Labor Unrest in South Africa, 1870-1990¶
Mark Beittel (1995). Labor Unrest in South Africa, 1870-1990. Review (Fernand Braudel Center), Vol. 18, No. 1, special issue: Labor Unrest in the World-Economy, 1870-1990. Research Foundation of SUNY for the Fernand Braudel Center.
| Type | journal article |
| Scope | Economics and economic history of apartheid |
| Method | descriptive quantitative |
| Period studied | 1870–1990 |
| Regions | South Africa, Witwatersrand, Kimberley, Natal, Eastern Cape |
| Themes | unions strikes, resistance social movements, methods data sources, labour markets wages, politics ideology elections, mining migrant labour |
| Citation key | beittel1995a |
Argument¶
Beittel gives a stylized periodisation of South African labour-capital conflict from the 1870s to 1990, in three broad periods (1900s-1920s, 1930s-1950s, 1960s-1980s). He then uses it to test two quantitative indicators of strike waves: the World Labor Research Working Group (WLG) indicator, based on newspaper mentions, and the official government strike statistics. Applying the Shorter-Tilly method to strike counts gives many false positives. A days-lost version (Bordogna-Provasi) fits the stylized facts fairly well, though it reflects only the magnitude of strike activity. The WLG indicator is better at picking out turning points and politicised or non-strike unrest, but it carries biases from its reliance on the Times (London). A comparison with the New York Times series shows broad consistency and also the NYT's own weaknesses.
Evidence and method¶
Government strike statistics 1910-1990 (National Manpower Commission 1986; Central Statistical Services 1991); WLG database of mentions in the Times (London) Index (2,306 mentions) and the New York Times Index (487 mentions), 1906-1990; secondary literature on South African labour history (Lewis, O'Meara, Moodie, Bonner, Davies and others).
Place in the debates¶
- How to measure waves of labour unrest: strike counts vs working days lost vs newspaper-mention indicators (see Data reliability and econometric method). Argues that strike counts give false positives. Working days lost is a fair proxy but tracks magnitude only. The WLG mentions indicator is the better guide to disruptive, turning-point unrest.
- Labour history (formal organisation and strikes) vs social history (informal and community protest) (see Worker consciousness, strikes and conflict on the mines and shop floor). Sees the richness of South African labour history as coming from combining both approaches, and notes that official strike statistics miss political stay-aways.
- Reliability of the official strike statistics (see Data reliability and econometric method). Sceptical: the series is poorly documented, incomplete before 1917, excludes the nominally independent homelands from the 1970s, and probably understates 1986 unrest.
Findings¶
-
Government strike statistics are the only long quantitative series, but they are designed to capture only conventional strikes and so exclude political stay-aways. (p. 7)
Hence, the statistics are designed to measure only conventional strikes, and not work stoppages called to protest, for example, government policies.
-
The 1986 official figure is implausibly low once stay-aways are counted; a consultant's conservative estimate is at least three times higher. (p. 7)
the government reports 1.3 million working days lost for that year while a reputable industrial relations consultant maintains that a conservative figure which took into account stay-aways would be at least three times as high
-
Homeland exclusion shows the state's ideology shaping the statistics. (p. 7)
Transkei, Bophuthatswana, Venda, and Ciskei, which were excluded from the statistics of strike activity since the 1970's
-
The original Shorter-Tilly technique applied to strike counts produces too many waves and misfits the stylized facts. (p. 8)
Applying the same technique to the South African data generates a list of 27 wave years- more than a third of the total years.
-
Both the WLG and days-lost indicators pick out the key moments in South African labour history. (p. 8)
Both the WLG "mentions" indicator and the days-lost indicator identify as wave-years 1922, 1947, 1973-74, 1980, and 1985-87
-
The 1950s politicised protest is visible only in the WLG indicator, because stay-aways were not counted as strikes. (p. 11)
the WLG indicator correctly identifies 1957 and 1959-60 as wave-years, while the days-lost indicator does not identify any waves during the 1950's.
-
The days-lost indicator over-identifies waves in the 1979-87 period and so cannot distinguish disruptive from routinized protest. (p. 12)
This proliferation of waves shows that the indicator reflects only the magnitude of strike activity, without distinguishing between innovative/disruptive and routinized forms of protest.
-
The WLG indicator misses the unrest after each world war and during the Second World War; the author's likely explanation is the Times' wartime paper shortage. (p. 13)
The most likely explanation for the immediate postwar underestimation of labor unrest is the paper shortage experienced by the Times (London) during those years
-
The WLG series for South Africa leans heavily on the Times (London), whose coverage is skewed by British interests, especially early in the century. (p. 15)
for the period 1906 to 1930, the ratio is more than 10:1 while for the period 1960 to 1990 it is 3:1.
-
Comparing the two newspaper series shows high consistency, but the NYT fails to separate labour unrest from other popular protest. (p. 15)
ten of the fourteen wave-years identified by each indicator are common to both of them, which demonstrates a high degree of consistency between the two in identifying wave-years.
-
The period 1985-87 is the peak of postwar labour-capital conflict and a pivotal shift in relations between labour, capital and the state. (p. 6)
the level, extent, and duration of Black labor unrest during 1985-87 mark these three years the peak of labor-capital conflict not only for the decade but for the entire postwar period.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| Government strike statistics (number of strikes, workers involved, working days lost), 1910-1985, in National Manpower Commission, Certain Aspects of Strikes in South Africa, RP 113/1986, Annexure C, Table 1 | official statistical series | 1910-1985 | Main quantitative series of strikes, used to compute wave-years by the Shorter-Tilly and days-lost methods and to compare with the WLG newspaper indicator. |
| South African Labour Statistics, 1991 (strike statistics for recent years) | official statistical series | c.1986-1990 | Extends the government strike series to 1990 for the wave-year calculations. |
| South African Labour Statistics, 1989 (p. 473, redefinition of strike incidents) | official statistical series | 1980s | Cited for the changed definition of strikes in the 1980s, which is then questioned against the 1986 days-lost figure. |
| Union of South Africa, Official Year Book No. 8, 1923 (p. 217) | official statistical series | to 1917 (published 1927) | Original publication noting that pre-1917 strike statistics are extremely incomplete. |
| Markham & Matiko (1987), 'State of the Unions', South African Labour Bulletin; industrial relations consultant's estimate of days lost incl. stay-aways | newspapers periodicals | 1986 | Alternative estimate of 1986 working days lost, at least three times the official 1.3 million, used to discredit the official series' claimed coverage of political strikes. |
| Private labor consultancy strike figures (Howe 1984, Indicator South Africa) | secondary literature compilation | late 1970s-1984 | Noted as alternative strike counts that differ from but partly derive from government statistics. |
| World Labor Research Working Group (WLG) database of mentions of labor unrest in the Times (London) Index | newspapers periodicals | 1906-1990 | Source of the WLG indicator (2,306 mentions), whose wave-years are compared to the stylized facts and the strike statistics. |
| New York Times Index mentions of labor unrest in South Africa | newspapers periodicals | 1906-1990 | Second newspaper series (487 mentions) used to test the reliability of the Times (London)-based indicator by calculating its own wave-years. |
| Table 1 / WLG Database wave-years for South Africa 1911-90 | newspapers periodicals | 1911-1990 | Table listing wave-years and major waves from the WLG indicator. |
| Secondary accounts of the 'stylized facts' of South African labor history (e.g., O'Meara 1975 on wartime wage gap; Bonner, Davies, Moodie, Keenan, Lodge, van Onselen) | secondary literature compilation | 1870-1990 | Provide the benchmark chronology of labor unrest against which the indicators are judged. |
Inference¶
Quantification: Descriptive tables.
Comparison of three time-series indicators (WLG Times mentions, government days-lost, strikes and workers involved) using the Shorter-Tilly wave-year rule (>=50% above mean of the previous five years), judged against a benchmark chronology ('stylized facts') drawn from the secondary literature; the New York Times index is used as a robustness check on the Times (London) bias.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Applying the Shorter-Tilly rule to the number of strikes yields many false positives and misses key years, so it is a poor guide to waves. (p. 93) | Government strike statistics (NMC 1986; CSS 1991) with wave-year rule applied by the author and compared with the stylized-facts chronology. | Partial data |
| The WLG indicator and days-lost indicator concur on the key waves 1922, 1947, 1973-74, 1980 and 1985-87. (p. 93) | Table 1 (WLG database) compared with days-lost wave-years from the government series. | Partial data |
| The WLG indicator captures politicized protest such as 1950s stay-aways that official strike statistics exclude, and distinguishes disruptive from routinized strikes. (p. 96) | Comparison of WLG wave-years with days-lost wave-years; the official series' definition of strikes as employer-reported work stoppages with specific demands. | Partial data |
| Official strike statistics are poorly documented, incomplete before 1917, politically shaped (homelands excluded) and likely understate 1980s unrest. (p. 92) | Notes to the statistics, the Official Year Book, and Markham & Matiko's estimate against the 1.3 million days-lost figure for 1986. | Partial data |
| The WLG indicator misses 1919-20, 1941-44 and 1946 largely due to Times (London) underreporting, including paper shortages and press bias. (p. 98) | WLG mention counts (1919: 50, 1920: 26, 1946: 25) versus wave thresholds, and the New York Times series, which does flag 1946. | Partial data |
| The Times (London) and New York Times indicators agree on ten of fourteen wave-years, but the NYT confuses popular protest (Soweto 1976) with labor unrest, making it a weaker indicator. (p. 100) | Wave-years computed from the two newspaper-index series and compared with the stylized-facts chronology. | Partial data |
| The Times (London) overreports South African labor unrest relative to the New York Times, especially early in the century (over 10:1 for 1906-30, 3:1 for 1960-90), reflecting British interests. (p. 100) | Annual counts of mentions in both indexes (means 27 vs. 6). | Systematic data |
The author's caveats¶
The author notes that the strike statistics are poorly documented, incomplete before 1917, excluded the nominally independent homelands, and changed definitions in the 1980s; three years (1915, 1925, 1929) are incomplete and were treated as zeros or missing. He also states that the newspaper indicators rest on index mentions rather than coded article content, and that biases in the Times (London) cannot be resolved without close examination of the articles.
Gaps in the evidence¶
The 'stylized facts' benchmark comes from the secondary literature and is not tested against any independent event dataset, so the comparison of indicators is partly circular. Coding of article content or events in the South African press, archival police and employer reports, and union records would test the indicators better; these existed and could be consulted. Stay-away and political-protest counts, such as consultancy data from the 1980s, would help assess how much unrest the official series omitted.
Replicability: Data described but not public.