Skip to content

South African Apartheid Economy Data Archive

Digitized historical datasets on the South African economy under apartheid, 1948 to 1994, with the record of how each was transcribed and checked, and a guide to the scholarship that uses the same sources. They were assembled for a paper on the 1974 mine-labour shock and a book on the political economy of apartheid's decline, and are described here so that others can use them.

Data

Agricultural, mining and manufacturing censuses, mine recruiting, population, forced removals, elections, strikes, wages and financial series, with the built panels.

Provenance

How each dataset went from printed page to panel, what has been checked and by whom, the corrections ledgers, and what is still open.

Literature

32 debates, 808 works and 897 primary sources in the scholarship on the apartheid economy. Reading notes written by an AI model, not yet read by a person.

The datasets

Five datasets share the 279 magisterial districts as their unit and can be joined into one district panel. The rest are national, sectoral or firm-level.

Dataset Years Unit Main variables
Agricultural census 1949–1990 District × year Farm employment by race, wages, tractors and combines
TEBA mine recruiting 1953–1975 District × year Mine recruits by district of origin
Population 1904–1996 District × census Population by race, urban and rural
Forced removals 1960s–1980s Event, district 646 rows from the Surplus People Project, 422 removals in the district panel
Elections 1938–1992 Division × election Votes by party, turnout, unopposed seats
Strikes c. 1970–1991 Incident Firm, union, workers, dates, from Work in Progress
Manpower Survey 1965–1994 Sector × occupation × year Employment by race and gender
Mining census 1964–1987 Mine, mineral Output, costs, profits, employment
Manufacturing census 1948–1985 Industrial region Output and employment by industry
Industrial wages 1973–1988 Industry × occupation × area Gazetted minimum wage rates
Sanctions and macro series 1968–2015 Daily, monthly Gold price, exchange rate, JSE indices
Share prices and company boards 1973–2015 Firm JSE market values, directors

The data catalog lists every dataset with its files, the built panels and the open quality items.

How the data were made

The tabular sources were typed into spreadsheets under contract in 2012–13, through two channels (oDesk freelancers and Rao's Infocube), and neither left a record of quality checks. For the agricultural census of 1961–83, Christopher Boone's research assistants later corrected the oDesk spreadsheets by hand against the printed volumes (2023–26), and the agricultural panel is built from their version. For the 1972–80 variables used in the 1974-shock paper, every cell on which the versions disagreed was read from the page image by an AI model; the 62 readings that change a value are rows in a corrections ledger, and a second AI reading agrees with 60 of them. No person has yet re-read them.

Each dataset carries a status term: transcribed, internally checked, machine-read against the page, checked against the printed source by a person, or reproduced in analysis. Provenance and checks gives the status of each, the checks, and the open items.

Data quality varies by dataset

The 2013 contract transcriptions contain digit misreadings, shifted rows and copied totals. Check a dataset's status on the Provenance and checks page before using it.

Associated research

  • Boone and Wilse-Samson, "Coalition Fracture Under Labor Scarcity: Evidence from South Africa's 1974 Mining Crisis". Between 1977 and 1987, right-wing vote share rose 14.2 percentage points more in electoral divisions east of the Eiselen Line than west of it (172 divisions, 434 division-elections), with a flat 1970 pre-trend (+0.51 pp, p = 0.54). The paper's numbers are reproduced by a scripted replication package; see Replication.
  • Wilse-Samson, The Cost of Apartheid: Labor Scarcity, White Politics, and the Fall of Racial Capitalism in South Africa, 1969–1994, a book manuscript in revision. Its strike and sanctions chapters draw in part on joint work with Suresh Naidu and Sebastien Turban.

Access and licence

The CSV files are not in the git repository. To request them, write to Laurence Wilse-Samson at NYU Wagner. The validation scripts, the panel builders, the corrections ledgers and the source of this documentation live in a private repository alongside the data and are supplied with it on request. The project's own transcriptions of public sources, its corrections ledgers, crosswalks, panels, code and this documentation are released under CC BY 4.0. Licensed and restricted material (TEBA records, DataFirst microdata, Datastream, Compustat, CEIC, McGregor BFA extracts and others) keeps its holder's terms and is not redistributed. Licensing and access gives the position dataset by dataset; Citation says how to cite.