Skip to content

Strike Panel

Strike records compiled from the strike reports in Work in Progress, a periodical of the Southern African Research Service (Braamfontein), covering industrial action from the late 1970s to 1990. Work in Progress reported only part of the strikes in the official statistics (see Coverage).

Key Files

File Rows Description
strikes_master_csv.csv 2,678 Master strike incident panel: 2,678 strikes from the Work in Progress strike reports (issues 6–69), dated c. 1970–1991; 1,308 dated 1984–90, 462 undated
strikes_locations.csv 2,678 Location/geographic crosswalk
strikes_with_macro_controls.csv — Strike data with macro controls
working_papers_converted/ 31 files Rao's Infocube transcriptions of the Work in Progress strike tables (issues 33–71)
working_papers_extracted/ ~480 files Machine-extracted table fragments
wip_extracted/ 25 issues AI extraction from Work in Progress issues 3–32 (1978–1984); not reconciled with the master file

Coverage

  • Period: dated c. 1970–1991, mostly 1977–1990
  • Incidents: 2,678 strikes in the master panel, from Work in Progress issues 6–69. 1,308 are dated 1984–90, 907 are earlier (mostly 1977–83, with some retrospective 1970s entries) and 462 are undated. Issues 3–5, 34, 50–51, 56–57, 62–63 and 70–71 are not represented. The 2026 AI extraction of issues 3–32 overlaps the hand-coded earlier rows and has not yet been reconciled with them.
  • Strike file for counts: data/derived/strikes_clean.csv in the strikes-apartheid-jpe repository (2,764 rows: the 2,678 master rows plus 86 page-read additions) adds record types (strike, dispute, strike and dispute, other action, uncoded), a stayaway flag, ISO dates with a precision field, Work in Progress issue and page references, and a corrections ledger. Use it for any count. In issues 6–19 the master file uses a bullet as a "not applicable" placeholder, which must not be read as union affiliation.
  • Coverage against official statistics: the WIP reports hold 26–29% of the stoppages recorded by the Department of Manpower for 1979–90, from about 50–64% in 1980–81 to about 10–20% in 1987–88. Draw WIP-based series next to the official series.
  • Firm matching: 294 strikes at 53 JSE-listed firms have usable returns. Datastream holds no pre-1990 history for the firms behind a further 161 matched strikes.

Pre-1984 Extension (2026-03-29)

wip_extracted/ — Claude API extraction from 25 Work in Progress PDFs (issues 3–32, Jan 1978 – Jul 1984). Uses text extraction where OCR layer exists, Claude Vision for image-only pages.

Output format: Company, Area, Region, Union, Issue, Workers, Date, WP_Number, WP_Date, Page

Early issues (1978–79) are primarily editorial with few structured strike entries. Issues 21 and later (1982–84) have consistent Company/Date/Workers/Union labeling yielding 20-40 incidents per issue.

Script: ops_admin/scripts/digitization/extract_strike_wps.py

Variables

  • Company, location, union, number of workers and date, as reported in Work in Progress
  • Work in Progress issue
  • Strike cause, coded for strikes matched to listed firms
  • Geographic identifiers for crosswalking (strikes_locations.csv)

The 2013 master file has no codebook. The rebuilt file in the strikes-apartheid-jpe repository documents its own columns.

What the market data show

These results are reproduced by build/run_all.py in the strikes-apartheid-jpe repository and reported in the book (Chapter 5).

  • Firm-level bound. Across 294 strikes at 53 firms, the average abnormal return over days 0 to +2 is −0.44% (95% interval −0.94% to +0.06%). An average three-day loss above about 1% of a struck firm's value is ruled out; most reactions are negative, but there is no measurable aggregate.
  • Sector returns. Across 980 day-dated stoppages in the seven sectors with a pre-1990 index, the striker-weighted effect is −0.0008% per 1,000 strikers (placebo p = 0.12).
  • The 1987 NUM strike (9–30 August, Work in Progress 50, p. 59). Gold-mine shares moved −2.1% against the market and gold (placebo p = 0.39), while the mining houses rose.

Provenance

  • Source: the strike reports in Work in Progress (Southern African Research Service), issues 6–69; source PDFs in raw_archives/05_labor/strikes/
  • Transcription: in 2013 Rao's Infocube typed 183 pages of strike tables, taking only company, union, workers and date. Research assistants at Columbia then filled in harmonized start dates from the issues (Quitze Valenzuela-Stookey in August 2013 and Marco Antonio Acosta from September 2013), coded strike causes in five categories for the strikes matched to listed firms (Quitze, August 2013, apparently extended by Marco), and helped match striking firms to listed companies: to Datastream names by Laurence Wilse-Samson and Quitze in 2013, and to McGregor names under matching rules written by Sébastien Turban in 2014. Who added the union-federation fields, and whether a planned recoding of political strikes was done, is not recorded
  • Rebuilt file: data/derived/strikes_clean.csv in strikes-apartheid-jpe, with page references and a corrections ledger (October 2026)
  • Licence: the incident records and coding are project work; share-price data matched to them are licensed and are not shared (see Licensing & Access)