The spatial persistence of population and wealth during apartheid: Comparing the 1911 and 2011 census¶
Waldo Krugell (2014). The spatial persistence of population and wealth during apartheid: Comparing the 1911 and 2011 census. ERSA working paper 459 (prepared for special issue on the Economics of Apartheid, Economic History of Developing Regions). Economic Research Southern Africa (ERSA).
| Type | working paper |
| Scope | Economics and economic history of apartheid |
| Method | descriptive quantitative |
| Period studied | 1911–2011 |
| Regions | South Africa, Gauteng/Witwatersrand, homelands/border regions |
| Themes | bantustans removals influx, demography urbanisation, inequality poverty unemployment, manufacturing industrial policy, origins segregation, methods data sources |
| Citation key | krugell2014 |
Argument¶
The paper asks whether apartheid's attempts to disperse people and industry (homelands, decentralisation, influx control) halted the agglomeration normally associated with development. Using place-level populations from the 1911 and 2011 censuses matched to 1911 district boundaries, it estimates a rank-size (Zipf) exponent q and finds it rose from 0.783 to 1.356, indicating agglomeration not stopped but carried to over-concentration. A Spearman rank correlation of vehicles per person (a wealth proxy) suggests wealth stayed concentrated in places of initial white settlement and big cities. It concludes that homeland development failed and that the roots of spatial inequality predate apartheid.
Evidence and method¶
Transcribed place/district-level population and vehicle-ownership data from the published 1911 census; 2011 census data from Statistics South Africa aggregated to 1911 district boundaries; a 1996 census check (q = 1.25); secondary literature on apartheid urbanisation and decentralisation policy (Posel, Davenport, Bos, Wellings and Black).
Place in the debates¶
- Effectiveness of apartheid decentralisation and homeland development in dispersing economic activity (see Homelands, separate development, decentralisation and removals). Argues these policies failed to stop agglomeration; border industries and homeland growth points had limited impact, and wealth did not spread to border regions.
- Origins of South African inequality (apartheid vs pre-Union roots) (see Long-run legacy and persistence of apartheid). Argues wealth was already highly unequally spread in 1911, so apartheid exacerbated but did not originate spatial inequality, aligning with Fourie and Von Fintel and Wilson.
- Optimal city size / over-concentration of South African urban system (see Urban segregation, Group Areas and the apartheid city). Supports Van der Merwe and Naudé and Krugell that there are too many small towns and too few, over-large cities; the counterfactual without apartheid is treated as speculative.
Findings¶
-
The 1911 rank-size exponent was well below 1, implying small, similar-sized towns. (p. 10)
Thus, in 1911 the q was 0.783 and notably smaller than one.
-
By 2011 q exceeded 1, which is unusual and implies apartheid could not stop agglomeration. (p. 10)
The 2011 results show a q value of 1.356.
-
Apartheid interventions failed to halt urbanisation despite homeland rural development efforts. (p. 10)
It is clear that apartheid’s interventions could not stop agglomeration forces.
-
A 1996 census check shows the high q is not only post-apartheid; agglomeration continued to over-concentration. (p. 10)
Agglomeration was not only an apartheid-era phenomenon, but has since continued to the point of over-concentration.
-
Wealth became concentrated in initial white settlements and large cities, with no spread to border regions, supporting the view that homeland development failed. (p. 11)
wealth became increasingly concentrated in places of initial white settlement and the large urban agglomerations.
-
Mean place population rose from about 29,000 to 252,000 between 1911 and 2011 (on 1911 boundaries). (p. 9)
The average number of people per place increased from approximately 29 000 to 252 000.
-
Literature reviewed holds that border industries and growth points failed. (p. 8)
point out that the border industries were a complete failure.
-
Urban labour preference and influx control could not slow rural-urban migration. (p. 7)
the system was unable to slow rural to urban migration
-
Without apartheid South Africa would plausibly have had more large cities (counterfactual). (p. 10)
Thus, without apartheid, South Africa would have had more ‘large’ cities.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| 1911 Census of the Union of South Africa, place/district-level population and vehicle ownership (published volume, transcribed by the author) | population census | 1911 | Supplies district populations for the 1911 rank-size regression and vehicle counts (total and motorised) as the wealth proxy for the Spearman correlation. |
| Census 2011 place-level data, digital (Statistics South Africa 2013), aggregated to 1911 district boundaries | population census | 2011 | Provides 2011 population and vehicle ownership, aggregated to 1911 boundaries, for the 2011 rank-size estimate and the wealth persistence comparison. |
| 1996 population census (place-level, used for robustness estimate of q) | population census | 1996 | A similar rank-size estimate (q=1.25) is used to show that over-concentration is not only a post-1994 phenomenon. |
| First industrial census (referred to via Bos 1992) | industrial mining census | c. 1915-1921 (not stated) | Cited via secondary literature to indicate high urbanisation and industrial development in the main cities early in the Union period. |
| Bos (1992) population growth rates for 1911-1921 rural and urban areas | secondary literature compilation | 1911-1921 | Provides annual population growth rates for rural versus urban areas as background for the take-off of urbanisation. |
| Apartheid-era commissions, reports and policy documents (Lagden, Fagan, Tomlinson, Riekert, Franzen, Klue reports; White Papers; physical planning laws) as described by secondary authors | commission or official report | 1902-1986 | Narrative context listing the policies and reports that shaped the apartheid spatial economy; the documents themselves are not analysed. |
| Native (Urban Areas) Act of 1923 and earlier segregation legislation (Native Reserve Locations Act 1902, Native Locations Act Natal 1904, OFS Municipal Ordinance 1903) | legislation gazette | 1902-1923 | Background on the legal basis of urban segregation, taken from Davenport's account. |
| Estimates of q for other countries and cities (Netherlands, USA) from Brakman et al. (1999) and Fujita et al. (1999) | secondary literature compilation | 1600-1990 | Benchmarks for interpreting South Africa's q values against Zipf's Law. |
New data¶
The author transcribed 1911 census data from the published volume for roughly 200 districts (population and vehicle counts, total and motorised). These were matched to 2011 census data aggregated to 1911 district boundaries, giving a place-level 1911-2011 comparison dataset of population and vehicles per person.
Inference¶
Quantification: Statistical analysis.
Rank-size (Zipf) regressions of log population on log rank for 1911 and 2011 (plus a 1996 check) on a fixed set of 1911 districts, judged against q=1. Wealth persistence is assessed with Spearman rank correlations of vehicles per person per place across 1911 and 2011. Policy context is drawn from secondary literature, and the counterfactual about apartheid is speculative.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| In 1911 q was 0.783, well below 1, meaning places were too small and similar in size, with urbanisation rather than localisation economies. (p. 9) | OLS regression of ln population on ln rank using the transcribed 1911 census districts (Table 3). | Systematic data |
| By 2011 q was 1.356, above Zipf's Law, indicating agglomeration continued to over-concentration despite apartheid's dispersal policies. (p. 9) | Rank-size regression on 2011 census data aggregated to 1911 boundaries, with a 1996 census check (q=1.25). | Systematic data |
| Apartheid's interventions (homelands, decentralisation) could not stop agglomeration, and rural development policy did little to alter economic forces. (p. 9) | Inferred from the change in q between 1911 and 2011 and from secondary literature on decentralisation failure; there is no counterfactual or policy-level data. | Partial data |
| Wealth, proxied by vehicles per person, is positively correlated across 1911 and 2011 at the district level, so wealth concentrated in places of initial white settlement and large urban areas. (p. 10) | Spearman rank correlations in Table 4 (0.377 and 0.465) between 1911 and 2011 vehicle ownership per person. | Partial data |
| Spatial inequality in wealth already existed in 1911, so the roots of inequality predate apartheid. (p. 10) | The 1911 vehicle proxy and its correlation with 2011, plus citation of other scholars; the correlations are modest (0.31-0.47). | Illustrative or anecdotal |
| Without apartheid, South Africa would have had more large cities and a q closer to 1. (p. 10) | Speculation from Zipf's Law and the place-dependence argument; no counterfactual modelling or data. | Asserted |
| All of today's large cities were already large in 1911, and the largest and smallest places show persistent geographic patterns. (p. 8) | Descriptive ranking of 1911 district populations (Table 1) and vehicle counts (Table 2). | Systematic data |
The author's caveats¶
The author notes that 1911 census districts are far fewer and coarser than 2011 enumerator areas, so 2011 data are aggregated to 1911 boundaries (for example Zoutpansberg is a large area, not the modern magisterial district). Information on individual wealth is scarce, so vehicle ownership per person is used as a proxy, and only this measure is available for both years. The 1911 census is noted as the only one to cover the black population comprehensively.
Gaps in the evidence¶
The study compares only two end-point censuses and omits intermediate ones (1921-1991), so it cannot show when or how apartheid policy affected the rank-size distribution, and there is no counterfactual or policy-exposure variable to isolate apartheid's effect. Vehicle counts in 1911 (mostly horse-drawn, with very few motorised) are a weak and changing wealth proxy; income, property, tax or consumption data would be better, though 1911 data of this kind are limited. Intermediate censuses, magisterial district series, urban/rural definitions, and homeland and border-industry data (which exist) could test the claims. The rank-size regression ignores the arbitrary size of 1911 districts, which makes the large q difficult to interpret.
Replicability: Data described but not public.