Skip to content

From Chimera to Prospect: Toward an Understanding of the South African Growth Absence

Literature › Works

Johannes Fedderke (2006). From Chimera to Prospect: Toward an Understanding of the South African Growth Absence. School of Economics, University of Cape Town.
Type working paper
Scope Economics and economic history of apartheid
Method econometric
Period studied 1970–1997
Regions South Africa
Themes macro fiscal finance debt, labour markets wages, mining migrant labour, manufacturing industrial policy, education human capital, inequality poverty unemployment, business and state, methods data sources, transition post apartheid
Citation key fedderke2006a

Argument

Fedderke surveys econometric evidence, much of it his own, to explain South Africa's long-run decline in growth and employment since the early 1970s. Weak investment is driven mainly by uncertainty, which has institutional and political roots and also affects capital flows. Labour-market rigidity (wages above productivity, strongly negative wage elasticities) and output-market pricing power form a 'pincer' of inefficiency. A poor-quality human-capital system, whose legacy lies in apartheid-era racial schooling and duplicated black universities, limits innovation and TFP growth. The tone is explicitly policy-oriented and favours wage moderation, competition policy and macroeconomic stability.

Evidence and method

Sectoral national-accounts and employment series 1970-97 (WEFA Southern Africa, SARB Quarterly Bulletin, Chamber of Mines skills data), plus the author's earlier time-series, panel (pooled mean group) and VECM/ARDL estimates. Also effective-protection rates computed from customs duty collections, and long-run education series by race 1910-93 (Fedderke, De Kadt and Luiz). Other inputs include a World Bank firm survey, Van der Merwe's remuneration-by-skill data and the international growth literature.

Place in the debates

Findings

  • Formal employment growth declined from the 1970s and turned negative overall in the 1990s, with mining losses most severe. (p. 15)

    TOTAL 2.05 1.12 -1.15

  • Growth accounting shows a shift from factor accumulation to TFP; labour's contribution turned negative in the 1990s. (p. 18)

    In the 1990’s growth in the labour force input contributed negatively, and growth in the capital input contributed relatively weakly to growth in GDP.

  • Uncertainty, sectoral and especially systemic, significantly lowers manufacturing investment. (p. 33)

    Both systemic and sectoral uncertainty appears to be pertinent for investment - though systemic uncertainty has a stronger impact than does sectoral uncertainty.

  • Capital flows and flight are sensitive to political risk, including political liberalisation and instability. (p. 45)

  • In gold and uranium mining, real wages rose after 1986 and the wage share of value added rose, which the author links to labour shedding. (p. 66)

    It increased thereafter to 67% in 1997.

  • Real wage elasticities of mining labour demand are consistently negative and strongly so. (p. 74)

    In all three aggregate mining sectors rising real per labourer remuneration is consistently and strongly related to declining labour usage.

  • Gold employment peaked in 1986 and fell to 60% of that peak by 1997, despite falling real output since 1970. (p. 57)

    Thus by 1997 employment in Gold & Uranium Mining had contracted to 60% of the levels attained at its 1986 high point.

  • Manufacturing price-cost mark-ups are about twice the US level where intermediate inputs are ignored. (p. 102)

    The average mark-up in South African manufacturing industry lies in the region of 80%, where intermediate inputs into production are ignored.

  • Separate black universities were an inefficient use of resources and produced poor, costly output. (p. 166)

    the black university system proved not only to generate output that was of poor quality, but that it proved to be poor output that was expensive.

Evidence ledger

Sources used

Source Kind Period How used
South African Reserve Bank real GDP series, 'old' and 'new' (revised) measures, 1970s-1990s national accounts macro 1970-1998 Establishes the long-run decline in real GDP and per-capita GDP growth and tests whether GDP revisions change the picture.
Formal sector employment series by major sector, 1970-98 (Figure 3, Table 1) official statistical series 1970-1998 Shows decelerating formal employment growth and 1990s job losses across sectors, especially mining.
Growth accounting decompositions of real GDP and sectoral output (Fedderke 2001c, 2002c), 1970-97 secondary literature compilation 1970-1997 Shows labour and capital contributions falling and TFP rising in the 1990s, with sectoral differences.
Fedderke (2003, 2004) sectoral panel investment study of South African manufacturing (PMG estimates; Table 4, Figure 5) secondary literature compilation 1970-1997 (implied) Provides long-run investment relation showing uncertainty (sectoral and systemic) lowers investment.
Fedderke, De Kadt and Luiz (2001a) indicators of political liberty, property rights and political instability, 1935-97 secondary literature compilation 1935-1997 Supplies systemic uncertainty and political instability/rights indices used in investment, capital flow and institutions estimations.
Fedderke, De Kadt and Luiz (2001b) time-series study of institutions and growth, 1935-97 secondary literature compilation 1935-1997 Source of the finding that institutional variables affect growth via capital accumulation (Figure 8).
Fedderke, Henderson, Kayemba, Mariotti and Vaze (2001) sectoral capital stock data (Machinery & Equipment), 1970-97 secondary literature compilation 1970-1997 Shows decade-average real capital stock growth by sector, indicating a 1990s reallocation toward manufacturing.
SARB Quarterly Bulletin series S-127, S-129, S-131, S-116 (corporate and household saving, private fixed capital formation, GNP at factor cost) official statistical series c. 1960-1990s Constructs private-sector savings and investment rates showing a persistent saving shortfall requiring capital inflows.
WEFA Southern Africa dataset on mining employment, wage bills and output (Fedderke and Pirouz 2002) company records or reports 1970-1997 Supplies employment, nominal earnings, real labour cost and productivity series for the mining sector analysis (Tables 7-10, Figures 10-13).
Chamber of Mines data on skills composition of mining labour force chamber of mines or recruiting records 1970-1998 Shows improved average skills in mining came from shedding unskilled workers, not from adding skilled ones.
Fedderke and Mariotti (2002) sectoral labour market dataset (46 sectors; 28 manufacturing sectors), 1970-97 secondary literature compilation 1970-1997 Basis for correlations between labour productivity and real wage and for the manufacturing labour demand panel (Table 11).
Van der Merwe (2000) Trends in South African Remuneration by Skill Level commission or official report 1970-1997 Provides real remuneration by skills category used to estimate wage elasticities by skill (Tables 12 and 14).
Fedderke, Shin and Vaze (2000) trade and labour usage panel, 28 three-digit manufacturing sectors secondary literature compilation 1970-1997 Supplies the manufacturing labour demand function (Table 13) and Stolper-Samuelson tests on trade liberalization and wages.
Edwards and Schor (2002) and Edwards and Golub (2002) trade composition and unit labour cost data secondary literature compilation 1990-1999 Shows South African trade is dominated by developed countries; supplies relative real unit labour costs for competitiveness.
Fedderke, Kularatne and Mariotti (2003) mark-up panel for South African manufacturing and two-digit sectors secondary literature compilation 1970-1997 Estimates price-over-marginal-cost mark-ups with Roeger's nominal Solow residual method, showing mark-ups roughly double the US.
Mariotti (2002) demand-side policy and growth study (VECM and threshold regression) secondary literature compilation post-war South Africa Provides findings that government consumption and inflation lower long-run output, with optimal levels below 12% of GDP and 3% inflation.
Kularatne (2002) financial deepening time series, 1952-92 secondary literature compilation 1952-1992 Provides VECM findings that credit extension and stock market liquidity raise per capita GDP via investment.
World Bank (2000) Constraints to Growth and Employment in South Africa: large manufacturing firm survey, 1998-99 household or labour survey microdata 1998-1999 Cited as supporting evidence of credit rationing constraining firm investment.
Fedderke (2001b) TFP spillover/Schumpeterian panel on manufacturing, with human capital and R&D indicators (Table 15) secondary literature compilation 1970-1997 Shows quality human capital (maths proportion, NES degrees) and R&D raise TFP growth, while capital stock growth gives no Romer spillover.
Fedderke, De Kadt and Luiz (2000) schooling dataset by race (matric pass rates, pupil-teacher ratios, per-pupil expenditure, teacher qualifications), 1910-93 secondary literature compilation 1910-1993 Documents persistent racial disparities in matriculation outcomes and school inputs under apartheid.
Fedderke and Luiz (2002) educational production function estimates for black and white schooling, 1910-93 secondary literature compilation 1910-1993 Time-series elasticities of pass rates to inputs show expenditure matters but governance and political instability dominate for black schooling.
Fedderke, De Kadt and Luiz (2003) university system dataset (student-lecturer ratios, per-student expenditure, degrees, research output), 1910-93 secondary literature compilation 1910-1993 Shows black universities had comparable inputs but far weaker degree and research outputs and higher unit costs.

Inference

Quantification: Econometric (causal design).

A survey and synthesis of the author's earlier econometric work: growth accounting (Solow residual), pooled mean group and ARDL dynamic heterogeneous panel estimates, Johansen VECM cointegration, descriptive correlations, effective protection rates and Roeger mark-up estimation. Causal claims rest on long-run coefficients from time-series and sectoral panels, supplemented by decade-average descriptive comparisons and cited international literature.

Key claims

Claim Basis Strength
South African GDP and per-capita GDP growth has been on a long-run structural decline since the early 1970s, with formal employment growth also falling and turning negative in the 1990s. (p. 10) Reserve Bank real GDP series (old and new measures) and sectoral formal employment series (Figure 1-3, Table 1). Partial data
The source of growth shifted from capital and labour accumulation in the 1970s-80s to TFP in the 1990s, with the capital contribution falling with the investment rate. (p. 18) Growth accounting decompositions in Tables 2 and 3 from Fedderke (2001c, 2002c), assuming constant returns and factor shares equal to prices. Systematic data
Capital flows and flight into and out of South Africa are sensitive to political risk, and higher instability and political liberalization both stimulated capital outflows. (p. 45) ARDL estimates (Table 5) with exchange rate-adjusted interest differentials, GDP growth and political rights/instability indices. Systematic data
Real wage increases outstripping productivity explain labour shedding in mining, with negative wage elasticities of labour demand across mining, manufacturing and sectors, and strongest for unskilled labour. (p. 74) Descriptive correlations, real labour cost and productivity tables for mining (WEFA, Chamber of Mines data); Johansen VECM and PMG labour demand estimates (Tables 10-14). Partial data
South African manufacturing displays high pricing power, with mark-ups around 80% versus 45% in the US, falling with trade openness and rising with concentration. (p. 102) Roeger nominal Solow residual estimation using PMG panel estimation on manufacturing and two-digit sector data (Fedderke, Kularatne and Mariotti 2003). Systematic data
Apartheid schooling produced large racial quality gaps in outputs and inputs, and the duplicated black university system gave poor, expensive output, undermining the quality human capital that drives TFP growth. (p. 147) Fedderke, De Kadt and Luiz (2000, 2003) long-run series on pass rates, pupil-teacher ratios, expenditure, degrees and research output; Fedderke and Luiz (2002) production functions; Fedderke (2001b) TFP panel. Systematic data

The author's caveats

Acknowledges that growth accounting is only broadly indicative (no quality-adjusted inputs, constant returns assumed). Notes the Chamber of Mines and WEFA mining data are not fully compatible, with discrepancies for coal. Mining VECMs have only 28 observations, so a panel check is run. Van der Merwe skills-wage data are not compatible with other series and results are preliminary. The highly skilled elasticity is 'suspiciously strong', the manufacturing wage elasticity is unstable across specifications, and point-to-point EPRs are noisy. The university student-staff ratio is not quality-adjusted, and the capital market restructuring claim is described as plausible and needing more research.

Gaps in the evidence

The paper relies on reduced-form aggregate and sectoral regressions, with limited identification of causal mechanisms, and its apartheid-era claims on institutions and uncertainty rest on constructed indices rather than direct evidence. Firm- or plant-level data (manufacturing census microdata, Chamber of Mines wage and employment records by mine, union wage agreements) would test the wage-productivity and mining job-loss claims better; some of this existed then and more is available now. Direct evidence on investment decisions (firm surveys, investment plan data) and on apartheid-era labour controls and discrimination is largely absent. The sectoral capital stock and mark-up claims depend on constructed capital stock and deflator series that are not independently verified.

Replicability: Data described but not public.