An Analysis of Economic Infrastructure Investment in South Africa¶
Peter Perkins; Johann Fedderke; John Luiz (2005). An Analysis of Economic Infrastructure Investment in South Africa. South African Journal of Economics, Vol. 73:2. Blackwell Publishing / Economic Society of South Africa.
| Type | journal article |
| Scope | Economics and economic history of apartheid |
| Method | econometric |
| Period studied | 1870–2002 |
| Regions | South Africa |
| Themes | macro fiscal finance debt, manufacturing industrial policy, methods data sources, business and state, mining migrant labour |
| Citation key | perkins2005 |
Argument¶
The paper assembles a long-run database of South African economic infrastructure (railways, roads, ports, air travel, telephones, electricity, and public-sector infrastructure investment and capital stock) covering 1870-2002. Using Pesaran-Shin-Smith bounds F-tests, it tests the direction of association between infrastructure and real GDP. It finds that the relationship runs in both directions: GDP mostly drives physical infrastructure services, while public infrastructure investment, fixed capital stock and roads appear to drive GDP. Infrastructure develops in phases, and the need for investment is permanent. Public infrastructure investment collapsed from the mid-1970s to 2002, which risks bottlenecks.
Evidence and method¶
Time series compiled from South African Reserve Bank national accounts data, Statistics South Africa and predecessors (Union Statistics for Fifty Years, Official Year Books, South African Statistics), railway general managers' reports, SARH/SATS annual reports, NATIS, the National Ports Authority, SAA, Telkom and Vodacom; secondary literature including De Kock (1936), DBSA (1998) and Department of Transport (1998). Series are documented in Appendix 2.
Place in the debates¶
- Does public infrastructure raise output (Aschauer vs Holtz-Eakin/Garcia-Milà on spurious findings) (see Did the apartheid economy fail? Growth, productivity, investment and the post-1970s crisis). Notes that estimates vary widely and OLS results on time series may be spurious. It uses bounds-test methodology suited to non-stationary data and finds bidirectional evidence, with a forcing relationship from public infrastructure investment and capital stock to GDP.
- Direction of causality between infrastructure and growth (see Did the apartheid economy fail? Growth, productivity, investment and the post-1970s crisis). Argues that causality runs both ways. GDP drives most physical infrastructure indicators, but roads and public infrastructure investment and capital stock drive GDP.
- Whether the post-1976 fall in public infrastructure investment reflected correction of prior over-expansion (see Did the apartheid economy fail? Growth, productivity, investment and the post-1970s crisis). Acknowledges the argument and cites overcapacity and lumpiness. Nonetheless judges the long decline relative to GDP probably undesirable, and links it to growing government consumption and deficits in the late apartheid years.
Findings¶
-
Public-sector infrastructure investment per capita collapsed between 1976 and 2002, falling well below an international benchmark share of GDP. (p. 3)
Investment per capita fell from R1 268 in 1976 to R356 in 2002 (1995 prices), a collapse of 72 per cent. Investment fell from 8.1 per cent of GDP to 2.4 per cent of GDP
-
After 1980 government shifted from investment to consumption, financed partly by growing deficits, possibly to prop up apartheid. (p. 4)
possibly reflecting the government’s attempts to prop up a political system that was socially and economically unsustainable.
-
The budget deficit widened in the late apartheid period. (p. 4)
the national budget deficit rose from an average of 2.9 per cent in 1980-986 to an average of 4.9 per cent in 1987-1994
-
Mineral discoveries (diamonds, gold) drove early railway construction, with very rapid early growth. (p. 5)
Between 1875 and 1885, the average annual growth rate in South Africa’s railway lines was 29 per cent.
-
Railway rolling stock and traffic grew briskly to 1980, then fell sharply as the investment rate dropped and roads took over. (p. 6)
After 1980 rolling stock fell sharply (Fig. 2) as South Africa’s overall investment rate fell and as roads came to play an increasingly important role
-
Paving of national and provincial roads expanded roughly thirty-fold from 1940 to 2001. (p. 7)
Between 1940 and 2001, South Africa’s national and provincial paved roads increased from around 2 000km to around 63 000km.
-
Port capacity was expanded abruptly in the late 1970s with the opening of Richards Bay and Saldanha. (p. 9)
between 1976 and 1979 the volume doubled from 38 million harbour tons to 76 million harbour tons, the result of opening Richards Bay and Saldanha harbours.
-
Air travel growth slowed sharply in the politically fraught 1975-1994 period and recovered after apartheid. (p. 9)
brisk growth before 1975, much slower growth during the politically-fraught 1975-1994 period (when economic growth was also low), and a substantial
-
Electricity grew from mining demand, and Eskom's 1990s electrification addressed areas neglected under apartheid. (p. 11)
In the 1990s Eskom embarked on an electrification programme to expand its services to areas which had been neglected during apartheid
-
Roads show a strong association with, and forcing effect on, GDP, while GDP drives most other infrastructure measures. (p. 13)
the correlation coefficient between paved roads and real GDP is 0.996 for the period 1938-2001, and the PSS F-tests indicate a long-run forcing relationship from roads to GDP.
-
Infrastructure development proceeds in phases that interact with the changing structure of the economy. (p. 12)
it is quite plausible that phases of infrastructure development took place in response to changes in the structure of the economy and also exerted some influence
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| South African Reserve Bank database: public-sector economic infrastructure gross fixed capital formation and fixed capital stock (1995 prices); also real/nominal GDP 1946-2002 | official statistical series | 1946-2002 (GDP); investment and capital stock to 2002, with Fig. 1 starting 1960 | Supplies the national accounts series for infrastructure investment, fixed capital stock and real GDP used in the PSS F-tests and Fig. 1. |
| South African Reserve Bank Annual Economic Report, 1996-2000 | official statistical series | 1996-2000 | Cited to explain that the 1990s recovery and later slump in infrastructure investment came from Telkom, Eskom and SAA expansion programmes. |
| Statistics South Africa and predecessors, South African Statistics (2002 and other years), incl. SSA 1980 value-added series | official statistical series | 1911-2002 | Source for GDP backcast to 1911, population, and most physical infrastructure series (rail, roads, vehicles, ports, air travel, electricity). |
| Union Statistics for Fifty Years, 1910-1960 jubilee issue | official statistical series | 1910-1960 | Source for early GDP deflator inputs (wholesale and retail price indices), population and early physical infrastructure series. |
| Official Year Book of the Union (1916, 1917, 1925, 1933) | official statistical series | 1916-1933 editions | Source for railway lines, roads, ports, phone lines; also for dates of the first telephone exchanges. |
| Population census figures for 1904 and 1911 | population census | 1904, 1911 | Used with SSA and US population figures to extrapolate the population series for per capita measures. |
| Railway administration annual reports: Cape of Good Hope, Colony of Natal, Natal Government Railways, Central South African Railways, and South African Railways and Harbours (Reports of the general manager) | company records or reports | 1873-1916 (colonial); 1910 onward (SARH) | Source for locomotives, coaching stock, goods stock, passenger journeys and revenue-earning traffic in the early period. |
| Spoornet and South African Transport Services annual reports | company records or reports | to 2003 | Recent railway, port and air data to complete the series to 2003. |
| Road data: OYBU, US, SSA, CSIR, SANRA (total and paved national and provincial roads) | official statistical series | 1920-2001 | Builds road length series for Table 4, Fig. 3 and the PSS tests, with interpolation for 1994-2000. |
| NATIS (National Traffic Information System) and SSA vehicle registration data | official statistical series | 1920-2002 | Numbers of passenger and goods vehicles used for traffic growth rates and PSS tests. |
| National Ports Authority cargo data (with OYBU, SSA) | official statistical series | 1911-2001 | Cargo handled in harbour tons used as a proxy for port infrastructure capacity and in PSS tests. |
| Air travel series: SAA passengers (SSA, SATS, SAA) and international passengers at South African airports (SSA) | official statistical series | c.1950-2001 | Proxy for air transport infrastructure, in the absence of a direct measure. |
| Telephone line data: OYBU, Canning (1998), Telkom, Vodacom | official statistical series | 1912-2003 | Fixed and mobile line series for growth rates and PSS tests; 1934-1960 interpolated. |
| Canning (1998) database of world stocks of infrastructure, 1950-95 | secondary literature compilation | 1950-95 | Supplements phone line data and is cited as justification for the unreliability of road data. |
| Electricity generated series (US, SSA) | official statistical series | 1920-2002 | Gigawatt-hours generated used for growth rates and PSS tests. |
| Eskom capacity and peak demand figures | company records or reports | 1993-2003 | Used to argue that capacity may be inadequate for peak demand by 2007. |
| Department of Transport (1998), Moving South Africa: a transport strategy for 2020 (draft) | commission or official report | 1998 | Evidence on investment shortfalls in roads and ports, truck overloading, and port customer dissatisfaction from interviews. |
| SANRA Horizon 2010 and SANRA Annual report 2002 | company records or reports | 2002-2010 | Estimates of road damage from overloading and constraints on road funding. |
| National Treasury, Medium Term Budget Policy Statement 2003 | official statistical series | 2000/01-2003/04 | Road spending growth figures. |
| SAA audited group results for year ended 31 March 2003 | company records or reports | FY2003 | Fleet size and upgrade programme. |
| Development Bank of Southern Africa (1998), Infrastructure: a foundation for development | secondary literature compilation | 1998 | Reports prior infrastructure-output elasticity estimates and gives a brief history of electricity (VFTPC, Eskom). |
| Young (ed.) (2003), Ports of Southern Africa and Mauritius, 54th edition | secondary literature compilation | 2003 | Evidence on Durban Container Terminal congestion. |
New data¶
Authors compiled a database of South African infrastructure series 1870-2002 (national accounts infrastructure investment and capital stock, railways, roads, vehicles, ports, air travel, telephones, electricity), splicing colonial and Union railway reports with Union and Statistics SA publications, standardizing units and interpolating gaps. Notes on each series are in Appendix 2.
Inference¶
Quantification: Econometric (causal design).
Time-series descriptive growth rates by period across infrastructure types, plus bivariate Pesaran-Shin-Smith bounds F-tests (ARDL error-correction) between per capita or aggregate real GDP and each infrastructure measure to infer the direction of long-run association, supplemented by historical narrative.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Public-sector infrastructure investment per capita collapsed 72 per cent between 1976 and 2002, from 8.1 to 2.4 per cent of GDP. (p. 213) | SARB national accounts series on public-sector infrastructure gross fixed capital formation, deflated to 1995 prices. | Systematic data |
| Long-run forcing runs from public infrastructure investment and capital stock to GDP. (p. 214) | PSS F-tests on SARB per capita series (Table 1); F-statistics of 6.39 and 7.69 against the 5.764 upper bound. | Partial data |
| For most infrastructure types (railway lines, coaching stock, passenger journeys, ports, SAA passengers, fixed phones) GDP drives infrastructure rather than vice versa; roads are the exception. (p. 223) | Bivariate PSS F-tests (Tables 3, 5, 7, 9) on Statistics SA and related physical series, plus a 0.996 correlation for paved roads and GDP. | Partial data |
| Infrastructure developed in phases: railway lines plateaued after 1930, roads after 1940 when paving began, with rolling stock falling after 1980. (p. 223) | Period average growth rates computed from the compiled railway and roads series (Tables 2, 4, 10; Figs 2-4). | Systematic data |
| Mining discoveries (diamonds, gold) drove early railway, telegraph and electricity development. (p. 215) | Historical narrative from De Kock (1936), DBSA (1998) and railway line growth rates. | Illustrative or anecdotal |
| Port congestion and under-investment, together with possible electricity capacity shortfall by 2007, indicate infrastructure bottlenecks that could constrain growth. (p. 221) | Department of Transport (1998) interviews and investment estimates, Young (2003), and Eskom peak demand and capacity figures. | Illustrative or anecdotal |
The author's caveats¶
Road data quality is described as questionable, with a structural break in 1956-57 when about 25,000km of OFS tertiary roads were reclassified, and no provision for lane numbers. Series are interpolated in many years; 1903-08 rail data are estimated from graphs; geographic coverage changes (Namibia, TBVC areas). Vehicle data have definitional changes in 1979, and cargo is acknowledged not to be a direct measure of infrastructure. Statistics SA discontinued many series.
Gaps in the evidence¶
The argument rests on bivariate tests, without controls for other determinants such as labour, private capital, mining output or policy, so causal direction is weakly identified. Physical quantities such as route-km or vehicle counts say little about quality, utilisation, spatial distribution or cost, and nothing is said about apartheid-specific racial and regional allocation of infrastructure. Project-level cost-benefit data, sectoral or provincial disaggregation and Transnet/Eskom archival records would test the claims better, and many of these exist.
Replicability: Data described but not public.