ files across 24 categories.
Origin and Quality¶
These datasets were digitized from archival PDFs — primarily South African government census publications, TEBA/NRC mine recruiting records, and academic working papers. The original digitization was performed by oDesk contractors in 2013-14, with extensive validation and correction in 2025-26.
All data traces back to source PDFs through a documented provenance chain. See Provenance & Validation for error rates, correction logs, and validation scripts.
Data quality varies by year
The oDesk contractors sometimes introduced transcription errors (1-23% error rates depending on year). The agricultural census years 1949-1976 have been validated; other datasets are at various stages. Always check the provenance documentation before using a dataset.
Associated Research¶
- Boone & Wilse-Samson — "Coalition Fracture Under Labor Scarcity: Evidence from South Africa's 1974 Mining Crisis." Documents differential-adjustment coalition fracture: a shock to a coerced-labor equilibrium that produces simultaneous radicalization and moderation within a single ruling coalition because constituents adjust on different margins. Flagship result: east-of-Eiselen constituencies shifted +14.2pp toward right-wing parties by 1987 (exact-line event study, N=426, p<0.001), with a clean 1970 pre-trend; the Conservative Party carries +15.83pp of that divergence and the older HNP declines. Within the eastern recruiting regions, open districts moved away from verkrampte parties relative to closed districts, consistent with Proposition 2 of the formal model (voters who have already shed labor face smaller absolute rent losses). The district-level combines DiD is specification-sensitive (null under TWFE; the tractor placebo fails) and is voluntarily downgraded; the national mechanization trend (+30% combines, +13% tractors) is robust.
- Naidu, Turban & Wilse-Samson — "Diffuse Pressure: How Domestic Labor Unrest Imposed Costs on Apartheid's Formal Economy, 1979–1990." Three robust findings on a database of 2,680 strikes and 222 matched JSE-listed firms: (1) aggregate announcement-day Rand value loss of R 606 million (~$259 million USD), with 1990 = 45% of the total; (2) targeting null — struck and non-struck firms had virtually identical Afrikaans board representation in 1983 (18.9% vs 18.8%, p = 0.94); (3) temporal pattern — strike effects were essentially zero during the 1985–1987 State of Emergency and amplified during the 1990 transition. Individual strike effects reported as a range across specifications: −0.156% Dimson-corrected (primary, p = 0.078) and −0.235% canonical (p = 0.011).
- Naidu, Turban & Wilse-Samson — "Sanctions, Dual Currency Systems, and the End of Apartheid." The September 1985 reintroduction of the financial rand successfully insulated the commercial rand from sanctions: across 36 post-1985 sanctions events, the commercial rand's cumulative abnormal return is precisely zero ($+0.00\%$, $p = 0.99$), while the financial rand depreciated by 6.7 percentage points on average (marginally significant at $p = 0.13$ pooled; $+9.2$ pp at $p = 0.08$ for trade sanctions). The strongest single empirical anchor is a Bai-Perron endogenous structural break in the SA–US credit spread dated August 1984, $+683$ basis points (Sup-$F = 2{,}758$), eleven months before the Chase Manhattan debt withdrawal. The financial-rand discount peaked at $+37\%$ in August 1986 and collapsed by $13.8$ percentage points around the February 1990 unbanning of the African National Congress.
- Wilse-Samson — The Cost of Control: Labor Scarcity and the Unraveling of Apartheid, 1969–1994 (book manuscript, ~79,500 words, 27 figures, 17 tables). Reframes apartheid's collapse as the resolution of a differential-adjustment coalition fracture triggered by the 1974 shock. Central argument: the 1974 shock did not end apartheid, but it changed the cost structure of maintaining it; the 1980s unraveling is legible as a system that had lost the ability to buy time. Includes a formal theory appendix, data appendix with voluntary specification sensitivity (Section D.0) and gold-vs-labor wage decomposition (Section D.5), and a prefatory note ("How to Read This Book") with an evidence-grading table that ranks every major claim by confidence tier.
Access¶
CSV data files are not in the git repository (too large). To request access, contact Laurence Wilse-Samson at NYU Wagner.
Code, documentation, and panel-building scripts are all in the GitHub repository.
License¶
CC-BY-4.0. See Citation for how to cite.
Source Copyright
The underlying source documents (government census publications, TEBA/NRC records) may be subject to separate copyright. This license applies to the digitized datasets, code, and documentation created by the project team.