The changing spatial structure of economic growth in South Africa over fifty years¶
McCarthy, Jeff (2000). The changing spatial structure of economic growth in South Africa over fifty years.
| Type | working paper |
| Scope | Economics and economic history of apartheid |
| Method | descriptive quantitative |
| Period studied | 1910–1994 |
| Regions | South Africa, Witwatersrand/Johannesburg, Durban, Cape Town, Port Elizabeth, bantustans |
| Themes | bantustans removals influx, demography urbanisation, manufacturing industrial policy, methods data sources, origins segregation, transition post apartheid |
| Citation key | mccarthy2000 |
Argument¶
McCarthy argues that the spatial structure of South African economic growth has changed only glacially since the early twentieth century. It was set mainly by mineral-led growth on the Witwatersrand, with coastal port cities as secondary nodes. Using Gross Geographic Product (GGP) data by magisterial district from 1968 to 1994, he shows that non-metropolitan areas did relatively better in the 1960s and 1970s, when apartheid decentralisation was at its height, than in the 1980s and 1990s. Past growth patterns are the best predictor of 1990s growth, so he projects a scenario of continued concentration in the metros and the core, with a slight coastal tilt. The paper is a draft policy aid and does not judge whether these trends are desirable.
Evidence and method¶
Statistics SA/HSRC Gross Geographic Product (GGP) data by magisterial district for 1968, 1972, 1981, 1991 and 1993/4, with 1996 data not yet available. Also the 1960 census and Union Statistics for Fifty Years, Board, Davies & Fair (1970), Fair (1983), and Abedian & Standish (1992). Methods are regression of growth on population and on past growth, rank-ordering of acceleration rates, and a gravity model. GIS mapping is used for the scenarios.
Place in the debates¶
- Effectiveness of apartheid-era decentralisation and growth-point policy in reshaping the space economy (see Homelands, separate development, decentralisation and removals). Finds only a modest effect. Non-metro areas kept 89-95% of metro growth in the 1960s-70s but 84-85% in the 1980s-90s. Overall spatial patterns persisted, and policy rarely deflects historically formed growth patterns.
- Metropolitan versus non-metropolitan performance as the core of spatial restructuring (urbanisation literature) (see Urban segregation, Group Areas and the apartheid city). Metro areas always accelerated faster. The non-metro share was somewhat higher while apartheid decentralisation was strong, and growth shifted to metros and coasts afterwards.
- Path dependence versus policy-induced spatial change (see Urban segregation, Group Areas and the apartheid city). Stresses path dependence. Radical locational shifts occurred only with primary activity (mining), and the paper expects future growth to elaborate on past patterns.
Findings¶
-
Mining and agriculture made up half of GDP in 1911, a far larger share than today. (p. 4)
by 1911 approximately 50% of the country’s G.D.P. was comprised by just two sectors – agriculture and mining (this is very different to today when they collectively contribute some 20%)
-
Witwatersrand mining was the main force shaping the space economy from 1910 to 1960, with Johannesburg overtaking Cape Town. (p. 5)
The mining of the complex of minerals on the Witwatersrand, and its associated stimulatory impacts upon related economic activities such as manufacturing, was the single most important driving force shaping the changing S.A. space-economy during the period 1910-1960.
-
The mid-zone (platteland and bantustans) led acceleration in the 1960s but ranked last afterwards, while the coast moved from last to first. (p. 7)
The mid-zone (including both the platteland and bantustans) did best in the 1960’s, and came last thereafter.
-
Non-metro growth relative to metro growth was higher during the peak of apartheid decentralisation than in the 1980s-90s. (p. 9)
during the 1960’s and 1970’s, when the application of apartheid and decentralisation policies was most pronounced, the non-metropolitan areas did somewhat better than in the 1980’s and 1990’s
-
Population growth explains little of the spatial variance in GGP growth, whereas past growth explains much of it. (p. 9)
the regressions revealed weak relationships. This was especially the case during the 1960’s and 1970’s when less than 10% of the variance in G.G.P. growth could be predicted by population growth
-
1980s growth explains 80% of the variance in 1990s growth, which shows strong persistence. (p. 9)
1980’s growth with 1990’s growth = 80% of variance explained
-
Manufacturing, construction and services are the best predictors of 1990s spatial growth, while mining and agriculture are the weakest. (p. 10)
The weakest predictors of the overall spatial distribution of growth are mining and quarrying, and agriculture, forestry and fishing.
-
Over 90% of top-performing districts lie within 120 km of four metropolitan centres. (p. 16)
over 90% of those districts performing in the top league fall within a 120 km radius of the centre of just four metropolitan places: Johannesburg, Cape Town, Durban and Port Elizabeth.
-
Growth in the 1990s was highly concentrated in a few core metropolitan districts. (p. 17)
58% of new growth during the 1990s was realised within a few magisterial districts (core metropolitan in nature) which comprised less than 4% of the country’s area.
-
The paper projects that roughly 80% of new growth will occur where growth has recently occurred. (p. 11)
probably 80% of all new growth will be in areas where it has been in the recent past
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| Gross Geographic Product (GGP) data by magisterial district, 1968, 1972, 1981, 1991 and 1993/4 (Statistics SA/HSRC) | official statistical series | 1968-1994 | Core dataset assembled into a district file to compute GGP change for four sub-periods and to run the coastal/core/mid-zone and metro/non-metro analyses, regressions and scenarios. |
| Statistics SA 1995: Gross Geographic Product 1994 | official statistical series | 1993/4 | Used to correlate sub-sector growth rates across districts to identify the 'predictor' sectors for the 1990s. |
| 1996 GGP data (Statistics SA/HSRC) | official statistical series | 1996 | Not used: it was unavailable at drafting and is expected to be incorporated in a fifth draft; the author says it should not alter the conclusions. |
| Union Statistics for Fifty Years (South Africa, 1960), Bureau of Census and Statistics | official statistical series | 1910-1960 | Supports the claims on relative city populations (Johannesburg overtaking Cape Town; Durban outstripping Pietermaritzburg). |
| 1960 Population Census (as used by Board, Davies & Fair 1970) | population census | 1960 | Is the primary data source for the Board et al. study of the 1960 space economy, which the author relies on for the 1960 baseline pattern (Figure 1, GDP per square mile). |
| Board, Davies & Fair (1970) 'The structure of the South African space economy', Regional Studies – GDP per square mile 1960 map | secondary literature compilation | 1960 | Provides the 1960 spatial distribution of GDP (Figure 1) that establishes the hub-and-spoke core-coast structure as baseline. |
| Abedian and Standish (1992) Economic Growth in South Africa – sectoral GDP shares | secondary literature compilation | 1911 | Used to show that agriculture and mining made up about 50% of GDP in 1911 versus about 20% today. |
| DBSA (1999) map of Spatial Development Initiatives (SDIs) | maps spatial data | 1999 | Figure 2 is used to compare the SDI corridors with the projected high-growth areas. |
| Ramos (1999) address on GEAR targets by the Director General of Finance | other | 3 August 1999 | Cited for the idea of 'incidence analysis' and variable returns to public expenditure, in support of the policy conclusion. |
| Author-constructed GIS maps of GGP growth by magisterial district (Figures 3-7) | maps spatial data | 1960s-2000s | Maps of past growth rates and the projected scenario are compared visually to argue the continuity of the growth pattern. |
New data¶
The author assembled a file of GGP by magisterial district for 1968, 1972, 1981, 1991 and 1993/4, classified districts as coastal/core/mid-zone and metro/non-metro, and derived growth-change measures (GGPC 1-4) and a district-level GIS scenario map. The district count and classification lists are not given.
Inference¶
Quantification: Statistical analysis.
Descriptive comparison of GGP growth across district groupings (coast/core/mid-zone; metro/non-metro) over four sub-periods, with rank-ordered 'acceleration' rates; OLS-style regressions of 1990s growth on earlier growth, population growth and sectoral growth (variance explained); a 'rules of the game' argument from history and international experience; and scenario construction by multiplying shares (80% persistence x past shares) with judgemental adjustments.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Spatial patterns of growth have changed only slowly since 1960; the 1960s and projected 2000s maps are nearly indistinguishable. (p. 14) | Author's GGP-by-district maps for different periods compared visually (Figures 3-7). | Partial data |
| Coastal districts moved from last to first in rank of growth acceleration, while mid-zone districts (platteland and bantustans) did best in the 1960s and last thereafter. (p. 5) | Rank-ordering of acceleration rates computed from the 1968-1994 GGP district file, Table 2. | Partial data |
| Non-metropolitan areas fared relatively better in the 1960s-70s, when apartheid decentralisation was at its peak, than in the 1980s-90s, when growth shifted to the metros. (p. 7) | Table 4 ratios of non-metro to metro GGP growth from the district GGP file; the policy link is asserted rather than tested. | Partial data |
| Past spatial growth is the best predictor of 1990s growth (61-80% of variance explained), whereas population growth explains little (under 10% in the 1960s-70s, about 20% later). (p. 7) | Regressions across magisterial districts of 1990s GGP growth on earlier growth and population growth; no details on specification or data given. | Partial data |
| Manufacturing, construction, trade, transport and finance are the strongest predictors of overall spatial growth in the 1990s, while mining and agriculture are weakest. (p. 8) | Cross-sector correlations of 1993/4 GGP growth rates by district (Statistics SA 1995). | Partial data |
| Mineral-led growth on the Witwatersrand was the main force that reshaped the space economy from 1910 to 1960, shifting the economy from the coast to the interior. (p. 3) | Union Statistics for Fifty Years (city populations), Board et al. 1970 and Fair 1983; illustrated with Johannesburg/Cape Town and Durban/Pietermaritzburg populations. | Illustrative or anecdotal |
| Growth over the coming decade will be about 80% in the same areas as before, with a mild shift to the coast and metros (Table 5: 62% metro; 32% coastal). (p. 12) | Extrapolated from regression results, the 1990s baseline and judgemental adjustments; not an estimated model. | Asserted |
| Growth in the 1990s was highly concentrated: 58% of new growth fell in a few core metropolitan districts covering under 4% of the area, and 24% in the Johannesburg-Pretoria channel of about 0.2% of the area. (p. 15) | Calculation from the author's district GGP data. | Partial data |
The author's caveats¶
The author notes that the 1996 GGP data were unavailable. GGP figures are imperfect measures of economic activity, especially the informal sector. Raw figures are not inflation-adjusted. Metro boundaries and the coast/core definitions are arbitrary, and raw totals depend on district counts. Scenario figures are treated as approximate, with deviations of up to about 10% possible.
Gaps in the evidence¶
The paper gives no specification, district counts or classification lists for the regressions and groupings, and uses nominal rather than real GGP. Pre-1968 spatial data are limited to a 1960 map from another study, so the 1910-1960 narrative is illustrative. Tests of the policy-effect claims (decentralisation, bantustan growth points) would need district-level data on state incentives and spending, which exist in archives and official reports. Real, deflated GGP series and the 1996 data (now available) would also allow a stronger test of persistence.
Replicability: Data poorly documented.