Green Growth and its Implications for Public Policy. The Case of South Africa by Agence Française de Développement

No 4 “Green growth and its implications for public policy The case of South Africa”

“Green growth and its implications for public policy The case of South Africa”

xxxx

Etudes de l’AFD

December 2015

Etudes de l’AFD

Green Growth and its Implications for Public Policy– The Case of South Africa Jules Schers, under the scientific coordination of Frederic Ghersi and Franck Lecocq Editor: Fabio Grazi (AFD)

Green Growth and its Implications for Public Policy– The Case of South Africa Jules SCHERS,1,2 under the scientific coordination of Frederic GHERSI2 and Franck LECOCQ2 EDITOR Fabio Grazi Agence Française de Développement (AFD)

Société des Mathématiques Appliquées et de Sciences Humaines (SMASH), 20 rue Rosenwald, 75015 Paris, France

Centre International de Recherche sur l’Environnement et le Développement (CIRED), UMR 8568 CNRS, ENPC, CIRAD, AgroParisTech, EHESS, 45bis avenue de la Belle Gabrielle, 94736 Nogent sur Marne CEDEX, France 2

The Etudes de l’AFD collection includes studies and research supported and coordinated by the Agence Française de Développement. It promotes the diffusion of knowledge gathered from both in-the-field experience and academic work. The papers are systematically submitted for approval to an editorial committee that draws on the opinions of anonymous experts.

All our publications are available at http://librairie.afd.fr

DISCLAIMER The analyses and conclusions presented in this document are the responsibility of the authors. They do not necessarily reflect the position of AFD or its partner institutions.

Publications Director: Anne PAUGAM Editorial Director: Gaël GIRAUD Designed and produced by: Flexedo, info@flexedo.com Printed by: Imprimerie de la Centrale Lens - ICL

Table of contents Executive summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1. Articulating environment and development objectives in South Africa: an introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2. Costs and impacts on other development goals of a carbon tax in SA: literature review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3. Original contribution of the present report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Methods: model description and calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4. The IMACLIM-SA model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. Main characteristics of IMACLIM-SA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Disaggregation of sectors, households and skills . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3. Model specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

17 17 23 27

5. Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. Overview of calibration data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. Demography, labour and general prospective settings. . . . . . . . . . . . . . . . . . . . . . . . . . .

30 30 32

Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

6. What dynamics for skill-segmented labour? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1. Reference projection with skills defined by educational attainment . . . . . . . . . . . . . 6.2. Alternative skill segmentation dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

39 39 42

7. Reference projection and policy scenarios results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1. The 2035 Reference Projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2. Carbon tax revenue recycling scenario definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3. The carbon tax policy scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.4. Sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.5. Comparison of parameterisation of SATIM runs to IMACLIM-SA variables . . . . . 7.6. General discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

44 45 55 58 67 71 74 3

General conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Appendix 1. IMACLIM-SA description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1. Producer and consumer prices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2. Households’ income, savings and investment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3. Consumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4. Production (institutional sector of firms) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Gross disposable income and investment decision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Production trade-offs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3. Gross operating surplus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

92 92 93 93

5. Public administrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. Tax, social security contributions and fiscal policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. Gross disposable income, public spending, investment and transfers . . . . . . . . . . . .

93 93 95

6. “Rest of the World” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1. Trade balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2. Capital flows and self-financing capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

96 96 96

7. Market balances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1. Goods markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2. Investment and capital flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3. Employment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

97 97 97 97

8. Model parameters and variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.1. Parameters calibrated on BY hybrid I-O table and other statistical data . . . . . . . . . 8.2. Parameter from other sources or exogenous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.3. Model variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

98 98 99 99

Appendix 2. Hybrid I-O tables and harmonised integrated economic accounts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 1. Building hybrid input-output tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1. Adjustment of uses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2. Adjustment of resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

103 104 104

2. Final hybrid I-O table for Base Year 2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

106

3. Quantities of uses and resources at Base Year 2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

108

4. Harmonised current and financial accounts for economic agents . . . . . . . . . . . . . . . . . . . . . . . . 4.1. 2005 current and financial accounts of agents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Building of current and financial accounts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3. Financial assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

109 109 109 112

5. Demography and labour force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. Demographic data for South Africa in 2005 and 2035 . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. Demographic data for South Africa in 2035 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

113 113 114

6. Elasticities for production, consumption and international trade . . . . . . . . . . . . . . . . . . . . . . . . . 6.1. Production function elasticities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2. Non-constrained consumption part of final consumption by households . . . . . . 6.3. Elasticities of international trade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

114 114 116 117

Appendix 3. Detailed numerical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 1. RP employment changes vs. BY by sector and by skill level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

119

2. Main settings and key outcomes for RP and all policy scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . .

120

3. Comparison of parameters and variables of SATIM and IMACLIM-SA runs . . . . . . . . . . . . . .

122

Appendix 4. Insights from similar exercises conducted in Brazil . . . . . . . . . 125 1. The Brazilian context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1. A rapidly growing economy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2. A very specific energy sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3. And a peculiar GHG emissions profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

125 125 126 128

2. Prospects at the interface of decarbonisation–energy–economic development . . . . . . . . . 2.1. The Brazilian National Plan on Climate Change 2009-2020 (PNMC) . . . . . . . . . . . . 2.2. Challenges beyond 2020 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. The oil question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

129 129 130 132

3. Presentation of two joint COPPE-CIRED studies on the articulation between decarbonisation and economic development in Brazil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Key results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

133 133 135 138

4. Conclusion: some insights from the comparison of the Brazilian and South African exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

142

Appendix references . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 List of abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 List of tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 List of figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

Executive summary South Africa is a rapidly growing middle-income economy with a coal-based energy system that generates high greenhouse gases emissions, on a par with the richest economies in the world. The country has pledged to significantly reduce its emissions (by 34% in 2020 and by 42% in 2025 relative to business as usual, conditional on financing and technical support from the international community), and it is actively discussing policies to achieve this goal, including a carbon tax. Yet climate mitigation is hardly the only challenge that South Africa faces. Despite significant progress in overcoming the inequalities inherited from the apartheid era and in improving quality of life since the onset of the democratic regime in 1994, economic growth has slowed down in recent years, poverty remains high and large inequalities persist. Notably, the South African economy is still experiencing very high unemployment, particularly amongst low-skill individuals, while there is a shortage of high-skill workers. The present report aims to provide some insights on the articulation between South Africa’s mitigation objectives and the key development challenges outlined above. It focuses in particular on economic growth and unemployment, with discussions about inequalities and education. It is relevant to the “green growth” conversation in that it aims at exploring the conditions under which environmental objectives (here, climate mitigation) can be achieved alongside other key development objectives for South Africa. To do so, it uses the development of IMACLIM-SA, a recursive computable general equilibrium model of the South African economy. IMACLIM-SA represents the South African economy as a small, open economy with 10 sectors (5 energy, 5 non-energy) and 5 income classes. Calibrated on base year 2005 (the most recent data available at the time of its construction), the model projects a balanced South African economy through 2035 based on assumptions about the evolution of key parameters (notably demography, labour and capital productivity, international prices, etc.). 2035 equilibria are computed both without (reference projection) and with mitigation policies, and analysed in terms of GDP growth, employment and income distribution. The 2005 calibration is performed on a revised social accounting matrix that has the original feature of bringing in monetary flows drawn from macroeconomic statistics and energy flows taken from energy tables. The model also captures differences in the prices of goods and services (notably energy goods) when sold to firms, households or the public sector, or when exported. It builds on specific runs of the South Africa Times model developed by the Energy Research Centre of the University of Cape Town to inform changes in the structure of electricity production from 2005 to 2035 (with or without mitigation policy). Finally, the model pays particular attention to the labour market, in terms of supply—with three skill classes—, demand and functioning. 7

In the reference scenario, given our assumptions, GDP grows at an average rate of 2.5% per year, and GDP per capita more than doubles over the period. It must be noted that to generate such a level of growth in the baseline, we must not only assume capital and labour productivity improvements, but also allow for an increase in international prices relative to domestic ones, thus improving the competitiveness of South African products on the export markets. Unemployment decreases markedly, though it remains high in 2035, and shortage of high-skilled labour persists. In the reference projection, emissions increase substantially, despite the implementation of the Longterm Electricity Investment Plan of 2011. We then explore seven policy packages based on the imposition of a carbon tax with different recycling schemes. We find that CO2 emissions in South Africa are sensitive to the carbon tax—even a carbon tax of 100 Rands (2005 Rands) per ton of CO2 equivalent (R100/tCO2), although “small” by international standards, is large relative to initial domestic energy prices—, and that R300/tCO2 might be able to achieve the country’s pledges. This is consistent with prior literature on mitigation in South Africa. Impacts on GDP growth and on unemployment reduction can be significant (GDP in 2035 can be 7% lower with policy than it would be without), but strongly depend on the revenue recycling mechanism. Amongst the recycling options we test, recycling carbon tax proceeds as lump-sum transfers to households underperforms the reference in terms of GDP and employment, but has a progressive impact on income distribution. Using carbon proceeds to reduce sales taxes results in GDP growth on a par with the reference projection and in a higher level of employment, without excessively degrading the mitigation impact of the tax (compared to scenarios with a less favourable economic impact), but at the cost of a slight increase of income dispersion. Finally, under the same economically efficient recycling option of sales tax reductions, diverting part of the proceeds to invest in education results in accelerated growth even for a conservative assumption of the induced productivity gains. However, it does not improve employment because the sales tax reduction has a specific positive impact on the relative cost of capital, which increases along with growth and the consecutive development of the carbon tax fiscal basis that fuels it.

Acknowledgements This research has been financially supported by the research department of Agence Française de Développement (contract RCH N°262-2011). It has also benefited from strong support from the Energy Research Centre of the University of Cape Town, notably on data provision, expert review and provision of and support for the SATIM model. The authors thank Harald Winkler, Bruno Merven, Alison Hughes, Tara Caetano, Bryce McCall, Alfred Moyo, Adrian Stone, Fadiel Ahjum and Hilton Trollip at the Energy Research Centre of Cape Town University for their support and comments. The authors also thank John Reilly and Francesco Ricci for very detailed and constructive reviews and discussions throughout the project. They thank Fabio Grazi, Cyril Bellier and Véronique Sauvat at the Research Department of AFD for their support, reviews and comments. They also thank present and former AFD staff Carl Bernadac, Jean-Michel Debrat, Damien Navizet, Katia Pascarella, Nicolas Rossin and Isabelle Vincent; Faaiqa Hartley and Konstantin Makrelov at the National Treasury of South Africa; Rob Davies and James Thurlow at UNU-WIDER; as well as participants of the MAPS Modelling workshop (Cape Town, 2013), the ECONLAB III Conference (Cape Town, 2014) and the 3rd Annual Conference of the Green Growth Knowledge Platform for very helpful comments on various aspects of this research. Finally, the authors acknowledge contributions from Henri Waisman and Emmanuel Combet from SMASH to previous versions of this work, and very helpful advice from Julien Lefèvre, Antonin Pottier and Jean-Charles Hourcade from CIRED. Of course, remaining errors are entirely the authors’.

Introduction 1. Articulating environment and development objectives in South Africa: an introduction South Africa is a rapidly growing middle-income economy that ranked 82nd in GDP per capita (in purchasing power parity—PPP—terms) in 2013.[1] Its coal-based energy system generates high greenhouse gas (GHG) emissions, on a par with the richest economies, and it ranked 40th in per capita CO2 emissions in 2011. South Africa GHG and CO2 emissions are comparable with those of France, an economy over three times as large in PPP terms (Table 1).

Table 1. A parallel between 2011 South Africa and France South Africa Population (million inhabitants)

France

51.6

65.4

GDP per capita (USD PPP)

11 848

36 248

GHG emissions (MtCO2e)

456

487

CO2 emissions (MtCO2)

374

338

CO2 from electricity generation (MtCO2)

229

Source: World Resources Institute CAIT 2.0.

Economic development in South Africa is indeed driven by abundant coal resources, highly subsidised electricity from coal, and coal-to-liquids technology. In this context, it is not surprising that the energy-intensive mining and industry sectors are historic drivers of the South African economy, and that they still contribute a significant share to its GDP. South Africa has adopted ambitious decarbonisation objectives. During the 15th Conference of the Parties of the United Nations Framework Convention on Climate Change (Durban, 2009), it pledged to reduce its GHG emissions by 34% in 2020 and by 42% in 2025 relative to business-asusual, conditional on financing and technical support from the international community. These objectives were reaffirmed in the 2011 Climate Change Response White Paper of the South African government (SAED, 2011). The White Paper also outlined a strategy to reach those, including carbon pricing, possibly in the form of a carbon tax (see Box 1). In 2013, a revised carbon tax plan was [1] Source: World Bank.

Box 1

Key elements of the South African mitigation strategy The 2011 Climate Change Response White Paper outlines the following key elements of the South African mitigation strategy (SAED, 2011; IEA, 2013): National GHG Emissions Trajectory Range. The objective is that emissions increase from • A 374 MtCO2 in 2011 to 398–583 MtCO2 between 2020 and 2025, then plateau at 398–614 MtCO2 between 2025 and 2035, and eventually decrease down to 212–428 MtCO2 by 2050. • Reduction targets for key sectors (based on cost-benefit analysis). • R equirement, for actors of such sectors, to submit mitigation plans, including actual carbon budgets for some companies in certain sectors and subsectors. • Implementation of mitigation policies that support job creation and economic development. • I mplementation of economic measures, including carbon pricing—which could come in the form of a carbon tax (or in the form of emission trading for the entities that are covered by carbon budgets). • Implementation of a monitoring system for GHG emissions.

released. To allow for further consultation, implementation of the carbon tax—initially scheduled for January 2015—has been postponed by a year to 2016. The carbon tax, however, is one of many instruments by which the country aims to meet its mitigation goals. It would complement and interact with many other plans and policies, notably in the energy sector. In particular, the Government of South Africa has adopted a Long-Term Electricity Investment Plan (DoE, 2011) that, following massive power outages in the late 2000s, aims to urgently ease the tension between supply and demand of electricity. On the demand side, the plan targets a 35% energy efficiency improvement while simultaneously envisaging further electrification. On the supply side, it calls for the addition of new capacity. Of these, renewables and primarily wind and photovoltaic (PV) would constitute the largest share (via price-competitive procurement), followed by nuclear.[2] Under the Long-Term Electricity Investment Plan, the share of coal would thus fall from 93% (current) to 46% of total electricity capacity in 2030.[3] Besides, the plan advocates further increases in electricity prices, which are regulated. These have risen from about 0.25 South African Rand per kWh (R/kWh) in 2005 to R0.65/kWh in 2013, and the National Energy Regulator of South Africa (NERSA) has already approved increases (IEA, 2013; Baker et al., 2014). Improving electricity transmission links with neighbouring countries could also reduce pressure on the South African power generation system and provide access to low-carbon electricity (notably hydro from Mozambique or beyond) and to liquefied natural gas (LNG). [2] The 2011 plan calls for 9.6 GW of new nuclear generation capacity by 2030, compared to installed capacity of 1.8 GW. [3] Nuclear and hydro each account for 12.7% of total installed capacity, wind 10.3% and PV 9.4%.

Given the high level of South African emissions, the set of policies capable of achieving the proposed mitigation targets are unlikely to be marginal. On the contrary, they are likely to have significant negative or positive implications on the achievement of other economic, social or environmental goals.[4] These side effects matter from a public policy standpoint, because they constitute other channels by which the mitigation policy impacts welfare—beyond the direct cost of the policy and beyond the benefits in terms of climate mitigation. Documenting side effects also matters from a political economy standpoint, as they matter strongly in the policy debate. Last but not least, the detailed design of mitigation policy packages—for example, as we will see, the way by which the proceeds of a carbon tax are recycled in the economy—is central to limiting adverse side effects and/or to maximising co-benefits.[5] In the case of South Africa, the priorities from the onset of the democratic regime in 1994 have been to overcome the inequalities inherited from the apartheid era and to improve quality of life. These overarching objectives have been embedded into successive development plans, notably the Reconstruction and Development Programme in 1994, the Growth, Employment and Redistribution Programme in 1996, the Accelerated and Shared Growth Initiative for South Africa in 2006, the New Growth Path Programme in 2009 and the National Development Plan in 2012 (Republic of South Africa, 2013). The country has experienced significant success. Growth has been substantial, with real GDP doubling over the past 30 years. Prevalence of extreme poverty has been significantly reduced, and other key millennium development goals have been achieved or are on the way to being achieved. However, despite significant progress, South Africa is still battling with important challenges. Growth rates have significantly decreased since 2009. Poverty remains high as very large inequalities persist, and key social objectives are still unmet (Republic of South Africa, 2013). In particular, despite the rapid economic growth experienced over the past decades, the South African economy has not been able to absorb the surplus of labour supply (at 51 million in 2010, the population of South Africa was 14.6 million higher than it had been 20 years before), and unemployment remains high. The official unemployment rate has remained around 25% (between 23% and 26%) of active population since 2009, and around 35% if “discouraged workseekers” are included (StatsSA, 2015).[6] Unemployment is particularly high among low-skill individuals, while there is a shortage of highskill workers. For example, according to the latest available statistics (Q4-2014), the absorption rate [4] The positive side effects are typically called “co-benefits” in the literature while the negative side effects are often denoted as “adverse side effects”. See Kolstad et al. (2014) and Ürge-Vorsatz et al. (2014) for recent discussions. [5] In fact, the expression “side effects” implicitly refers to a distinction between “climate” policies, the primary goal of which is climate mitigation, and “non-climate policies”, adopted mainly for other reasons. As climate mitigation objectives become more stringent, and as the links between mitigation objectives and other goals become more apparent, this distinction between “climate” and “non-climate” policies is becoming increasingly blurry. This is important because it allows thinking about mitigation with a broader set of policy instruments. For example, the response of transportation emissions to carbon pricing in a city depends strongly on the shape of the city, which itself results from a broad set of urban policies (e.g., transportation, zoning, fiscal policies, housing, etc.). Playing on this set may thus have significant implications for mitigative capacity in the medium term, even though the underlying policies would not be classified as “climate” policies. [6] Source: http://beta2.statssa.gov.za/?page_id=737&id=1 (last viewed Feb. 16, 2015).

(i.e., the ratio between employed population and total population) is 84.4% for men with tertiary education (73.3% for women) against 59.6% (44.2%) for men (respectively women) with high school (known as “matric”) education, and 38.8% (resp. 26.7%) for men (resp. women) with less than high school education (StatsSA, 2015). High unemployment is one indicator of persisting inequalities in the society. For example, absorption rates of the black population are lower than national rates, at all educational levels (StatsSA, 2015). In fact, South Africa still ranks among the countries with the highest Gini indexes in the world, and this index has not evolved significantly over the past 15 years (Republic of South Africa, 2013). The present report aims to provide some insights on the articulation between South Africa mitigation objectives and the key development challenges outlined above. It focuses on two major dimensions: economic growth and unemployment, with some discussions about inequalities and education. It is thus relevant to the “green growth” conversation, in that it aims at exploring the conditions under which environmental objectives (here, climate mitigation) can be achieved alongside other key development objectives for South Africa. To do so, it builds on an original dynamic computable general equilibrium of the South African economy, which it uses to explore the possible consequences of the imposition of a carbon tax—as is currently envisioned—on the dimensions above. In this “prospective” approach, the goal is to explore consistency between plausible pictures of the future state of the South African economy, and key assumptions about the functioning of the economy and about exogenous parameters and trends (related inter alia to demography, technology deployment, behaviour of the Rest of the World, etc.). The modelling approach allows building internally consistent pictures of the future of relatively high complexity; and exploring numerous scenarios.

2. Costs and impacts on other development goals of a carbon tax in SA: literature review The introduction of carbon taxes in South Africa has already been thoroughly analysed in the literature. In fact, few countries (both developed and developing) have had such deep analysis. This is comparable, for example, to the depth of analysis of the potential consequences of the introduction of a carbon tax in France.[7] The earlier analysis is by Van Heerden et al. (2006), who ask whether revenue recycling from a carbon tax can generate a double dividend in South Africa. To do so, they use a static Computable General Equilibrium (CGE) model of the South African economy. To capture the relative scarcities of skills in the South African labour market, the model features a supply of high-skill labour inelastic to wages, while the supply of low-skill labour is perfectly elastic to real wages. Calibrated on 1998 [7] Analysis of the economic consequences of the introduction of a carbon tax in France has been conducted by Vielle et al. (2009) using the GEMINI-E3 computable general equilibrium model; Combet et al. (2010) and Combet and Hourcade (2014) using the IMACLIM-France model; and Callonnec et al. (2011) using the Three-ME macro-econometric model. Other analyses conducted with the MESANGE macro-econometric model of the French Treasury (Klein and Simon, 2010) and with the NEMESIS CGE model (Brécard et al., 2006) have been published in expert reports (notably Comité Trajectoires 2020-2050, 2011) but not in peer-reviewed literature.

data, the paper tests four policy instruments (a carbon tax, a fuel tax, an electricity tax and a broad energy tax) under three recycling schemes (a decrease in direct taxes, a decrease in indirect taxes, and a decrease in taxes on food). The main findings are that irrespective of the nature of the tax that is introduced, a food tax breaks recycling results in lower emissions, higher GDP and lower poverty than the reference, what the author call a “triple dividend”. De Pauw (2007) explores the potential impact of various mitigation scenarios on the South African economy. To do so, he develops a dynamic CGE of the South African economy. He analyses the implications of three policy scenarios over the 2005 to 2015 period, labelled respectively “start now”, “scale up”, and “use the market”, with various recycling options. Unlike Van Heerden et al., de Pauw finds no double dividend: introduction of a carbon tax reduces GDP and employment relative to the baseline—for instance, employment in 2015 is 2% to 7% lower in 2015 under a R250/ tCO2 tax than it would be without the tax—regardless of recycling options. Devarajan et al. (2009, 2011) ask how a carbon or an energy tax will fare in a distortion-laden economy such as South Africa’s. To do so, they first build a static CGE of the South African economy with a perfect labour market (in which wages adjust to leave aggregate employment fixed). Using 2003 as base year, they test three ways to reduce CO2 emissions by 15%: a carbon tax, a sales tax on energy inputs, and a sales tax on energy-intensive sectors; all with lump-sum recycling. Then they explore the sensitivity of their results to different assumptions about economic rigidities and labour market distortions. Their main result is that market rigidities are critical in the choice of the policy instrument. Precisely, “labour market distortions such as labour market segmentation or unemployment will likely dominate the welfare and equity implications of a carbon tax for South Africa” (Devarajan et al., 2011, p.18). Alton et al. (2012, 2014) explore the likely impacts of domestic and/or foreign carbon taxes on the South African economy. They develop a dynamic CGE model of the South African economy, linked to the South Africa Times energy Model (SATIM)—the bottom-up, energy sector model developed by the Energy Research Centre at Cape Town University. Using 2005 as the base year, they test three policy scenarios: a domestic carbon tax, a domestic carbon tax plus border tax adjustments, and a foreign carbon tax plus border tax adjustments in South Africa’s main trading partners. Three recycling options are explored: uniform reduction in sales tax, a reduction in the corporate tax imposed on the capital earnings of domestic firms, and a scaling-up of existing transfer programs. They find that a phased-in domestic carbon tax that reaches $30/tCO2—i.e., R210/tCO2—in 2022 would achieve the 2025 national emissions target of a 42% cut in emissions compared to baseline. They also find a trade-off in the choice of the recycling option between revenue distribution and economic growth. Finally, Musango et al. (2014) provide a multicriteria analysis of “green economy” policies in South Africa. They use a dynamic, differential equations model of the South African economy derived from the Threshold 21 (T21) platform of the Millennium Institute.[8] The model notably includes a representation of the dynamics of major natural resources. The authors test a business-asusual scenario over the 2001-2030 time period, and three alternatives. BAU2% assumes that an additional 2% of GDP is invested following BAU investment patterns. The other two scenarios [8] See http://www.millennium-institute.org/integrated_planning/tools/T21/

assume the same increase in investment, but directed towards green economy activities acrossthe-board (GE2%) or towards mitigation in the energy sector (GETS). The main finding of the paper is that all scenarios result in higher GDP and higher employment than in BAU (albeit with small variations), while GETS and GE2% also result in lower environmental pressure, notably on CO2 emissions. It is not fully clear from the paper why additional investments would be systematically positive on all dimensions. A probable explanation is that, given implicit rates of return on investment built into the technical model that represent investment outcome, the level of investment in BAU, which derives from past trends, is suboptimal.

3. Original contribution of the present report Relative to this profuse literature, the value-added of the research presented in this report is twofold. First, in methodological terms, it uses a “hybrid” dynamic computable general equilibrium model of the South African economy that features (i) consistent energy/economy accounting; (ii) “second-best” market features and explicit integration of technical constraints that make it possible to test a broad range of views/beliefs about the functioning of the economy; and (iii) explicit representation of the secondary distribution of income across economic agents, with differentiated pricing (notably energy pricing) (see below for a detailed description). Evidence from analysis in France and Brazil suggests that integration of the above-mentioned characteristics have significant implications for model outcomes. Second, in terms of policies tested, our research focuses on a different set of recycling schemes than in the papers outlined above. In particular, we explore recycling tax revenues through investments in education as a way to unlock growth potential. To do so, we pay particular attention to the representation of skills supply and demand and to their evolution in time. To our knowledge, this is the first attempt to link climate policies and investment in education via a carbon tax reform. The report is organised as follows. Its second part describes the model, its Base Year (BY) calibration and the set of assumptions shared by all prospective scenarios including the projection in the absence of a carbon tax, referred to as the Reference Projection (RP). Part 3 describes the set of mitigation policies that are tested, and analyses the induced departures from the reference projection outcomes. Finally, Part 4 concludes and opens perspectives for next steps. The analysis is complemented with 4 Appendixes. Appendix 1 describes the IMACLIM-SA model developed for the study. Appendix 2 reports key input and scenario data for the model runs. Appendix 3 reports detailed results pertaining either to the policy runs, to sensitivity analysis or to the comparison of IMACLIM-SA and SATIM runs. Finally, Appendix 4 reports on research conducted on the Brazilian economy with the same objective and using the same IMACLIM model, although differently linked to bottom-up modelling of the corresponding energy systems.

Methods: model description and calibration 4. The IMACLIM-SA model IMACLIM-SA is a two-period, dynamic computable general equilibrium (CGE) model of the South African economy. Although constructed specifically for the present research, it shares the key features of the existing IMACLIM models developed for France and Brazil notably. It thus builds on “hybrid” data—i.e., on a dataset reconciling money-metric national accounts and physical sectoral data; it features “second-best” characteristics of the equilibrium structure, notably mark-up pricing and rigidities in labour market adjustments; and it embraces a mixed top-down/bottom-up vision of technical change. Developments specific to IMACLIM-SA include (i) the hybridisation of national accounts and energy data for the particular case of South Africa; (ii) the implementation of skill-differentiated labour varieties, with varying degrees of rigidities and limited substitutability; and (iii) the import of bottom-up information on the South African electricity sector from the SATIM energy model (ERC, 2013). Appendix 1 provides a detailed description of the IMACLIM-SA model. The present section outlines the main features of the model. We first discuss the hybridisation approach, the use of technical coefficients and the second-best features of the IMACLIM approach to CGE modelling in section 4.1. Then we report on the granularity and specific features of our representation of the South African economy (4.2). Finally, we conclude with an overview of the behavioural specifications of the model (4.3).

4.1. Main characteristics of IMACLIM-SA 4.1.1. Dual accounting of economic and energy data through hybridisation The data hybridisation process at the basis of our modelling approach consists of 3 steps: (i) Creating Supply and Use tables in physical units, (ii) creating an Input-Output (I-O) equivalent for energy sectors from “energy bills”, and (iii) hybridising monetary and physical I-O matrices. Appendix 2 details steps 1 and 2. In this section, we describe the assumptions made for the third step, namely the hybridising of the I-O table on the basis of economic data, with the matrix of energy expenses at the level of sectors (of production) and economic agents (for consumption). We present the resulting hybrid I-O table in Appendix 2, section 2. 17

General principles The objective of hybridisation is to simultaneously and consistently track Input-Output tables in monetary terms and their counterpart in physical terms, while respecting two basic principles. First, both physical and money descriptions must respect the conservation principle, namely that each use of a good assumes the availability of the resource, and conversely that each resource must be used—with the possibility of transforming it or capitalising it as a stock. The second principle is that physical and monetary flows are linked by a system of prices. The total economic value associated with the production, trade or consumption of a good should thus follow from multiplying the aggregate volume of its use times a (weighted average) price for this use. These principles can be formalised mathematically into a system of accounting rules in the following manner: ∀ i,

∑Q

= ∑ Qi,R

∑V

i,E

∀ i,

i,E

Supply and use balance (quantities)

∑V

i,R

Supply and use balance (values)

∀ (i,0),

Vi,0 = Pi,0 * Qi,0

Consistency between quantities, values and prices

∀ (i,0),

CO2i,0 = e i,0 * Qi,0

Physical consistency (carbon balance)

With: i, index of economic operation Q, quantities (in physical units, e.g., PJ for energy) V, values (in monetary units) R, set of operations linked to resources (production and importation) E, set of operations linked to uses (consumption, investment, exportation) P, set of prices associated to each type of exchange CO2, CO2 emissions

Procedure The “hybridisation procedure” consists of steps that ensure the consistency between the macroeconomic data from national accounts and datasets from material balances and/or tangible physical indicators at the calibration date. Energy flows are particularly suitable for hybridisation because they are critical for the analysis of climate-related issues, and because data is available in the form of energy balances, which is the first ingredient necessary for hybridisation (see Box 2 below). However, the methodology can be applied to other material flows as well (calories, tons of cement, etc.), provided quantity and price data is available. After energy balance, the second ingredient to build a hybrid I-O table is the supply-use table in monetary values, which can be found in national accounts, and is built as follows (Figure 1): Columns report resources, and distinguish between intermediate consumption in monetary value (Vt), value-added (VA) and imports (M). Rows detail uses and distinguish between intermediate 18

consumption (CI), final consumption (CF) by households and public administration, gross fixed capital formation and exports. Finally, the third ingredient necessary to build hybrid I-O Tables is a vector of energy prices, possibly differentiated between actors (energy prices of households and firms are typically quite different). In general, this vector can also be derived from energy statistics. Based on the three ingredients above, the hybridisation procedure is as follows (Figure 1): from the energy balances and the price vectors, purchase bills of energy Vt can be derived by multiplying observed quantities Qt and prices Pt. These bills are then substituted for the monetary value in the supply-use Table, and the rest of the Table (row and column) is adjusted to balance again. In particular, excess value in the supply-use table is attributed to the composite sector, since it represents value attached to goods and services other than energy flows (e.g., services attached to energy, etc.).

Box 2

Energy balances Energy flow data are available in standardised energy balances that detail the processes of energy production, transformation and consumption measured according to their energy content (here expressed in petajoules, PJ). This type of statistical system makes it possible to reorganise datasets in a supply-use format similar to the one adopted for national accounts. Energy balances are typically organised as follows (see IEA/OECD energy balances):

1. The top part reports the supply of all energy carriers,[9] through imports minus exports, domestic production and stock changes. It also signals aggregated statistical discrepancies. Domestic production mainly matches mining and extraction industries in economic InputOutput tables as far as fossil fuels are concerned.

2. The middle part reports energy uses and transformations in the energy industry. It relates to some of the industries of economic Input-Output tables, like refineries and electricity production and distribution. This is also the place in energy statistics where own use or “autoproduction” by industries can be found.

3. The bottom part consists of energy use or final consumption of energy. It is split in three: (i) the industrial and (non-energy) mining energy use, (ii) other uses, i.e., agriculture, services, residential and transport uses, with transport encompassing both freight and passenger transport in its public as well as private varieties; (iii) non-energy uses. From this description it is clear that all elements are present to link energy balances to I-O elements like the intermediate energy consumptions of productive sectors (CI), the final consumption by households and exports (CF), whereas supply in the energy balance concerns volumes produced (P) and imported (M).

[9] Energy carriers typically included in an energy balance are coal (in different varieties), oil (in different varieties), natural gas, biomass (in different varieties), different transformed solid, liquid and gaseous “fuels” and other refined products, heat (mostly energy content of steam) and electricity.

The result is a modified input-output table in which the consumption and the value-added of energy products are isolated from those corresponding to non-energy products in the energy sectors. After this the non-energy part of energy sectors is aggregated, depending on the degree of sectoral disaggregation targeted, into either one large composite sector or into suitable (industry) sectors, as is the case for the final version of IMACLIM-SA reported here. With this re-arrangement in nomenclature we conserve the total value-added (VA) of the economy.

Figure 1.â&#x20AC;&#x201A; Schematics of the hybridisation procedure Statistical sources

Step 1 Matrix in volumes (physical units) FC E prod

Qij Step 2

FC Pij

E prod

Matrix in prices (monetary unit/physical unit)

â&#x20AC;&#x153;Energy billsâ&#x20AC;?

E prod

Vij = Pij . Qij

Other

Energy

Vij Insertion of energy bills into I-O table (Vij)

& adjustment of the accounting tables

E prod

Other

Step 3

An important advantage of hybrid I-O tables concerns own use of energy. For instance, coke ovens produce natural gas that are used to fuel industrial processes again. As these are not economic transactions, they are not registered in national accounts. Of course, the (mostly) primary energy that is consumed by a sector is registered, just like any energy carriers that are sold by a sector (e.g., refinery products and electricity). Thanks to the hybridisation, own use (and transformation) of energy becomes a measurable activity for the I-O table too. The advantage is that it is now possible to model changes in energy technology and for instance energy efficiency more realistically. 20

Application to South Africa In applying this procedure to South Africa the first difficulty was to establish a balanced monetary I-O table with the level of detail needed for our IMACLIM-SA model. In France, for example, this I-O table is directly available from national statistics. For South Africa this is also largely the case, but South Africa’s social accounting matrix (SAM) lacks some detail, and there is no precisely corresponding data for Integrated Economic Accounts for the principal economic agents (firms, public administrations, households and the rest of the world). Therefore, the hybridisation procedure started with adding detail to the I-O table and matching it with Integrated Economic Accounts. An overview of sources can be found in section 5.1. The resulting hybrid I-O table is presented in Appendix 2.

4.1.2. Technical coefficients and integration of bottom-up modelling insights Thanks to the dual accounting of energy and monetary flows, it is possible to properly introduce explicit constraints on the development of energy use in consumption and production. This is especially important in the case of the South African electricity sector. With only a few power plants largely controlled by public authorities and the paramount role of public investment in the development of a notoriously lagging electricity generation capacity, the development of the electricity sector’s technology can be assumed to depend more on political choices than on the development of factor prices. In combination with this, electricity has a regulated price that further reduces the market dimension of the sector. To acknowledge this combination of factors, which effectively act as many market biases outside the grasp of the standard CGE approach, we systematically turn to the SATIM (South Africa Times energy Model) simulations to constrain our prospective development of electricity generation technology. Developed by the Energy Research Centre of Cape Town University, SATIM is a partial equilibrium, dynamic linear optimisation model with a bottom-up (BU) representation of energy production, transformation and use in South Africa (ERC, 2013). Introducing BU insights in a CGE such as IMACLIM-SA raises methodological questions (Ghersi and Hourcade, 2006). The main condition to a consistent linking between the two models is the convergence (to some extent) of the trajectories they share in common, i.e., trajectories of energy supply and demand (in connection to the income of economic agents), of energy prices, and of energy-related investment (both on the supply and end-use sides). In theory at least, it is possible to control the consistency between these trajectories by coupling the BU and the CGE models in one consolidated simulation architecture—an operation referred to as “hard coupling”—, or by performing manual iteration of model runs—“soft coupling”. However, besides being very time- and resource-demanding, coupling (whether hard or soft) naturally results in combined models that are less flexible in terms of the breadth of assumptions they can explore. Also, convergence in prices and investments does not guarantee that the BU model generates unbiased behaviour for the CGE model. In fact, because most BU models (including SATIM) assume a social planner minimising the overall costs of the energy system over the entire time-horizon explored, BU model results are likely to diverge from the aggregation of individual energy-related decisions made by many economic agents in reality (and on which a CGE is calibrated). The relevance of the combined modelling architecture will also depend on the simulation period and on the time-step of interaction between the two models. 21

In our case, we use a simpler linking technique, and set in IMACLIM-SA primary and secondary factor intensities in the electricity sector—i.e., the consumption of physical energy, of other intermediate inputs, of capital and of labour per unit of electricity produced—obtained from SATIM model runs. As a result, IMACLIM-SA reference projection for 2035 embarks the consequences of the Updated Integrated Resources Plan for the electricity sector (DoE, 2013), which SATIM explicitly describes. Similarly, in all the IMACLIM-SA runs with a carbon tax, we use the vector of primary and secondary factor intensities in the electricity sector drawn from a run of SATIM under the same level of carbon tax. This way, IMACLIM-SA captures BU information on the way electricity supply could react to the introduction of a carbon tax.[10] Section 7.5 below discusses the comparability of relevant parameters and variables in SATIM runs vs. derived IMACLIM-SA runs.

4.1.3. Second-best features of IMACLIM-SA The IMACLIM family of models was initially started as an input-output framework, firmly anchored in national accounts data, meant as a macroeconomic shell in which partial equilibrium analyses of energy markets could be embedded to access general equilibrium consistency—not in the textbook sense of markets defined by the optimising behaviour of producers and consumers, but in the more restrictive, although quite fundamental sense, of a comprehensive coverage of economic activity allowing proper assessment of the feedbacks between energy and non-energy markets. As in the previous section, this modelling approach was grounded in a strong intuition that the smooth mechanics of micro-founded consumption and production functions are ill-suited to representing the complex and manifold dynamics at play behind the substitution of capital, labour or other non-energy factors or goods, to energy. Engineering approaches to energy systems modelling based on explicit supply and end-use technology descriptions are key to such a representation, and combining such bottom-up approaches with some top-down comprehensive coverage of economic markets is thus the necessary condition for pertinent energy/economy modelling (Ghersi and Hourcade, 2006). Capital plays a particular role in this combination. There is indeed a huge gap between the standard neoclassical notion of a capital stock, both regularly scraped and augmented through investment accumulation but fundamentally homogeneous through time and the entire economy, and the wide array of successive machine vintages, which when combined with energy consumptions provide energy services—basic ones such as heat, light or motion, but also more complex ones such as those attached to ITCs. To allow better identification of this machine-related reality hiding in the non-labour remainder of value-added, IMACLIM-SA, like most IMACLIM models, distinguishes between “fixed capital consumption” (capital amortisation) and profits or net operating surplus (NOS) in the national accounting sense. [10] A possible extension of the method we used to link SATIM and IMACLIM-SA is to develop “reduced forms” of the SATIM model output, in which price-and-demand-response behaviour of the BU model are synthesised on the basis of a large number of scenario runs to create a “space” of possible future production technologies (vectors of primary and secondary factor intensities) for a sector in the CGE model (Ghersi and Hourcade, 2006). The advantage of this procedure is that the BU model is no longer necessary in the end, as it is encapsulated into the “reduced form” function—making the resulting model much easier to use. Exploration of the space of BU model response can also help identify sets of parameters that are critical for model response, resulting in a multi-parameter “reduced form”. Issues of consistency between trajectories, however, still exist with this method.

Fixed capital consumption coefficients are the entry points for bottom-up information on the capital intensity of energy productions whenever such information is available, as is the case for the electricity sector of IMACLIM-SA (see section 4.1.2 above). Isolated from this physical consumption of capital, profits in the accounting sense must be given a specific dynamic. From a microeconomic perspective, under the assumption of competitive markets and thus marginal cost pricing they could be interpreted as stemming from aggregate decreasing returns to scale. The case seems however too weak and we rather use these profits to calibrate sheer mark-up pricing, which we then consider roughly constant through time—meaning that we almost strictly apply the mark-up rates calibrated on 2005 data to the economy of 2035. We deviate slightly from this rule for reasons specific to South Africa’s economic development, which we further discuss section 5.2.2 below. Another central second-best feature of the IMACLIM models regards the labour market. The transitional nature of both the climate change and the fossil reserve depletion challenges calls indeed for a description of some market imperfections, lest some adjustment costs on the more inert markets be utterly overlooked. At the very least, the possibility of unemployment variations should be modelled for its far-reaching consequences on both economic efficiency (growth) and equity (income distribution). IMACLIM-SA does so by resorting to a “wage curve” approach, which describes an inverse correlation between the real wage and the unemployment rate (see Equation 58 in section 7.3 of Appendix 1). Note that the causality between one variable and the other is not settled in any way: the curve can be interpreted both in negotiation terms—any pressure on labour demand results in an increase of the average wage—and in “wage moderation” terms—any lag in the progression of wages behind consumer prices induces higher labour demand through substitution for other inputs in the production processes.

4.2. Disaggregation of sectors, households and skills 4.2.1. Sectoral disaggregation Our choice of sectoral breakdown reflects the obvious trade-off between an increased detail of modelling outcomes on the one hand, and data requirement, model complexity and transparency of results on the other. Our selection of those sectors most relevant to our research effort rested on the following five criteria: • Value-added (i.e., weight in the South African economy) must be large enough to significantly

impact aggregate growth. • Energy use is a proxy for the sensitivity of a given sector to energy transition policies such as

the carbon tax scenarios we mean to test. • With labour a critical issue in South Africa, total employment in the sector is extremely

important. Data permitting, it would be useful to distinguish between the sectors that primarily demand skilled labour and those that demand unskilled labour. • To distinguish between the sectors that primarily demand skilled labour and those that

demand unskilled labour we compare relative shares of different skill levels in total employment by sector (on the basis of labour data in the SAM 2005 (StatsSA, 2010b). 23

• Finally, the International Panel on ASGISA (Hausmann et al., 2008), among others, finds that

South Africa’s employment partially depends on exports, and thus on tradable goods. Some measure of “tradability”, such as the share of exports in each sector’s uses, should therefore be taken into account.

Table 2 presents a pre-aggregated group of industries from the South African social accounting matrix (SAM) characterised on the above criteria.[11] On the basis of our criteria we eventually decided to distinguish 10 aggregate sectors (Table 4), namely: Coal mining (COA), Oil (imports) (OIL), Gas (extraction, imports and in the form of industrial by-products) (GAS), Production of refined products (oil refineries, CTL and GTL) (REF), Electricity production and distribution (ELC), Energy Intensive Industries (EIN), Manufacturing (MAN), Low-Skill Sectors (LSS), High-Skill Services (HSS), and Transport services (TRA).

[11] The pre-aggregation in which the table is presented is based on the maximum number of industries available from energy balances. The number of industries in the SATIM model is smaller though (ERC, 2013), which poses an inherent constraint—although with additional assumptions some of the SATIM energy model industries can be disaggregated.

1,1% 45

0,4% 0,3

billion Rand

thousands of persons

pct of SA jobs over pct of SA VA

Gross wages

as % of total wages

Number of employees

Percentage high-skill job types

Percentage medium-skill job types

Percentage low-skill job types

Jobs as % of total employment

Jobs over VA ratio

0,0

Ratio of imports over domestic output

Ratio of exports over domestic output

Gold mining 0,9

0,0

1,1

1,6%

17%

69%

14%

203

2,1%

1,5%

Other mining 0,6

0,4

11%

0,3

1,6%

17%

65%

17%

201

2,7%

4,7%

Petroleum 0,1

0,2

33%

0,3

1,3%

22%

49%

29%

156

3,5%

4,5%

Other non-metallic minerals 0,1

0,2

1,5

1,0%

19%

62%

19%

128

0,5%

0,7%

Basic iron/steel 0,4

0,4

0,8

3,1%

20%

57%

23%

380

3,8%

0,1

0,3

0,9

0,4%

14%

53%

32%

0,6%

0,5%

Radio 0,4

3,3

2,7

0,7%

20%

44%

36%

0,3%

0,2%

Transport equipment 0,2

0,5

1,0

1,8%

15%

58%

26%

218

2,0%

1,7%

Textiles 0,1

0,6

1,4

0,9%

23%

57%

20%

111

0,9%

0,7%

1,3

Footwear 0,0

0,9

1,0

0,1%

18%

67%

15%

0,1%

2,1%

0,1

1,0 0,0

2,8% 0,0%

19%

55%

26%

341

Wood, paper, pulp & products 1,6%

0,7%

1,0%

Furniture, Jewelry, & Other

0,1

0,7

2,1%

29%

54%

18%

258

2,7%

3,2%

Note: For VA, Gross wages and energy content the Education sub-sector is included in government, but for employment, skills and imports and export data it is included in “Other activities” for reason of categorisation within the SAM 2005.

Sources: authors own calculation with data from: for Value-Added (VA), Gross wages, and Intermediate Consumption (IC) of energy: Supply and Use tables 2005, StatsSA (2010a); and for the number of jobs, jobs by job type level of skill, and ratios of imports and exports over domestic output: SAM 2005, StatsSA (2010b).

0,5

IC of energy products as % of total IC

19%

61%

20%

1,3%

billion Rand

as % of total VA

Coal mining

Value-added

SAM 2005 sector

Electrical machinery

Table 2. Characteristics of sectors according to sectoral disaggregation criteria Food, Beverages & Tobacco

1.7% 830

6.7% 2.5

billion Rand

thousands of persons

pct of SA jobs over pct of SA VA

Gross wages

as % of total wages

Number of jobs

Percentage high-skill job types

Percentage medium-skill job types

Percentage low-skill job types

Jobs as % of total employment

Jobs over VA ratio

0.07

Ratio of imports over domestic output

Ratio of exports over domestic output 0.12

10%

IC of energy products as % of total IC

38%

50%

12%

2.7%

billion Rand

as % of total VA

Agriculture

Value-added

SAM 2005 sector

Construction 0.00

0.00

2.7

5.9%

21%

61%

18%

734

2.2%

140

Trade 0.01

0.00

1.0

10.3%

24%

50%

26%

1 268

11.7%

10.0%

Hotels & Restaurants 0.22

0.15

3.3

3.3%

30%

49%

22%

413

0.8%

1.0%

Transport services 0.09

0.21

17%

0.6

3.1%

14%

61%

25%

379

4.9%

5.6%

Communication 0.06

0.04

24%

0.2

0.8%

18%

37%

44%

2.1%

3.9%

131

Financial intermediation 0.04

0.02

0.3

2.8%

35%

55%

341

9.5%

9.3%

0.00

0.01

0.1

0.7%

13%

32%

55%

0.8%

6.7%

Real estate

Table 3. Characteristics of sectors according to sectoral disaggregation criteria (continued)

0.02

0.06

1.8

7.3%

16%

51%

33%

907

5.6%

4.0%

Business activities

26 253

Government + Health 0.00

0.00

0.4

7.6%

17%

39%

43%

941

28.9%

202

18.1%

Water 0.00

0.00

0.5

0.3%

18%

48%

34%

0.3%

0.5%

Electricity 0.00

0.00

57%

0.3

0.5%

14%

57%

29%

1.5%

1.9%

122

Other activities 0.04

0.05

3.8

33.0%

62%

16%

21%

4 074

7.1%

8.7%

0.12

0.14

10%

1.0

100%

35%

40%

25%

12 364

100%

699

100%

1 401

Total domestic production

Table 4. Correspondence of IMACLIM-SA to national accounts sectors Sector Code

Sector Name

Corresponding national accounts sectors

COA

Coal mining

Coal*

OIL

Oil

Oil resources*

GAS

Gas

Gas resources, Gas distribution*

REF

Refined products

Refineries*

ELC

Electricity

Electricity*

EIN

Energy-Intensive Industries

Gold, Other mining, Petrochemical industry, Other NMM products, Basic iron/steel, Non-ferrous metals

MAN

Manufacturing

Food, Footwear, Metals basic manufacturing, Electrical machinery, Radio, Transport equipment, Other manufacturing

LSS

Low-Skill Sectors

Agriculture, Construction, Trade, Hotels and restaurants, Domestic and other services, Direct purchases by residents abroad

HSS

High-Skill Sectors

Water, Communications, Financial intermediation, Real estate, Business activities, General government, Health and social work, Education

TRA

Transport

Transport services

*Goods resulting from the national accounts and energy data hybridisation process.

4.2.2. Household and skill disaggregation We disaggregate the current and financial accounts of households between five social groups. To do so, we use information from the SAM 2005 (StatsSA, 2010b) on the distribution of income and its uses (energy and non-energy consumption, investment), as well as data on demographics (see Appendix 2, section 5). Households are also divided into skill classes. On the basis of job type descriptions in the SAM 2005, three skill classes are introduced: high-, medium- and low-skilled. Calibration of skill classes and the linkage between skill classes, social groups and sectors of activity is detailed in section 5.2.1.

4.3. Model specifications 4.3.1. Production trade-offs For all other production sectors save the electricity sector we use KLEM production functions connecting capital (K), labour (L), energy (E) and “materials” (M) intensities to relative price variations. Following Van der Werf (2008), we opt for a nested structure combining capital and labour to form value-added (VA), then VA and energy to form a KLE aggregate, ultimately tradedoff with a “materials” aggregate of non-energy goods and services to produce output (Figure 2). We use fixed coefficients (Leontief) for both the production of the energy aggregate (E) and for the production of the materials aggregate out of non-energy goods. All other substitution possibilities follow a constant elasticity of substitution (CES) assumption—section 6 of Appendix 2 reports on the retained elasticity values.

Figure 2. Nested production function Y EIN COA

KLE

MAN

Mat

TRA

OIL

LSS

GAS HSS REF

KL23

ELC KL3

Constant Elasticity of Substitution (CES)

Fixed production coefficients (Leontief)

As mentioned in section 4.2.2 above, labour is split into three substitutable skill classes, which enter production functions as follows: First, we assume complementarity between capital and high-skill labour. Support for this hypothesis, though restricted to equipment-capital, can be found in Krusell et al. (2000), who demonstrate a skill premium for equipment-capital on the basis of historic data for the US.[12] Next, we assume that the capital-high-skill labour aggregate (KL3) is a substitute for medium-skill labour (L2), and a KL23 aggregate substitutable by low-skill labour (L1) (see Figure 2). The final relation between capital and the three skill levels depends on the elasticities, as discussed in Appendix 2. It is important to note that, as explained above, the capital coefficient of the production tree is calibrated on the fixed capital consumption share of the gross operating surplus only, while net operating surpluses (profits in the accounting sense) are modelled as quite stable mark-ups (cf. section 4.1.3 above).

[12] Krusell et al. have only two types of labour for their estimation model whereas we have three. But our medium-skill category roughly overlaps with Krusell et al.’s unskilled category.

4.3.2. Labour markets We model the labour market as three strictly separated (segmented) markets. In particular, we assume no possibility for high-skill individuals to get jobs requiring lower skills. Since there is currently a shortage of high-skilled workers, and since this tension is expected to persist over time, the strict labour markets separation assumption we make is probably not too problematic. This assumption also limits the complexity of the model and makes it easier to solve. Finally, representation of a job market in which jobseekers can move between skill levels raises both conceptual (what does a “skill” mean in this case?) and practical (how does one calibrate inter-skill mobility?) issues. Further discussion of the notion of skills and of the way they are modelled in IMACLIM-SA is provided in section 6 below. For each skill level, as discussed above, unemployment is endogenised through a wage curve linking wage level and unemployment level (see Equation 57). In production, the sector-average net wage for each skill level is assumed proportional to the base-year ratio of sectoral wage over average wage. For the economy as a whole, the average net wage for each skill is specified by a wage curve in which wage responds to unemployment, with elasticity . We set at 0.2 for all skill levels. Average wage is also corrected proportionally to the Consumer Price Index for all skill levels, and to labour productivity gains accruing to workers (LPgain, see section 5.2.2). In terms of wages by household classes, we use a specific approach to keep the model solvable. We calculate the theoretical revenue of each household class on the basis of old wages and the new number of employed by household class. For each skill level the share of total net wage by skill then becomes the share in new net wage by household class.

4.3.3. Consumption trade-offs We use nested-CES type specifications for consumption trade-offs. However, considering the strong inertias in the consumption of energy and transport by households—e.g., inert housing location choices massively determining households’ demand for energy for private transport—, CES flexibility is allowed only beyond basic needs, i.e., exogenously set floor volume consumptions of electricity, refined products and coal (see Appendix 2).

Figure 3. Households’ consumption decision tree U

COMP

TRA EIN (above basic need)

MAT

EAG

LSS

HSS

REF (above basic need)

ELC (above basic need)

4.3.4. Savings and investment Households’ savings and investment behaviour Households’ savings ratio is exogenous. In the BY 2005 the saving rate of households was very low according to SAM data, at an average of 0.1% (StatsSA, 2010b). This is not a sustainable situation, as household savings fuel domestic investment (see debt positions in our hybrid I-O table in Appendix 2). We therefore assume different savings rates by household class on the basis of the additional assumption that low-income households tend borrow to consume, whereas high-income households save more than lower-income households. We assume an approximate median savings rate of 2% and then define fixed differences between household class specific savings rate and the median savings rate. This results in the savings rates of Table 5. The effective average savings rate for all households will depend on the development of income for each household class in our model projections.

Table 5. 2035 savings rates assumptions, 5 household classes Household class

Class 1

Class 2

Class 3

Class 4

Class 5

Savings rate

0.5%

1.0%

2.5%

3.5%

Investment by households is assumed to be a fixed rate of gross disposable income. Since savings and investment behaviour are fixed relative to gross disposable income, households’ capacity to self-finance their consumption and investment (AFC) is fixed as well. Households of classes 1 to 3 are set to be net debtors and classes 4 and 5 net savers.

Government and firms' investment behaviour Public investment is computed as a fixed share of GDP (as are public expenditures). Firms’ investment is calculated as a fixed share over their gross disposable income.

5. Calibration IMACLIM-SA can be summarised as consisting, on the one hand, of the hybrid I-O table and corresponding current and financial accounts for the economic agents (firms, public administrations, five household classes and the rest of the world); and, on the other hand, of coefficients for the behavioural specifications, notably elasticities. Both require calibration or estimation. The approach on how we combine energy and economic Input-Output data to arrive at the hybridised I-O table has already been explained section 4.1.1. In the present section, we provide an overview of the data used for base-year (BY) calibration, and we present assumptions common to all projections at year 2035.

5.1. Overview of calibration data The model is calibrated on 2005 data. At the time when our research started, these were the latest available SAM data with disaggregation of labour costs and jobs by job type and educational level, for all sectors. 30

We use the following data sources to build the accounting system of IMACLIM-SA—the first two sources together make up the complete Social Accounting Matrix of South Africa.

Monetary I-O table and current and financial accounts • “Final Supply and Use Table 2005” (SU tables), Statistics South Africa, Report No: 04-04‑01

(2005), Pretoria, Statistics South Africa, 2010 (Stats SA, 2010a): This report presents the resources (including taxes and subsidies) of 104 goods and services and their uses by 171 productive sectors, households and government, and for exports and capital formation. It provides a global breakdown of value-added as well.

• “Final Social Accounting Matrix, 2005 (Updated Version)” (SAM 2005 report), Statistics South

Africa, Report No: 04-03-02 (2005) (and accompanying tables), Pretoria, Statistics South Africa, 2010 (Stats SA, 2010b): gives the generation, primary and secondary distribution of income at main levels and sometimes its distribution over industries, products, professional categories and ethnic groups.

• “Financial Statistics of consolidated general government 2010/2011”: Appendix B: Statistical

Tables, Statistics South Africa, Pretoria, 2012 (Stats SA, 2012): details the breakdown of product taxes by type of tax.

• “Quarterly Bulletin March 2007” (QB/SARB): Statistical Tables: National Accounts, South

African Reserve Bank, 2007, Pretoria, South African Reserve Bank, 2007 (SARB, 2007): details the breakdown of generation and distribution of income, social contributions and benefits and other transfers lacking in the SAM for companies, households and government. Numbers differ slightly from the final supply and use table and the SAM.

• “Integrated Economic Accounts”, presentation for the “Regional Seminar on developing a

programme for the implementation of the 2008 SNA and supporting statistics in Southern Africa (17-19 October 2012, Pretoria, South Africa)”, South African Reserve Bank, 2012 (SARB, 2012): gives a more detailed account of the breakdown of generation and distribution of income for households, with numbers corresponding to and more detailed than in the SAM.

Energy volumes • South Africa’s 2005 energy balance from the DoE/RSA (DoE, 2009a): reports data on the

production, import, export, transformation and end use of energy for about 50 fuels or types of energy and for several energy subsectors, 14 industries, 7 types of transport and 4 other sectors (agriculture, services, households, and “non-specified (other)”).

• South Africa’s 2005 energy balance from International Energy Agency’s (IEA) “Energy

statistics of non-OECD countries—2012 edition”, International Energy Agency/OECD, Paris, 2012 (IEA, 2012): provides the same data as the balance by the DoE/RSA above, but with some divergences and generally with more detail.

• Total 2005 final energy consumption data by end use from ERC’s (calibrated) SATIM

model (ERC, 2013): details the end use of energy of 10 types of fuels for 10 industry sectors, agriculture, services, 7 types of residential uses for 3 levels of income, 18 types of transport, as well as exports. 31

• Electric power generation data for 2006 from ERC’s SATIM model obtained from ERC: data

following from the ERC’s research and modelling assumptions on South Africa’s electricity production.

• Coal-to-Liquids (CTL) and Gas-to-Liquids (GTL) data for 2006 SATIM model obtained from

ERC: data following from the ERC’s research and modelling assumptions on South Africa’s electricity production.

Energy prices • 2005 Electricity prices from ESKOM, obtained from ERC (see ESKOM, 2005). • 2005 prices for oil- and CTL- or GTL-derived fuels estimated on the basis of fuel price data

from SAPIA for 2006, obtained from ERC (see SAPIA, 2013).

• Coal prices from the Chamber of Mines for South Africa, obtained from ERC. • 2007 natural gas prices from NERSA (NERSA, 2009). • 2005 petroleum prices for Dubai and Brent crude oil for 2005 published by the Department

of Energy in the “Digest of South African Energy Statistics 2009” (DoE, 2009b).

5.2. Demography, labour and general prospective settings 5.2.1. Current and prospective demography, educational attainment and labour supply Base year demography and labour force For IMACLIM-SA an important aspect of the demographic data used is its implication for the labour market. The SAM gives the number of employed people broken down by job type and educational attainment, but does not report this type of data for the unemployed. This can be found in the Quarterly Labour Force Survey (QLFS) of September 2005 (StatsSA, 2008), but total numbers are not the same. For the number of employed we use the SAM 2005 as the principal number. We use the QLFS to approximate the official unemployment rate of 2005 (26.7%), as well as the number of discouraged workseekers (3.31 million, leading to a total unemployment level of 38.8%). In the QLFS we also find estimates of the number of actives by educational degree (Table 2.5.1 of StatsSA, 2008). Discouraged workseekers are a category normally considered inactive, but for the purpose of this study we treat them as unemployed. In other words, we assume that these persons would return to the labour market, South Africa’s employment situation allowing. We combine these numbers to arrive at an estimation of the number of employed, unemployed (including discouraged workseekers), and inactive by level of education (Table 6).

Table 6. Distribution of 2005 working-age population (15 to 64 years old) by educational attainment and labour market status, thousand individuals Educational attainment None Up to grade 3 / Std 1 Grade 4 / Std 2 Grade 5 / Std 3 Grade 6 / Std 4 Grade 7 / Std 5 Grade 8 / Std 6 Grade 9 / Std 7 Grade 10 / Std 8 Grade 11 / Std 9 Grade 12 / Std 10 NTC I to NTC II Dipl/Cert with less than Gr12/ Std10 Dipl/Cert with Gr12/Std10 Degree and higher Total

Employed Unemployed (A) (B)

Discouraged workseekers (C)

“Broadly” unemployed (B + C)

Active (A + D = E)

Inactive (G – E)

Total (G)

699

153

276

429

1 128

825

1 953

440

112

144

257

697

432

1 129

303 366

111 133

102 125

213 259

516 625

306 376

822 1 001

505

213

193

406

911

578

1 489

786 942 769

299 366 433

300 385 423

599 751 855

1 386 1 694 1 624

899 1 152 1 264

2 284 2 846 2 888

1 078

515

427

942

2 020

1 281

3 301

977

633

365

998

1 975

1 097

3 072

3 311

1 302

466

1 768

5 079

1 430

6 509

123

171

221

166

200

226

1 094

127

180

1 274

167

1 441

754

814

910

12 315

4 486

3 312

7 798

20 113

9 978

30 091

Source: authors’ calculations and assumptions on the basis of SAM 2005 and QLFS 2005 data.

The total (47.6 million) and working age population (30.1 million) counts are mid-year population estimates of Statistics South Africa (StatsSA, 2013).

Base year labour force by skill level Whereas educational attainment is one possible key to skill disaggregation, another possible key is the categorisation of existing jobs: StatsSA provides a categorisation of job types by skill level compatible with the 2005 SAM and details the number of people employed and total net salary paid by job type and by sector (StatsSA, 2010b). The link between job type and skill level, and the assumed link to educational attainment for the SAM 2005 and for IMACLIM-SA are provided in Table 7. This link between salaries and number of jobs is very useful for our labour market and constitutes the basis for our choice to use this skill level categorisation for our Base Year skill level definition. Table 7 presents the number of active (employed and broad unemployed) and inactive population by job type skill levels, and also gives the previously given educational attainment levels in more aggregate categories. 33

Table 7. Classification of job types by skill level and corresponding educational attainment in the SAM 2005 and IMACLIM-SA Skill level, 2005 SAM

Skill level, IMACLIM-SA

Legislator Senior management, Professional

3 / High

University (graduate) or post-graduate degree

Technician

3 / High

Beyond high school education lasting 1 to 4 years, starting age 17/18, except university

Clerk, Service worker, Skilled ag. worker, Craft worker, Plant/machine operator

2 / Medium

Secondary education lasting 5 years, starting at the age of 13/14

Elementary occupation, Domestic worker

1 / Low

No education to primary education

Job type, 2005 SAM

Corresponding educational attainment, StatsSA (StatsSA, 2005)

The comparison demonstrates that the StatsSA linking of job types and educational attainment does not hold in 2005 South Africa: educational attainments are too low to provide all job types with correspondingly educated people. Whether this is a question of survey bias in the QLFS, or reflects a more concrete mismatch of jobs and education attainment in South Africa is unknown. Given South Africa’s history of Apartheid leaving the majority of South Africa’s population without access to (proper) education, it is very likely that many people that gained skills outside formal education occupy high- and medium- skill jobs.

Table 8. Population by educational attainment and IMACLIM-SA skill, 2005[13] Age group

Educational category

Population

Active population

Inactive population

0-14

15 465

No education

1 953

Primary education

6 726

4 691 (low-skilled)

3 058 (low-skilled)

Lower secondary education

12 106

Upper secondary education

8 396

10 386 (medium-skilled)

5 774 (medium-skilled)

5 036 (high-skilled)

1 146 (high-skilled)

20 113

9 978

15-64

Post-secondary education

910

65+

2 084

Together

47 640

Source: authors’ calculation and assumptions, on the basis of SAM 2005 and QLFS Sept 2005 data.

[13] This table aligns educational attainment and skill for the sake of data calibration only. It thus disregards the possibility of people with higher education occupying low-skill jobs—and indeed vice-versa.

Prospective population and labour force by educational attainment We build our prospective labour force disaggregation on the expectation that rising educational attainment will induce a more systematic match of educational attainment and job skill. This allows us easier use of projections of educational attainment. The only such existing projection for South Africa that we know of is that by IIASA (K.C. et al., 2013), who estimate future attainment levels on the basis of a historic link between present and future educational enrolment levels by type of education and age group. As we considered it necessary to have flexibility in demographic scenarios,[14] and as the IIASA’s educational projections lack detail in this respect, we decided to apply their prospective educational attainment split to UN population prospects. However, this resulted in a rapid increase of South African educational attainment, but one that did not appear to be in line with the country’s current concerns about the quality of primary and secondary education, and high public budget deficits. We therefore developed an alternative scenario in which we assume enrolment levels by type of education to remain constant from 2010 onwards. We call this the Low Educational Progress (LEP) scenario and use it as the baseline of educational attainment projection in this study. Projected numbers for 2035 are given in Table 9.

Table 9. Population by educational attainment in IMACLIM-SA’s Low Educational Progress (LEP) 2035 population scenario Age group

Educational category

Population

0-14

14 407

No education

15-64

287

Primary education

3 764

Lower secondary education

15 075

Upper secondary education

18 989

Post-secondary education

2 241

65+

4 765

Together

59 528

Source: authors' assumptions using UN population scenario (Reference) disaggregated by educational class and projections of educational attainment levels from K.C. et al (2013).[15]

[14] One can think of the role of the HIV/AIDS epidemic in southern African countries, but also of the role of immigration and emigration. [15] For the development of scenarios of total population by age group and working age population by educational attainment level, as well as for estimations of expenditure by student we are grateful to Louis de Franclieue for the work he did during his internship at CIRED.

Current and prospective demography and labour supply by household Another issue is the distribution of population by educational attainment, job type or labour market status over the household classes built from StatsSA. To do so, we use three basic principles: 1. T he number of people working at a certain skill level increases with the (expenditure-based) household class. 2. U nemployment of the different skills decreases with the household class—while of course the average wage of the employed increases. 3. Within a household class the average wage increases with the skill level. With the additional assumption that for each class the percentage of inactive workers decreases as the skill increases, we arrive at a base year distribution of wages, employment, unemployment and inactivity over the household classes: see Appendix 2.[16] For the 2035 distribution of labour over household classes, the starting point is the LEP scenario of population by skill level. We simply assume that the distribution of actives over classes is the same in 2035 as it was in 2005, for each of the three skills. Then, to the resulting class-specific working age population, we apply the cross-class (total population) ratios of below-15 and above-65 to working age population to derive class-specific below-15 and above 65 population counts. These numbers can also be found in the above-mentioned Appendix 2.

5.2.2. Parameter settings related to productivity and growth Assumptions about productivity and the labour market Next to demographic changes and education, another crucial driver of economic development is productivity improvements. In IMACLIM-SA these translate into decreased factor intensity of the various productions. In our reference projection as in all our carbon scenarios, we assume the following exogenous improvements in productivity: • Capital productivity: +2% a year • Labour productivity (skill-undifferentiated): +1% a year • “Materials” productivity[17]: + 0.25% a year

Such gains can be interpreted in different ways, as allowed by the capital (and technology) used, by the skill of the worker operating it, by progress in the organisation of production, etc. For this reason we cannot systematically attribute labour productivity improvements to the action of the worker—which is one possible “bottom-up” way of explaining why wage increases should not mirror labour productivity gains, as it appears they have ceased to do for some time in various economies (Cotis, 2009).[18] Another explanation of the same increased wedge between wage and

[16] A more elaborate way of matching skills, wages and social classes might be to exploit the Income and Expenditure Survey (IES) or QLFS data. [17] Materials “productivity” is the inverse of the intensity of production in non-energy goods and services. [18] http://www.insee.fr/fr/publications-et-services/dossiers_web/partage_VA/rapport_partage_VA.pdf

productivity dynamics is increased globalisation: workers see their wage bargaining power weakened by the ever more credible prospect of a reallocation of their production units to foreign countries, which amounts to opening the labour supply competition to countries of potentially lesser wage expectations at similar skill levels. Notwithstanding the comparative merits of these two quite different explanations, we resolved to model such a lag of real wages behind labour productivity improvements. We thus enforce in the wage curves governing each skill market “LPgain” coefficients defined as the (below 1) percentages of labour productivity improvements that become concrete as real wage improvements under the assumption of an unchanged unemployment rate (see Equation 58 in Appendix 1). We relate the values of these lag parameters to the workers’ negotiation positions on the skill markets and hence to the relative abundance or shortage of skills. Even in our conservative LEP demography scenario, the number of people with high school or higher degrees increases sharply, whereas the number of people with primary school or lower education, who naturally target low-skill jobs, decreases both in relative and absolute terms. We therefore assume the following values to the wage expectations lags: • LPgain for High-skilled labour:

33%,

• LPgain for Medium-skilled labour: 50%, • LPgain for Low-skilled labour:

75%[19]

Assumptions about international trade and borrowing and lending to the Rest of the World One last important parameter to our runs is the trend increase of export volumes (the dX of Equation 51 of Appendix 1). It reflects the growth of South African export markets linked to the economic growth of South Africa’s trading partners, independent of terms-of-trade variations. We rather conservatively set this parameter at +1.5% a year. Several runs with the above parameters showed the importance of international borrowing and lending. Without going into the details of our Reference Projection yet (see section 7.1), let us stress that that we additionally assume South Africa’s public deficit to arrive at -2% of GDP in 2035 and investment by firms and households to be constant shares of their gross disposable incomes (GDI). These rules, combined with constant savings rates, induce a structural surplus for households and a structural deficit for firms. The Rest of the World (ROW) mechanically makes up for these deficits. As a consequence, South Africa on the whole (public and private agents aggregated) is assumed to structurally borrow from the Rest of the World, which, in combination with price elasticities of international trade, leads to a trade balance surplus and a devaluation of the South African Rand (see section 7.1).

[19] This reads as: skill 3/2/1 workers expect real wage increases of 33%/50%/75% of their respective labour productivity gains if their specific unemployment rate is unchanged. On this basis, the wage curves of the three skill markets settle what real wage variations are compatible with what unemployment variations.

Adjustment of mark-up rates and specific margins These outcomes became slightly problematic under our original assumption of constant mark-up rates (see section 4.1.3). Due to a decrease of domestic prices and revenues compared to international prices, the price of imported goods increased relative to that of other inputs to production. With our approach to mark-up rates and specific margins being constant rates of the total intermediate and factor input costs, this resulted in an increase of the share of net operating surplus in value-added. This was an unintended effect of constant profit rates, which therefore led us to modify our mark-up rate assumption. We eventually opted to link the mark-up rate to capital productivity, thereby linking operating surplus to capital intensity and thus historic investmentâ&#x20AC;&#x201D; although our approach was purely pragmatic and did not rest on any formalised relationship between profit margins and returns on investment. We thus ended up reducing the rate of markups and specific margins, tNOS and tMS, by a factor amounting to 10% of the capital productivity growth (see EquationÂ 9 of Appendix 1). At this level, the share of NOS in VA remains close to (within 1 percentage point of) the 34.9% calibration data value.

Results and discussion This third part of our report presents and discusses modelling results. In section 7 we explore our reference projection (RP) and the implications of a carbon tax for economic growth, employment and distribution for various recycling schemes, including investment in education. Prior to this, we present in section 6 a methodological discussion of the representation of the dynamics of skill-segmented labour markets. An initial projection obtained with the assumptions described in section 6.1 yielded results that were inconsistent with experts’ expectations of persisting high-skill shortage. As a result, we opted for another approach to labour market segmentation, which we implement in our Reference Projection (RP) and policy scenarios of section 7. Such trials and errors in the construction of models are usually not put forward when reporting results. But there is valueadded to this particular experience, because there is no consensus in modelling of labour market representation, and because it also illustrates how a model can be used to raise the question of the consistency (or lack thereof) of different assumptions.

6. What dynamics for skill-segmented labour? In this section we discuss the dynamics of skill segmentation and its impact on our reference projection. We started exploring a disaggregated and segmented labour market for South Africa based on the seemingly innocuous assumption of a definition of skills as (constant) levels of educational attainment common to our calibration year and our projection horizon. Yet these runs turned out to be inconsistent with development experts’ expectation about the persistence of a high-skill labour shortage. For illustrative purposes we open this section with a presentation of one such run focusing on its results in terms of employment. We then discuss possible amendments to our modelling of supply or demand of skills that could bring model results more in line with experts’ expectations. Finally, we retain one of these options, which we implement for all subsequent scenarios and variants reported in section 7 and further.

6.1. Reference projection with skills defined by educational attainment As exposed above, in our initial trial runs we assumed a constant relationship between skill and educational attainment in 2005 and 2035—e.g., a medium-skill worker would be defined as a worker with completed secondary education, whether in 2005 or 2035. We will further refer to this approach to skill segmentation dynamics as the “constant educational attainment” (CEA) approach. In section 5.2.1 we detailed the link between job type, educational attainment and skill level in BY data. Applying the CEA approach to the LEP scenario for educational attainment, we could produce a prospective 2035 skill segmentation (Table 10), which exhibits increases in both the absolute endowment of high skill labour and the share of high skill labour in total labour. 39

Table 10. Active population by skill level in 2005 and 2035, CEA approach Thousand individuals Low skill Medium skill High skill Total active population

[20]

2005

2035

4 691

2 159

10 386

15 592

5 036

10 759

20 113

28 510

Source: for 2005, author’s estimations on the basis of StatsSA (2008). For 2035, author’s assumptions using UN population scenario (Reference) disaggregated by educational class and projections of educational attainment levels from K.C. et al. (2013).[20]

To illustrate the consequences of the CEA approach to skill dynamics we choose to run it under the exact same set of parameters as that of our eventual reference projection (cf. section 7 below), notwithstanding the arguable double-counting of productivity improvements this entails.[21] This “CEA projection” delivers a real GDP per capita increase of 2.8 times the 2005 level (Figure 4), from ZAR2005 33k to 91.5k.[22] This amounts to an average annual GDP growth rate of close to 4.3%. This is well above the 1.4% a-year (VA-share) weighted average of assumed capital and labour productivity improvements. The gap between the rates is partly explained by a lower dependency ratio, which causes GDP per capita to increase faster than GDP per occupied individual, but also by a 7-point reduction in total unemployment, from 38.8% in 2005 to 31.9% in 2035 (Figure 4). However, the decreasing aggregate unemployment rate masks contrasting results by skill level. Whereas the unemployment rate of low skill workers drops substantially (-20 points) and that of medium skill labour also recedes (-7 points), that of high skill workers registers a slight increase (+1 point), although starting from a substantial 26.5% level in 2005. Such a shift in the hierarchy of unemployment rates is at odds with the fact that South Africa is currently experiencing a pervasive high-skill labour shortage (e.g., Daniels, 2007), and the expectation that such a shortage is likely to persist in the future. This result demonstrates that there is an inconsistency between one or more of our assumptions on the supply and demand of labour in a projection under a CEA approach, and experts’ views on South Africa’s labour market and its foreseeable development. On the supply side of the skill markets, our 2035 skill endowments build on an educational attainment trend from the Wittgenstein Centre, see Table 10 and section 5.2.1 above. Using this educational trend, and under CEA assumption, skill-3 labour force population roughly doubles up, while skill-2 labour force population is stable and skill-1 labour force population is cut by half (Table 10). This already goes a long way towards explaining the differentiated unemployment

[20] Let us recall that we use a broad definition of unemployment and therefore also of active population by extending them to “discouraged jobseekers”. [21] Assuming both a general labour productivity increase and a shift of the skill segmentation in favour of the higher skills amounts double-counting productivity gains compared to an approach with productivity increases only. We disregard this inconsistency for illustrative purposes. [22] This roughly corresponds to an increase in GDP per capita from USD2013 6k to 16.7k, bringing South Africa to the present-day GDP per capita levels of Latvia and Slovakia. Source: current USD GDP per capita, World Bank’s World Data Indicators: http://data.worldbank.org/indicator (accessed February 2015).

Figure 4. Growth and unemployment projections under constant educational attainment (CEA) definition of skills 45.0% 37.5%

300 Unempl. skill 1

38.8%

250 31.9%

30.0%

Unempl. skill 2 200 Unempl. skill 3

22.5%

150

15.0%

100

Average broad unemployment

7.5%

GDP/capita index (right axis)

0.0% Base Year (2005)

CEA Projection (2035)

trends across skill levels—especially considering that all 3 skill markets are depicted by wage curves of identical elasticity (cf. section 4.3.2 above). On the supply side of the skill markets, our 2035 skill endowments build on an educational attainment trend from the Wittgenstein Centre, see Table 10 and section 5.2.1 above. Using this educational trend, and under CEA assumption, skill-3 labour force population roughly doubles up, while skill-2 labour force population is stable and skill-1 labour force population is cut by half (Table 10). This already goes a long way towards explaining the differentiated unemployment trends across skill levels—especially considering that all three skill markets are depicted by wage curves of identical elasticity (cf. section 4.3.2 above). Demand for skills results from a combination of assumptions. Firstly, we assume that labour has uniform productivity gains across skills and sectors (+1% per year). In other words, we do not exogenously force the skill-intensities of the various sectors to evolve in differentiated ways. Secondly, we do not implement any specific form of income elasticity that could allow growth in income to translate into an increased budget share for some skill-intensive goods—one of the plausible explanations for the skill-intensive nature of economic growth. Indirectly, because we consider a constant distribution of the active population among household classes according to skills (Table 11), our model does mechanically record the impact on the aggregate household budget shares of a relatively higher share of high-skill workers in the active population. It turns out that the consumption of household class 5 indeed mobilises relatively more skill 3 workers and relatively fewer skill 1 workers than that of lower (expenditure) household classes, and especially of class 1 (Table 12).

Table 11. Distribution of active population across household classes according to skill, 2005 base year and all 2035 projections Household class 1

Household class 2

Household class 3

Household class 4

Household class 5

Low-skill workers

22%

46%

20%

Medium-skill workers

18%

49%

28%

High-skill workers

29%

68%

Note. due to rounding up, lines may not sum up to 100%.

Source: data from Statistics South Africa and author’s assumptions.

Table 12. Share of skills in the labour content of household classes’ consumption, 2005 (BY) Class 1

Class 2

Class 3

Class 4

Low-skill labour

28%

26%

25%

Class 5 22%

Medium-skill labour

49%

50%

47%

High-skill labour

23%

24%

25%

26%

31%

Reading note: the 2005 consumption of class 1 mobilises labour that is 28% skill 1 labour, 49% skill 2 labour, etc. Source: authors’ computation on base year calibration data.

The projected unemployment rates by skill demonstrate, however, that this increase of high-skill intensity of consumption with household class is not strong enough for the increase of the share of high-skilled in the population to absorb the additional supply of high-skill labour force. The conclusion is that there is a need to change some element of the previously outlined approach to skill supply and demand to allow our projection to have a more plausible comparative development of skill-specific unemployment rates. In the next section we identify three different perspectives for reconsidering our approach to skill segmentation dynamics: (i) revising the link between qualifications of the worker and skill level in the labour market—following on from our previous example: it is possible that by 2035 a high school diploma will no longer be sufficient to qualify for a medium-skill job; (ii) adding income elasticities to consumptions differentiated according to their skill contents; (iii) differentiating labour productivity trends by skill level.

6.2. Alternative skill segmentation dynamics Skills as constant shares of the labour force Looking first at the supply of skills, the unexpected unemployment distribution of our CEA projection could be resolved by changing the representation of skill segmentation. The CEA approach amounts to assuming that skills are strictly related to educational degrees, and thus that skill dynamics can be summarised by the evolution of educational attainment. In this view, firms look for sets of skills—each worker’s educational degree being the indicator of the set of skills he or she has acquired. 42

The opposite approach is to consider education as a positional good. In this view, what matters for firms is the relative, not the absolute educational attainment of individuals. There is indeed some evidence that education has become increasingly “positional” over time (see e.g. Bol, 2015). A simple way to model this second approach is then to define skills as a constant share of the labour force (CSLF). The shares are still calibrated on educational attainment of the calibration year, but are then kept constant in the projection, irrespective of how nominal degree attainments evolve.[23],[24]

Differentiating income elasticities of consumption On the skill demand side, one possibility is to introduce differentiated income elasticities of consumption. The rationale, as mentioned above, is that, as people get richer, they tend to spend a higher share of their income on goods and services other than basic needs, and that these (non-basic needs) goods and services are high-skill intensive. However, the limited product disaggregation of our model does not allow this option to be properly exploited. The Low-Skill Sectors and High-Skill Services aggregates, in particular, are too encompassing for such an exercise.

Differentiating labour productivity trends by skills Differentiating labour productivity trends by skill is another way of increasing pressure on the highskill market. As noted above, we postulate a uniform 1%-a-year labour productivity improvement across skills and sectors, which amounts to an exogenous trend forced upon the labour intensity of sectors. However, beyond sheer productivity gains, some composition or quality effects could indeed translate into increased high-skill intensitites for some sectors. For example, shifting from producing basic equipment to producing technically complex equipment could require higher engineering costs in manufacturing industries (composition effect). Similarly, better enforcement of building regulations could require more consultancy work in building companies (quality effect). We could thus consider forcing exogenous positive trends on the skill 3 intensities of some if not all of our productive sectors.[25]

Bottom line: acknowledging the heterogeneity of 2005 vs. 2035 skills and products Taking a step back, it appears that the issue we are faced with stems from the fundamental heterogeneity of skills and products modelled in 2005 versus their counterparts a distant thirty years ahead. In the abstract framework of CGE modelling, both heterogeneities hide behind identical naming conventions in 2005 and 2035 (naming of skills and products). From this perspective, all three alternatives we came up with allow a similar acknowledgment of the hidden heterogeneities through a changed definition of either the skills themselves, their use in production or the complex [23] Representing skill distribution with constant share of labour force is also consistent with hypotheses about “educational inflation” or “credential inflation”, according to which some academic degrees tend to lose value on the labour market over time. [24] Intermediate approaches between CEA and CSLF are also possible. In particular, the link between educational attainment and skill could be severed only partially, so that an increase in the number of people with degrees attached to skill 3 in 2005 would only partially translate into an increase in the supply of skill 3 people in 2035. The issue then becomes calibration of this “weaker” link between educational attainment and skill. [25] From a sheer mathematical viewpoint this would amount to implementing lower labour productivity gains. For the sake of clarity we would however strongly advocate explicitly distinguishing the two opposite forces.

nature of systems (or technologies) of production that use them. It is thus probable that either of these alternate approaches to skill segmentation could lead to quite similar if not identical projections if properly calibrated (Box 3).

Box 3

Potential numerical equivalence of the CEA and CSLF approaches to skills Consider the CEA skill projection above. Combining its 2035 skill endowments (Table 10) with the assumed uniform 1% annual labour productivity increase makes it possible to compute 2035 efficient labour endowments of 2,910, 21,016 and 14,902 thousand low-, medium- and high-skill “efficient individuals”. An approximated CSLF approach (applied at the level of total labour force)[26] splits the 28,510 thousand labour force of 2035 into 6,649, 14,722 and 7,138 thousand low-, medium- and high-skill individuals. Applying differentiated productivity improvements of -2.72%, +1.19% and +2.39% a year to the low-, medium- and high-skilled makes it possible to exactly match the efficient labour endowments of the CEA projection. The CES approach to production only considers aggregate efficient labour supply and does not distinguish between fewer individuals with higher productivity or more individuals with lower productivity: the employment rate and aggregate wages of all skills cannot but turn out identical. Nonetheless, the number of unemployed individuals will come out differently, which has an impact on the income of household classes and public budgets through unemployment benefits. This distributional impact is likely to be of second order compared to the overall economic outcome, though, and could be cancelled out by playing on the unemployment benefits dynamics.

Within the scope of the present project, we decide to focus on the most straightforward (if not most explicit) treatment of skill segmentation dynamics: we shift the definition of skills from given educational attainment to the positional interpretation, which leads us to define them as constant shares of the labour force. This of course has an impact on our exploration of the consequences of investment in education and training, which we comment upon further below.

7. Reference projection and policy scenarios results We now turn to the results of different scenario analyses performed with IMACLIM-SA. In section 7.1 we discuss the reference projection. In 7.2, we present the policy scenarios, the outcomes of which we discuss in 7.3. Finally, in 7.4 we conclude.

[26] This is an approximation because the CSLF rule applies at sectoral level rather than at the level of the total labour force, which implies that output composition effects should be accounted for. The output composition of the CEA projection is easily used to pinpoint the CSLF skill populations that allow it to be replicated.

7.1. The 2035 Reference Projection In this section we describe key characteristics of the Reference Projection (RP), which we then thoroughly develop by examining the link between employment and structural change, the role of factor intensities, that of exports and the trade balance, impacts on income distribution and on per capita consumption by households. This detailed discussion of the RP helps to set the basis for the interpretation of policy scenarios in the next section.

7.1.1. GDP per capita and CO2 emissions Under the set of assumptions outlined above, and given the Constant Share of Labour Force (CSLF) approach to skill segmentation dynamics, our RP sees real[27] GDP per capita grow from ZAR2005 33K in 2005 to ZAR2005 71K in 2035 (a 116% increase, see Figure 5). This is approximately equivalent to 13K USD2013 per capita in 2035, in line with present-day GDP per capita of Poland or Hungary.[28] With a population increase of 25% this means that in our reference projection total GDP grows 2.7 times compared to 2005 (Figure 5). Total unemployment decreases from 38.8% in 2005 to 28.7% in 2035, whereas CO2 emissions increase by a factor of around 1.8 from 443 to 801 Mt, i.e., from 9.3 to 13.5 tCO2 per capita.

Figure 5. Main performance indicators, base year (BY) and reference projection (RP) 300

45.0% 37.5%

38.8%

250

Unempl. Low skill Unempl. Medium skill

30.0%

28.7%

200 Unempl. High skill 150

22.5%

100

15.0%

7.5%

Average broad unemployment GDP/capita index (right axis) CO2 index (right axis)

0.0% Base Year (2005)

Reference Proj. (2035)

The impact of a carbon tax revenue recycling scheme on employment and inequality will be one of the key criteria for its success. Impacts on employment by skill level will play a big role in these outcomes. It is therefore important to understand RP outcomes in this respect, and to understand to what extent employment impacts are driven by structural change, or by technological change [27] Real GDP per capita is GDP per capita corrected by Fischer GDP price indexation. [28] Source: current USD GDP per capita, World Bank’s World Indicators Data: http://data.worldbank.org/indicator/NY.GDP.PCAP.CD (accessed February 2015)

in production as induced by prices or assumed through productivity gains. We therefore begin by exploring outcomes in employment, before going into changes in output, structural change of GDP and finally impacts on distribution and public budgets. Alongside, we also reflect on the role of international trade.

7.1.2. Employment by sector and skill In relative terms, employment increases most in EIN and ELC industries, but in absolute terms Low-Skill Sectors (LSS) and High-Skill Services (HSS) contribute the most (Table 13). Furthermore, employment growth exhibits a slight shift towards medium- and high-skill labour, whose shares in total labour both increase by 2 percentage points, to the detriment of low-skilled labour (Table 14). In Appendix 3 we report on sectoral changes in employment disaggregated by skill level. These numbers show that it is mainly the LSS sector, and only to a lesser extent the HSS, EIN and MAN sectors that contribute to the growth of medium-skill employment. Growth in high-skill employment mainly takes place in the HSS sector and is fastest in the ELC and EIN sectors, whereas low-skill employment growth comes almost entirely from the LSS sector. From this we can conclude that structural change from LSS to HSS, EIN and MAN only explains part of the shift in jobs by skill level, as within the LSS sector employment grows most for high- and mediumskilled jobs.

Table 13. Sectoral employment, base year (BY) and reference projection (RP) COA

GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

All

802

1 174

5 918

3 923

349

12 315

Share in total

0.3%

0.0%

0.3%

0.6%

10%

48%

32%

2.8%

100%

Growth in jobs (thousands)

+32

+14

+72

+847

+826

+2 756

+2 216

+151

+6 916

Relative growth

+75%

+74%

+40%

+103%

+106%

+70%

+47%

+56%

+43%

+56%

RP employment (thousands)

143

1 649

2 000

8 674

6 139

500

19 231

0.4%

0.0%

0.3%

0.7%

10%

45%

32%

2.6%

100%

BY employment (thousands)

Share in total

Note: we do not report on the OIL sector, for which we assume no domestic production in both BY ad RP.

Table 14. Skill disaggregation in BY and RP BY 2005

Change

RP 2035

Total employed (thousands)

12 315

+56%

19 231

High skill

30%

+2 406

32%

Medium skill

47%

+3 594

49%

Low skill

22%

+917

19%

7.1.3. Changes in volumes of output, imports and uses Output volumes evolve unsystematically in line with GDP or employment, and in contrasting ways. Domestic output rises most in Energy-intensive industries (EIN), Gas supply (GAS) and Coal mining (COA), and least in the Refineries sector (REF) (Table 15). Furthermore, domestic production rises much more strongly than imports, especially in sectors with high elasticities of imports: EIN, MAN and HSS.

Table 15. Change in volumes of resources (Y, M) in RP (2035) as percentage of the respective resources and uses in BY (2005) GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

Output Y

+144%

COA

id.

+154%

+76%

+103%

+186%

+137%

+100%

+117%

+92%

Imports M

+59%

+70%

+62%

+36%

+71%

+16%

-5%

+82%

+3%

+79%

+143%

+70%

+123%

+74%

+103%

+164%

+95%

+99%

+115%

+90%

Total Resource

OIL

On the uses side of national accounts and for most sectors the volumes of exports (X) and intermediate consumption (IC) grow fastest (Table 16). Exceptions are the Low-Skill Sector (LSS) and High-Skill Services (HSS) which increase most through household final consumption (FC). For Coal (COA) it is exports that are responsible for the increase in total uses. The 100% drop in household final consumption of coal is a direct result of our exogenously defined phase-out of coal use by households. Finally, for the Electricity (ELC) sector we observe that it is intermediate consumption (IC) that drives the lion’s share of growth in its uses.

7.1.4. GDP shares of sectors For GDP impacts the results are different from changes in volumes. The sectors that see the strongest rises in output are not always the sectors that see their Value-Added (VA) rise most. Some sectors, especially Refineries (REF), see an increase in value-added that is relatively high (+94%, see Table 17) compared to the increase in domestic output (+76%, see Table 15). On the other hand, sectors like Manufacturing (MAN) and High-skilled Services (HSS) have much lower increases in Value-Added (resp. +103% and +80%) than in their physical Output (Y) (resp. +137% and +117%). As a consequence, the energy sectors, and MAN and EIN see their share in total VA increase.

Table 16. Uses variations in volumes from BY (2005) to RP (2035) Intermediate consumption Household consumption Government consumption Investment

COA

OIL

GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

+83%

+70%

+123%

+83%

+123%

+124%

+110%

+111%

+102%

+114%

-100%

id.

+45%

+48%

+75%

+56%

+129%

+121%

+33%

id.

+122%

id.

+41%

id.

Exports

+277%

id.

+98%

+68%

+256%

+220%

+72%

+224%

+66%

Total Uses

+143%

+70%

+123%

+74%

+103%

+164%

+95%

+99%

+115%

+90%

Table 17. GDP shares of VA by sector in 2005 and 2035 RP COA Share in 1.2% 2005 VA Sectoral VA +234% growth Share in 2035 2.1% VA

OIL

GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

All

0.09%

0.7%

1.9%

11%

21%

48%

5.1%

100%

+111%

+94%

+128%

+145%

+103%

+72%

+80%

+91%

0.1%

0.8%

2.3%

14%

12%

19%

45%

4.8%

100%

7.1.5. Structural change in value of resources The difference between the growth in physical output and that in VA can be linked to changes in the composition of the cost structure of different products. These changes affect Value-Added in a different ways depending on the sector. In ELC, the VA increase relates to an increase in capital and labour intensity that stems from the technical coefficients derived from SATIM (Appendix 3) This is reflected in the rise of the share of labour and consumption of fixed capital (CFC) in the value of total resources; from respectively 21% and 29% to respectively 26% and 34% (comparing Table 18 to Table 19). In EIN, VA grows rapidly compared to other sectors (+234% vs. +91% on average, see Table 17). This is partly due to an increase of input prices. NOS, the profit element of VA, increases along with these input costs, due to the almost fixed mark-up (and specific margins) assumption.[29] These inputs mainly consist of intra-sectoral IC, but also of MAN and TRA products (see Tables 18 and 19)—with the last two facing increasing relative import prices: MAN resources make up 23% of imports, for TRA this is 17%. The combination of increasing input prices (and their consequence for profits)—and a moderate shift away (substitution) from intermediate inputs towards labour and capital factors—explains why the composition of “resources” of the EIN sector does not show much change compared to other sectors (compare Tables 18 and 19). Explanations for the relative growth or decline of VA for other sectors are similar: HSS and LSS sectors have above-average VA as a share of total resources (like COA and ELC), but in contrast to other sectors with lower elasticities for their production functions (see Appendix 2 for a description of elasticities), they substitute more labour and capital (CFC) for intermediate inputs. This is explained by labour and capital benefiting from productivity gains. The consequence is that the HSS and LSS sectors see their costs go up less relative to other sectors. For other sectors inelastic input intensities and quasi fixed profit margins cause VA to grow relatively despite productivity gains for labour not being entirely compensated by wage increases. Coal (COA) also has above average growth of VA, which can be explained by a high growth of the contribution by specific margins, becoming 18% of total resources in the RP, and about 36% of [29] See sections 4.1.3 and 5.2.2. Note that this increase is also not completely proportional, due to the fact that the profit margins are calculated over all costs up to the factory gate, which include the labour and fixed capital part of VA which might rise less than IC costs due to productivity gains. The “proportionality” is thus related to the share of the mentioned intermediate inputs in total costs and how they develop vis-à-vis costs of labour and consumption of physical capital.

total VA in 2035. Almost half the value-added consists of “profit margins” in the instance of coal: the rents on the natural resource, though for a part it also hides the difference in cost structure between domestic low-grade coal and exported high-grade coal.[30] The interpretation of the negative specific and total margins for electricity (ELC) is not straightforward either: the specific electricity margins are negative for electricity supplied to EIN, mining and heavy industries like metals and chemicals. In fact, this is not so much a true negative margin as an issue of aggregation. More precisely, the electricity supplied to mining and heavy industries is in reality cheaper due to lower distribution and administration costs. Since ELC sees its sales (in volume) especially to EIN go up, it yields a total negative margin, which should be interpreted as a consequence of the aggregation of sectors and products, and of the fact that we do not distinguish production technology between customers.[31]

Table 18. BY (2005) resource structure of sectors COA

OIL

GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

All

Energy inputs

1.3%

12%

30%

18%

4.6%

0.5%

2.6%

0.6%

31%

4.4%

EIN, MAN inputs

10%

38%

14%

25%

26%

32%

10%

19%

LSS, HSS, TRA inputs

33%

11%

15%

39%

33%

27%

24%

Labour

19%

21%

11%

29%

16%

18%

Capital - NOS (incl. TY)

21%

10%

14%

28%

16%

13%

Specific margins

0.0%

Capital - CFC

29%

4.8%

Imports

100%

32%

11%

23%

17%

12%

28%

12%

14%

-49%

-16%

14%

Trade & transport margins Product taxes & subsidies

[30] The specific margins are a correction for BY aggregation. The rise of exports versus domestic consumption of coal means that the weight of the higher-priced, but also more costly to produce (to transport) exported (high-grade) coal in the aggregate coal product goes up, whereas this is not reflected in the cost structure. Total margins are therefore probably overestimated, whereas inputs of, for instance, transport from mines to ports are underestimated. In the extreme case where all specific margins (no resource rent) were eliminated (disrespecting equilibria effects), this would represent about 0.8% (36% of 2.1%) of total SA VA in the RP. In reality, a large part would be countered by added VA from other sectors through the higher inputs, therefore the total impact on 2035 GDP (close to total VA) is estimated to be less than 0.5% of 2035 GDP. [31] If the cost of production per unit of output were modelled to represent shifts between types of customers (and thus “types” of products within a sector), then the relative shift in sales from small (residential) to big (industrial) customers would entail an increase of the average profit margin for electricity and a decrease of the average costs per volume unit sold, so that one finally retains a net profit.

Table 19. RP (2035) resource structure of sectors COA

OIL

GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

All

1.2%

11%

40%

15%

5.9%

0.5%

3.3%

0.6%

37%

6.1%

32%

16%

26%

29%

37%

12%

21%

LSS, HSS, TRA inputs

30%

11%

15%

14%

39%

32%

19%

22%

Labour

14%

26%

11%

32%

31%

12%

17%

Capital - NOS (incl. TY)

16%

13%

27%

15%

11%

12%

Specific margins

18%

-6%

0.2%

Capital - CFC

34%

3.4%

Imports

100%

42%

11%

24%

22%

29%

15%

Trade & transport margins

30%

13%

15%

-69%

-16%

Product taxes & subsidies

-1%

Energy inputs EIN, MAN inputs

7.1.6. Structural change in resources and factor intensities In the RP, labour and capital intensity decrease in physical terms (units of inputs per unit of output) except for the electricity sector (Table 20). The main drivers for these changes in intensities are the factor-specific productivity gains that we assumed. Secondly, relative price variations induce trade-offs within the structure of the nested-CES production function, following sector-specific elasticities of substitution (see Appendix 2). Because labour and capital prices, and productivity, change proportionally for all sectors, it is these elasticities of substitution that explain the differences between sectors. Under our assumed elasticities the Refineries (REF) and Transport (TRA) sectors cannot reduce their energy intensities much, despite a strong relative increase of energy prices. For example, refineries energy inputs (mainly from OIL and COA) roughly see a doubling in price relative to other inputs and production factors.

Table 20. Change in physical intensities of production by sector from BY to RP COA

OIL

MAN

LSS

HSS

TRA

Energy

-6%

-2%

-3%

-5%

-7%

-18%

-25%

-30%

-5%

Materials & services

-8%

-7%

-2%

+7%

-7%

-5%

Labour (total)

-28%

-32%

-21%

-0%

-28%

-27%

-28%

-25%

Capital (CFC)

-38%

-37%

-31%

+16%

-38%

-40%

-41%

-37%

Factor or Input

GAS

REF

ELC*

EIN

* Changes in physical intensity for the electricity sector follow exogenous intensities derived from SATIM.

7.1.7. Structural change in uses and the trade balance On the Uses side of GDP we observe that the trade balance (exports minus imports) becomes a bigger part of GDP, whereas investment loses share in GDP (Figure 6).

Figure 6. GDP shares of uses in BY and RP 70% 60% 50%

FC Households

40%

FC Government

30%

Investment

20%

Exports - Imports

10% 0% -10%

BY (2005)

RP (2035)

Contrary to the standard assumption of an exogenous trade balance, the trade balance of IMACLIM results from the closure of the model, i.e., the balancing of GDP expressed from the side of uses and from the side of resources. This is a consequence of the model describing the secondary distribution of income between households, firms and public administrations, which leads to making independent assumptions on the final consumption by households (pC C) and public administrations (pG G) and on investment by households, public administrations and firms (GFCF). With value-added (VA) evolving as labour and capital payments and taxes (T) (indexed on either expenses or volumes), the trade balance (pX X – pM M) is the remaining account that is endogenously set to close the model according to the accounting equilibrium: VA + T = pC C + pG G + FBCF + pX X – pM M. Considering that both export and import volumes (X, M) are elastic to terms of trade, it is through the adjustment of terms of trade that the balance reaches its equilibrium value. This adjustment is obtained by uniformly scaling up or down the set of domestic prices (in fact, the capital and labour prices, on which all domestic prices are built) relative to international prices.[32] As we have hinted, domestic prices are substantially scaled down compared to international prices to allow the trade balance to increase to a positive 5% contribution to GDP. Barring changes in official reserves, this means a need for foreign investment, the reason for which lies in our assumption of maintained deficits of domestic economic agents at their 2005 values.[33] For South Africa as a [32] As international prices do not change among themselves in our current analysis there is not one good whose price can be pinpointed to be the numeraire of the model. [33] To understand this, one has to look at the way consumption, savings and investment are modelled in relation to GDI per agent. There, one can see that this means that auto-financing capacity over GDI by agent (or over entire GDP) is fixed, meaning that the evolution of debt position is fixed too, as well as that of the Rest of the World which needs to balance returns on capital/interest payments and debt positions.

country this means a continued aggregate net deficit, i.e., the inability of the country to finance all its investment domestically (see the discussion in section 5.2.2.).

7.1.8. Distribution of income between household classes, and per capita consumption To evaluate changes in income distribution and inequality we look at the following indicators: • Per capita gross disposable income (GDI) by type of revenue and class • Changes in employment by skill and by household class • Changes and differences in per capita consumption levels

For income, we observe that classes 3 and 4 (which both have gross disposable incomes below the national average) register the highest increases (Table 21). This is due to: • Classes 3 to 5 all benefit at comparable levels from increases in net wages. When looking at

employment impacts, the employment gains, which increase with skill level, are by definition more beneficial to classes 4 and 5, which have a high percentage of high-skill workers in their (working age) population (see Appendix 2).

• Furthermore, all classes benefit from increasing gross operating surplus (GOS) accruing to

households, but class 5 sees little increase in its returns on capital/interest receipts, which make up a large part of its revenue (19%).

The explanation for the second point is that, although we assume 2035 savings rates on average 3 points higher than those calibrated on 2005 SAM data for the BY (average HH savings rate = 0.1%), it is clear that this is too low to sustain households’ relative financial asset positions, which in aggregate decrease from 2.5 times to only 1.3 times total household annual gross disposable income. If it were not for this diminished role of income from returns on capital then income inequality would have grown more, with the income of class 5 growing thanks to their relatively better capital position (classes 1 to 3 are net debtors at the BY) while, due to differentiated saving rates and different historic asset positions, returns on capital do not contribute as much to income for classes 1 and 2. In Table 22 we show what impact the previously described change in household income has on volumes of per capita final consumption. For the increases in consumption per household classes, we observe similar patterns as for household income, meaning class 3 and 4 see the biggest relative increases compared to their BY levels. If we then differentiate the picture by type of product, we notice that in our RP energy consumption goes up only slightly, consumption of industrial and manufactured goods (EIN and MAN: the latter includes food, clothes, electronics, cars, etc.) increases significantly, but consumption of LSS (e.g. agricultural products, construction, domestic workers, retail, and hotel and restaurant services) and HSS (insurance, finance, real estate, health, education, public services) grows most.

Table 21.â&#x20AC;&#x201A; Population, total gross disposable income and per capita incomes per class and the composition of income for BY and RP Household class

Class 1

Class 2

Class 3

Class 4

Class 5

All

Base Year (2005) Total population (million)

4.9

9.3

9.4

146

692

935

1 828

3 849

5 517

9 751

76 642

19 623

Share of income from Net wages

13%

21%

36%

67%

66%

62%

Share from GOS and imputed rents

3.5%

6.1%

11%

20%

18%

Share from returns on K / interest

0.3%

1.1%

1.6%

3.1%

25%

19%

Share from pensions, social security, other transfers

84%

73%

53%

19%

14%

-0.9%

-1.3%

-1.8%

-9.4%

-16%

-13%

Gross Disposable Income (billion ZAR2005) Income per capita (ZAR2005)

Share paid in income taxes

Reference Projection (2035) Total population (million)

5.8

Gross Disposable Income (billion ZAR2005)

114

341

1 328

1 868

Income per capita (ZAR2005)

2 720

6 226

9 959

17 809

111 196

31 387

Change vs. BY

+49%

+62%

+81%

+83%

+45%

+60%

14%

20%

35%

65%

72%

66%

Share of income from Net wages Share from GOS and imputed rents

14%

26%

29%

26%

Share from returns on K / interest

-1%

13%

Share from pensions, social security, other transfers

84%

75%

53%

20%

14%

Share paid in income taxes

-1.1%

-1.6%

-2.2%

-12%

-20%

-16%

A caveat on these outcomes might be that if electricity consumption increases only slightly (+21% and +24% for class 3 and 4 respectively) while sales of manufacturing products, which include home appliances, increase substantially (respectively +49% and +46%), this might signal either an underestimation of electricity consumption or an overestimation of price elasticity for the latter two goodsâ&#x20AC;&#x201D;although their role in electricity consumption might be low compared to demand for heating/cooling and cooking.

Table 22. Per capita consumption of household classes, RP[34] Class 1 COA (GJ) COA change vs. BY* REF (GJ)

Class 2

Class 3

Class 4

Class 5

Total

-100%

0.9

1.9

2.7

4.7

24.8

7.4

Share of basic need

28%

14%

10%

REF change vs. BY

+8%

+13%

+19%

+24%

+8%

+16%

0.4

0.9

1.5

4.2

14.2

4.7

ELC (GJ) Share of basic need

76%

63%

50%

40%

37%

ELC change vs. BY

+6%

+11%

+21%

+24%

+9%

+18%

EIN (units)

0.02

0.05

0.08

0.3

3.9

0.9

+40%

+55%

+75%

+72%

+28%

+40%

0.9

2.1

3.4

5.4

22.8

7.5

+20%

+32%

+49%

+46%

+9%

+25%

EIN change vs. BY MAN (units) MAN change vs. BY LSS (units) LSS change vs. BY HSS (units) HSS change vs. BY TRA (units)

1.0

1.7

2.5

4.3

21.4

6.5

+77%

+95%

+121%

+117%

+62%

+84%

0.3

0.7

1.3

3.2

47.2

10.9

+78%

+96%

+121%

+117%

+62%

+77% 1.1

0.2

0.5

0.8

1.3

2.2

Share of basic need

30%

38%

45%

36%

53%

TRA change vs. BY

-1%

+5%

+11%

+12%

-5%

+6%

* The complete phase-out of coal as a fuel in household consumption is an exogenous assumption that is meant to reflect successful policy on indoor air pollution to ban the use of solid fuels for cooking and heating.

7.1.9. Government revenue and expenses In government revenue and expenses no major changes take place (Table 23), except for two: 1. S ocial security expenditures do not grow as much as other expenses, because per capita social security is only indexed on CPI, and not on net wages, for example. This assumption has strong distributional implications: it amounts to having the social aid-dependent population (the unemployed, the retired) lose ground on living standards compared to the working population, which, at least partly, captures the gains from higher productivity. 2. T he second significant change is that of an increasing weight of interest payments: although public income grows slightly faster than total public expenses, allowing a reduction of the public budget deficit from 3% of GDP in 2005 to 2% in 2035, this is obviously insufficient to prevent the public debt from rising through deficit accumulation.

[34] OIL and GAS are not represented because we did not register nor assume for the future any household consumption of oil or natural gas or other gas products (e.g. coke oven gas) by households.

Table 23. Government revenue and expenses in 2005 and in 2035 for the Reference Projection (RP) In million ZAR(2005)

BY – 2005

RP - 2035

Change

Government revenue GOS and imputed rents

30 471

124 358

+308%

187 843

36%

470 234

32%

+150%

38 908

93 858

+141%

Social contributions

10 204

31 185

+206%

Taxes on production

28 498

77 084

+170%

Corporate, income and property taxes

223 292

43%

669 691

46%

+200%

Total revenue

519 216

100%

1 466 410

100%

+182%

10%

158 718

10%

+180%

VAT (sales tax) Fuel levy

Government expenses Product subsidies (net, minus taxes)

56 736

Social security*

61 330

11%

108 303

+77%

Other transfers

54 735

10%

147 554

10%

+170%

305 732

56%

824 190

53%

+170%

Administration and Consumption Investment

25 702

69 287

+170%

Interest payments

46 498

243 077

16%

+423%

550 733

100%

1 551 129

100%

+182%

Total expenses

* Note that under our assumptions pensions are entirely a matter for firms, though they can be public firms. In reality South Africa has a mixed system, with government pensions for which no contributions are needed alongside employee and/or private pensions.

7.2. Carbon tax revenue recycling scenario definitions Six policy scenarios for carbon tax revenue recycling Our policy scenarios are analysed for the implementation of a carbon tax at two different levels, namely 100 ZAR2005/tCO2 and 300 ZAR2005/tCO2 (see Table 24 for a comparison with present-day values). These levels have been chosen with respect to the South African carbon tax policy proposal of a tax of 120 ZAR2016 per tonne CO2 and for reasons of availability of comparable scenarios in SATIM.

Table 24. 2035 carbon tax levels explored in policy scenarios Unit

in ZAR 2005

in ZAR 2013

in USD 2013

Low level

100 ZAR/tCO2

170 ZAR/tCO2

18 USD/tCO2

High level

300 ZAR/tCO2

510 ZAR/tCO2

55 USD/tCO2

We explore the following options for fiscal or government budget reform in the wake of the carbon tax implementation, i.e. for the recycling of carbon tax proceeds: 1. Reduction of public deficits (possibly leading to a budget surplus) 2. Reduction of sales taxes[35] 3. U niform reduction of income and corporate tax rates relative to the reference projection value 4. C ombination of options 2 and 3: sales tax and income and corporate tax rates are uniformly reduced compared to their reference projection rates 5. Increase of government expenditure 6. Lump-sum transfer to households on a per capita basis The precise rule of thumb in all these recycling options is that of budget neutrality interpreted in the particular sense of a public deficit maintained at the level of 2% of GDP, as targeted in RP. The reasons underlying the different recycling options are as follows: 1. R ecycling option 1 (R1) is the option closest to “preference-neutral”. In CGE models with one aggregate household (class), lump-sum recycling is often considered the reference carbon tax recycling option, as it is neutral in terms of consumption preferences. With 5 household classes however, “preference neutrality” no longer exists, for the reason that preferences and carbon tax burdens differ between classes. By recycling the carbon tax proceeds to the public deficit, the recycling scheme itself does not have a direct impact on the tax burden and household preferences (e.g., by changing relative prices beyond the carbon tax impact).[36] Additionally, public deficit is a significant issue in South Africa, which is looking for ways to increase government revenue in the long term. 2. R ecycling option 2 (R2) explores the economic inefficiencies associated with a carbon tax vis-à-vis that of a sales tax. R2 can also be seen as an overcompensation of households, who, through the sales tax cuts, receive back not only their own carbon tax payments but also those of the polluting firms (for which they of course partly pay indirectly through higher prices). 3. R ecycling option 3 (R3) is an alternative option to compensate firms and households for their increasing energy costs: Because it benefits economic agents according to their non-CO2 emission-related revenue taxes and corporate tax payments, it implicitly transfers some of the total tax burden from the conventional tax payers to the CO2 emitters. 4. T otal carbon tax proceeds potentially exceed total sales tax proceeds for the highest carbon tax level. Hence the introduction of recycling option 4 (R4) as a combination of R2 and R3. [35] South Africa officially has a VAT, but for simplification of model calibration we have replaced it with an ordinary sales tax on final consumption only. [36] Indirectly, the change in budget deficit that results from R1 recycling of carbon tax revenues will change the balance of debts and assets and therefore probably the current account between South Africa and the Rest of the World. This will impact relative price developments as well as exports and import prices. For this reason the R1 scenario should be considered as a slightly different perspective from other policy scenarios.

5. R ecycling option 5 (R5) explores increased public spending and thus, theoretically, increased provision of public services, although the feedback of such services on economic activity are not modelled—except for the option additional to the 6 recycling schemes discussed here, namely that of investment in education and training (discussed next). 6. R ecycling option 6 (R6) is meant as a more progressive version of R3: rather than benefiting households in proportion to their consumption budget it equally shares the total carbon tax proceeds among all residents of South Africa. In this way, South Africa’s large economic inequality might be addressed while reducing CO2 emissions.

Investment in education and labour force training Next to the six revenue recycling options discussed, we additionally explore the possibility that part of the carbon tax proceeds be invested in education and training over the entire time trajectory to 2035. As a result of this investment, we assume one million additional students enrolled in high schools relative to the constant enrolment numbers of the LEP demography scenario, starting in 2015. We then translate these increased enrolment numbers into increased educational attainment by use of the success rates that have been derived from IIASA’s GET (K.C. et al., 2013) and in the LEP demography scenario.[37] At an estimated cost of ZAR2005 6K to 9K per secondary student, this translates into an average additional investment in education of 7.5 million ZAR2005 per year. The sum amounts to about 10% of the 2035 proceeds of the 100 ZAR2005/tCO2 tax rate across the different scenarios, and equals 2.5% of government expenditures (305 billion ZAR2005) on final consumption and administration in our Base Year 2005 data. In reality, such an investment would of course not be so monotonously aimed at one type of education, and would rather impact all levels and types of education, including technical and professional colleges and vocational training. Under a Constant Educational Attainment (CEA) approach for skill definition, this increased spending could translate into a downward shift of skill segmentation thresholds.[38] Under the Constant Share of Labour Force (CSLF) definition of skills (see above), we assume that such spending positively impacts the productivity of primary factors. Based on the estimated educational expenses of our reference projection, and relating this expense trajectory to the productivity gains of our reference projection, we assume that the increased spending triggers a 25% increase in productivity gains, which thus shift from 1% to 1.25% a year for labour, and from 2% to 2.5% a year for capital.

[37] Calculations are notably based on the internship report by Louis de Franclieu: Exploring socio-economic scenarios for South Africa 2005-2035, Louis de Franclieu, Ecole des Ponts/ParisTech, 2014. [38] Note however that the literature does not support a direct link between public education spending and educational enrolment, but rather points to the inefficiency of education spending on secondary education enrolment (see e.g. Grigoli, 2014).

7.3. The carbon tax policy scenarios 7.3.1. Main outcomes Most policy scenarios turn out to have comparable outcomes in terms of growth, employment and carbon emissions (Table 25): At a 100 ZAR2005/tCO2 carbon tax rate the R1, R3, R4 and R6 scenarios see GDP/capita increase from BY to RP between 100% and 105%. Unemployment goes down by 6.2 to 10.8 percentage points, whereas CO2 emissions go up between 36% and 39%. R2 stands out, with a GDP/capita increase of +188% compared to BY, unemployment dropping 13.7 points, but CO2 emissions increasing by 45% compared to BY. R2 thus equals or even outperforms RP in terms of GDP and employment, with still considerable reductions of CO2 emissions relative to RP (20%). R4 finds itself between R2 and the other scenarios, which is understandable since R4 is a combination of R2 and R3. A 100 ZAR2005/tCO2 carbon tax rate is insufficient to achieve a 42% reduction compared to baseline emissions—the 2025 objective of South Africa’s Copenhagen Pledges of 2009 (Janoska, 2014). The pledges aim at an absolute stabilisation of CO2 equivalent emissions between 2025 and 2035, meaning that decarbonisation relative to baseline should even be greater than 42% by 2035. With a tax level of ZAR2005 300/tCO2 this goal is achieved in policy scenarios R4, R5 and R6, with emissions respectively at 44%, 46% and 46% below RP. At this tax rate, R2 achieves a 42% reduction relative to RP. Besides, economic outcomes are still better than for the other scenarios at a 100 ZAR2005/ tCO2 reduction, with GDP per capita going up 109% and unemployment going down 12.8 points. At this level of carbon tax, the reduction in sales taxes for final consumption[39] allowed by recycling the total carbon tax revenue goes from 14% to 4.3%. To explain the success of these carbon tax levels in environmental terms, one should consider the price impact of the carbon tax. In Table 26 and Table 27 we show the ex ante direct price impact of a carbon tax on energy sector product prices for different users for the low and the high carbon tax levels studied. The relative price impact is based on comparing RP prices by sector or end-user (household final consumption) with the calculated sum of carbon tax per unit. This is calculated on the basis of emission coefficients of energy inputs to a sector. For instance, oil is mainly transformed by the refinery sector and little CO2 is released in the transformation process, therefore the emission coefficient of oil to refineries is low and the price impact of a carbon tax for this specific use also. On the other hand, the high price impact for coal can be related to both the high emission coefficient of coal and the very low price that most industries pay for their consumption of domestic (and mainly low-grade) coal. We have not so far commented upon the results of the scenario with partial use of the carbon tax revenue for investment in education and training: R2+. We analyse this for the carbon tax scenario with the best macroeconomic outcomes, that of a carbon tax of 100 ZAR2005/tCO2 with recycling of tax revenues beyond the investment through a sales tax cut (as in scenario R2). The added investment in education, which we assume to accelerate productivity gains of labour and capital by 25%, has a positive impact on GDP compared to the RP, with GDP/capita going up 148%

[39] The sales tax excludes “other product taxes and subsidies”. See Appendix 2 for total numbers in the hybrid I-O table in BY and RP.

compared to BY (versus 116% for the RP). It also has a positive impact on unemployment, which falls by 13.5 percentage points compared to BY (versus 11.1 percentage points in the RP). This comes at the cost of the environmental impact, as the increase in economic activity makes the emission reduction less than in for instance R2 without investment in education and training, namely only 15% less emissions compared to the RP.

Table 25. Key outcomes for tax revenue recycling scenarios with carbon tax levels of ZAR2005 100/tCO2 and ZAR2005 300/tCO2 BY (2005)

Scenario

RP (2035)

R2+

Results for a CO2 tax of 100 ZAR2005/tonne CO2 GDP/capita (ZAR2005) change vs. BY Unemployment change vs. BY CO2 emissions (Mt)

33k

72k

66k

72k

66k

69k

68k

66k

82k

+116%

+100%

+118%

+101%

+108%

+105%

+101%

+148%

38.8%

28.7%

32.6%

25.1%

31.4%

27.8%

28.0%

31.6%

25.3%

-11.1pt

-6.2pt

-13.7pt

-7.4pt

-11pt

-10.8pt

-7.2pt

-13.5pt

443

801

604

644

610

629

615

610

683

change vs. BY

+81%

+36%

+45%

+38%

+42%

+39%

+38%

+54%

change vs. RP

-25%

-20%

-24%

-21%

-23%

-24%

-15%

n/a

Results for a CO2 tax of 300 ZAR2005/tonne CO2 64k

63k

60k

change vs. BY

GDP/capita (ZAR2005)

n/a

+109%

69k

n/a

+95%

+91%

+83%

Unemployment

26.2%

30.3%

30.7%

35.6%

change vs. BY

-12.8pt

-8.5pt

-8.1pt

-3.2pt

CO2 emissions (Mt)

467

449

434

428

change vs. BY

+5%

+1%

-2%

-3%

change vs. RP

-42%

-44%

-46%

Table 26. Ex ante sectoral price impact of 100 ZAR2005/tCO2 per energy product COA

OIL

GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

FC by HHs

COA price

+194%

+195%

+256%

OIL price

+0.1%

GAS price

+10.3%

+9.5%

+4.5%

REF price

+3.4%

+3.0%

+3.4%

+3.9%

+3.2%

+2.8%

Table 27. Ex ante price impact of 300 ZAR2005/tCO2 per energy product by consuming sector COA

OIL

GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

FC by HHs

COA price

+583%

+586%

+768%

OIL price

+0.2%

GAS price

+30.9%

+28.4%

+13.4%

REF price

+10.1%

+9.0%

+10.1%

+10.2%

+11.7%

+9.7%

+8.5%

A caveat on the results of R5 (increased public spending) is that we have assumed no GDP multipliers, productivity increases, or cost reductions as a result of additional public expenditures (for example, transport costs could decrease thanks to improved infrastructure), nor have we translated this into in-kind transfers to households. As a consequence, the economic growth and distributional impacts we report are conservative numbers. For a more detailed presentation of the policy scenario results, we focus on R2 and R6 at 100 ZAR2005/tCO2 tax. We select R2 because it might seem a politically attractive option for the short term; we compare it to the R2 recycling option, which has the best general economic outcomes. Then we analyse R2 with the higher 300 ZAR2005/tCO2 tax to see what happens at higher carbon tax levels, and, because it almost achieves South Africa’s Copenhagen pledges while producing social and economic outcomes that can be considered reasonable to good when compared to RP and other scenarios. Obviously, we also discuss the outcomes of the R2+ scenario in detail.

7.3.2. Employment by sector and skill level In all policy scenarios, employment systematically grows between 2005 and 2035 for all sectors, including the energy sectors. The introduction of a carbon tax causes a slight shift in the distribution of employment away from energy sectors (COA, GAS, REF, ELC) and energy intensive industries (EIN), and towards High-Skill Services (HSS) (see Table 28). The manufacturing sector (MAN) remains largely unaffected by the carbon tax. The distribution of employment by sector is not very sensitive to the different recycling schemes. We also notice that in the R2+ scenario some sectors obtain higher shares in total employment compared to R2 (without the additional investment in education). These sectors are: ELC, EIN, and HSS. The sector that loses share in employment due to the R2+ policy change compared to R2 is the LSS sector (from 45.7% of jobs for R2 to 45.1% of jobs for R2+). In the following we analyse what causes differences in results between the four scenarios. When differentiating the carbon tax impacts by skill level, the picture we obtain is that of a relative shift of employment towards high-skill jobs (Table 29). This is a logical outcome of the shift between sectors that we observe, because the energy sectors and EIN sector together have below average highskill shares in labour, whereas the HSS sector has far above average high-skill employment intensity. Similar to the shift in jobs per sector shown in the previous table, here we see that our modified R2 scenario with investment in education and training causes a shift from low- to high-skill jobs. 60

Table 28. Employment by sector for BY, RP and 4 policy scenarios Thousand jobs BY

COA

OIL

GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

All 12 315

0.0

2.0

802

1 174

5 918

3 923

349

0.34%

0.00%

0.02%

0.28%

0.57%

6.5%

9.5%

48.1%

31.9%

2.8%

100%

0.0

3.5

143

1 649

2 000

8 674

6 139

500

19 231

Share in total

0.38%

0.00%

0.02%

0.25%

0.74%

8.6%

10.4%

45.1%

31.9%

2.6%

100%

C tax 100 R6

0.0

3.4

123

1 532

1 925

8 322

5 948

477

18 438

Share in total

0.33%

0.00%

0.02%

0.25%

0.67%

8.3%

10.4%

45.1%

32.3%

2.6%

100% 20 192

Share in total RP

C tax 100 R2

0.0

3.6

130

1 621

2 098

9 218

6 498

511

Share in total

0.32%

0.00%

0.02%

0.24%

0.64%

8.0%

10.4%

45.7%

32.2%

2.5%

100%

3.5

115

1 540

2 068

9 106

6 471

497

19 900

Share in total

0.27%

0.00%

0.02%

0.23%

0.58%

7.7%

10.4%

45.8%

32.5%

2.5%

100%

C tax 100 R2

138

1 661

2 100

9 077

6 550

506

20 148

Share in total

0.32%

0.00%

0.02%

0.24%

0.68%

8.2%

10.4%

45.1%

32.5%

2.5%

100%

C tax 300 R2

7.3.3. Change in volumes of resources As discussed above, GDP growth in policy scenarios is in most cases only slightly lower than in RP (this is of course without taking mitigation benefits into account). Also, we observe that employment shifts from energy and energy-intensive sectors towards the High-Skill Services sector (HSS). Both observations can be witnessed in terms of output volumes too: With a carbon tax of 100 ZAR2005/tCO2, domestic output (Y) of the MAN, HSS, LSS and TRA sectors goes up more for R2 than it does in the RP. For the energy sectors and EIN this is the contrary.

Table 29. Employment by skill level for BY, RP and 4 selected 2035 C tax scenarios Thousand jobs

skill 1

skill 2

skill 3

Total

2 768

5 845

3 702

share in total

22.5%

47.5%

30.1%

12 315 100%

3 685

9 439

6 108

19 231

share in total

19.2%

49.1%

31.8%

100%

C tax 100 R6

3 450

9 010

5 978

18 438

share in total

18.7%

48.9%

32.4%

100% 20 192

C tax 100 R2

3 975

9 941

6 276

share in total

19.7%

49.2%

31.1%

100%

C tax 300 R2

3 881

9 774

6 245

19 900

share in total

19.5%

49.1%

31.4%

100%

C tax 100 R2+

3 837

9 967

6 344

20 148

share in total

19.0%

49.5%

31.5%

100%

Imports benefit slightly from the carbon tax regime, and in most cases (for most sectors) they are higher under a C tax regime than without one (keep in mind that we assume no border tax adjustment). Compared to the original R2 scenario, the R2+ variant (with additional investment in education) induces increases for both domestic production and imports for all sectors except the MAN sector where imports stabilise at their 2005 level. The explanation is most likely a combination of the high price elasticity of imports of manufacturing products and the fact that Manufacturing’s production benefits from productivity gains and substitution of inputs by labour and capital.

Table 30. Changes of resource volumes versus BY for RP and 4 C tax scenarios GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

+144%

COA

id.

+154%

+76%

+103%

+186%

+137%

+100%

+117%

+92%

RP – imports (M)

+59%

+70%

+62%

+36%

+71%

+16%

-5%

+82%

+3%

+79%

RP - total resource

+143%

+70%

+123%

+74%

+103%

+164%

+95%

+99%

+115%

+90%

C tax100 R6 – Y

+103%

id.

+112%

+65%

+75%

+163%

+126%

+93%

+110%

+84%

C tax100 R6 – M

+37%

+60%

+67%

+36%

+62%

+19%

-2%

+77%

+5%

+72%

C tax100 R6 – total

+102%

+60%

+96%

+63%

+75%

+144%

+88%

+92%

+108%

+82%

C tax100 R2 – Y

+109%

id.

+123%

+74%

+85%

+174%

+141%

+107%

+127%

+94%

C tax100 R2 – M

+41%

+68%

+73%

+42%

+70%

+22%

+4%

+90%

+13%

+81%

RP - output (Y)

C tax100 R2 – total

OIL

+108%

+68%

+106%

+72%

+85%

+154%

+101%

+106%

+125%

+92%

C tax300 R2 – Y

+72%

id.

+86%

+62%

+64%

+155%

+133%

+104%

+125%

+86%

C tax300 R2 – M

+19%

+56%

+83%

+46%

+60%

+25%

+7%

+88%

+16%

+76%

C tax300 R2 – total

+71%

+56%

+85%

+61%

+64%

+139%

+96%

+103%

+123%

+85%

+130%

id.

+143%

+81%

+97%

+203%

+160%

+120%

+147%

+105%

C tax100 R2+- M

+46%

+74%

+78%

+44%

+76%

+20%

-0%

+99%

+11%

+91%

C tax100 R2+ total

+129%

+74%

+121%

+79%

+96%

+179%

+113%

+118%

+145%

+103%

C tax100 R2+ – Y

7.3.4. Change in GDP shares of sectors In RP, as discussed above, changes in VA shares do not follow automatically from changes in output. One reason is that productivity operates only on labour and consumption of fixed capital (CFC) and the share of these two components in VA is not the same for all sectors. A second reason is that the profit part of the capital component is assumed to largely be a fixed margin. This means that a cost increase in inputs, in the case of inflexible demand, can also translate into VA growth in absolute terms. On the other hand, an increase in price of intermediate inputs can also cause a (relative) increase of capital and labour intensity. In our discussion of VA impacts of the carbon tax scenarios, we focus only on the overall outcome of these different mechanisms (see Table 31). The general conclusion from Table 31 is that in our policy scenarios COA, ELC, EIN and TRA have reduced shares of total VA compared to the RP, whereas LSS and HSS have higher shares when compared to the RP. The output gains under our R2+ scenario seem to act against this trend, and VA shares in the R2+ scenario resemble those of the RP, except for HSS, which also sees its share in VA increase in R2+ like in the other scenarios, whereas LSS has a decrease in VA compared to the RP. 62

Table 31.â&#x20AC;&#x201A; Sectoral shares in total Value-Added, BY, RP and 4 policy scenarios COA

OIL

GAS

REF

ELC

1.21%

0.00%

0.09%

0.74%

1.93%

2.12%

0.00%

0.10%

0.75%

2.29%

C tax 100 R6

2.06%

0.00%

0.11%

0.75%

2.22%

C tax 100 R2

1.94%

0.00%

0.11%

0.74%

C tax 300 R2

1.88%

0.00%

0.13%

0.74%

C tax 100 R2+

2.04%

0.00%

0.11%

0.75%

2.20%

EIN

MAN

LSS

HSS

TRA

11.0%

11.2%

20.7%

48.0%

5.12%

14.0%

11.9%

18.7%

45.3%

4.82%

13.8%

11.9%

18.8%

45.5%

4.84%

2.14%

13.4%

11.9%

18.9%

46.1%

4.76%

2.20%

13.1%

11.9%

19.0%

46.3%

4.78%

13.7%

11.9%

18.5%

46.1%

4.74%

7.3.5.â&#x20AC;&#x201A; Structural change in uses The shift in output and VA between sectors under the influence of the carbon tax is, by construction, reflected in changes in uses, both in consumed volumes of products in intermediate consumption and final consumption, as in the expenditure shares of the different types of consumption. Table 32 shows how consumption in volume terms grows compared to BY for the RP and the four selected carbon tax scenarios. We restrict ourselves to presenting volume changes for Intermediate Consumption (IC), Household Final Consumption (FC) and Exports. Changes in volumes of Government final consumption and Investment are uniform for all sectors and smaller in size compared to these other uses and therefore left out. The main observations of Table 32 are that while a 100 ZAR2005/tCO2 C tax under R6 mostly reduces uses of IC, under R2 it forces a relative shift of IC from energy products to non-energy products when compared to the RP. FC sees an even stronger relative shift than for IC, away from energy products and also EIN, MAN and TRA towards LSS and HSS. The same can be observed for R2 at a 300 ZAR2005/tCO2 C tax, but total volumes of uses grow less than for lower tax levels, also for HSS and LSS sectors. Under the R2+ scenario the already-mentioned increased activity (GDP) goes with an increase in all uses when compared to the R2 scenario, and the same for all sectors but COA, GAS, ELC and TRA in comparison to the RP. Volumes of exports clearly benefit from the productivity gains in R2+ and are in the cases of COA, EIN, MAN, LSS and HSS even higher than in the RP. The introduction of carbon taxes affects the composition of GDP from the point of view of expenditures (in value terms) to some extent: The introduction of a carbon tax has an influence on the trade balance, as it has a lower share in GDP when compared to the RP in the R2 and R2+ scenarios (Table 33). Instead, in these scenarios we see a bigger role for household final consumption, which under R2 and R2+ scenarios makes up 63.5% to 64.4% of GDP in contrast to only 61% for the RP and R6 scenarios. Apparently our R2 and R2+ scenarios show untapped potential for growth from domestic consumption. This is encouraged by the sales tax reduction, whereas under the R6 scenario of a lump sum transfer to households we might not observe this effect because a part of carbon tax revenue is saved.

Table 32. Changes in volumes of uses compared to BY of IC, Household FC and Exports for RP and four selected carbon tax policy scenarios

COA

OIL

GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

+143%

+70%

+123%

+74%

+103%

+164%

+95%

+99%

+115%

+90%

Intermediate consumption

+83%

+70%

+123%

+83%

+123%

+124%

+110%

+111%

+102%

+114%

Household FC

-100%

id.

+45%

+48%

+75%

+56%

+129%

+121%

+33%

Exports

+277%

id.

+98%

+68%

+256%

+220%

+72%

+224%

+66%

C tax 100 R6 – Total Uses

+102%

+60%

+96%

+63%

+75%

+144%

+88%

+92%

+108%

+82%

+36%

+60%

+96%

+71%

+88%

+111%

+101%

+102%

+94%

+102%

Household FC

-100%

id.

+39%

+38%

+60%

+54%

+125%

+114%

+33%

Exports

+251%

id.

+87%

+61%

+222%

+200%

+71%

+209%

+65%

C tax 100 R2 – Total Uses

+108%

+68%

+106%

+72%

+85%

+154%

+101%

+106%

+125%

+92%

+43%

+68%

+106%

+80%

+98%

+123%

+115%

+108%

+114%

RP – Total Uses

Intermediate consumption

Intermediate consumption Household FC

-100%

id.

+48%

+47%

+84%

+71%

+147%

+141%

+40%

+254%

id.

+87%

+62%

+225%

+201%

+71%

+210%

+65%

+71%

+56%

+85%

+61%

+64%

+139%

+96%

+103%

+123%

+85%

-1%

+56%

+85%

+70%

+71%

+114%

+109%

+105%

Household FC

-100%

id.

+40%

+74%

+70%

+146%

+142%

+36%

Exports

+232%

id.

+72%

+58%

+197%

+186%

+70%

+200%

+64%

C tax 100 R2+ – Total Uses

+129%

+74%

+121%

+79%

+96%

+179%

+113%

+118%

+145%

+103%

+52%

+74%

+121%

+90%

+113%

+143%

+132%

+133%

+125%

+132%

-100%

id.

+51%

+95%

+78%

+167%

+165%

+39%

+300%

id.

+92%

+63%

+264%

+234%

+73%

+244%

+66%

Exports C tax 300 R2 – Total Uses Intermediate consumption

Intermediate consumption Household FC Exports

Table 33. Composition of GDP by different uses for BY (2005), and for RP and four selected carbon tax policy scenarios in 2035 BY (2005)

C tax 100 R6

C tax 100 R2

C tax 300 R2

C tax 100 R2+

Households FC

GDP share

63.1%

61.2%

61.1%

63.6%

63.5%

64.4%

Government GC

19.5%

19.7%

Investment

18.0%

13.4%

13.2%

13.3%

12.3%

Trade balance

-0.5%

5.9%

6.1%

3.7%

3.8%

3.6%

7.3.6.â&#x20AC;&#x201A; Change in factor intensities Another part of the explanation of key results (the differences in GDP growth, employment and CO2 emissions discussed in section 7.3) lies in the changes in factor intensities. Combined with the changes in output, they explain for instance the decrease in CO2 emissions versus the RP, as well as the (lack of) change in sectoral shares of employment. Labour intensities go down less in the R2 scenarios compared not only to the RP, but also the R6 scenario (Table 24). The reason must be that labour as a factor is more attractive, thus that its output comes at a lower cost. The reason lies in the consumer price indexing of salaries, which is lower (in relative terms) under R2 than for R6 thanks to the sales tax cuts. Also noteworthy, and expected, are the decreases in energy intensity of production with a C tax. When one compares these for a C tax of 100 ZAR2005/tCO2 and one of 300 ZAR2005/tCO2, one

Table 34.â&#x20AC;&#x201A; Changes in factor intensities vs. BY for RP and 4 selected policy scenarios Variation vs. BY

COA

OIL

GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

-18%

-25%

-30%

-5%

RP Energy Materials

-6%

-2%

-3%

-5%

-7%

-8%

-7%

-2%

+7%

-7%

-5%

Labour

-28%

-32%

-21%

-0%

-28%

-27%

-28%

-25%

Capital

-38%

-37%

-31%

+16%

-38%

-40%

-41%

-37%

-13%

-12%

-3%

-18%

-12%

-40%

-31%

-48%

-5%

-8%

-4%

-2%

+34%

-7%

-5%

Labour

-28%

-20%

-21%

-0%

-27%

-28%

-26%

Capital

-37%

-25%

-30%

+34%

-36%

-38%

-39%

-40%

-37%

-13%

-11%

-3%

-18%

-12%

-39%

-30%

-47%

-5%

Materials

-8%

-4%

-2%

+34%

-7%

-5%

Labour

-27%

-19%

-0%

-26%

-25%

-27%

-24%

Capital

-38%

-25%

-31%

+34%

-37%

-39%

-41%

-37%

C tax 100 R6 Energy Materials

C tax 100 R2 Energy

C tax 300 R2 Energy

-16%

-4%

-24%

-16%

-55%

-36%

-59%

-6%

-8%

-0%

-1%

+43%

-7%

-5%

Labour

-27%

-6%

-18%

-0%

-25%

-24%

-27%

-23%

Capital

-37%

-12%

-29%

+67%

-36%

-37%

-40%

-41%

-36%

-13%

-12%

-4%

-18%

-13%

-41%

-34%

-50%

-6%

Materials

-8%

-4%

-1%

+34%

-7%

-5%

Labour

-33%

-26%

-24%

-0%

-32%

-31%

-30%

-32%

-29%

Capital

-45%

-33%

-38%

+34%

-44%

-46%

-48%

-44%

Materials

C tax 100 R2+ Energy

Note: capital intensity variations report consumption of fixed capital variations.

notices that the marginal impact of a C tax seems to be decreasing with a higher C tax level. This, of course, is linked to the CES specification for production. In reality, we might expect more discrete changes with “threshold” tax levels that could function as economic tipping points for some technological changes. For the Electricity sector (ELC), the changes in intensities follow outcomes of SATIM with some additional assumptions. We note an increase in capital intensity, but only a relatively moderate decrease in energy intensity when moving from a tax 100 ZAR2005/tCO2 to one of 300 ZAR2005/tCO2. Two explanations are possible: either energy efficiency gains of electricity production come at higher marginal costs beyond a 100 ZAR2005/tCO2 price signal. Or, it could also be an artefact from the Updated IRP scenario that remains to be explored. For the additional investment in education version of R2+, we observe further decreases in factor intensities of production. Especially (and logically), labour and capital intensities of production go down, which is due to the productivity effect. For the rest, no divergence between sectors seems to be caused by using a part of the carbon tax proceeds to invest in education and training.

Table 35. Gross disposable income (after taxes) per household class and per capita, for RP and four selected carbon tax scenarios Class 1

Class 2

Class 3

Class 4

Class 5

Together

114

341

1 328

1 868

RP GDI (billion ZAR2005) GDI per capita (ZAR2005)

2 720

6 226

9 959

17 809

111 196

31 387

Variation from BY

+49%

+62%

+81%

+83%

+45%

+60%

C tax 100 R6 GDI (billion ZAR2005)

120

336

1 265

1 818

GDI per capita (ZAR2005)

3 576

6 920

10 436

17 510

105 910

30 535

Variation from BY

+96%

+80%

+89%

+80%

+38%

+56%

C tax 100 R2 GDI (billion ZAR2005)

123

374

1 433

2 021

GDI per capita (ZAR2005)

2 935

6 675

10 698

19 496

120 048

33 952

Variation from BY

+61%

+73%

+94%

+100%

+57%

+73%

C tax 300 R2 GDI (billion ZAR2005)

121

367

1 419

1 996

GDI per capita (ZAR2005)

2 874

6 556

10 525

19 157

118 813

33 534

Variation from BY

+57%

+70%

+91%

+96%

+55%

+71%

C tax 100 R2+ GDI (billion ZAR2005)

129

397

1 526

2 147

GDI per capita (ZAR2005)

2 998

6 899

11 260

20 735

127 815

36 066

Variation from BY

+64%

+79%

+104%

+113%

+67%

+84%

7.3.7. Distributional impacts To evaluate the distributional impact we focus on the growth in gross disposable income per class. The gross disposable income per capita (deflated on the basis of the CPI) is obviously different for the R2/R2+ and R6 scenarios. The lump-sum transfer clearly benefits the lower income classes, for whom an equal sum per capita means more income increase than for the higher classes (Table 35), which is caused by the fact that the relative size of the lump-sum transfer for lower income classes is bigger in relative terms when compared to their previous income. Conversely, we observe that class 4 households benefit most from recycling through sales tax reduction, with their gross disposable income growing 17 percentage points more than in the RP. The reason probably lies in a combination of employment effects (especially growth in medium-skill employment; see Table 29 on employment by skill level), salary growth, and consumer price index effects.

7.4. Sensitivity analysis Considering the coverage and depth of our IMACLIM-SA model, producing and commenting upon a comprehensive sensitivity analysis of the above results are beyond the scope of our effort. In this section we nevertheless develop some central parameterisation variants that contribute to a better qualification of our main modelling results. We focus first on our reference projection (RP) to highlight the weight of a series of central assumptions on the growth and unemployment results of the model. We then turn to our most favourable R2+ carbon tax scenario, for which we test the specific assumption of the productivity gain induced by diverting 10% of the carbon tax proceeds into increased educational spending.

7.4.1. Reference Projection parameterisation We group changes to central elements of our RP parameterisation (see Appendix 2 for details) as follows: 1. Growth drivers • Higher growth drivers (GD+): 50% higher export volume trend (from +1.5% to +2.25%

a year) and annual growth of capital (from +2% to +3% a year) and labour productivity (from +1% to +1.5%);

• Lower growth drivers (GD-): 50% lower export volume trend, capital and labour

productivity increases.

2. Labour supply • More flexible labour supply (LS+): higher elasticities of real wage to unemployment for all

three skill levels (a swu of 0.35 instead of 0.20).

• Less flexible labour supply (LS-): lower elasticities of real wage to unemployment for all

three skill levels (a swu of 0.05 instead of 0.20 in the RP).

3. Labour demand • More flexible labour demand (LD+): higher substitutability between capital (or its

composite with L3) and the three (or two) skill levels of labour in comparison to the RP: 67

an elasticity of 0.4 between K and L3, of 2.25 between KL3 and L2, and of 6 between KL32 and L1. • More rigid labour demand (LD-): 50% lower substitutability between capital (or its

composite with high-skill labour) and the skill levels of labour through halving of elasticities: an elasticity of substitution of 0.05 instead of 0.1 between capital (K) and high skill labour (L3), of 0.75 instead of 1.0 between KL3 and L2, and one of 2.0 instead of 4.0 between KL32 and L1.

4. International trade: • Higher price elasticity of exports and imports (IT+): spM and spX 3 times higher than in the

RP (see Table 48 p.158 for RP values);

• Lower price elasticity of exports and imports (IT-): spM and spX 3 times lower than in the RP.

Results of this partial sensitivity analysis of our RP can be summed up as follows (Table 36): • A 50% higher growth in productivity and in export trends (export growth ex ante relative

price corrections) leads to a 44% higher growth of real GDP (+274% vs. +170%), whereas 50% lower growth drivers linearly induce 50% less growth (+85% vs. +170%). The higher growth variant thus appears to increase tensions on the labour markets, which divert more of the growth potential into higher wages. This is linked to the higher absolute impact of our variant on capital productivity than on labour productivity, which tends to make labour a relatively less abundant factor in the variant with increased productivity.

• A rigid labour supply, with 75% lower wage-to-unemployment elasticity, turns out positive

for economic growth with a 2035 GDP 12.2% higher than in RP (303% vs. 270% of BY GDP)— and a total unemployment level 4 points lower. Indeed, a lower swu moderates real wage demands at any employment level, which favours employment and thus growth. Conversely, a more flexible labour supply means increased pressure on real wages at any unemployment level. This is detrimental to growth, although the impact turns out asymmetrical at those points of the wage curves where the RP lies: a 75% higher elasticity of real wages to unemployment only triggers a 5.9% drop of 2035 GDP (254% vs. 270% of BY GDP) and a 2.2 point increase in unemployment.

• Labour demand rigidity or flexibility—lower or higher elasticities of substitution between

capital and labour—has little effect on economic growth, but an interesting impact on employment. Lower elasticities mainly benefit medium and low-skill employment, as they increase complementarity of these two skill levels with capital (and its composite with highskill labour).

• F inally, higher price elasticity of exports and imports leads to significantly more economic

growth compared to the reference projection, whereas lower price elasticity has strong negative consequences for growth. This might be an obvious result in regard of the fact that our RP requires a devaluation of the South African Rand compared to foreign currencies.

The conclusions of this sensitivity analysis therefore are: • The identified drivers of growth truly function as such in IMACLIM-SA and therefore need

careful definition.

• The elasticities of the wage curve have a significant but not excessive impact on model

outcomes. The only exception being that they affect wage moderation or wage increases for high-skill labour (which is in short supply). Considering that our RP wage curve elasticities are rather high (0.2 where 0.1 seems common among CGE modellers), it might mean that we are underestimating economic growth for the given set of drivers of growth. Lowering these elasticities might also help increase the capital intensity in the reference projection, as it turns out to be low compared to BY (and to other current economies).

Capital-labour elasticities of substitution in our nested-CES production functions are •

relevant for the employment effects for medium- and especially low-skill labour, and thus for the distributive outcomes of the model, but not so much for the overall perspective of economic growth and CO2 emissions. In view of our scenario with investment in education, it might be worth studying (and re-defining) the link between our educational approach (CSLF) and the impact of education and training on the capital-labour complementarity.

• Finally, sensitivity to price elasticities of trade is very high, hinting that the devaluation of the

South African Rand is a central element in the economic growth outcome of the RP. This points to a need to re-examine the assumed price elasticities of imports and exports. It also points to a high sensitivity of projections to assumptions about international prices, for which we assume no divergence between different goods. This needs further exploration.

Table 36. Sensitivity analysis of the Reference Projection Sensitivity analysis run

RP (CSLF)

GD+

801

1 007

597

844

778

804

802

382

1 031

Real GDP (vs. BY)

+170%

+244%

+85%

+203%

+154%

+171%

+172%

+4%

+192%

Trade balance-to-GDP ratio (+ is surplus)

+5.9%

+4.8%

+7.9%

+5.7%

+6.1%

+5.9%

+14.2%

+4.3%

116%

95%

151%

112%

118%

116%

262%

86%

CO2 emissions (Mt)

Public-debt-to-GDP ratio

GD-

LS-

LS+

LD-

LD+

IT-

IT+

Investment-to-GDP ratio

13.4%

10.5%

16.5%

13.9%

13.2%

13.1%

15.0%

15.4%

10.4%

L in VA

53.2%

57.5%

49.9%

52.1%

53.8%

53.4%

51.9%

49.7%

59.9%

NOS in VA

34.2%

32.2%

35.2%

34.9%

33.9%

34.2%

34.3%

35.6%

30.6%

CFC in VA

10.6%

8.3%

12.9%

11.0%

10.3%

11.8%

12.7%

7.7%

Unemployment (broad)

28.7%

19.7%

40.8%

24.7%

30.9%

26.9%

31.4%

72.3%

9.0%

9.6%

3.0%

25.1%

1.5%

13.6%

9.7%

10.2%

59.6%

1.0%

Medium skill

32.2%

22.0%

44.9%

26.9%

34.9%

31.3%

33.7%

74.5%

10.9%

Low skill

41.4%

32.7%

48.8%

45.0%

40.6%

35.6%

49.2%

81.2%

13.5%

High skill

7.4.2. Productivity gains from educational investment in the R2+ scenario We base the R2+ scenario of section 7.3 above on assumptions about increasing productivity through investment in education and training that are rough estimates, based on a linear extrapolation of a surmised link between educational investment and productivity growth in our RP scenario. In reality, the impact of such investments will depend on the type of intervention in education and training, as well as on the interaction with technical progress. To further our assessment of the R2+ option, we thus develop sensitivity analysis on the particular parameter that governs the impact of the scenario on productivity. More precisely, we explore what minimum productivity improvement is necessary for R2+, which diverts 10% of carbon tax proceeds into education and training investment, to equal the economic outcomes of R2, which recycles 100% of carbon tax proceeds into sales tax reduction. This minimum productivity improvement turns out to be an arguably modest 1% increase over the thirty years of our projection (see “R2+ 1%” scenario Table 37 below), far below the 25% increase that our rough computation envisages (whose main outcomes are recalled in column “R2+ 25%” Table 37 below). This tends to confirm education and training expenses as a relevant option for carbon tax recycling, as any productivity improvement higher than the 1% threshold thus defined is bound to induce economic performance above that of R2. To complement this exploration and give a notion of how the GDP dividend develops as the educational investment gains in efficiency, we additionally explore the consequences of an induced 5% increase in productivity growth (“R2+ 5%” scenario Table 37 below). Similarly to RP, R2+ GDP growth is less than linearly impacted by increased productivity.[40] Contrary to RP however, a remarkable result is that overall unemployment is hardly affected by the productivity gain assumption and ensuing extra GDP growth: the sales tax recycling option turns out to block the substitution effect of decreased labour costs, probably because its direct impact on “machine” capital costs strengthens as emissions, i.e. the carbon tax fiscal base, increase with growth—a feedback absent from the RP framework. By skill level, employment outcomes do differ, with regressive distributional consequences: the aggregate stable unemployment rate masks a decrease of high-skill unemployment vs. a “bellshaped” impact on medium-skill unemployment—a 0.4-pt increase for the low 1% productivity gain, a 0.3-pt increase for the 5% productivity gain turning into a 0.4-pt decrease for the 25% productivity gain—vs. a systematically negative impact on unemployment of low-skill labour, which moreover increases with the productivity gain. This is however quite conditional upon our specific choices of nested structure of value-added and corresponding elasticities (see Figure 2 p.40 and Table 46 p.157). Indeed, for the simple reason that our RP assumption of productivity gains is 2% for K vs. 1% for all L types, our sensitivity tests, by scaling up the growth rates of productivities rather than their absolute levels, impact K-productivity more than L-productivity (Table 37). As a consequence, the cost of K relative to L3 decreases, which (i) induces moderate substitution of K to L3 (0.1 elasticity) and hence a decrease of the cost of KL3 [40] For R2+ 5% vs. R2+ 1% compare +178%/+173% = 2.9% faster GDP growth to +1.05%/+1.01% = 3.9% faster productivity increases. For R2+ 25% vs. R2+ 5% compare +209%/+178% = 17.4% faster GDP growth to +1.25%/+1.05% = 19% faster productivity increase.

aggregate relative to L2; which in turn (ii) induces a more pronounced substitution of KL3 to L2 (elasticity of 1.5) and hence a decrease of the cost of KL3L2 aggregate relative to L1; which (iii) induces even more substitution of KL3L2 to L1 (elasticity of 4). Then the drop of the cost of the total VA aggregate induces substitution to non-VA inputs that benefits all labour skills. Combined with higher growth, this is enough to systematically compensate the slight decrease of L3 intensity of KL3 but not the strong decrease of L1 intensity of the VA aggregate—for L2 the resulting impact depends on the productivity gain. We base both our nested VA structure and its internal elasticities on the only paper by Krusell et al. (2000), which we must largely interpret to fit our SA context and further labour disaggregation (see discussion section 6.1 of Appendix 2). Besides, sensitivity analysis of RP demonstrates that lower substitution elasticities in our nested VA structure strongly benefit L2 and L1 unemployment (see LD test of Table 36 above). We could additionally ask whether, under our CSLF approach, training and education investment could make low- and medium-skill labour more compatible with new (more productive) physical capital. For all these reasons, the regressive employment impact of our R2+ option is not a robust enough result to find its way into our conclusions (or executive summary).

Table 37. Sensitivity of R2+ results to impacts on productivity growth Scenario Annual K productivity growth K productivity, 2035 (BY = 100) compared to R2 Annual L productivity growth L productivity, 2035 (BY = 100) compared to R2 GDP index (BY = 100) compared to R2 Unemployment (broad) High skill

R2+ 1%

R2+ 5%

R2+ 25%

2.00%

2.02%

2.10%

2.50%

181

182

187

210

+0.6%

+3.0%

+15.8%

1.00%

1.01%

1.05%

1.25%

135

137

145 +7.7%

+0.3%

+1.5%

273

278

309

+0.1%

+2.1%

+13.5%

25.1%

25.3%

7.1%

6.9%

6.1%

Medium skill

28.6%

28.8%

28.7%

28.4%

Low skill

36.8%

37.1%

37.4%

39.0%

CO2 emissions (Mt) compared to R2

644

650

683

-0.0%

+1.0%

+6.0%

Note: “R2+ 25%” is the R2+ scenario of section 7.3.

7.5. Comparison of parameterisation of SATIM runs to IMACLIM-SA variables In the mobilised version of IMACLIM-SA we applied exogenous technical coefficients to the BY inputs and factors intensities of the electricity production and distribution sector (ELC). These intensities were derived from runs of the South Africa TIMES model (SATIM) of the University of Cape Town Energy Research Centre (ERC, 2013). In section 4.1.2 we explained that we integrate such BU information to incorporate certain expectations of, and specific behaviour for, technological 71

development. In this section we discuss the extent to which the use of these parameters is acceptable, as they originate from a different model with its own parameterisation and trajectories. We have not defined strict criteria to measure acceptability.[41] We rather reflect on whether comparability of parameters and outcomes between the two models can be explained by differences in approaches. We also try to assess how adjustment of these parameters would alter outcomes in either model. We must bear in mind, lastly, that the two models must be expected to differ in some dimensions because of fundamentally different approaches: SATIM models a “central planner” with perfect foresight that minimises the total cost of energy systems (within constraints) over the entire modelling period, but without macroeconomic feedback; conversely, IMACLIM has full macroeconomic feedback but operates in a second-best setting and under myopic behaviour. Our comparison of price and demand trajectories and other central assumptions in SATIM versus the 2005-2035 change in IMACLIM-SA shows that there are some similarities, but also a few strong differences between the runs of the two models. BY to 2035 changes for energy prices and total electricity production and other relevant parameters for electricity production and demand are shown in Table 50 and Table 51 in Appendix 3. More than the absolute price levels, which are a question of calibration, it is relative price changes that matter for the projection, for which reason the tables in the appendix show price changes compared to the price index for physical capital (pK in IMACLIM notations).[42] A comparison is made between a full SATIM run that models all energy use of South Africa (and not only electricity), and the “IRP updated” investment plan for electricity production (DoE, 2013).[43] Furthermore, a comparison is made between SATIM runs with as the only difference a carbon tax of 100 and 300 ZAR2010/tCO2 and the R2 scenarios with a 100 and 300 ZAR2005/tCO2 carbon tax level.[44] We observe the following divergences in the development of prices and other parameters/variables between SATIM and IMACLIM-SA runs: [41] In our version of IMACLIM-SA we do not use complete integration of BU modules (like in IMACLIM-R), soft-linking (as in IMACLIM-Brazil, with the MESSAGE model, see Appendix 4), or reduced forms to represent engineering system “behaviour”. We apply estimates of South Africa’s future electricity production technology as an “engineering vision”. Social sciences literature probably has more to say about this kind of scenario building or “knowledge integration”. Possibly, more experience with bottom-up model integration would be needed to define criteria and measurement procedures. For instance, through comparative studies of model uses with different degrees of integration of technology-specific modules, reduced forms, or coefficients. [42] SATIM implicitly assumes constant prices for all goods except energy goods, including capital, labour or materials. Measuring prices relative to capital costs is better than relative to the international oil price or the price of coal as it does not suffer, respectively, from devaluation of the currency or variations of the extraction cost of coal. [43] See: ERC (2013) for more detail about how this is modelled. The version of SATIM that produced technical coefficients for IMACLIM-SA was the version with 5-year aggregated investment made available by the Energy Research Centre of the University of Cape Town on November 12, 2014. The full model and latest versions of input data files can be downloaded from ERC’s website: http://www.erc.uct.ac.za/Research/esystems-group.htm. [44] In the SATIM runs that we use, the carbon tax increases linearly starting from 0 Rand/tCO2 in 2014 to the given carbon tax level in 2020, after which it remains constant. The single time-step of IMACLIM-SA forbids explicit modelling of how the carbon tax develops between 2005 and 2035; it only assumes that the tax has been present long enough for the 2035 economy to have adjusted to it. Furthermore, SATIM uses ZAR2010 as the monetary unit whereas IMACLIM-SA uses ZAR2005. We did not correct for this. Applying an inflator of 1.45, it would result in the ZAR2010 100 and ZAR2010 300 carbon taxes equalling respectively 69 and 207 ZAR2005/tCO2.

• In SATIM, without a carbon tax, the price of coal goes up vs. the unit cost of physical capital

whereas this does not happen in the IMACLIM-SA runs. The rise in costs in SATIM is linked to a shift from cheap to more expensive coal (due to extraction and transport costs);[45] this does not appear in the technical coefficients of IMACLIM-SA, which evolve as functions of relative prices according to their CES specifications.

• With a carbon tax, the relative price increase of coal (compared to pK) is higher in IMACLIM-SA

than in the SATIM runs. This can partly be related to the higher BY coal price in SATIM, which is a question of calibration (due to differences in assumptions about the heating value of coal).

• Difference in the prices of gas and diesel/refined fuels for electricity production are not

very important because of the minor role these fuels play in SATIM and consequently in IMACLIM-SA. Again, SATIM price developments seem more realistic as they capture price expectations of natural gas imports from neighbouring countries and as LNG, whereas in IMACLIM-SA the GAS sector comprises cheap but unsuitable for electricity production coke-oven and gasworks gas, which go to industry.

• Whereas SATIM assumes electricity produced by neighbouring countries to remain relatively

constant in price (compared to pK), in IMACLIM-SA imported electricity increases in price due to the devaluation of the South African Rand that follows from our scenarios.[46]

• Overall, the quantity of imports is underestimated in IMACLIM-SA. This adds to an already

lower amount of electricity made available in South Africa in IMACLIM-SA compared to SATIM, where it follows endogenously on the basis of exogenously determined useful energy demand linked to assumptions about development GDP and household income (ERC, 2013). This might indicate that the price elasticity of electricity consumption is too high in IMACLIM-SA, though in the end such a judgement can only be made based on a comparison with SATIM outcomes for electricity consumption by sector and final consumption by households and other institutional agents.

• The international price of crude oil is another point of divergence. In IMACLIM-SA it rises

strongly due to the currency devaluation effect. This is a difference of perspective on the development of the South African economy. Crude oil is not of direct importance for electricity production, but in the long run it could lead to more substitution of electricity to refined fuels (or CTL fuels). The latter effect is not captured, because of the Leontief CES between energy products for the industry sectors but the “end-point” of this comparison, electricity production, was already found to have possibly been underestimated in IMACLIM-SA.

• Finally, SATIM does not assume a divergence between unit costs of capital or labour costs,

whereas in IMACLIM-SA unit labour costs go up because of decreasing unemployment.

[45] Based on personal communications with Bruno Merven, ERC, University of Cape Town, November 2014. [46] This devaluation is likely compared to main international currencies, and probably less so compared to the currencies of neighbouring countries. On the other hand, due to a low price elasticity of electricity imports in IMACLIM-SA, a lower price would not have altered the balance between domestically produced and imported electricity, arguing for the application of a technical coefficient for electricity imports instead of using an elasticity of substitution.

For the estimation of technical coefficients, the underestimation of the coal price in IMACLIM-SA compared to the engineering-based estimations incorporated in SATIM would, in a CES setting, have led to higher energy intensity versus capital intensity in the RP, whereas with a carbon tax in a “CES case” this would then have probably been reversed too strongly by the excessive relative cost increases of coal consumption.[47] In both cases the use of technical coefficients seems justified. It partly compensates for the shortcomings in IMACLIM-SA, rather than combining seemingly incompatible visions of the future. More problematic might be the divergence in the international oil price between the two models. This would be a problem for total electricity demand and production, rather than for the values of the technical coefficients. It would be logical to expect that under such a macroeconomic development, industry would show a shift to electricity and coal and coal-related fuels (CTL within refined fuels), whereas households would also abandon refined fuels faster and possibly electrified transport would gain too. We have not estimated the order of magnitude of such possible shifts (or in other words: the equivalent price elasticity between oil/refined fuels and coal/CTL and electricity). It could be interesting to test what consequence for electricity use this has under the partial equilibrium settings of SATIM. One possible consequence, a higher demand for electricity, could change the technical coefficients depending on the load curve. If higher demand leads to a relatively higher peak generation capacity requirement, then this would increase the capital intensity of electricity production. If higher demand means a more even distribution in time, then the contrary would be true. Overall we judge the use of technical coefficients from SATIM to be appropriate.

7.6. General discussion 7.6.1. Key results and their interpretations The analysis above can be summarised in five key results, which we interpret below. 1. Tax rates of R100/tCO2 have a significant impact on South Africa’s 2035 GHG emissions (from -15% to -25% relative to RP). Tax rates of R300 have an even greater impact (from -42% to -46% relative to RP), allowing the country to almost meet its emissions stabilisation pledge. Interpretation: though it might appear small in absolute value, the carbon tax represents a significant increase in energy prices. In particular, R100/tCO2 represents an ex ante 256% increase of the price of coal to the electricity sector in our reference projection (see Table 26 p.84). As the carbon tax increases the price of energy, the model substitutes KL aggregate to E in the production functions, thereby reducing emissions. Since long-run elasticities of substitution are relatively large, notably [47] From the presentation of parameter and variable values in Appendix 3, one can observe that the coal intensity of electricity generation is higher in IMACLIM-SA than in SATIM. This is related to how we have calibrated these intensities (relatively to BY). This means that we slightly overestimate CO2 emission intensity of electricity production. In view of the underestimation of electricity consumption discussed in this section as well, we seem to be in the right benchmark of electricity-related coal consumption and CO2 emissions for the RP (see Appendix 3), whereas for the carbon tax scenarios, the impact on coal use by ELC of the (price elasticity-related) electricity demand (low compared to SATIM) is bigger than the overestimate of coal intensity of ELC.

in the HSS sector, the shift is significant. Finally, in most scenarios, economic slowdown (relative to RP) induced by higher aggregate manufacturing costs further limits demand and thus increases emissions reductions. 2. Most policy scenarios lead to lower CO2 emissions relative to RP, but underperform RP both in terms of unemployment (though the rate of unemployment still decreases substantially relative to BY) and of growth of GDP per capita (though GDP per capita increases in all policy scenarios). The only exceptions are R2, which presents a slight double dividend with lower emissions and higher GDP at R100/tCO2, and the ”R2 + investment in education“ policy scenario, which we discuss below (see result No.5). Interpretation: this result would be straightforward if the model assumed a first-best economy. Then environmental gains would necessarily be compensated with economic losses. Within a second-best economy, however, it is possible in theory at least to get both environmental and economic gain if the policy instrument(s) that internalise the carbon externality also reduce pre-existing distortions (see, e.g., Hourcade et al., 1996).[48] Our results simply suggest that the recycling mechanisms we targeted—barring reduction in sales tax that is discussed below—have not corrected distortions large enough to produce a double dividend. 3. The negative impact of a carbon tax on GDP growth can be significant (up to -7.4% relative to RP in 2035 for a R100 tax). However, recycling through sales tax reduction (R2 policy scenario) allows the impact of the tax to be offset (in R2 under a R100 tax, GDP in 2035 is 0.8% above RP). R2 is also the policy scenario which performs best for unemployment, but the least efficient in terms of CO2 emissions. Interpretation: R2 is the only policy scenario in which the price increase associated with higher energy cost is fully compensated (via the decrease in sales tax) from a consumption point of view. Consumption dampening is thus limited compared to the other recycling options, thereby leading to higher demand and higher activity—and consequently higher emissions. Moreover, sales tax cuts are the only recycling option that induces a decreased consumer price index (CPI), which puts downward pressure on nominal wages under our wage curve assumption. As a consequence, conditions for trade improve as well as for employment and real wages. Under our scenario settings, this effect is large enough to lead to a double dividend at the low carbon tax level of ZAR2005 100/tCO2. Other recycling scenarios do not benefit from such an impact on trade and labour: In R1 the lower budget deficit only has indirect implications for consumption through South Africa becoming a net lender to the Rest of the World, this being balanced by a larger trade surplus. The same is true for R5, though in this case, as noted above, our modelling is conservative on the economic implications of additional public spending (e.g., no Keynesian multiplier effect). Lump-sum transfers (R6) do not necessarily boost the income of those households with a higher propensity to purchase high valueadded goods—and lead to higher savings. A similar effect occurs with R3 (reduction in corporate and income tax), but most importantly, these measures do not induce a wage moderation that [48] Note that differentiated sales taxes and international trade modelled according to the Armington specification are “second best” features significant enough to create a double dividend potential (see Ghersi, 2014, commenting on Böhringer et al., 2010).

can stimulate a virtuous circle of employment and consumption, contrary to R2. This finding urges further exploration of alternative labour market specifications, to gain insight on its robustness. 4. In terms of income distribution, RP is more favourable to quintiles 3 or 4 than to higher or lower quintiles insofar as income per capita increases more rapidly in those classes than in the others. In policy scenario R2 all class incomes increase more, but distribution remains skewed towards groups 3 and 4. Lump-sum recycling of carbon tax proceeds (policy scenario R6), on the other hand, has a major progressive impact. Interpretation: as discussed above, this result is straightforward by construction of the lump-sum recycling mechanism. The difference with R2 points to potential political economy problems in the deliberation over policy packages, since different income groups will have different preferences (at least based on the expected outcome in income per capita). It also suggests a difficulty in articulating distributional, growth and environmental considerations in the same policy package. To meet the latter objective, policy packages should be carefully designed to avoid increasing inequality, a result likely if the policy package reduces taxes that are mainly paid by the middle and upper classes of the South African society, or results in investments in education that mostly benefit middle- and upper-class jobs. 5. Recycling part of the tax proceeds of R2 through investment in education results in higher GDP per capita than the full R2 option, even under restrictive views on the impact of such investment on general productivity. However, it also results in higher CO2 emissions (as a direct consequence) and does not benefit employment. Interpretation: the fact that an increase in factor productivity would lead to an increase of GDP is not surprising. This comes at the expected cost of higher CO2 emissions, although detailed results reveal that the growth in emissions is checked well below that of GDP thanks to the further decarbonisation induced by increased substitution of value-added to other inputs.[49] What comes more as a surprise is that this increase of value-added happens to the entire benefit of capital and does not entail unemployment reduction. This is the consequence of the direct decreasing impact of the sales tax recycling option on “machine” capital costs, which strengthens as emissions, i.e. the carbon tax fiscal base, increase with growth—a feedback absent from the RP framework, where, more expectedly, higher growth through productivity gains implies higher employment.

7.6.2. Comparison of model outcome with other analyses Finally, this section compares model outcome with other analyses. It is the occasion to come back to the apparent strong impact of carbon tax on the South African emissions and economy, relative to studies published in other contexts. As noted in section 1.2 above, there has already been substantial analysis of carbon tax implementation in South Africa. The present study is closest to Alton et al. (2012), who find that a carbon tax of R210/tCO2 leads in 2025 to an approximately -42% decrease in GHG emissions relative to BAU—i.e., consistent with South Africa objectives—, with GDP loss in 2025 between 0.68% [49] Compare the impact of the three tested R2+ options on R2 GDP vs. R2 emissions, Table 37.

and 1.23% relative to BAU, depending on the recycling option. Thus, Alton et al. also find that the South African economy is very responsive to seemingly modest carbon taxes, notably because of low base prices of energy (as also outlined e.g. in Pauw, 2007, Table 6). On the other hand, they find a narrower range of impacts of a carbon tax on GDP than our own findings, all the more so given that in their study GDP is growing at 3.9% per annum over 2010-2025 (against 2.5% per annum over the 2005-2035 period in ours).[50] This may be due to the fact that their recycling schemes all include a tax on imports and a rebate on exports on top of the domestic carbon tax, which decreases domestic activity to the detriment of foreign producers. Also, Alton et al. note that the elasticity of mitigation costs to level of mitigation is very high, around the -42% level (Figure 3).[51] Another interesting comparison point is with van Heerden et al. (2006). They also obtain strong impacts of limited carbon taxes on South African emissions. The particularity, as noted in 1.2, is that they find strong double dividend—i.e., an increase in GDP with the carbon tax relative to baseline— when recycling carbon tax revenues into food tax breaks. The mechanism, they contend, is that “when energy is complementary to capital [as they assume], and when tax revenue recycling can be used to increase [infinitely supplied] unskilled labour demand [as the food tax break does by increasing demand for agricultural products and thus for agricultural, mostly low-skilled labour], a double dividend may materialize in South Africa as in the model of Bovenberg and van der Ploeg (1996; 1998).” We do not test a similar recycling mechanism, so the comparison is not direct, but the results would likely be different because in our model, energy and capital-labour aggregate can be substituted, and because the wage curve limits the increase in demand for unskilled labour.[52] Turning to the numerous studies on other countries, let us simply underline that our findings echo their IPCC synthesis, which is most detailed in the institution’s 2001 report (IPCC, 2001):[53] any given economy’s most efficient recycling option is that of reducing its most distortionary pre-existing tax. Faced with the South African economic and social context, IMACLIM-SA purposely lends great importance to imperfect segmented labour markets where wages cannot freely adjust to absorb labour supply because they are also required to preserve purchasing power.[54] It is an unexpected [50] We find GDP losses for a R200/tCO2 tax in the R2 scenario around 5% relative to RP in 2035. Assuming GDP growth rates are constant over the 2005-2035 period in both RP and R2 scenario—a simplifying assumption—then GDP in 2025 is around 4.7% below RP in 2025, to be compared with only (-0.68% to -1.72%) in Alton et al. (2012). [51] A second difference with the result of our study is that amongst the recycling schemes they test, the corporate tax has significantly better impacts in terms of GDP relative to the sales tax. They argue that this is because firms are major providers of savings in the South African economy, and thus that increased savings result in higher investment. Though our results are not directly comparable (R3 includes both reductions in income and corporate taxes), the difference seems significant. It might again be due to the presence of a border tax adjustment (absent from our analysis) and from a higher propensity of firms to invest in their model. [52] Finally, comparison with Devarajan et al. (2011) is more difficult because they use static CGE with a different mitigation target. [53] Comparison with any precise analysis conducted in other countries is difficult. Not only are models, key assumptions and policy packages different—as is the case with the other studies conducted on South Africa—, but the metrics of comparison is also less obvious. A carbon tax expressed in PPP USD per tCO2, for example, may still correspond to widely different increases in domestic energy prices (or in the prices of other goods) depending on pre-existing price levels and cost structures. [54] Note that, conversely, the model’s treatment of capital supply as demand-driven does not allow for any distortion of capital supply to be corrected by e.g. recycling in income tax reduction. This assumption and its consequences are arguably adapted to the South African economy, where investment shortage mainly concerns infrastructure investment by the public sector, i.e. little sensitivity to capital income taxation.

consequence of this central second-best feature of our model that the most efficient recycling option should turn out to be that which best preserves purchasing power, namely the reduction of sales taxes. Note that Van Heerden et al. (2006) distinguish between skilled and unskilled labour on the basis of job types in the South African context. They model high-skill labour as inelastic and low-skill labour as perfectly elastic.[55] By this means, revenue recycling through a reduction of taxes on food (VAT) increases demand for low-skill labour more than general VAT or direct tax breaks, but does not impact wage setting for high-skill labour. As a consequence, they might underestimate economic benefits from a VAT tax break through wage-setting and demand for skilled labour (equivalent to our medium- and high-skill labour), whereas we might slightly overestimate them by overestimating the CPI impact on low-skill wage-setting.

[55] Unfortunately their paper lacks further detail on what inelastic supply of high-skill labour implies for wage-setting.

General conclusion In this report we explore the conditions under which environment and development goals can be articulated in the context of a rapidly growing economy such as South Africa’s. Precisely, we explore the articulation between climate mitigation, unemployment reduction and economic growth. To do so, we use the development of IMACLIM-SA, a hybrid CGE model of the South African economy, to evaluate several policy packages at the 2035 horizon. We show that a moderate carbon tax (R300/tCO2), as envisioned by the South African Government, could allow the country to reach its mitigation pledge. However, we find only one instance of a double dividend, and mitigation comes at a cost in terms of less rapid economic growth and a lower decrease in unemployment relative to baseline. Recycling carbon tax proceeds through sales tax appears to be by far the most efficient in terms of GDP and unemployment, but with no improvement in distribution relative to baseline, while recycling the carbon tax proceeds through lump-sum transfers has progressive implications (again relative to baseline) but does impact growth performance. Finally, investing part of the carbon proceeds in the education system—to the extent that it leads to an increase in labour and capital productivity—leads to an increase in GDP, but does not improve unemployment for quite specific reasons when the rest of proceeds are used to cut sales taxes. From a methodological standpoint, the study illustrates two ways in which a prospective economywide model such as IMACLIM can be used to inform policy making, even in a context of limited information. It makes it possible to explore the links between issues that are not necessarily connected in actual policy making: here education policies and environmental policies.[56] And even with limited information—here without consensus over the way skills should be modelled—the model can test the consistency between various skills modelling and experts’ assumptions about the persistence of skill shortage. We find one model of skill segmentation dynamics inconsistent with experts’ views, thus leading us to use another one to articulate our scenarios. From a policy perspective, the present analysis suggests that the South African economy has a strong potential for mitigation, but that careful attention must be paid to the design of the recycling scheme to limit implications for GDP growth. Out of the set of recycling options we test, a reduction in sales tax rate appears to be the most promising. Investment in education can have an even better outcome in strict efficiency terms, but does not improve employment when combined with sales tax reduction, which has a quite specific positive impact on the relative cost of capital. Finally, the set of instruments we have tested suggests that there is some form of trade-off between the impact on GDP and unemployment and distributional issues.

[56] In studies conducted in France, environmental policies and the long-run sustainability of pension financing (Combet et al., 2014).

As noted throughout the text, the analysis has limitations that should be overcome in future work. In particular, additional information could be drawn from the SATIM BU model, possibly through soft coupling of the two models. Also, the calibration of production and consumption functions could be improved, data permitting. Third, a wider range of labour market specifications could be tested beyond the two that we have relied upon in this exercise. Finally, but in a longer-term perspective, an important element missing from the current model and very relevant to South Africa is a representation of the informal economy, and its relationship with the formal economy. In terms of policy issues, the present analysis has focused on four main development objectives: economic growth, employment, equity across income groups and carbon mitigation. These are certainly critical, but others such as health or other environmental issues such as water management could be explored in future exercise. Similarly, the range of policies explored could be expanded in a future exercise, to include on the one hand other forms of public investment (in transport infrastructure, for example) as well as industrial development policies.

References Alton, T., C. Arndt, R. Davies, F. Hartley, K. Makrelov, J. Thurlow, D. Ubogu. 2012. The Economic Implications of Introducing carbon taxes in South Africa. WIDER Working Paper No.2012/46. Alton, T., C. Arndt, R. Davies, F. Hartley, K. Makrelov, J. Thurlow, D. Ubogu. 2014. Introducing carbon taxes in South Africa. Applied Energy (116):344–354. Aron, J., J. Muellbauer, J. Prinsloo. 2007. Balance sheet estimates for South Africa’s Household Sector from 1975 to 2005. South African Reserve Bank Working Paper/01/07, Pretoria. Baker, L., P. Newell, J. Phillips. 2014. The Political Economy of Transition: The Case of South Africa. New Political Economy 1–28. Blanchflower, D. G., A. J. Oswald. 1995. An introduction to the wage curve. The Journal of Economic Perspectives: 153–167. Bol, T. 2015. Has education become more positional? Educational expansion and labour market outcomes, 1985-2007. Acta Sociologica. doi: 10.1177/0001699315570918 Böhringer, C., A. Löschel, U. Moslener, T. Rutherford. 2009. EU Climate Policy up to 2020: An Economic Impact Assessment. Energy Economics, 31(S2), S295–S305. Brécard, D., A. Fougeyrollas, P. Le Mouël, L. Lemiale, P. Zagamé. 2006. Macro-economic consequences of European research policy: Prospects of the Nemesis model in the year 2030. Research Policy 35(7):910–924. Callonnec, G., F. Reynès, Y. Y. Tamsamani. 2011. Une évaluation macroéconomique et sectorielle de la fiscalité carbone en France. [A macroeconomic and sectoral assessment of carbon taxation in France]. Revue de l’OFCE / Débats et Politiques (120):123–154. Combet, E., F. Ghersi, J.-C. Hourcade, C. Thubin. 2010. La fiscalité carbone au risque des enjeux d’équité [Carbon tax at risk of equity]. Revue Française d’économie 25(2):59–91. Combet, E., F. Ghersi, J.C. Hourcade, D. Théry. 2010. Carbon Tax and Equity: The Importance of Policy Design, in: Dias Soares, C., J. Milne, H. Ashiabor, K. Deketelaere, L. Kreiser (eds.), Critical Issues In Environmental Taxation. USA. Combet, E., H. Waisman, J.-C. Hourcade, F. Lecocq. 2012. AFD research programme “Green Growth and its Implications for Public Policy – The Case of South Africa”. Intermediary report #1. SMASH: Nogent-sur-Marne. Comité « Trajectoires 2020-2050 vers une économie sobre en carbone ». 2011. Trajectoires 2020 – 2050 vers une économie sobre en carbone. Rapport du Comité présidé par Christian de Perthuis. Paris. 81

Daniels, R.C. 2007. Skills Shortages in South Africa: A Literature Review. DPRU Working Paper 07/121. University of Cape Town. Death, C. 2014. The Green Economy of South Africa: Global Discourses and Local Politics. Politikon: South African Journal of Political Studies 41(1):1–22, Devarajan, S., D. S. Go, S. Robinson, K. Thierfelder. 2011. Tax Policy to Reduce Carbon Emissions in South Africa. World Bank Policy Research Working Paper No.4933. Washington D.C.: World Bank. Devarajan, S., D. S. Go, S. Robinson, K. Thierfelder. 2011. Tax Policy to Reduce Carbon Emissions in a Distorted Economy: Illustrations from a South Africa CGE Model. The B.E. Journal of Economic Analysis and Policy 11(1). DoE (South Africa Department of Energy). 2009a. Energy Balances, 2005. Department of Energy, Republic of South Africa, Pretoria. Accessed at: http://www.energy.gov.za/files/media/media_ energy_balances.html. DoE (South Africa Department of Energy). 2009b. Digest of South African Energy Statistics 2009. Directorate: Energy Information Management, Department of Energy, Republic of South Africa, Pretoria. DoE (South Africa Department of Energy). 2011. IRP 2011. Pretoria. DoE (South Africa Department of Energy). 2013. Integrated Resource Plan for Electricity (IRP) 20102030 Update Report 2013, Department of Energy, Republic of South Africa, Pretoria, 21 November 2013. EIA (South Africa Energy Information Agency). 2013. South Africa. Country profile, version January 17, 2013. Washington, D.C. ERC (Energy Research Center). 2013. Assumptions and Methodologies in the South African TIMES (SATIM) Energy Model. Version 2.1, 2013/04/08. Systems Analysis & Planning Group, Energy Research Centre, University of Cape Town, Cape Town. ESKOM. 2005. Tariffs and Charges. Effective from 1 January 2005 – 31 March 2006. ESKOM, Johannesburg. Available at: http://www.eskom.co.za/c/article/141/tariff-history/. Ghersi, F. 2014. Low-carbon policy making vs. low-carbon policy modelling: State-of-the-art and challenges. Environmental Modeling and Assessment, 19, 345-360. Ghersi, F., J.-C. Hourcade. 2006. Macroeconomic Consistency Issues in E3 Modelling: The Continued Fable of the Elephant and the Rabbit. Energy Journal, Special Issue No.2: 39–62. Grigoli, F. 2014. A Hybrid Approach to Estimating the Efficiency of Public Spending on Education in Emerging and Developing Economies. IMF working paper 14/19. Hausmann, R. 2008. Final Recommendations of the International Panel on Growth, Pretoria: National Treasury. Hourcade J.-C., R. Richels, J. Robinson (convening and principal lead authors). 1996. “Estimating the costs of mitigating greenhouse gases”, in Bruce J., H. Lee, E. Haites (eds.) Climate Change 1995: 82

Economic and Social Dimensions of Climate Change (Chapter 8), Cambridge University Press, Cambridge, pp.263–296. IEA (International Energy Agency). 2012. Energy statistics of non-OECD countries – 2012 edition. International Energy Agency/OECD, Paris. IEA (International Energy Agency). 2013. Integrating Carbon Pricing with Existing Energy Policies: Issues for South Africa. IEA Workshop Report 2013. Paris: IEA. IPCC (Intergovernmental Panel on Climate Change). 2001. Climate Change 2001: Working Group III: Mitigation. See particularly section 8.2.2.1: Aggregate Assessment of Revenue-Raising Instruments. Third Assessment Report of the Intergovernmental Panel on Climate Change, Geneva. Available at: http://www.grida.no/publications/other/ipcc_tar/. IPCC (Intergovernmental Panel on Climate Change). 2006. Guidelines for Greenhouse Gas Inventories. Volume 2: Energy. Intergovernmental Panel on Climate Change, Geneva. Available at: http://www.ipcc-nggip.iges.or.jp/public/2006gl/index.html. Janoska, P. 2014. Package of measures to deal with climate change and the role of a carbon tax, Presentation, Department of the National Treasury, 29 May 2014. Jorgenson, D.W., R.J. Goettle, M.S. Ho, P.J. Wilcoxen. 2013. Double Dividend: Environmental Taxes and Fiscal Reform in the United States. MIT Press, Cambridge, USA. K.C., S., M. Potancokova, R. Bauer, A. Goujon, and E. Striesnig. 2013. Summary of Data, Assumptions and Methods for New Wittgenstein Centre for Demography and Global Human Capital (WIC) Population Projections by Age, Sex and Level of Education for 195 Countries to 2100. International Institute for Applied Systems Analysis (IIASA), Laxenburg. Klein, C., O. Simon. 2010. Le modèle MESANGE – nouvelle version ré-estimée en base 2000 [The MESANGE model: new version re-estimated on year 2000 data] Document de travail de la Direction générale du Trésor et de la politique économique No.2010/02. Kolstad C., K. Urama, J. Broome, A. Bruvoll, M. Cariño Olvera, D. Fullerton, C. Gollier, W.M. Hanemann, R. Hassan, F. Jotzo, M.R. Khan, L. Meyer, and L. Mundaca, 2014: Social, Economic and Ethical Concepts and Methods. In: Climate Change 2014: Mitigation of Climate Change. Contribution of Working Group III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Edenhofer, O., R. PichsMadruga, Y. Sokona, E. Farahani, S. Kadner, K. Seyboth, A. Adler, I. Baum, S. Brunner, P. Eickemeier, B. Kriemann, J. Savolainen, S. Schlömer, C. von Stechow, T. Zwickel and J.C. Minx (eds)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. Krusell, P., L.E. Ohanian, J.-V. Rios-Rull, G.L. Violante. 2000. Capital-Skill Complementarity and Inequality: A Macroeconomic Analysis, Econometrica 68(5):1029–1053 Musango J. K., A. C. Brent, A. M. Bassi. 2014. Modelling the transition towards a green economy in South Africa. Technological Forecasting & Social Change 87:257–273. NERSA (National Energy Regulator of South Africa). 2009. Aggregated (average) piped-gas prices per province for the various categories of customers – 2007 Calendar year. Piped-Gas Subcommittee 18 June 2009. Energy regulator of South Africa, Pretoria. 83

Okagawa, A., K. Ban. 2008. Estimation of substitution elasticities for CGE models: Discussion Paper 08-16. Discussion Papers in Economics and Business, Graduate School of Economics and Osaka School of International Public Policy (OSIPP), Osaka University, Osaka. Pauw K. 2007. Economy-wide Modelling: An input into the Long Term Mitigation Scenarios process, LTMS Input Report 4, Energy Research Centre, Cape Town, October. Republic of South Africa. 2013. Millennium Development Goals: Country Report 2013. Pretoria. SAED (South Africa Department of Environmental Affairs). 2011. National Climate Change Response White Paper. Pretoria. SAPIA (South African Petroleum Industry Association). 2013. Industry overview – Old fuel prices. Petroleum products prices in cents per litre 2006-2013. Sandton. Available at: http://www.sapia.co.za/ industry-overview/old_price.html. SARB (South African Reserve Bank). 2007a. Quarterly Bulletin March 2007. Statistical Tables: National Accounts, Pretoria. SARB (South African Reserve Bank). 2012. Integrated Economic Accounts. Presentation for the “Regional Seminar on developing a program for the implementation of the 2008 SNA and supporting statistics in Southern Africa (17-19 October 2012, Pretoria, South Africa)”. Pretoria. SARB (South African Reserve Bank). 2013a. Annual Economic Report 2012. South African Reserve Bank, Pretoria. SARB (South African Reserve Bank). 2013b. Government Finance Statistics of South Africa: 19942012. Supplement to the South African Quarterly Bulletin, March 2013. South African Reserve Bank, Pretoria. StatsSA (Statistics South Africa). 2005. Social Accounting Matrix, Labour accounts for South Africa. Discussion Document – D0403. Statistics South Africa, Pretoria. StatsSA (Statistics South Africa). 2008. Labour Force Survey. Historical Revision March Series 2001-2007. Statistics South Africa, Pretoria. StatsSA (Statistics South Africa). 2010a. Final Supply and Use Table, 2005. Report No: 04-04-01 (2005) (plus tables)., Pretoria. available at: http://www.statssa.gov.za/publications/statsdownload. asp?PPN=Report-04-04-01&SCH=4764. StatsSA (Statistics South Africa). 2010b. Final Social Accounting Matrix, 2005 (Updated Version). Report No: 04-03-02 (2005) (plus tables). Statistics South Africa, Pretoria. available at: http://www. statssa.gov.za/publications/statsdownload.asp?PPN=Report-04-03-02&SCH=4432. StatsSA (Statistics South Africa). 2012. Financial Statistics of consolidated general government 2010/2011. Annex B: Statistical Tables. Statistics South Africa, Pretoria. StatsSA (Statistics South Africa). 2013. Mid-year population estimates, 2013. Statistical release P0302. Statistics South Africa, Pretoria. StatsSA (Statistics South Africa). 2015. Quarterly Labour Force Survey, Q4-2014. Pretoria. 84

Ürge-Vorsatz, D., S. Tirado Herrero, N. K. Dubash, F. Lecocq. 2014. Measuring the Co-Benefits of Climate Change Mitigation, Annual Review of Environment and Resources 39: 549–582. Van der Werf. 2008. Production functions for climate policy modeling: An empirical analysis, Energy Economics, 30:2964–2979 van Heerden, J., R. Gerlagh, J. Blignaut, M. Horridge, S. Hess, F. Mabugu and M. Mabugu. 2006. Searching for Triple Dividends in South Africa: Fighting CO2 pollution and poverty while promoting growth. The Energy Journal 27(2):113–141, Vielle, M., J.-M. Moulinier, A. Bernard. 2009. Evaluation du paquet Energie-Climat à l’aide du modèle GEMINI-E3 [Evaluation of the Energy-Climate Package with the help of Gemini-E3 Model]. Revue de l’Energie (589):153–173. Winkler, H., A. Marquand. 2009. Changing development paths: From an energy-intensive to low-carbon economy in South Africa. Climate and Development 1(1):47–65.

Appendix 1. IMACLIM-SA description From a mathematical point of view IMACLIM-SA is a comparative statics model that boils down to a set of simultaneous equations:

f1 (x1,..., xn, z1,..., zm) = 0 f2 (x1,..., xn, z1,..., zm) = 0 ... fn (x1,..., xn, z1,..., zm) = 0

with: • xi, i ∈ [1, v], a set of variables (as many as equations), • zi, i ∈ [1, p], a set of parameters, • fi, i ∈ [1, v], a set of functions, some of which are non-linear in xi.

The fi constraints are of two quite different natures: one subset of equations describes accounting constraints that are necessarily verified to ensure that the accounting system is properly balanced; the other subset translates various behavioural constraints, written either in a simple linear manner (e.g. households consume a fixed proportion of their income) or in a more complex non-linear way (e.g. the trade-offs of firms and households). It is these behavioural constraints that ultimately reflect, in the flexible architecture of IMACLIM-SA, a certain economic “worldview”. In what follows we list the equations of the reference projection (RP) of IMACLIM-SA. Carbon tax scenarios differ (for a few equations and in their split between parameters and variables) in ways that are easily deducible from the scenario presentation of section 7.3 p.81). For reference purposes, variables and parameters are listed and described. Any variable name indexed with a “0” designates the specific value taken by the variable in the 2005 equilibrium (i.e. the value calibrated on the 2005 hybrid I-O table or the harmonised current and financial accounts of agents described); it thus indicates a parameter of the equation system.

1. Producer and consumer prices The producer price of good i, pY,i, is built following the cost structure of the production of good i, that is, as the sum of intermediate consumptions pCIji aji of goods j = 1 to 10 for the 10 sectors described in section 4.2.1, labour costs for skill sk = 1 (low) to 3 (high), capital costs, a tax on production, tYi (kept at BY value) and a mark-up rate, tNOSi (corresponding to the net operating surplus):

pYi = ∑ j =1 pCIjiα ji + ∑sk =1 p Lsk,i lsk,i + pK ki + τ Yi pYi + τ NOSi pYi .

(1)

The price of imported good i, pM,i, is good-specific, but we do not assume divergence of international prices in this study, and collectively treat international prices as the numéraire of the model. They are consequently assumed constant: pM,i = pM,i0 . (2) The average price of the resource of good i, pi , is the weighted average of the domestic and import prices: p Y + pMi M i . (3) pi = Yi i Yi + M i The domestic and foreign varieties of the energy goods are assumed homogeneous in the sense that they can be exchanged, but as we will see in the discussion of the terms of trade, there is no perfect substitutability between the two goods. The alternative assumption of product differentiation, adopted by many CGEMs through their use of an Armington specification for international trade (Armington, 1969), has the disadvantage of creating “hybrid” good varieties, whose volume unit is independent from that of the foreign and national varieties they hybridise; this complicates maintaining an explicit accounting of the physical energy flows and thus an energy balance. For the sake of simplicity, the non-energy goods are treated similarly to the energy ones. The price of good i consumed in the production of good j, pCIij, is equal to the resource price of good i plus trade and transport margins, agent-specific margins, a domestic excise on oil products (tTIPPCIi),[57] an aggregate of other excise taxes (tAIPi) and a carbon tax (tCI) calculated over the CO2 emission factor (kg CO2 per PJ) of the use of a quantity input i (a fuel, measured in PJ) in the production of a unit of good j ( γ CIij ).

p CIij = p i (1 + τ MCi + τ MTi + τ MSCIij ) + tTIPPCIi + t AIPi + t CI γ CIij .

(4)

The consumer price of good i for household h (pChi), public administrations (pGi) and investment (pIi), and the export price of good i (pXi), are constructed similarly and only differ on whether they are subject to the sales tax ( τ TVAi ; the same rate is applied to all consumptions of one good) and the carbon tax (tCF γ CFi )or not. The latter tax applies to household prices only, as national accounting makes households the only final consumer of energy goods.[58]

[ (

)

pChi = pi 1 + τ MCi + τ MTi + τ MSChi + tTIPPCFi + t AIPi + t CF γ CFi

] (1 + τ ) . (5) TVAi

[57] The TIPP levied on the intermediate and the final fuel consumptions is differentiated to take account of the underlying fuel mixes. [58] Public administrations consume a “public service”, whose energy content appears in the energy consumption of the production in which it is aggregated—and is taxed for its carbon content at this level.

[ ( = [ p (1 + τ + τ p = p (1+ τ

) )+ t

] (1 + τ ). (6) ] (1 + τ ) . (7)

p Gi = p i 1 + τ MCi + τ MTi + τ MSGi + tTIPPCFi + t AIPi

pIi

MCi

MTi

MCi

+ τ MSIi

TIPPCFi

+ t AIPi

TVAi

)

+ τ MTi + τ MSXi + tTIPPCFi + t AIPi . (8)

The general mark-up rate tNOSi and the agent-specific margins tMS*i respond to scenario settings for the change in capital productivity (see section 5.2.2). They are defined as: ⎛

1 ⎜ τ NOS or MS * = τ NOS or MS * ⎜ 1 + ((1 + KPF ) tREF − 1)τ NOS / MS

⎝

⎞ ⎟, (9) ⎟ setting ⎠

with the base year (calibrated) mark-up rate or specific margin (for consumer*), KPF being the annual capital productivity growth factor, tREF the number of years of the projection period (30), and the factor that determines the extent to which capital productivity growth is taken into account in the correction (see section 5.2.2). Trade margins tMCi and transport margins tMTi , identical for all intermediate and final consumptions of good i, are calibrated on the base year (2005) equilibrium and kept constant, with the exception of those on the productions aggregating transport and trade activities—respectively the TRA and LSS goods, which are simply adjusted, in the projected equilibrium, to have the two types of margins sum up to zero:

∑τ

MCLSS

p LSS α LSSj Y j + τ MCLSS p LSS (C LSS + GLSS + I LSS + X LSS )

j =1

∑ ∑τ

i ≠ LSS

MCi

pi α ij Y j +∑τ MCi pi (C i + Gi + I i + X i ) = 0 i ≠ LSS

(10)

and similarly:

∑τ

MTTRA

pTRA αTRAj Y j + τ MTTRA pTRA (CTRA + GTRA + ITRA + X TRA )

j =1

∑ ∑τ

i ≠ TRA

MTi

pi α ij Y j + ∑ τ MTi pi (C i + Gi + I i + X i ) = 0. (11) i ≠ TRA

Labour costs of skill sk in sector i, pLsk,i , are equal to the prevailing net wage wsk,i plus payroll taxes (both employers’ and employees’ social contributions), which are levied following a skill-specific rate tCS,sk common to all sectors, and pension contributions (both public and employees’ private pension contributions) tACS,sk also following a uniform rate by skill level sk calibrated on the hybrid I-O table (harmonised with current and financial accounts data) for 2005:

p Lsk,i = ( 1 + τ CSsk + τACSsk ) wsk,i .

(12)

The average wage of skill sk in production i, wsk,i , varies as the average wage of skill sk across all sectors wsk: w wsk,i = sk wsk,i0 , (13) wsk0 89

which is subject to variations linked to unemployment following the wage curves (see below). The cost of capital pK is understood as the cost of the “machine” capital. It is obtained as the average price of investment goods: n

∑p

pK =

i =1

. (14)

∑I

i =1

CPI the consumer price index is computed following Fisher, i.e. as the geometric mean of a Laspeyres index (variation of the cost of the BY basket of goods from the BY to the future set of relative prices) and a Paasche index (variation of the cost of the future basket of goods from the BY to the future set of relative prices):[59] n

∑p

Ci 0

∑p

Ci 0

CPI =

i =1 n

Ci 0

i =1

∑p

i =1 n

Ci 0

(15)

i =1

2. Households’ income, savings and investment GDIBT,h the gross domestic income of class h before taxation is defined as the addition and the subtraction of the following terms: • A share wLh,sk of the sum of aggregate endogenous net wage income by skill level of labour sk n

for all sectors i, ∑ wsk ,i lsk ,i Yi . This share of net wage is calculated over the share in a “reference i =1

revenue” on the basis of the previous average net wage by skill level and household class and the new number of actives by skill level in each household class. • A share wGOS,h of the total gross operating surplus accruing to households GOSH, which

responds to the share of gross operating surplus accruing to companies, wGOS,S. This share could also be considered as a part of “ownership income” (the gross operating surplus of national accounting) that goes to households, GOSH, which are based on the SAM 2005 (StatsSA, 2010b) and normally in the System of National Accounts they should correspond to the real and imputed rents that accrue to households, but it is not excluded that income from direct ownership of enterprises by households is part of GOS in the South African SAM as well, as it also includes mixed income. The wKh (the distribution of GOSH across household classes) are endogenous.

• Social transfers, in 3 aggregate payments (pensions rPh NPh , unemployment benefits rUh NUh ,

other social transfers rAh NAh ), the calculation of which is similarly based on the product of a per capita income r and a target population N.

[59] Class-specific indexes can similarly be constructed using class-specific prices (differentiated thanks to specific margins) and consumptions and applied for instance to the computation of class-specific real gross disposable income variations.

• An exogenous share wATh of (small) residual transfers ATH . • A “debt service” iH Dh , which corresponds to property income (interests, dividends, real

estate revenues, etc.).

Hence:

sk = 1

i =1

GDI BT,h = ∑ (ω Lh,sk ∑ wsk,i l sk,i Yi ) + ωGOS,H GOS H

+ ρ Ph NPh + ρUh NUh + ρAh Nh + ωATh ATH − ih D h (16)

with ATH a constant share ωATH of AT (see Equation 53):

ATH = ωATH AT

(17)

The gross disposable income of class h GDIh is obtained by subtracting from GDIBT,h the income tax TIRh levied at a constant average rate (Equation 36), and other direct taxes Th that are indexed on the CPI (Equation 37). Rh , the consumption budget of class h, is inferred from disposable income by subtracting savings. The savings rate tSh is exogenous (calibrated to accommodate the values of GDIh and Rh in the BY equilibrium). On top of the GDI before taxes minus tax levies, we also account for the build-up of pension equity, PEQ, which we define as the net difference between pension contributions and pension benefits, though in reality it should be corrected for gains or losses in pension equity value. This is equity that is managed by financial corporations (pension funds, be they private or public) (see below), but which is counted by the SAM 2005 (StatsSA, 2010b) as part of household gross disposable income.

GDIh = GDI BT,h − TIRh − Th + PEQ

(

(18)

)

Rh = 1 − τ Sh GDIh (19)

A further exploration of the data available in the current and financial accounts gives households’ investment GFCFh (Gross Fixed Capital Formation) as distinct from their savings; GFCFh is assumed to follow the simple rule of a fixed ratio to gross disposable income (Equation 20). The difference between savings and investment gives the auto-financing capacity (AFC) of class h, AFCh.

GFCFh GDI h

GFCFh 0 GDI h 0

(20)

AFCh = τSh GDIh − GFCFh (21)

3. Consumption The consumption trade-offs of households are thoroughly described section 4.3.3.

4. Production (institutional sector of firms) 4.1 Gross disposable income and investment decision Similar to that of households, the firms’ disposable income GDIS is defined as the addition and subtraction of: • An endogenous share (explained below) ωGOS,S of Gross Operating Surplus i.e. GOS (see

Equation 27),

• A “debt service” (interests, dividends) iS DS , which is strongly positive in the Base Year

equilibrium (firms are net debtors in 2005), and served at a fixed interest rate iS = iS,0,

• Corporate tax payments TIS , on the basis of corporate tax rate tIS and calculated over their

GDI before taxation,

• And an exogenous share ωATS of other transfers AT, which are assumed to be a constant share

of GDP (Equation 53).

GDI S = ωGOS,S GOS − iS DS − TIS − ρ Ph N Ph − PEQ + RP + ω ATS AT . (22)

Additionally PEQ and RP refer to the build-up of pension fund value by households. South Africa has a mixed pension system. There is a state pension which is completely independent from social contributions to reduce poverty at old age. There also are employees’ contributions to employee pension funds, and people are free to make use of private pension funds. As pension funds fall under financial corporations, thus firms, and pension equity is registered as a debt of firms to households in the SAM 2005 (StatsSA, 2010b), we decided to make pension contributions add to company income, and to subtract pension benefits paid to pensioners, and the change in pension equity (to lift it from taxation). Thus we define pension contributions RP, next to pension benefits rPh NPh and pension equity build-up PEQ m

R P = ∑ ρPh N Ph (23) h =1

and indeed constrain the rates of private social contributions tACS to balance the private social insurance system RP = PEQ (24) The ratio of the gross fix capital formation of firms GFCFS to their disposable income GDIS is assumed constant; similar to households and in accordance with national accounting their self-financing capacity AFCS then arises from the difference between GDIS and GFCFS. The net debt of firms DS is then calculated from their AFCS following the same specification as that applied to households.

GFCFS GFCFS 0 . (25) = GDI S GDI S 0

AFC S = GDI S − GFCFS . (26)

4.2. Production trade-offs The trade-offs between production factors follow the nested-CES structure described section 4.3.1, except for the electricity (ELC) sector, whose coefficients we derive from the South African TIMES model of the Energy Research Centre of the University of Cape Town, as described section 4.1.2.

4.3. Gross operating surplus Capital consumption, almost constant rates of operating margin pi and specific margins MS (see Equation 9) determine the gross operating surplus (GOS):

GOS =

∑( p

)

ki Yi + π i pYi Yi + MS

(27)

i =1

This GOS, which corresponds to capital income, is split between agents following endogenous shares, which serve to solve the income balance to in fact solve the capital market. The reason is that with fixed interest rates and fixed positions of investment over gross disposable income a variable is needed to balance revenue and investment. By construction, the specific margins on the different sales MS sum to zero in the Base Year equilibrium (this is a constraint of the hybridisation process). However, they do not sum in the future equilibrium, in which their constant rates are applied to varying prices. Their expression is then:

⎛ ⎞ M S = ∑ ⎜ ∑τ MSCIij pi α ij Yj + ∑τ MSCh i pi C hi + τ MSGi pi Gi + τ MSXi pi X i ⎟ (28) ⎜ ⎟ i ⎝ j h ⎠

5. Public administrations 5.1. Tax, social security contributions and fiscal policy Tax and social security contributions form the larger share of government resources: ∀ X ∈ [CI, CF] tTIPPXi = CPI tTIPPXi 0 , (29) tAIPi = CPI tAIPi 0 . (30)

The various tax revenues are defined by applying these rates to their respective bases: n

TY = ∑τ Yi pYi Yi , (31)

i =1

TTIPP = ∑ ∑ tTIPPCI ji α ji Yi + ∑ tTIPPCFi (Ci + Gi + I i ) , (32) i =1 j =1 n n

i =1 n

TAIP= ∑ ∑ t AIPjα ji Yi + ∑ t AIPi (Ci + Gi + I i ), (33) i =1 j =1

i =1

TTVA = ∑ i =1

τ TVAi

1 + τ TVAi

( pCi Ci + pGi Gi + pIi I i ) , (34)

TIS = τ IS GOSS , (35)

TIRh = τ IRh GDI BTh . (36)

The rates tIS and tIR together serve as a variable to adjust government budget at -2% of GDP in the RP. They remain at RP levels in carbon tax scenarios except when targeted as recycling options. Similar to excise taxes and for lack of more detail on its composition, the sum of households’ direct taxes other than the income tax TIR is assumed to be constant in real terms: Th = CPI Th 0 . (37)

The sum of social contributions TCS follows the same logic as other tax revenues again, and pension contributions TACS are calculated in a similar way, with their skill specific rates:

sk = 1 3

i =1 n

TCS = ∑ (τ CS,sk ∑ wsk,i lsk,i Yi ) . (38) T ACS = ∑ (τ ACS,sk ∑ wsk,i l sk,i Yi ) . (39) sk = 1

i =1

So does the carbon tax on intermediate consumptions (tCI) and on final consumptions (tCF)—which in fact is not differentiated in any of our scenarios. It is calculated over the emission intensity, U, of the use of input i in sector j or respectively in final consumption by households:

TCARB = ∑∑ tCI γ CIji α ji Yi + ∑ tCF γ CFi Ci . (40) i =1 j =1

i =1

Contrary to other dynamic IMACLIM-P frameworks, which apply budget neutrality in terms of constant DG over GDP, IMACLIM-SA applies a “constant budget position” rule as a constraint on public deficit (AFCG), enforced as a targeted ratio (pct_deficit) dependent on the reference setting, but in the present study maintained at -2% of GDP. The variables that adjust to meet this constraint are tIS and tIR, which are scaled up or down by a common factor. In the case of carbon tax revenue recycling scenarios these variables are fixed and other variables change to maintain budget deficit. (41) AFC G = pct_deficit GDP Lastly, T is the sum of taxes and social contributions:

h =1

T = TCS + TY + TTIPP + TAIP + TTVA + TIS + ∑ TIRh + ∑ Th + TCARB (42)

5.2. Gross disposable income, public spending, investment and transfers Similar to households and firms, the gross disposable income of public administrations GDIG is the sum of taxes and social contributions, of an endogenous share wGOS,G (responding to wGOS,S) of GOS, and of an exogenous share wATG of “other transfers” AT , from which are subtracted public expenditures pG G , a set of social transfers RU and RA , and a debt service iG DG : n

GDI G = T + ωGOS,G GOS + ω ATG AT − ∑ pGi Gi − RU − R A − iG DG

(43)

i =1

Public expenditures pG G are assumed to keep pace with national income and are therefore constrained as a constant share of GDP: n

∑ pGi Gi

i =1

GDP

∑p

Gi 0

i =1

Gi 0

GDP0

, (44)

Social transfers RU and RA are the sum across household classes of the transfers defined as components of their before-tax disposable income (Equation 16): m

RU = ∑ ρUh N Uh

(45)

h =1 m

RA = ∑ ρAh Nh , (46)

h =1

Lastly, the interest rate iG of public debt is fixed at Base Year rate as are iH and iS. Public investment GFCFG, the same as public expenditures pG G, is supposed to mobilise a constant share of GDP. Subtracting it from GDIG produces AFCG, which determines the variation of the public debt. For the calculation of the accumulation of debt we assume that gross auto-financing capacity GAFCG, i.e. AFCG net of interest payments, evolves linearly over time between our base year and our projection year, n years into the future (n=30): GFCFG GFCFG 0 (47) = GDP GDP0 AFCG = GDI G − GFCFG (48)

DG = (1 + iG ) D0

n −1

−

k n −1 ∑k =0 (1 + iG ) GAFCG, 0 n −1

− ∑k =0 (1 + iG ) (n − 1)dGAFCG + ∑k =0 k (1 + iG ) dGAFCG (49)

Note that we assume debts to be held entirely in South African Rand. This is an approximation of the base year situation in which only 15% of total South African debts are financed by the Rest of the World, and the rest by household savings. The value of base year debt is then inflated by the consumer price index. For companies and households we apply a similar calculation of the accumulation of debt (or gross savings).

6. “Rest of the World” 6.1. Trade balance Competition on international markets is settled through relative prices. The ratio of imports to domestic production on the one hand, and the “absolute” exported quantities on the other hand, are elastic to the terms of trade, according to constant, product-specific elasticities: σ

⎛ pMi 0 pYi ⎞ Mpi ⎜⎜ ⎟⎟ (50) ⎝ pYi 0 pMi ⎠ σ Xpi X i ⎛ pMi 0 p Xi ⎞ ⎟⎟ (1 + δ Xi ) (51) = ⎜⎜ X i 0 ⎝ p Xi 0 pMi ⎠ The different treatment of imports and exports merely reflects the assumption that, notwithstanding the evolution of the terms of trade, import volumes rise in proportion to domestic economic activity (domestic production), while exports are impacted by global growth. The latter fact is captured by assuming an extra, exogenous dXi increase of volumes exported. In total, as far as exports are concerned, South Africa is depicted as supplying a terms-of-trade elastic share of a dXi expanded export market. Mi Mi0 = Yi Yi 0

6.2. Capital flows and self-financing capacity Capital flows from and to the “Rest of the World” (ROW) are not assigned a specific behaviour, but are simply determined as the balance of capital flows of the three national institutional sectors (households, firms, public administrations) to ensure the balance of trade accounting. This assumption determines the self-financing capacity of the ROW, which in turn determines the evolution of DROW , its net financial debt. n

AFC ROW = ∑ p Mi M i − ∑ pXi X i + i =1

i =1

∑i

K = H,S,G

DK −

∑A

. (52)

K = H,S,G

By construction the auto-financing capacities (AFC) of the 4 agents clear (sum to zero), and accordingly the net positions, which are systematically built on the AFCs, strictly compensate each other in the projected as in the present equilibrium. The hypothesis of a systematic “compensation” by the ROW of the property incomes of national agents without any reference to its debt DROW may seem crude, but in fine only replicates the method of construction of the current and financial accounts. Indeed, in the 2005 calibration equilibrium the effective interest rate of the ROW (ratio of net debt to its property income), which ultimately results from a myriad of debit and credit positions and from the corresponding capital flows, is negative—unworkable for modelling purposes. Finally, as previously mentioned other transfers AT are defined as a fixed share of GDP:[60] A T = T0 GDP GDP0

(53)

[60] The sum across agents of other transfers being nil by definition, AT is in fact calibrated on the sum of the net transfers that are strictly positive. As a consequence the shares wATH, wATS, wATG and wATROW , summing to 0 by construction, are ratios properly speaking.

7. Market balances 7.1. Goods markets Goods market clearing is a simple accounting balance between resources (production and imports) and uses (intermediate consumption, households and public administrations’ consumption, investment, exports). Thanks to the process of hybridisation, this equation is written in Mtoe for energy goods and consistent with the 2005 energy balance of the IEA at the base year (notwithstanding that the G and I of energy goods are nil by definition). n

Yi + M i = ∑ α ij Yj + Ci + Gi + I i + X i (54) j =1

Of course the aggregate consumption of households Ci sums up the consumptions of all classes Cih.

7.2. Investment and capital flows The demand of investment is constrained by the assumption that the ratio of each of its real components Ii to total fixed capital consumption (the sum of ki Yi) is constant. In other words, the capital immobilised in all productions is supposed homogeneous, and all its components vary as the total consumption of fixed capital.

∑ GFCF = ∑ p K

∑k

Ii (55)

i =1

K = H,S,G

= Yj

j =1

Ii 0

∑k

Yj 0

(56)

j =1

As mentioned, the closure of the market for investment is indirect through adjustment of the distribution of GOS by wGOS of the three domestic agents firms, government and households. It therefore serves as a closing variable for model resolution, though constrained by the different other balances in the model.

7.3. Employment Labour endowment , proportional to total active population, is split into three skill levels, L2, L3. For each skill sk the level of unemployment usk is endogenous, and real sector-average wage wsk /CPI corrected for share wLPsk of labour productivity gains Φsk accruing to workers is elastic to unemployment (wage curve):

(1 − u sk ) Lsk = w sk

∑

l Y . (57) i =1 sk,i i

⎛u ⎞ = Wsk 0 ⎜⎜ sk 0 ⎟⎟ (1 − ω LPsk )CPI + ω LPsk φ sk CPI ⎝ u sk 0 ⎠

σwu

(58)

8. Model parameters and variables Calibration consists in providing a set of values to all variables and then determining the values that should be given to the parameters so that the set of equations defining the model holds. The exercise is therefore to determine what values the parameters must take in order for the values drawn from national accounts to be linked by the set of equations. However, all parameters do not receive their values from the calibration: the carbon tax, for instance, is a purely exogenous parameter; other parameters have their values set according to some econometric estimation on data beyond the national accounts. As a result of these distinctions, the notations below are presented in three categories, (i) the parameters of the model that are calibrated on our hybrid national accounts, (ii) the exogenous parameters and (iii) the variables of the model properly speaking. Within each of these categories the notations are listed in alphabetical order (the Greek letters are classified according to their English name rather than according to their equivalent in the Latin alphabet). Note that the split is that which prevails in the reference projection. It differs in the scenarios where taxes, expenses (G) or transfers (lump-sum transfers) vary. Notably, the corporate and income tax rates tIS and tIR are fixed in most scenarios, not variable as in the RP where they are adjusted to target a 2% public deficit.

8.1. Parameters calibrated on BY hybrid I-O table and other statistical data We do not report the (too-numerous) demography and labour supply parameters, which are presented elsewhere in this report. γ CIji

CO2 emissions per unit of good i consumed in the production of good j

γ CFi

CO2 emissions per unit of good i consumed by households

ωATh

Share of the other transfers accruing to households received by household class h

ωATS

Share of other transfers accruing to firms

ω ATG

Share of other transfers accruing to public administrations

ωKh

Share of the capital income of households accruing to household class h

ωKS

Share of capital income accruing to firms

ωKG

Share of capital income accruing to public administrations

τ CS,sk

Social contribution rate applicable to net wages by skill level sk

tAIPi

Excise taxes other than the TIPP per unit of consumption of good i

tTIPPCIi

Fuel tax per PJ of household consumption

tTIPPCFi

Fuel tax per PJ of intermediate consumption

τ MSCIij

Specific mark-up rate on intermediate consumption of good i by sector j

τMSChi

Specific mark-up rate on household h’s consumption of good i

τ MSGi

Specific mark-up rate on public energy consumptions

τ MSXi

Specific mark-up rate on energy exports

τ MCCOM,i≠LSS

Commercial mark-up on the commercial good or on the aggregate encompassing it, excluding that of the LSS sector which receives all trade margins

τ MCCOM,i≠TRA

Transport mark-up on the transport good or on the aggregate encompassing it, excluding that of the TRA sector which receives all transport margins

τTVAi

VAT rate applying to the final consumption of good i[61]

Effective interest rate on the net debt of households

Effective interest rate on the net debt of firms

Effective interest rate on the net debt of public administrations

8.2. Parameter from other sources or exogenous bAh

Share of household class h 2005 consumption of good A that is a basic need of good A in the projected economy

bji

Technical asymptote of the technical coefficient aji

bKi

Technical asymptote of the capital intensity of good

bLi

Technical asymptote of the labour intensity of good i

Substitution elasticity of the variable shares of production factors

Substitution elasticity of the variable shares of products or aggregates forming aggregate A

sMpi

Elasticity to terms-of-trade of the share of imports in good i resource

sXpi

Elasticity to terms-of-trade of good i exports

tCI

Carbon tax on the carbon emissions of intermediate consumptions

tCF

Carbon tax on the carbon emissions of households’ consumptions

tPROJ

Number of years projected

8.3. Model variables aij

IO coefficient, good j’s intensity in good i

rAh

Per capita not-elsewhere-included transfers benefiting to household class h

wLsk,h

Share of labour income from skill sk accruing to household class h

tMCCOM

Commercial mark-up on the commercial good or on the aggregate encompassing it

[61] In the IO, investment is conventionally valued at prices that include the sales tax. Treating the VAT as a sales tax cancels some distributive effects between sectors, all the more negligible as the good aggregation is high. Our experience is that in most policy runs it is virtually without discernible effect on macroeconomic results or those concerning the distribution of income between households.

tMCTRANS

Transport mark-up on the transport good or on the aggregate encompassing it

rPh

Per capita pensions to the retirees of class h

rUh

Per capita unemployment benefits to the unemployed of household class h

AFCh

Self-financing capacity of household class h

AFCROW

Self-financing capacity of the rest of the world

AFCS

Self-financing capacity of firms

Other transfers

ATG

Other transfers to public administrations

ATH

Other transfers to households

ATS

Other transfers to firms

Total household consumption of good i (all classes)

Cih

Final consumption of good i by household class h

COMPh

Total composite consumption (beyond basic needs) of household class h

CPI

Consumer Price Index

Net public debt

DHh

Net debt of household class h

DROW

Net debt of the rest of the world

Net debt of firms

GDIBTh

Before-tax gross disposable income of household class h

GDIG

Gross disposable income of public administrations

GDIh

Gross disposable income of household class h

GDIS

Gross disposable income of firms

GFCFG

Gross fixed capital formation of public administrations

GFCFh

Gross fixed capital formation of household class h

GFCFS

Gross fixed capital formation of firms

Final public consumption of good i

GOSG

Gross operating surplus accruing to public administrations

GOSH

Gross operating surplus accruing to households

GOSS

Gross operating surplus accruing to firms

Final consumption of good i for investment

Capital intensity of production i

li,sk

Skill sk labour intensity of production i

Imports of good i

Sum across goods and uses of specific margins

MSCC

Sum of specific margins on household consumption

wGOS,G

Share of Gross Operating Surplus accruing to government

100

wGOS,H

Share of Gross Operating Surplus accruing to households

wGOS,S

Share of total Gross Operating Surplus accruing to firms

wKH

Share of capital income accruing to households (all classes)

pCih

Price of good i for household class h (index i extends to aggregates specific to household consumption

pCIij

Price of good i for the production of good j

PEQ

Pension equity value

pGi

Public price of good i

Average price of good i resource

pIi

Investment price of good i

Cost of capital input (weighted sum of investment prices)

pLi

Cost of labour input in the production of good i

pXi

Export price of good i

pYi

Production price of good i

RAS

Social transfers to households not elsewhere included

Consumed income of household class h

Sum of retirement pensions

Sum of unemployment benefits

Total taxes and social contributions

TAIP

Fiscal income of excise taxes other than the â&#x20AC;&#x153;TIPPâ&#x20AC;?

tACS,sk

Rate of pension contributions applicable to net wages by skill level sk

tNOS,i

Mark-up rate of production i

TCARB

Carbon tax revenues

TCS

Sum of social contributions of employers and employees

Other direct taxes paid by household class h

TIRh

Total income tax of household class h

TIS

Total corporate tax income

TTIPP

Fiscal income from the fuel levy and other fuel-related taxes

TTVA

Total sales tax proceeds

Average overall unemployment rate

uh,sk

Unemployment rate of skill level sk by household class h

usk

Unemployment rate of skill level sk

wsk

Average net wage across sectors for skill level sk

wsk,i

Average net wage in the production of good i for skill level sk

Good i exports

Good i output

101

Appendix 2. Hybrid I-O tables and harmonised integrated economic accounts 1. Building hybrid input-output tables Once the input-output tables that describe the economic circuit of energy flows in quantity, value and price have been built, it remains to integrate them into the national accounts input-output tables without changing the variables important for the empirical analysis. This is the hybridisation step as such (Figure 7) that can be analysed in two stages: a work on the rows of the matrix currency (sub-step 1—Adjustment of uses) to insert the monetary matrix derived from step 2, and informing the energy bills paid; and a work on the columns (2—adjustment of resources) which provides a description of the contents of these bills: the cost structure of one litre of fuel purchased, a kWh, etc. These columns describe the fixed and variable costs of industries that supply, process and distribute energy to consumers. In Appendix 2 the end result of these steps, the final hybrid I-O matrix is shown.

Figure 7. Principles of alignment of material balances and monetary flows Energy

E prod

Other

1. Adjustment of Uses Values from energy sectors not linked to material flows are allocated to other (composite) sectors.

Vij

Energy bills (Vij)

2. Adjustment of Resources (a) Disaggregation of cost structures and margins (b) Values not linked to material flows are allocated to other (composite) sectors

103

1.1. Adjustment of uses Starting from input-output in national accounts, we replace the values of energy branches by the values of reconstructed energy bills. Differences between the old and the new consumption of energy products are added to the uses/consumption of non-energy goods of suitably affiliated sectors. For imports, a similar change is made. These operations do not affect the total value of uses, but change the different sub-totals. Therefore, without adjustment of the components of Valueadded, the margins and taxes on products, the uses and the resources (supply) of goods and services would no longer be balanced.

1.2. Adjustment of resources Balances between uses and resources are restored by adjusting the cost structure of the productive sectors (columns of the I-O table). Values of imports and intermediate consumption are fixed, because they follow from the energy bills. Other cost components—value-added, margins, and taxes on products—are not fixed and are adjusted to restore the equality of resources with uses. We apply the following assumptions: • Value-added (excluding net operating surplus and specific margins): these are re-allocated

on the basis of the new total intermediate consumption of the specific energy sectors. For example, if IC by an energy sector (column) is now 15% higher, we multiply the wages, contributions, CFF and taxes on production by a factor 1.15. The total amounts over all sectors of wages, contributions, CFF and taxes on production are kept constant; therefore the change in energy sectors has to be mirrored by an opposite change in the composite sector. The net operating surplus is partly adjusted already, but will serve as the last item to be adjusted through recalculation of specific inter-sectoral margins.

• Taxes on products: fuel levies remain unchanged and are all allocated to (products from)

the refineries sector; VAT and other taxes (minus subsidies) on products are re-allocated in the same way as the VA items (excluding net operating surplus).

• Trade and transport margins: are also adjusted according to the ratio of change in

Intermediate Consumption by the energy sectors. Trade and transport margins reflect the consumption of trade and transport services, which are part of the composite sector. Therefore, margins paid by the energy sectors, are earned by the composite, thus a negative trade and transport margin and a total for all sectors equalling zero.

• Net operating surplus serves as the closing piece of the restructuring: to do so, we

recalculate the inter-sectoral specific margins, relative to the composite sector (and the net operating surplus already counted). In this calculation, the consumption and taxes paid by the different end-use sectors are taken into account.[62] This we do for the energy sectors and the net operating surplus of the composite sector serves as the last number to be adjusted. After this last step, all accounting identities of the hybrid description are satisfied.

[62] An additional assumption applied to simplify this calculation was that VAT was only applied to Final

Consumption by households and government, investment and exports.

104

105

2. Final hybrid I-O table for Base Year 2005 Table 38. Final hybrid I-O table for Base Year 2005 Unit: million ZAR 2005 Grey shaded areas origin from “energy bills” IC

Intermediate Consumption (IC) REF ELC EIN

954

2 750

9 082

37 168

Gas & gas products

3 312

4 137

Refineries

116

5 154

Electricity

513

1 112

14 596

Energy intensive ind. & other mining

1 471

2 951

10 022

597

100 723

Manufacturing

883

2 465

2 275

6 823

32 190

Low-skilled sectors

968

126

2 944

210

20 019

High-skilled sectors

2 230

3 719

4 284

34 671

Transport services

9 675

1 668

1 261

27 853

17 322

4 298

64 969

22 372

240 227

5 867

206

3 869

8 976

51 538

105

161

904

Pension contributions

1 224

822

1 862

9 386

Cons. of Fixed Cap

1 573

140

1 775

14 790

16 459

Tax on production

373

-312

1 672

7 833

837

4 176

1 172

73 663 393 849

Social contributions

NOS (excl. T on prod.) Subtotal of Domestic Production Imports Commercial Transport Specific Margins on sales to:

GAS

Oil

Net salaries

Margins

OIL

Coal mining

Subtotals VA

COA

34 298

5 528

75 368

49 353

749

37 168

2 595

9 304

57 223

2 506

35 831

58 507

367

5 250

8 573

Coal mining

-149

Oil

-742

Refineries

-2 139

-22

-487

Electricity

-7 066

-4

Energy intensive industries & other mining

-687

-28

339

-5 292

Manufacturing

Gas & gas products

-740

-3

-16

-847

Low-skilled sectors

-19

-455

590

High-skilled sectors

-208

-350

952

Transport services

-1 213

-80

2 592

4 668

Specific margins on sales to Gov.

Specific margins on capital goods

11 681

-894

535

TVA

2 196

6 333

Fuel levy on IC

24 221

Fuel levy on FC

14 687

611

-18 252

-230

17 220

38 530

37 168

8 126

146 409

51 411

541 705

Specific margins on final consumption

Specific margins on exports Taxes and subsidies on products

Other taxes and subsidies TOTAL RESOURCES

Source: authors’ own calculations and assumptions and data based on SAM 2005, SU 2005, energy balances, energy use and price data from various sources.

106

MAN

LSS

HSS

Subtotals

TRA

Final Consumption (FC) Households Gov Investment

Exports

TOTAL USES

952

267

14 911

102

23 517

38 530

37 168

449

158

8 126

401

8 571

1 944

64 551

80 738

48 955

16 716

146 409

3 439

3 841

5 885

748

30 133

17 885

3 392

51 411

98 473

58 879

35 556

4 980

313 652

43 366

8 200

176 486

541 705

183 952

93 482

91 227

12 063

424 574

384 016

162 545

117 523

1 088 657

87 546

20 827

60 021

18 112

210 773

126 587

99 143

39 270

475 774

55 927

133 937

327 922

31 269

594 052

322 201

305 732

12 241

45 213

1 279 439

20 357

30 705

33 263

8 023

132 883

47 662

32 434

212 980

451 496

350 425

556 155

139 747

1 847 011

990 776

305 732

282 129

454 551

3 880 199

63 052

113 722

304 485

27 966

579 681

1 105

2 005

5 355

495

10 204

11 407

21 357

56 190

5 360

107 651

18 723

22 401

100 031

9 900

185 793

991

5 245

18 877

1 633

28 498

61 989

125 787

187 599

26 425

489 481

608 763

640 942

1 228 692

211 525

3 248 318

252 798

43 478

21 840

36 627

461 865

135 934

-233 393

608

19 919

-34 200

-149

-742

-2 649

-7 070

-5 668

-1 606

129

435

-1 183

7 181

11 322

67 121

27 721

78 618

5 853

187 843

24 221

14 687

4 121

-2 975

-50 407

-6 826

-56 736

1 088 657

475 774

1 279 439

212 980

3 880 199

107

108

Gas & gas products

Refineries

Electricity

Energy intensive ind & other mining

Manufacturing

Low-skilled sectors

High-skilled sectors

Transport services

Reference units

665

129

1 108

1 310

REF

4 374

ELC

221

908

251

656

259

146

687

184

154

128

113

819

343

1 251

681

752

581

1 359

1 027

411

1 108

7 646

651

389

444

280

443

620

187

169

147

103

318

547

279

7 192

472

2 684

1 360

1 683

1 192

1 721

1 750

411

1 108

14 837

407

2 661

1 281

1 442

1 125

1 717

1 660

308

14 741

241

102

1 108

472

2 684

1 360

1 683

1 192

1 721

1 750

411

1 108

14 837

Final Consumption (FC) SubTOTAL Domestic TOTAL HouseInvest- Exports USES Production Imports RESOURCES HSS TRA totals Gov holds ment

Source: Authors’ own calculations and assumptions and data based on or from (for energy) energy balances and various other sources (for non-energy: indicative theoretical units).

128

294

340

179

144

520

EIN MAN LSS

Note: energy volumes figure in PJ thanks to the hybridising process.

Oil

Coal mining

COA OIL GAS

Intermediate Consumption (IC)

Table 39. Quantities of uses and resources at base year 2005

3. Quantities of uses and resources at Base Year 2005

4. Harmonised current and financial accounts for economic agents 4.1. 2005 current and financial accounts of agents Table 40. Base Year agent specific income distribution (top) and expenditures (bottom) Underlined means: part of primary distribution of income (part of VA)

Firms

Government

Households

Total domestic

GOS and imputed rents

500 576

30 471

172 724

703 771

Returns on (financial) capital

-159 308

-46 498

177 097

-28 709

Social benefits and pensions

-50 620

-61 330

111 950

697 536

Social and pension contributions

107 651

10 204

-117 855

Income and property taxes

-98 779

223 292

-124 513

Other taxes (on production and products)

-28 498

198 513

Gross salaries

Other transfers

Rest of the World 28 709

170 015

16 195

-54 735

17 904

-20 635

Increase pension fund value HHs

-57 031

57 031

Gross disposable income (GDIS,G,H)

20 635

230 187

299 917

991 875

1 521 978

Final consumption

305 732

990 776

1 296 508

Savings

1 099

240 925

25 702

15 502

282 129

-10 738

-31 517

-14 403

-56 658

56 658

-1 975 370

-938 238

2 483 608

-430 000

430 000

Gross Fixed Capital Formation Trade balance Auto-financing capacity (AFCS,G,H) STOCK: Estimated net financial assets (negative means a debt)

49 344

7 314

4.2. Building of current and financial accounts To complete the integrated hybrid accounting of IMACLIM-SA, the building of a second macroeconomic table is needed: the current, financial and capital accounts (TEE) that describe the balance of sources and uses of income for each institutional agent (households, firms, government and Rest of the World). Again, we start from the data of the SAM 2005, which gives a more consolidated picture of the distribution and uses of income among economic agents. In this section we present the data and assumptions used to refine the description and to obtain the final format needed for the calibration of IMACLIM-SA. Due to the previously discussed lack of detail that is required for IMACLIM in South Africa’s SAM 2005, we need to apply some assumptions to obtain a TEE that is balanced with the I-O table. Table 42 below shows the items of the TEE are required for IMACLIM in darkish blue, and whether it is available in the SAM or other documents. Items (in light blue) are available from the SAM (either the SU tables or from the TEE (Stats SA, 2010a; 2010b)) and can be used in calculating the desired 109

numbers of the TEE. The last column shows the items where there is a link between the TEE and the I-O for IMACLIM.

Table 41. South Africa current and financial accounts 2005 Million Rands

Firms

Gross wages (as in SAM) Social contributions paid

Government

Net op. surplus (SAM 2005) Consumption of fixed capital

Total Domestic

699 018

-117 855

581 163

312 523

30 252

171 484

514 259

187 790

Taxes (-subsidies) on production3

-28 498

Gross Operating Surplus

471 815

30 252

171 484

673 551

Total VA accruing to sectors

500 313

30 252

870 502

1 401 067

Trade balance

Rest of the World

Net wages

Households

7 357

-159 308

-46 498

177 097

-28 709

28 709

107 651

10 204

170 015

Fiscal revenue, besides ICRP

107 651

208 717

316 368

Social and pension benefits

-50 620

-61 330

111 950

Taxes on income and property

-98 779

223 292

-124 513

18 456

-54 516

17 664

-18 396

18 396

289 215

299 917

934 845

1 523 977

305 732

990 578

1 296 310

240 925

25 702

15 502

282 129

48 290

-31 517

-71 235

-54 462

Capital / property income Social contributions received1 Taxes (-subsidies) on production Taxes less subsidies on products

Other transfers Gross available income Final Consumption5 Gross Fixed Capital Formation Net surplus (+) or deficit (-)

28 498

54 462

Social contributions are not part of gross disposable income, but do explain the difference between gross wages in the SAM 2005 and the net wages as used for the TES and the TEE. In the TEE social contributions are part of government’s and (for pensions) companies’ “fiscal revenue” (FISC). 1

CCF should be allocated to the other sectors as well, but the additional assumptions for the distribution of EBE that should go along seemed needlessly complicating and values have therefore been set to zero for these sectors. For CFF, the SAM 2005 also contained a mismatch with the TES (likely a rounding error) and was therefore corrected downward 0.212 compared to the SAM. There is no need for a correction in “other transfers” here as it showed that there were other imbalances in the SAM that were cancelled by this adjustment. 2

Taxes less subsidies on production is not part of the Gross Operating Surplus of government, but does lower the GOS of companies compared to the net operating surplus + CFF as presented in the SAM 2005, which considers these transfers secondary. It is part of “fiscal revenue”. 3

The trade balance 2005 in the SAM is originally -7 390 million Rand (more imports than exports), which is 32.87 million Rand higher than in the Supply and Use tables. We have corrected the SAM number so the TEE matches the TES. There is no need for a correction in “other transfers” here as it showed that there were other imbalances in the SAM that were cancelled by this adjustment. 4

Final consumption in the SAM 2005 is 196.51 million Rand higher than in the Supply and Use tables (the TES) with, which is partially caused by the error in the trade balance and partially it has been solved in the SAM by introducing a “residual item” of 164 million Rand in primary income. By correcting the SAM number for the difference with the Supply and Use tables, the residual item can be dropped. 5

110

Table 42. Data need for current and financial accounts for IMACLIM, availability from SAM 2005, and correspondence with I-O tables Item no.

Item in Current and Financial Accounts

Link to I-O for IMACLIM-SA

Source for data

Gross wages

Available in SAM

Social (and pension) contributions paid/received

Estimated before on the basis of assumptions

Equals social contributions paid in VA

Net wages

Can be calculated by subtracting social contributions from gross wages

Equals total wages in VA

Net operating surplus

Equals net operating surplus in ValueAdded

Consumption of fixed capital

Total for all sectors available in SAM

Part of Value-Added

Taxes (-subsidies) on production

Available from the SU tables

Part of VA

Gross Operating Surplus

Can be calculated by combining net operating surplus, CFF and taxes on production

(indirectly through net operating surplus and CFF)

Total VA accruing to sectors

Equals net wages, plus social (and pension) contributions, plus gross operating surplus; and equals the value in the I-O

Trade balance

Available in SAM from SU tables and current account of goods and services

Capital / property income

Available in SAM from primary income from property in current accounts

Taxes less subsidies on products

Available in SAM from SU tables and current account of goods and services

Equals total of product taxes

Fiscal revenue, besides ICRP

Can be calculated from total of social (and pension) contributions, taxes on production and taxes on products.

Social benefits

Hidden in “secondary distribution of income transfers” in current, financial and capital accounts

Taxes on income and property

Hidden in “secondary distribution of income transfers” in current, financial and capital accounts

Other transfers

Hidden in “secondary distribution of income transfers” in current, financial and capital accounts

Gross available income

= 3 + 7 + 9 + 10 + 12 + 13 + 14 + 15

Final Consumption

Available in SAM from SU tables and current account of goods and services

Equals consumption by Households and Government in I-O

Gross Fixed Capital Formation

Available by product but not by sector in SAM from SU tables and current account of goods and services

Equals Investment

Net surplus (+) or deficit (-)

= 16 - 17 - 18

Equals exports minus imports

Following these observations of data availability we estimate the missing detail on the basis of other data sources and additional assumptions. In particular, two important steps of data processing have been done: 1) the calculation of social benefits and other transfers, 2) the calculation of fiscal 111

revenues, and 3) the calculation of the amounts of money used by each economic agent to finance the fixed capital formation.

4.3. Financial assets IMACLIM, as will be presented in the next section, makes use of interest rates in its description of the economy as well. To calculate them we need data on property income, and on financial assets. To estimate the assets of each sector we make use of several sources and assumptions. We list them by institutional sector below. Data on foreign assets were available in detail for 2012 instead of 2005, which is about 2.5 times higher. We therefore overestimate net foreign assets, net government and net corporate debt, but this does not have an important impact on the results. Therefore, these figures will be improved in the next steps of the project. • Households: SARB has published a working paper in which appears an overview of the

balance sheets of assets and liabilities of households (Aron et al., 2007b). Assets and liabilities are presented as a share in personal disposable income, which is also presented in the overview. Households’ total financial assets minus liabilities amount to R 2 484 billion, mostly in pension funds and shares.

• Companies: for companies, only partial information about financial assets and liabilities

was found. From the SARB we know their liabilities to households through pension funds totalling R 1 574 billion (Aron et al., 2007). In the SARB Annual Report 2012 (SARB, 2013a) we find information on the outstanding balance of banks to households and corporations for 2011 and 2012 and of annual growth in total lending from 2006 to 2012. The average share of corporations in banks’ outstanding balance in 2011 and 2012 is used to estimate their part in the total outstanding balance at the beginning of 2005. The latter is derived on the basis of the numbers on annual growth in total lending. We thus estimate the outstanding balance of banks to corporations to be R 401 billion. We assume no (other) financial assets for companies and thus arrive at a total net debt of companies of R 1 975 billion.

• Rest of the World: from the same SARB annual report (SARB, 2013a) we get a number for

net foreign assets (visual estimation of figure data). For 2012 this was R 430 billion, and for 2005 it was R 170 billion.

• Government: this account serves as a balancing piece for the other accounts of financial

assets and liabilities. Total assets should equal total liabilities, therefore the net government debt is calculated as: Household assets + RoW assets – Corporate debt = R 938 billion. In 2004 this amounts to approximately R 678 billion, which is still higher than the SARB estimate of total gross and net debt of government, which approximately amounts to respectively R 550 to 500 billion for 2005 (SARB, 2013b).

112

5. Demography and labour force 5.1. Demographic data for South Africa in 2005 and 2035 Table 43. Demography and distribution of actives, employed and broad unemployed by skill level and household class for the Base Year*

Total population Pensioned / 65+

Class 1

Class 2

Class 3

Class 4

Class 5

Total population

4 950

9 316

9 378

14 964

9 033

47 640

123

463

634

738

125

2 084

Of working age / 15-64

2 609

4 335

4 639

10 315

8 192

30 091

Children / <15

2 217

4 518

4 105

3 910

715

15 465

Share in total occupied active

10%

14%

33%

41%

100%

Employed

349

1 227

1 723

4 022

4 995

12 315

Unemployed

874

1 295

1 276

2 913

1 440

7 798

Active, at working age

1 223

2 522

2 999

6 935

6 435

20 113

Share of working age active

47%

58%

65%

67%

79%

67%

Unemployment

71%

51%

43%

42%

22%

38.8%

1 386

1 814

1 640

3 381

1 758

9 978 3 702

Inactive, at working age Working, skill 3

100

900

2 702

Working, skill 2

100

780

2 722

2 193

5 845

Working, skill 1

299

1 127

842

400

100

2 768

Unemployed, skill 3

550

709

1 334

Unemployed, skill 2

120

246

1 116

2 333

726

4 541

Unemployed, skill 1

754

1 049

1 922

Unemployment, skill 3

43%

38%

21%

26%

Unemployment, skill 2

71%

59%

46%

25%

44%

Unemployment, skill 1

72%

48%

41%

Active, skill 3

175

1 450

3 411

5 036

Active, skill 2

170

346

1 896

5 055

2 919

10 386

Active, skill 1

1 053

2 175

927

430

105

4 691

Inactive, skill 3

330

776

1 146

Inactive, skill 2

123

250

1 369

3 051

982

5 774

Inactive, skill 1

1 264

1 563

231

3 058

* Skill 3 = high skill, skill 2 = medium skill, skill 1 = low skill. Class 1 is the lowest expenditure class in BY data, and class 5 the highest. Source: authors’ own calculations and assumptions and data based on various other sources.

113

5.2. Demographic data for South Africa in 2035 Table 44. RP demography and distribution of actives by skill level and household class for 2035* Class 1

Class 2

Class 3

Class 4

Class 5

Total / Avg

Total population

5 846

11 081

11 496

19 165

11 940

59 528

Pensioned / 65+

281

1 058

1 451

1 689

287

4 765

Of working age / 15-64

3 500

5 814

6 221

13 834

10 987

40 356

Children / <15

2 066

4 209

3 824

3 642

666

14 407

Share in total occupied active Active, at working age

14%

35%

39%

100%

1 640

3 382

4 022

9 300

8 630

26 974

Share of working age active

47%

58%

65%

67%

79%

67%

Inactive, at working age

1 859

2 432

2 199

4 534

2 357

13 382

Active, skill 3

235

1 945

4 575

6 754

Active, skill 2

228

465

2 543

6 779

3 915

13 929

Active, skill 1

1 412

2 917

1 244

577

141

6 291

Inactive, skill 3

442

1 041

1 537

Inactive, skill 2

165

336

1 836

4 091

1 316

7 744

Inactive, skill 1

1 695

2 097

309

4 101

* Skill 3 = high skill, skill 2 = medium, skill 1 = low; Similarly, class 1 is the lowest expenditure class in BY data, and class 5 the highest. Source: authors’ own calculations and assumptions and data based on various other sources.

6. Elasticities for production, consumption and international trade The nested-CES production and consumption functions require elasticities to determine the response to relative price changes of inputs, factors or goods consumptions. The timeframe issues surrounding the use of econometrically estimated elasticities will not be discussed here. In any case, the estimation of such parameters on South African economic data was outside the scope of our research. We therefore had to fall back on estimates provided by the existing literature, or indeed on mere assumptions.

6.1. Production function elasticities Van der Werf (2008) and Okagawa and Ban (2008) provide recent international estimations of production function elasticities for KLEM-type nested-CES production functions (Table 45). Because of differing sectoral disaggregation we can only use these estimates as guidance for our own assumptions. At the very least, the large set of sectoral values reported by both authors provides a range in which we may reasonably confidently place the values we pinpoint for each one of IMACLIM-SA sectors. 114

Basic metals

Non-metallic minerals

Food & Tobacco

Transport equip.

EIN/MAN

EIN

MAN

LSS

V. d. Werf

K/L elasticity

0.62

0.45

0.46

0.41

0.27

0.22

KL/E elasticity

0.65

0.25

0.40

0.17

0.45

0.29

O. & B.

Table 45.â&#x20AC;&#x201A; Estimations of KLE elasticities of substitution by Van der Werf (2008) and Okagawa and Ban (2008)

K/L elasticity

0.64

0.41

0.39

0.52

0.21

0.64

0.53

KL/E elasticity

0.22

0.36

0.38

0.14

0.38

0.16

0.07

Industry Corresponding IMACLIM-SA sector

Paper Textile industry industry

Construction

On this basis we decided to assume low KL/E elasticities for heavy industries and energy-intensive sectors (Table 46). Besides KL/E elasticities our nested-CES structure also requires estimates for substitution elasticities between the different levels of K and L substitution. Krusell et al. (2000) provide rare estimates of capital substitutability[63] to skilled vs. unskilled labour based on US time series between 1963 and 1992. We estimate our KL elasticities on the basis of their insights. To understand the choice of elasticities we first compare their definition of labour. Krusell et al. define skilled labour as College completion or higher (16 years of schooling or more) and unskilled labour as below completed high school education. Their skilled labour category is thus slightly more selective than our skill 3 category, which for a large part consists of people with only high school degrees. Their unskilled labour covers the rest of the labour force. This means that their unskilled labour category partly consists of people with a higher degree than in our skill 2 and 1 categories at base year. Furthermore, in 1962 the US population with high school or higher education was about 50% of people over 25 years old,[64] whereas in 2005 South Africa it was approximately 33% of the population of working age only. This means that the starting point of the analysis by Krusell et al. is an economy with already more skilled labour than 2005 South Africa, and therefore the potential substitutability of high-skilled labour to capital might be lower. Ultimately, Krusell et al. (2000) find substitution elasticities of 0.67 between skilled labour and equipment capital, and of 1.67 between unskilled labour and equipment capital. As, to our knowledge, no such substitution estimates are available for South Africa, and also as Krusell et al. acknowledge that the skill premium in the US was not constant over time, we resort to reasonable estimates about the substitution between labour and capital. In initial runs we applied their unskilled labour-equipment elasticity for our substitution function of low-skilled labour to capital & medium-to-high-skill labour aggregate, and their skilled labourequipment elasticity to our high-skill to capital substitution elasticity. These runs showed low [63] They more precisely focus on equipment capital separated from infrastructure capital. For lack of more compatible estimates we ignore this refinement. [64] US Census, see: http://www.census.gov/prod/2012pubs/p20-566.pdf (accessed March 2015).

115

response of the production function to the availability of labour at different skill levels, and in regard to the caveat of the already large supply of high-skill labour in Krusell et al. (2000) we considered it necessary to introduce more flexibility in this part of our South African nested-CES production functions. Therefore we assumed a very low elasticity of 0.1 for high skill (L3) and capital (K) substitution, a high elasticity for our low-skill category (4) and an in-between value for the elasticity of substitution between the KL3 aggregate and medium skill labour (L2) (Table 46).

Table 46. Nested-CES production function elasticities of IMACLIM-SA sectors Elasticity of substitution of... Sector*

KLE to M

KL to E

KL23 to L1

KL3 to L2

K to L3

COA

0.1

0.25

1.5

0.1

GAS

0.1

0.25

1.5

0.1

REF

0.1

0.2

1.5

0.1

ELC

0.1

0.2

1.5

0.1

EIN

0.1

0.25

1.5

0.1

MAN

0.1

0.64

1.5

0.1

LSS

0.1

0.64

1.5

0.1

HSS

0.1

0.99

1.5

0.1

TRA

0.1

0.18

1.5

0.1

* We report no estimates for the OIL sector, whose output is systematically projected at nil in 2035.

6.2. Non-constrained consumption part of final consumption by households IMACLIM-SA treats part of the consumption of energy goods and transport services as basic needs that are irresponsive to prices. This is the case for COA, REF, ELC and TRA. Beyond such needs three CES functions rule (i) the energy mix of the EAG energy aggregate, (ii) the “material mix” of the COMP composite material aggregate, and (iii) the trade-off between the two aggregates to ultimately produce utility. Because this nested structure is quite specific to our study, and again due to the lack of time to inform it in any way with the flexibilities revealed by SATIM runs, we had to resort to fully exogenous assumptions on the prevailing elasticity of substitution, which we simply chose in the usual broad range of comparable studies (Table 47).

Table 47. Nested-CES household consumption elasticities of IMACLIM-SA Class 1

Class 2

Class 3

Class 4

Class 5

Elasticity of substitution of EAG, COMP in U

Household class

0.5

Elasticity of substitution of ELC, REF in EAG

0.5

0.4

0.3

0.2

0.1

Elasticity of substitution of EIN, MAN, LSS, HSS, TRA in COMP

1.5

116

6.3.â&#x20AC;&#x201A; Elasticities of international trade For international trade we rely on exogenous assumptions within the range of the available literature, taking account of the exposure to trade of our IMACLIM-SA sectors.

Table 48.â&#x20AC;&#x201A; Terms-of-trade elasticities of exports and imports of IMACLIM-SA Sector

COA

OIL

GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

sMp

0.50

0.10

0.50

0.75

0.25

1.00

0.10

0.75

0.10

sXp

1.00

0.10

0.75

0.10

1.00

0.10

0.75

0.10

sMp is the elasticity to terms-of-trade of the share of imports in total resource (volume), sXp is the elasticity to terms-of-trade of sheer exports. See model formulary in Appendixes.

117

Appendix 3. Detailed numerical results 1. RP employment changes vs. BY by sector and by skill level The table below gives the numbers with the discussion in section 7.1.2 of the development of employment in South Africa in our Reference Projection (RP) vs. our base year calibration of 2005 data (BY).

Table 49. Growth in jobs of RP compared to BY by sector and skill Units: Thousands of jobs or percentages

COA

OIL

GAS

REF

ELC

EIN

MAN

LSS

HSS

TRA

Total SA

31.8

1.5

13.9

847

826

2 756

2 216

151

6 916

High skill

8.6

0.6

5.8

217

253

591

1 252

2 406

Medium skill

21.7

0.9

577

511

1 449

882

3 594

Job growth RP vs. BY

Low skill

1.5

0.0

0.1

715

917

Share of total job growth

0.5%

0.0%

0.2%

1.0%

12.2%

11.9%

39.9%

32.0%

2.2%

100%

High skill

0.1%

0.0%

0.1%

0.4%

3.1%

3.7%

8.6%

18.1%

0.7%

34.8%

Medium skill

0.3%

0.0%

0.1%

0.7%

8.3%

7.4%

21.0%

12.8%

1.4%

52.0%

Low skill

0.0%

0.8%

0.9%

10.3%

1.2%

0.0%

13.3%

75%

74%

40%

103%

106%

70%

47%

56%

43%

56%

High skill

95%

107%

56%

124%

127%

82%

53%

64%

55%

65%

Medium skill

77%

73%

41%

103%

108%

74%

53%

56%

44%

61%

Low skill

31%

12%

56%

36%

35%

21%

33%

Growth from BY

119

120

+1.0%

10%

23.9%

tIS

-86%

-83%

-111%

DROW/GDP

-6%

82%

+7.4%

+33.1%

+100%

+150%

+3.11%

604

IRPupCT100

+1.00%

+2.0%

100

2035

2005

Ct100 R1

DH/GDP

79%

116%

DG/GDP

+5.9%

Trade balance-to-GDP ratio (2035)

DS/GDP

+116%

+36.8%

+170%

Real GDP (vs. BY)

Real wage (vs. BY)

+3.36%

Average annual growth rate, GDP

Real GDP per capita (vs. BY)

801

CO2 emissions (Mt CO2)

Results

IRPupdated

Annual L productivity gain

IRP scenario

+2.0%

Annual K productivity gain

Investment in educ. & training

2035

Projection year

Carbon tax in ZAR(2005)/ tCO2

2005

Calibration year

Hypotheses

RP (CSLF)

23.9%

-67%

-81%

72%

76%

+3.7%

+41.7%

+118%

+173%

+3.40%

644

IRPupCT100

+1.00%

+2.0%

100

2035

2005

Ct100 R2

21.6%

-113%

-85%

118%

80%

+6.1%

+34.4%

+101%

+151%

+3.12%

610

IRPupCT100

+1.00%

+2.0%

100

2035

2005

Ct100 R3

19.9%

-68%

-82%

73%

77%

+3.8%

+38.5%

+108%

+160%

+3.24%

629

IRPupCT100

+1.00%

+2.0%

100

2035

2005

Ct100 R4

23.9%

-68%

-81%

72%

76%

+3.8%

+38.1%

+105%

+156%

+3.18%

615

IRPupCT100

+1.00%

+2.0%

100

2035

2005

Ct100 R5

24.7%

-114%

-85%

118%

81%

+6.1%

+34.2%

+101%

+151%

+3.11%

610

IRPupCT100

+1.00%

+2.0%

100

2035

2005

Ct100 R6

Table 50. Main settings and key outcomes for RP and all policy scenariosRuns

+1.00%

+2.0%

300

2035

2005

Ct300 R2

+1.00%

+2.0%

300

2035

2005

Ct300 R4

+1.00%

+2.0%

300

2035

2005

Ct300 R5

+1.00%

+2.0%

300

2035

2005

Ct300 R6

23.9%

-64%

-77%

70%

72%

+3.6%

+50.9%

+148%

+209%

+3.84%

683

23.9%

-68%

-81%

72%

76%

+3.8%

+40.8%

+109%

+161%

+3.25%

467

18.0%

-69%

-83%

74%

78%

+3.9%

+36.3%

+95%

+144%

+3.01%

449

23.9%

-69%

-82%

73%

78%

+3.8%

+35.7%

+91%

+139%

+2.94%

434

25.7%

-118%

-87%

122%

84%

+6.3%

+31.2%

+83%

+129%

+2.80%

428

IRPupCT100 IRPupCT300 IRPupCT300 IRPupCT300 IRPupCT300

+1.25%

+2.5%

Yes

100

2035

2005

Ct100 R2+

2. Main settings and key outcomes for RP and all policy scenarios

121

2.0%

TY in VA

9.6%

32.2%

41.4%

Of skill 2

Of skill 1

Unemployment (broad)

Of skill 3

10.6%

28.7%

Share of CFC in VA

34.2%

53.2%

Gross wages in VA

Share of NOS in VA

40.3%

Share of VA in output

44.8%

13.4%

GFCF/GDP

Share of K in VA

42.5%

GOS/GDP

RP (CSLF)

46.2%

36.3%

12.3%

32.6%

10.8%

34.5%

45.4%

2.0%

52.6%

39.8%

13.5%

42.2%

Ct100 R1

36.8%

28.6%

7.1%

25.1%

10.0%

34.3%

44.3%

2.0%

53.6%

40.1%

13.2%

43.6%

Ct100 R2

44.8%

35.1%

11.3%

31.4%

10.7%

34.3%

45.0%

2.0%

52.9%

40.1%

13.3%

41.9%

Ct100 R3

40.4%

31.4%

8.7%

27.8%

10.3%

34.1%

44.4%

2.0%

53.5%

40.3%

13.2%

42.0%

Ct100 R4

40.4%

31.7%

8.9%

28.0%

10.4%

33.7%

44.1%

2.1%

53.8%

40.8%

13.0%

41.0%

Ct100 R5

45.2%

35.3%

11.5%

31.6%

10.7%

34.4%

45.1%

2.0%

52.9%

40.0%

13.4%

41.9%

Ct100 R6

39.0%

28.4%

6.1%

25.3%

9.4%

34.1%

43.5%

2.1%

54.4%

39.2%

12.3%

42.8%

Ct100 R2+

38.3%

29.8%

7.5%

26.2%

10.1%

34.6%

44.7%

2.1%

53.2%

39.6%

13.3%

44.0%

Ct300 R2

43.6%

34.0%

10.1%

30.3%

10.5%

34.3%

44.8%

2.1%

53.1%

40.0%

13.2%

41.8%

Ct300 R4

43.8%

34.6%

10.4%

30.7%

10.6%

33.7%

44.4%

2.1%

53.6%

40.7%

13.0%

40.3%

Ct300 R5

50.1%

39.4%

14.3%

35.6%

11.0%

34.5%

45.5%

2.0%

52.4%

39.7%

13.3%

41.2%

Ct300 R6

122

2010

2020

2006

Gas in electricity

Diesel in electricity

Imported electricity

International oil

100

2006

Materials cost index

GDP index°°

2006

Electric output(PJ)

Electricity imported (PJ)

Coal in electricity (PJ)

Diesel/Gas in elec. (PJ)

2 608

904

97.8

4 377

185

1 791

110.2

312

100

102.1

21.1

114.1

85.8

12.9

2035 value

1.3

1.7

7.7

2.0

1.1

3.1

1.0

1.5

1.0

1.4

2.2

1.5

No carbon tax

2035 vs. start year

+13%

id.

+52%

id.

+38%

+122%

+53%

Relative to pK

2035 vs. start year Relative to pK

3 269

186

1 731

131.4

312

100

102.1

21.1

119.2

89.6

24.0

1.3

7.7

1.9

1.3

3.1

1.0

1.5

1.0

1.4

2.3

2.8

+34%

+52%

id.

+44%

+132%

+183%

Carbon tax 100 ZAR(2010)/tCO2

2035 value

2035 vs. start year

Relative to pK

1 911

186

1 690

154.2

312

100

102.1

21.1

129.4

73.8

49.5

1.3

0.7

7.7

1.9

1.6

3.1

1.0

1.5

1.0

1.6

1.9

5.9

+58%

id.

+52%

id.

+56%

+91%

+486%

Carbon tax 300 ZAR(2010)/tCO2

2035 value

Notes: ° Conversion factor 1 ZAR(2005) = 1.45 ZAR(2010). °° ERC documentation gives GDP growth rates by 5-year period; on this basis we estimated an overall GDP growth for 2006-2035; * Energy prices by fuel are set exogenously, but prices here are weighted average values for given categories of fuels, and weighing is endogenous; ** Note on coal price: by proxy of CLE, as CLD is negligibly small and has approximately the same development.

2006

Electricity cost

Outcomes

100

2006

100

2006

100

67.1

21.1

82.9

38.7

8.5

Labour cost index

Starting value

Capital cost (pK) index

Exogenous parameters

2006

Starting year

Coal in electricity**

Energy prices*

SATIM run

Run/item

Table 51. Parameters and variables of the SATIM runs used to derive technical coefficients for the ELC sector in IMACLIM-SA

3. Comparison of parameters and variables of SATIM and IMACLIM-SA runs

123

57.1

2005

Imported electricity

2005

Materials cost

GDP index

Electricity cost

Electricity output (PJ)

Electricity imported (PJ)

2005

Cons. of COA by ELC (PJ)

Cons. of REF&GAS by ELC (PJ)

Variables following from tech coef.

844

2005

2 541

58.4

100

1049.4

156.3

2005

Labour cost

1128.7

84.0

Capital cost pK

Other variables (endogenous)

International oil

38.8

2005

Refined fuels in electricity

44.1

2005

Gas in electricity

5.7

Starting value

2005

Starting year

Coal in electricity*

Energy prices (endogenous)

IMACLIM run

Run/item

4 374

1 717

95.6

270

1545.2

286.0

1739.0

185.2

126.0

198.2

78.4

8.0

2035 value RP

7.3

1.7

2.0

1.6

2.7

1.5

1.8

1.5

3.2

2.4

1.8

1.4

2035 vs. start year RP

+6%

-4%

+19%

id.

+110%

+53%

+15%

-9%

Relative to pK

2035 vs. start year

Relative to pK

3 071

1 560

130.1

273

1575.9

290.5

1666.8

177.6

120.8

211.1

107.6

29.1

6.6

1.2

1.7

1.8

2.2

2.7

1.5

1.9

1.5

3.1

2.5

2.4

5.1

+51%

+2%

+26%

id.

+111%

+70%

+65%

+244%

2035 vs. start year

1 630

1 382

153.0

261

1553.0

279.8

1712.3

165.4

112.5

234.3

163.4

70.6

5.9

0.6

1.6

2.6

1.5

1.8

1.5

2.9

2.8

3.7

12.3

+73%

-2%

+18%

id.

+91%

+84%

+144%

+711%

Relative to pK

C tax 300 ZAR(2005)/tCO2 + R2

2035 value

(Coal, Gas, and Ref incl. CO2 tax, see report section 7.3.1)

C tax 100 ZAR(2005)/tCO2 + R2

2035 value

Table 52.â&#x20AC;&#x201A; Variables of selected IMACLIM-SA runs for comparison to parameters and variables from SATIM runs

Appendix 4. Insights from similar exercises conducted in Brazil The International Research Centre on Environment and Development (CIRED) and the Alberto Luiz Coimbra Institute for Graduate Studies and Research in Engineering (COPPE) of the Federal University of Rio de Janeiro, Brazil have been conducting joint research since 2011 on the articulation between decarbonisation, energy policy and economic development in Brazil. Despite obvious differences between the Brazilian and South African contexts, this ongoing programme is of interest to the present study because it has the same objectives, namely exploring how decarbonisation objectives can be articulated with other key development objectives at a national level, and because it uses the same methodology, i.e., a multi-criteria assessment of various policy packages using a tailor-made IMACLIM_Brazil model to inform policy debate. The present Appendix briefly describes the Brazilian context, recaps the main questions that have been explored so far by COPPE and CIRED and the results obtained, and concludes with some comparative insights with the case of South Africa.

1. The Brazilian context 1.1. A rapidly growing economy Ranking fifth by area (8.5 Mkm2) and population (201 Mhab - 2013) and seventh by GDP (US$ 2,246 bn – 2013), Brazil is one of the most powerful economies in the world today. It is one of the five major emerging economies usually identified as the BRICS (Brazil-Russia-India-China-South Africa). Despite stagnation in 2009 due to the economic crisis, the Brazilian economy has experienced a 3.5% average GDP growth rate over the last decade. Nevertheless, GDP per capita (US$ 11,208 – current, PPP, 2013) remains more than four times smaller than in the US. The Brazilian economy is dominated by services, which make up 66.8% of GDP in 2011 (PDE2022, EPE, 2013), while industry and agriculture represent 27.8% and 5.4% respectively. Main industrial productions include mineral and fossil energy resources, steel, petrochemicals or automobile production. Agriculture and agro-industry—which together represent one quarter of GDP—are powerful sectors in Brazil owing to generous natural endowments of 340 Mha of arable land. 125

Dynamic agribusiness has made Brazil one of the major producers and exporter of agricultural goods and the country is currently the largest exporter of coffee, soybeans, beef, sugar cane, ethanol and frozen chickens. The share of trade in GDP, around 25 %, is quite low compared to other emerging economies. This is partly explained by the large and growing domestic economic market. Among other macroeconomic indicators, the unemployment rate is low (around 5% in 2014), and so is the saving rate (average 18.8% of GDP, 2007-2011). The trade balance has been positive in recent years, around 0.5-1% of GDP. Public deficit and debt are rather high and amounted to 2.6% and 40.3% of GDP respectively on average over the 2007-2011 period. As an emerging economy, Brazil has a young population and a high share of active population. The rate of urbanisation is also particularly high (84% of total population lived in cities in 2010). Brazil is known for significant social inequalities, although the situation has improved rapidly in recent years. In fact, the Gini coefficient of Brazil decreased from 0.596 in 2001 to 0.543 in 2009 (IPEA - www.ipeadata.gov.br). Persisting inequalities can be explained by multiple factors, including very heterogeneous levels of education, an important urban-rural contrast, unequal land ownership and a regressive tax system. Specific income transfer policies (such as the Bolsa Familia—Family Allowance—programme), made it possible for around 25M people to join the middle-income group during the 2003-2009 period only (SAE 2013).

1.2. A very specific energy sector Brazil stands out with a particularly enviable endowment of natural and energy resources and its energy potential largely exceeds the country’s current needs. Such favourable conditions have made it possible for Brazil to develop a strong energy system to sustain its economic growth. Close to global self-sufficiency, the Brazilian energy system has notably provided almost universal access to electricity. Moreover, the Brazilian energy matrix is actually one of the most “renewable” and least carbon-intensive in the world with a share of almost 45% of renewables in the primary energy demand. Salient features of the Brazilian energy system include large demand by industry, a power generation system dominated by hydropower, a large share of bioenergy use (30% of the primary mix) especially in industry and in the form of liquid fuels for transportation, and the fact that oil represents the bulk of fossil fuel consumption (41% of primary energy demand)—mainly in transportation. Over the last two decades, energy demand in Brazil has closely followed GDP growth with even a small increase of the global energy intensity of domestic product. Overall, energy demand has doubled since 1990 to reach 270 Mtoe in 2011. With the exception of natural gas, which now amounts to 10% of primary energy demand, the structure of the primary energy mix has not changed significantly over this period. Fast per capita GDP growth coupled with a decrease in inequality have boosted demand for end-use sectors energy, especially for specific electricity in residential and commercial sectors and for liquid fuels for passenger transportation. Regarding the latter, the growth in mobility demand and in private vehicle ownership have driven a 4% per year increase in liquid fuel demand. The industrial sector remains the largest end-use energy sector though, with a 3.5% per annum growth of final energy demand, driven by steel and paper industries among others (IEA, 2013). 126

In terms of power supply, hydro still accounted for 70% of installed capacities and 81% of power generated in 2011. However, the share of other generation technologies is growing. Bioenergy now represents 6% of power generation—mainly through auto-production in the sugar industry through the combustion of “bagasse”, a by-product of sugar cane. Thermal generation represents 10% of the mix (half of which is natural gas) and the rest is composed of nuclear (two plants) and other renewables. Although still small in share, wind and other renewables are expanding rapidly and should take a substantive part of the power generation mix in the coming decades. A third singularity of the Brazilian energy system is the prominent share of bioenergy. This is the result of voluntary public policies initiated with the so-called “ProAlcool” plan in the 70s. In the wake of the first oil shock, the Plan aimed at developing domestic production of ethanol from sugar cane for energy security purposes. Nevertheless, bioenergy today takes varied forms in the energy matrix. First, traditional biomass still accounts for a significant share of household energy consumption—25% of firewood demand—but is decreasing fast with rising average income per capita and the adoption of modern energy sources. Bioenergy is also a major supplier of industry energy needs. In fact, 75% of Brazil firewood production is used in industry, including 35% that is transformed into charcoal for steel production. Agricultural residues like bagasse are also extensively exploited, especially in the paper industry. Liquid fuels from biomass today account for a steady share of transportation fuel demand. Beyond ethanol from sugar cane, the use of biodiesel made from soybean oil is increasing—it is incorporated with traditional diesel through a rising blending mandate. However, ethanol remains by far the main biofuel. It has been supplying between 13 and 21% of Brazil’s total demand for fuel in road transport in the last two decades. In practice, ethanol final consumption can assume two forms: either blended with gasoline in an anhydrous form at 25% share (gasoline C as opposed to pure gasoline A) or in a pure hydrated form (AEHC). In parallel, the fleet of light vehicles has been adapted: so-called “flex-fuel” cars, which can accommodate both gasoline C and AEHC or any blend of the two, represent around 60% of the total fleet of light duty vehicles and most of the new sales. One consequence of this flexibility on the demand side is the ratio between the demand for ethanol and for gasoline, which is highly sensitive to relative prices. In recent years, price competition has been detrimental to ethanol for several reasons.[65] Better conditions are expected for the future with fiscal conditions improving, lower production costs owing to productivity gains and right market pricing for gasoline. A final salient trait of the Brazilian energy system is that Brazil is an important oil consumer and producer. On the demand side, oil represents 41% of primary energy demand and petroleum liquid fuels make up more than 80% of total transport fuel demand. The two main fuels are diesel essentially used by trucks and buses and gasoline used by light-duty vehicles. On the supply side, Brazil has become an important oil producer since the 1980s, notably owing to the development of off-shore deep-water fields. Since 2006, Brazil has even been more or less self-sufficient (despite temporary imports of liquid fuels depending on domestic demand). Ninety per cent of Brazil's proven reserves are off-shore deep-water reserves. Since 2006, the huge deep-water discoveries [65] Government has artificially maintained low gasoline prices to contain inflation and thus boosted gasoline consumption. At the same time, there have been constraints on ethanol supply with high sugar prices—which have incited sugar cane producers to produce sugar instead of ethanol—combined with weak harvests.

127

qualified as “pre-salt”[66] in the Santos basin have revived the ambition of the country towards the oil sector. The exploitation of pre-salt fields is just beginning and nurtures Brazil’s ambition to be a major exporter of crude oil and a possibly significant gas producer in the coming decades.

1.3. And a peculiar GHG emissions profile Brazil is characterised by a rather low level of GHG emissions per capita—around 6.5 tCO2eq/cap in 2010—compared to other emerging economies with similar GDP per capita. For example, China and South Africa respectively emitted around 7 and 11 tCO2eq/cap in 2010 whereas their GDP per capita PPP were 20% and 17% lower than Brazil. This below-average level of emissions comes together with a very specific emissions profile. Indeed, the bulk of Brazil emissions are not energyrelated, but come from agriculture processes, forestry and other land-use sectors. At the same time, energy-related CO2 emissions per capita are very low—less than 2 tCO2/cap in 2010—owing to the low carbon intensity of the energy system (see above). Until recently (2009-2010), the largest share of emissions came from land use, land-use changes and forestry (LULUCF), especially from deforestation in the Amazon region—resulting from the expansion of the agricultural frontier (pastures for cattle-raising especially) to the detriment of the rainforest. However, deforestation rates have decreased significantly in recent years (LULUCF accounted for 23% of total emissions in 2010), and the agriculture and livestock sectors have become the main emitting sectors with 35% of the total in 2010, an illustration of the importance of the agricultural and agro-industrial sector in Brazil. Agriculture and livestock emissions are not energyrelated.[67] They come mostly from production processes and from the enteric fermentation of one of the largest bovine herds in the world (CH4 emissions). The low level of energy-related emissions (375 MtCO2 in 2010—around 30% of the total) and its decomposition across sectors is directly related to the structure of the energy system described above. As expected, owing to the large hydropower capacities, power generation contributes little to emissions, contrary to the situation in most countries. It amounts to only 10% of total energyrelated emissions in Brazil, against 40% and more than 50% in China and South Africa respectively. Emissions from the building sector are also limited because of the still substantial use of traditional biomass and the low need for heating owing to the favorable climate. As a result, the bulk of energyrelated CO2 emissions comes from industry (128 Mt, or 37%, in 2010) and from transportation (162 MtCO2, or 46%, in 2010). Yet even in these sectors the widespread use of bioenergy leads to carbon intensities below the world average. For example the carbon intensity of Brazilian industrial energy is close to 1 tCO2/toe whereas it is higher than 1.5 tCO2/toe in most other countries. In the transport sector, the impact of biomass use is less clear-cut. First, biofuel mainly concerns light vehicle passenger transportation. On-road freight transportation typically uses trucks running on (traditional) diesel. Second, the share of road transport in total transportation is very high. As a result, the carbon intensity of passenger transportation energy is well below average whereas the carbon intensity of freight transport energy is close to average.

[66] This is because hydrocarbons are trapped deep underground (5,000 m under the sea floor) below a thick layer of salt. [67] Emissions from fossil fuel combustion for trucks and farm equipment are accounted in the energy sector.

128

To sum up, the recent drop in deforestation rates coupled with an energy system of historically low carbon intensity has enabled Brazil to improve its position as a low-carbon economy. Nevertheless, the share of fossil fuel combustion in total GHG emissions is increasing fast—from 16% to 32% during the period 2005-2010. As we will see in next section, Brazil faces now the challenge to build on its historical low level of GHG emissions while continuing to improve the living standards of its population.

2. Prospects at the interface of decarbonisation–energy– economic development 2.1. The Brazilian National Plan on Climate Change 2009-2020 (PNMC) At COP15 in Copenhagen, Brazil announced voluntary mitigation goals for the 2009-2020 period. These pledges further became mandatory in law in December 2009 through the National Plan on Climate Change (PNMC) coordinated by the Ministries of the Environment and of Science and Technology. Federal Law No. 12187 establishes the target to maintain total GHG emissions in 2020 from 36.1% to 38.9% below a business-as-usual scenario (BAU) built under the supervision of the Brazilian Forum for Climate Change (BFCC). This voluntary pledge falls into the framework of NAMAs (National Appropriate Mitigation Actions) under the United Nations Framework Convention on Climate Change (UNFCCC). Federal Decree No.7390 (2010) enforces the mitigation goals of avoided GHG emissions until 2020.[68] Meeting the PNMC pledge would result in emissions in 2020 lower than those in 2005.[69] If emissions from deforestation are maintained at the historically low level observed in 2010 until 2020, the LULUCF sector alone will be sufficient to meet the pledge. Similarly, the BAU assumes a decrease in the share of renewables in the energy mix between 2005 and 2020 with, in particular, no new expansion of hydropower or other renewables—new generation is assumed to be gas-fired power plants—, and no expansion in the demand for biofuels—additional demand is assumed to be for gasoline.[70] The mitigation goal assigned to the energy sector (234 MtCO2 abatement compared to BAU) corresponds to the implementation of the Ten-Year National Energy Plan (PDE 2019).[71] To sum up, Brazil appears in good position to meet its PNMC objectives. Yet the growth of energyrelated emissions announces further challenges for decarbonisation beyond 2020. [68] To reach the mitigation goal, the Decree creates a process for developing action plans to reduce the deforestation rates in the Amazon Basin by 80% and in the Cerrado by 40%, and to restore 35 million hectares of degraded land. The Decree also sets a deadline of December 15, 2011, for each of twelve major greenhouse gas-producing sectors to submit action plans for emissions reductions. The resulting targets may form the basis for emissions trading. [69] As a matter of comparison, China and India have respectively pledged to lower their CO2 emissions per unit of GDP by 40-45 and 20-25 per cent in 2020 relative to 2005. Assuming GDP growth continues at comparable rates over the coming decade, China and India’s GHG emissions should still be higher in 2020 relative to 2005. On the contrary, if Brazil meets its pledge, its GHG emissions in 2020 will be 6-10% lower than in 2005. [70] The BAU contemplates a 15% increase of energy-related emissions intensity of GDP between 2005 and 2020. [71] Released every year, ten-year plans represent official guidelines for the development of the Brazilian energy system. The 2010 plan includes the development of added hydro-power capacities, investments in energy efficiency and additional penetration of ethanol for transportation.

129

2.2. Challenges beyond 2020 Beyond 2020, the country will face the challenge to control fast growing energy-related emissions linked to dynamic economic growth and the rise of living standards: the size of the Brazilian economy is expected to double over the 2015-2030 period, with an economic development partly driven by oil and gas. Yet, the margins of manoeuvre to decarbonise the energy system seem narrower in Brazil than in other countries—since Brazil is already starting from a low base. The three scenarios developed in La Rovere et al. (2013) illustrate the mitigation challenges that Brazil is expected to face post 2020. The first scenario (Scenario A) is close to the PNMC BAU extended to 2030: persistence of high deforestation rates and no development of renewables beyond 2009. The second scenario (Scenario B) corresponds to the achievement of the PNMC, constant deforestation rates from 2020 to 2030 and 2020-2030 energy trends resulting from a combination of PDE 2019 and the 2030 National Energy Plan (PNE 2030). Released by the Energy Planning Agency, the PNE 2030 embodies the long run strategy of expansion of the Brazilian energy system. Finally, the third scenario (scenario C) includes further mitigation efforts relative to B in all sectors: afforestation, further penetration of renewables, energy efficiency programmes, etc. In short, scenario B can be considered as a reference scenario until 2030 (Figure 8).

Figure 8. Total GHG emissions of Brazil, 2015–2030 per sector in Scenario B 3000

2500

MtCO2e

2000 Waste

1500

IPPU Energy

1000

Agriculture LULUCF

500

2015

2020

2025

2030

Source: La Rovere et al. (2013)

Scenario B results in a shift in emission patterns around 2020. Total GHG emissions decrease from 2005 to 2020 owing to the PNMC, but starts rising again afterwards, driven by energy related emissions which become the main source of emissions (around 45% of total in 2030). This increase of energy-related emissions—around 5% per year for the 2020-2030 period—is embedded in current energy plans and reflects expected trends. 130

Figure 9 details the contribution of each sector to energy-related emissions in Scenario B. Power generation and the residential sector remain small contributors, as a result of the development of renewables embedded in energy plans. And energy-related emissions beyond 2020 are mainly driven by transportation and the industry (including the fossil-fuel sectors) sectors fossil-fuel consumption.

Figure 9. Energy-related GHG emissions in Brazil per sector, 2015–2030, in Scenario B 1400 Fugitive Emissions

1200

Industrial

MtCO2e

1000

Transports Agriculture/Husbandry

800

Public Administration 600

Commercial Residential

400

Power Generation 200 0

Energy Sector

2010

2015

2020

2025

2030

Source: La Rovere et al. (2013)

Future energy-related emissions are primarily driven by population and economic growth through the rise of energy demand. Existing energy-economy scenarios indeed contemplates a rate of GDP growth around 4% per year on average until 2030.[72] In addition, the rise of living standards will drive up some structural changes that may aggravate emissions on top of the growth driver. First of all, rising per capita incomes are expected to increase demand for passenger mobility and to result in a strong increase in private vehicles ownership. One key aspect will be the extent to which low-carbon urban transportation modes develop in a country with a weak public transportation systems and a domination of private car use. Likewise, without specific actions to introduce low-carbon interurban freight transportation systems (e.g., railways or waterways), freight transport is likely to [72] Such dynamic economic growth—which would make it possible for Brazil to partially catch-up with developed countries GDP per capita and living standards in 2030—is not warranted and Brazil will have to overcome specific structural weaknesses. Those include the deficit of modern infrastructures of transport and sanitation, lack of a sufficiently large skilled labour force and the institutional constraints and complex regulatory environment known as the “custo Brasil”. However, investments in education and in innovation should bring the needed productivity gains in the medium and long run and improve the competitiveness of domestic productions. Moreover, the improved stability of institutions and the macroeconomic policy should help to secure the productivity gains and the efficiency of investments in the future (EPE 2014).

131

continue to rely massively on diesel trucks. Eventually, emissions growth will be driven by global industrial production growth needed to satisfy a fast growing domestic market. If this scenario is indeed the reference, then a key variable for GHG emissions is the extent to which bioenergy replaces fossil fuels in heavy industry and road transportation. In the heavy industry sector, the trade-off will concern for instance the level of penetration of natural gas and the level of penetration of charcoal in the steel industry. The issue is even more crucial in the road transportation sector. On the one hand, on-road freight transport is more or less locked-in diesel, despite the expected 5% to 10% blending of biodiesel. On the other hand, important margins of freedom exist for light passenger vehicles given the massive penetration of flex fuel cars (85% of the total fleet in 2030). The penetration of bioethanol to replace gasoline will depend upon numerous variables such as the production constraints of bioethanol and the domestic oil price. As explained above, the government is currently imposing artificially low oil and gasoline prices to control inflation, a move that is detrimental to ethanol demand. Government intervention towards domestic oil and gasoline prices will be a decisive driver of the future relative share of ethanol and gasoline in transportation fuel demand and further an important driver of future energy-related emissions levels in Brazil.

2.3. The oil question The trade-off between bioenergy and fossil fuels is a crucial junction point for the articulation between energy-related emissions mitigation policy and a development policy based on the exploitation of oil resources. As noted above, with the recent discovery of huge deep-water “pre-salt” reserves in the Santos Basin, Brazil seeks to become a major player in the global oil market. On average, existing forecasts bet on a production of around 5 Mba per day in 2030—that is a doubling of the production for the 2010-2030 period per day in 2030 (e.g. Goldemberg, 2014). Natural gas production is expected to be multiplied by four to reach around 80 Gm3 in 2030. Hydrocarbon resources exploitation thus represents an important source of economic development in the coming decades, and is expected to drive the development of a whole industry.[73] Potential revenues from oil exports are very difficult to anticipate, given the combination of uncertainties on production, domestic demand and international oil price. The most optimistic forecasts mention that oil exports could generate up to US$100 billion in 2030 (IEA, 2013). Yet even in this case, contrary to most large oil exporters, this would still represent a small share of GDP (2.5%). This figure leads IEA experts to believe that the risk of “Dutch disease” is low unless the effect could be amplified in conjunction with the other export commodities. However, the risks associated with the volatility of international oil markets remains. Oil activity is eventually expected to generate important fiscal resources which are planned to be invested in education and health programs. At this stage large uncertainties remain concerning the pace of production that will be reachable for the next twenty years. These uncertainties are about the deep-water extraction that should make up 90% of oil production in 2030 and coming mainly form the Santos basin. Uncertainties also concern the whole supply chain of a very capital-intensive activity. Delays could happen in the context of technical difficulties, supply chain constraints and availability of investments for projects that are of the most complex ever undertaken in the oil industry. Almost the entire responsibility falls on Petrobras, which will have to manage the major [73] For example, oil exploration and production will require machines and equipment—potentially offering a strong boost to domestic machinery production.

132

part of the US$60 billion per year needed for oil production. Eventually there is a risk that a lasting context of low international oil prices may delay the current plans. How this oil is used, and in particular to what extent it is used domestically, is critical for the country GHG emissions. As recent years show, the temptation exists to set low oil prices to curb inflation, thus driving domestic consumption up. Using the majority of the newfound oil on the domestic market would add to this pressure. Today the main scenarios and plans (EPE, 2014; IEA, 2013) suggest the contrary. The bulk of new resources is planned to be exported and renewables should increase shares in domestic demand. And it is very important for the feasibility of decarbonisation in Brazil to avoid the domestic use of the newfound resources because it would weaken the efforts to develop energy efficiency and renewable energy use.

3. Presentation of two joint COPPE-CIRED studies on the articulation between decarbonisation and economic development in Brazil 3.1. Objectives In light of the context framed above, CIRED and COPPE have conducted two studies on the articulation between deeper decarbonisation of the domestic energy system (beyond national plans), general macroeconomic goals (GDP, economic structure, trade, employment, public budget, household welfare) and other specific socio-economic considerations. In general, the studies contemplate no significant technical and structural changes and focus on the substitution between existing technologies. Both studies address the medium run, i.e., the 2020-2030 period. The first study (La Rovere et al., 2015) focuses on the macroeconomic and sectoral impacts of two different mitigation plans compared to a governmental scenario at horizon 2030 (based on the 2050 Energy National Plan: EPE, 2014). Each plan includes a carbon tax and a set of additional mitigation measures. The additional mitigation measures concern all GHG gases and all emitting sectors (AFOLU, energy, industry, transport, residential, services and waste), with a focus on industry. The study builds on a coupling between the estimation of mitigation costs with a bottom-up methodology (MACCs) and the analysis of the consequences of these mitigation programmes when embarked in a hybrid CGE model—namely the IMACLIM-Brazil model (see below). The scenarios were built based on several meetings with key stakeholders: academic experts, government, industry, civil society, etc.

Table 53. Scenarios – Study 1 Scenario

Description

CPG

Reference Scenario – Follow trends from Copenhagen Pledges

MA1-T

Mitigation Scenario 1 – (20 US$ carbon tax + mitigation measures)

MA2-T

Mitigation Scenario 2 – (100 US$ carbon tax + mitigation measures)

133

Table 54. Mitigation measures included in Scenarios M1 and M2 of Study 1 Mitigation scenario 1 (MA1) SECTOR

MITIGATION MEASURE

WASTE

Methane destruction > Reference scenario

SERVICES

More efficient light bulbs

ENERGY

Sugar Cane Bagasse Wind Power Traffic management Bikeways Energy efficiency programmes for light vehicles

TRANSPORT

Ethanol: + 6 Billion litres Biodiesel – B10 BRT Energy efficiency programs for heavy vehicles Biological nitrogen fixation (Corn) Pasture restoration

AFOLU

Crop-livestock-forestry integration systems(+2M ha) Afforestation – 12.5 M ha Reforestation – Atlantic forest Sugar cane area increase (M1)

INDUSTRY

Cement: Reduction from 3,50 GJ/t clinker in 2010 to 3,1 in 2030 and increase co-processing in 75% from 2010 level Steel: 2% reduction in energy intensity related to 2010

Mitigation scenario 2 (MA2) SECTOR

MITIGATION MEASURE Solar heating of water

RESIDENTIAL

Efficient light bulbs Efficient refrigerators

WASTE

Methane destruction > Mitigation scenario 1

SERVICES

Efficient light bulbs > Mitigation scenario 1 Bagasse = MA1

ENERGY

Wind power = MA1 Refining efficiency > MA1 Hydro power increase > reference scenario

134

SECTOR

MITIGATION MEASURE Traffic management Ethanol: + 16 billion litres (6 billion litres from 2G) Bikeways Energy efficiency programmes for light vehicles

TRANSPORT

Energy efficiency programmes for heavy vehicles Modal shift to rail and hydro for load transport VLT Subway Biodiesel - B15 Biological nitrogen fixation (Corn) Pasture restoration

AFOLU

Crop-livestock-forestry integration systems (+2M ha) Afforestation – 15 M ha Reforestation – Atlantic forest Sugar cane area increase (M2)

INDUSTRY

Cement: Reduction from 3,50 GJ/t clinker in 2010 to 2.85 in 2030 and increase co-processing in 75% from 2010 level Steel: 5% reduction in energy intensity related to 2010

3.2. Method Both studies use the IMACLIM-Brazil (hereafter IMACLIM-BR) model, developed by CIRED in collaboration with COPPE. IMACLIM-BR is a computable general equilibrium model (CGEM) (in the broad sense of Thissen, 1998) of the Brazilian economy designed to build consistent projection of the economy–energy–environment system in Brazil over the medium to long term. In particular, it makes it possible to assess the macroeconomic implications of aggregate price or quantity-based carbon and energy policies over the medium- to long-term. It departs from standard neoclassical CGE models in several crucial features. As for most hybrid CGE models, IMACLIM-BR is based on an energy-economy hybrid accounting framework in which economic flows and physical flows (with a special focus on energy balances) are balanced. However, it first departs from the standard models in that its description of the consumer and producer trade-offs, and the underlying technical systems, are specifically designed to facilitate calibration on bottom-up expertise in the energy field, with a view to guaranteeing technical realism for the simulations of even large departures from the reference equilibrium. Confirming observations already made with the IMACLIM-France model, forcing consistency between economic and energy data has significant implications for the model reactions to price shocks, notably carbon taxes. Secondly, IMACLIM-BR represents an economic system in a second-best setting in that it computes general equilibria characterised by production factors underemployment and imperfect markets (in 135

labour, goods and capital). Globally and in spirit, IMACLIM-BR follows the empirical method of original CGE models like Johansen’s model (1960) more than adhering to the standard neoclassical paradigm. Thirdly, IMACLIM-BR relies on comparative-static analysis. It first projects equilibrium of the economy in the medium to long term and simulates the resulting deformation of this reference projection caused by a given policy package. The resulting insights are valid under the assumption that the policy-induced transition from the reference equilibrium to its policy-constrained counterpart is completed, after a series of technical adjustments whose duration and scope are embedded in the response functions of production and consumption retained for the studied time horizon. The transition process itself is not described, but is implicitly supposed to be smooth enough to prevent multiple equilibria, hysteresis effects, etc. The sectoral breakdown of IMACLIM-BR can vary according to the study and ranges in its current version from 4 to 18 sectors of production. The model considers three primary production factors: labour, capital and land. Each sector produces one single good so that commodities and activities match and the Input-Output table is square. Some goods are further desegregated for the representation of household final consumption. Productive sectors generate income through productive factors. National income is further distributed to four representative institutional agents—households (HH), firms (FIRMS), public administrations or government (GOV) and the rest of the world (ROW)—in the same pattern as in national accounts. On the whole, the accounting system used in the model keeps the orthogonal logic between productive sectors and institutional agents in order to keep the detail of primary income distribution. This makes it possible to identify the shares of capital, land income and generated profits that accrue specifically to autonomous producers (Households: family farm, individual entrepreneurs, landlords (agriculture and housing), etc.), corporate firms or public administrations (public companies). Furthermore, through secondary income distribution, institutional agents break down their income between goods consumption, investment, tax payments and transfers. The model considers a detailed system of taxes and transfers. The Rest of the World classically interacts with domestic agents through trade of goods and capital balance. In its current version, IMACLIM-BR is calibrated at base year 2005 on a SAM built on data from two synthesis tables produced by the Instituto Brasileiro de Geografia e Estatistica (IBGE): • The MIP (Matrix Insumo Produto, input-output table) balances the uses and resources of

products—up to 110 of them in its most disaggregated version.

• The CEI (Contas Econômicas Integradas) details the primary and secondary distribution of income

between 6 “institutional sectors”, i.e. aggregate economic agents: financial firms, non-financial firms, households, non-profit organisations, public administrations, “Rest of the World”.

Raw MIP data are processed to obtain a description of production and consumption in a square “product × product” system, with no accumulation of stocks. The CEI is aggregated into four institutional sectors (households, firms, public administrations and “Rest of the World”), and its many entries are simplified into a set of transfers at a level of aggregation comparable to that of the MIP. Basically the primary distribution of income is composed of productive factor remuneration (labour, productive capital and land) and income from property (financial income). The secondary distribution of income comprises indirect taxes and social 136

transfers. Its use ultimately allows extending the traditional framework of general equilibrium modelling to the distribution of national income between economic agents, the resulting changes in the financial positions of those agents, and the corresponding debt payments. MIP and CEI data are finally combined in a unique SAM framework. The energy-economy data hybridisation is based on the national energy balance (Balanço Energétic National, BEN) compiled by EPE (Empresa de Pesquisa Energética). It is also possible to gather end-user specific energy price data from various sources (BEN, ANE – Agência National do Petroleo, etc.). The term-by-term product of energy balances and agent-specific prices (the single-price assumption is abandoned) defines a matrix of energy consumption in monetary terms, which does not match that embedded in the MIP for energy products, for a variety of reasons (the inclusion of services beyond sheer energy products, the heterogeneity of products, biases from the statistical balancing methods, etc.). Hybridisation of the MIP then consists in imputing the differences between the values found in the MIP, and those computed from energy statistics, to some non-energy good—in the model with 12 products, to the aggregate non-energy composite good. For lack of a better hypothesis, the value-added of the energy products is corrected pro-rata for this imputation. In this way, the product disaggregation is amended, while the cross-sectoral totals are kept consistent with the original statistics. The calibration of the model on this hybrid MIP (which is included in the final benchmark SAM) eventually leads it to depict (i) volumes of the non-energy goods that are traditionally derived from the single (normalised) price assumption, and (ii) volumes and prices of the energy goods that are strictly aligned on the available statistics. The differences in price of the same energy good from one agent to the other (e.g. the variable average prices of a kWh of electricity) are accounted for by calibrating “specific margins” to the different uses. For the first study, the IMACLIM-BR was used in conjunction with sectoral model simulations to capture the potential implications and costs of the different mitigation measures identified to be included in the two alternative scenarios. Three sectoral models were used: a Brazilian version of MESSAGE for the energy sector (IAEA, 2007), the LEAP (Long-range Energy Alternatives Planning) model (Heaps, 2013) for transport and residential sectors, and the BLUM (Brazilian Land-Use Model) model (ICONE, 2014) for AFOLU sectors. The sectoral simulations made it possible to estimate the costs as well as the technical characteris tics (mainly energy and capital content variations from reference) of a set of mitigation measures. These measures were sorted into two groups according to their mitigation costs: those under 20 US$/t and those under 100 US$/t. Furthermore the two mitigation scenarios were run with the respective simulation of these two groups of measures in IMACLIM-BR. In practice for one given scenario, the inclusion of mitigation measures at sectoral level consisted in (i) adjusting the technical coefficients according to the technical characteristics (as estimated from sectoral simulations) of the package of mitigation measures involved and (ii) enforcing the payment of the respective carbon price (20US$/t or 100 US$/t). The resulting change of technical and cost structure of sectors triggers a change of market equilibrium estimated by the model. Finally, the disaggregation of the “representative household” into six income classes (second study) is based on an extrapolation of the 2002-2003 Pesquisa de Orçamento Familiar (Household budget) 137

survey by IBGE, which extensively covers the resources and uses of Brazilian households. It is largely documented in Grottera (2013).

3.3. Key results The two studies, carried out independently, show complementary results towards the search for an articulation between decarbonisation goals and socio-economic development in Brazil in the medium term. The first study focuses on the macroeconomic impacts as well as on the detailed sectoral impacts of two mitigation policy packages in the context of a multilateral implementation of a carbon price at world scale. Sectoral mitigation options are specifically informed by bottom-up analysis as described above. The second study is more top-down in nature and looks at the trade-offs and synergies between decarbonisation goals, global economic performance, income inequality and poverty reduction in a context of unilateral climate action in Brazil. To do so, different climate policy packages based on a universal carbon tax and different recycling schemes are simulated. Both studies simulate decarbonisation scenarios at time horizon 2030 compared to similar refe rence cases that range between a 16% and 35% GHG emissions decrease. This corresponds to the general objective of maintaining emissions at their current levels (see Figure 10 for Study 1).

Figure 10. GHG Emissions (Mt CO2 eq.) – Study 1 1700 1600

Mt CO2eq.

1500 1400 1300 1200 1100 1000 900

2010

2020

2030

CPG

1214

1272

1666

MA1-T

1214

1147

1308

MA2-T

1214

1047

1031

Nevertheless, the macroeconomic impacts of similar decarbonisation levels are very contrasted between the two studies: real GDP losses relative to business-as-usual are around 1.5% for a 33% decrease of emissions in Study 1 whereas GDP losses amount to 6% for a 30% decrease of emissions in Study 2 (see Table 54 for Study 1). The two cases implement a carbon tax recycled through a decrease of payroll taxes, which has a positive effect on employment. 138

Table 55. Macroeconomic results – Study 1 2005

2013

Population (10^6)

185

196

GDP (10^12 R$2005)

2.14

3.24

CPG-2030 223

MA1-T

MA2-T

223

5.55

5.54

5.46

3.88%

3.87%

3.81%

GDP variation in relation to CPG in 2030

-0.2%

-1.5%

GHG emission reduction in relation to CGP in 2030

-15,8%

-33,4%

GDP Growth per year (%)

GHG emissions (Mt CO2eq./year)

2351

GDP per capita (10^3 R$2005)

11.57

16.52

Total Jobs (10^6)

91.21

100.06

127.3

127.7

128.1

Unemployment rate (%)

9.9%

6.7%

4.4%

4.2%

3.9%

16.6%

20.1%

17.2%

Price index in relation to 2005 (%) Trade Balance (10^6 R$2005)

1539

1296

1025

24.87

23.83

22.50

78.8

3.90

40.1

38.6

77.9

Trade Balance (% GDP)

3.70%

0.12%

0.72%

0.70%

1.43%

Investment rate (% GDP)

15.5%

18.2%

20.8%

20.2%

20.7%

0.33

0.59

1.15

1.12

1.14

174.9

537.7

20.7

82.9

0.37%

1.52%

Total Investments (10^12 R$2005) Total Mitigation Investments 2015-2030 (10^9 R$2005) Mitigation investments in 2030 (Bi of R$2005) Mitigation invest./GDP in 2030

Study 1 even shows a strong employment dividend with an increase of net employment compared to the reference (Table 54). There are at least two explanations for the differences: (i) the perimeter of GHG gases is wider in study 1 and includes non-energy emissions, with lower mitigation costs; (ii) the multilateral implementation of climate policy enhances Brazilian competitiveness in Study 1, whereas unilateral action is detrimental to the competitiveness of Brazilian exports in Study 2. Mitigation options informed by bottom-up information may also be cheaper than in the top-down formulation of production and consumption trade-offs. Study 1 further explores the heterogeneity of sectoral impacts of mitigation scenarios (Figure 11, Figure 12). These results reflect the diversity of mitigation potential of the sectors as well as the difference of carbon intensities among domestic sectors and between domestic sectors and foreign industries. For instance, paper and steel industries are much less carbon-intensive in Brazil— because they are heavily reliant on bioenergy—compared to the rest of the world so that such industries gain competitiveness in the context of a global carbon price. It is also the case for the aluminium industry, which benefits from a very low-carbon domestic electricity (hydropower and other renewables). Without surprise, fossil fuel sectors record a decrease of domestic production and consumption to the benefit of bioenergy sectors that expand in the energy matrix. More interestingly, the impacts are very contrasted as far as industrial sectors are concerned. The cement sector is negatively 139

impacted whereas paper, steel and non-ferrous industries are better off in mitigation cases with superior production and net export levels.

Figure 11. Sectors output levels – Study 1 5.00 4.50 4.00 3,50 3.00 2.50 2.00 1.50 1.00 0.50 -

a iva s tiv Ele es ct ric Tr ity an s Ag por t Pu ricu ltu lp an re dP ap e Ce r m en t No St n- eel Fe rro Ch us em ic Re M al ini st n of Ind g us try Se rv ice s

il O

il D

al Co

Bio

BY-2005 CPG MA1-T MA2-T

Figure 12. Net export levels of selected industry sectors – Study 1 8 000 6 000 4 000 2 000 Pulp and Paper -2 000

Cement

Steel

Non-Ferrous

BY-2005 CPG MA1-T MA2-T

-4 000 -6 000

Study 2 compares the impacts of three contrasted recycling schemes of carbon revenues (Figure 13): (i) carbon revenues increase public budget, (ii) payroll tax rates are decreased in the same proportion for all economic sectors and (iii) total carbon revenues are transferred to households through a universal green cheque. One first interesting outcome is that the green cheque option leads to the smallest GDP loss amongst three options. One reason is that the green cheque promotes the lowest income groups which have lower savings rates in the model. Among others effects, this boosts final consumption and partially offsets the negative effect of the carbon tax over the economy.

140

Figure 13. Trade-offs between environmental, economic and social performance in Study 2 Real GDP 1.05 1 0.95

Decarbonisation index (scale 1/8)

Employment

0.9

Reference No recycling Payroll tax decrease Green cheque

0.85

Poverty alleviation (bottom class income - scale 1/3)

Equality index (GINI inverse)

Figure 14. Real disposable income variations from reference for the 6 income groups – Study 2 20% 15% 10%

Reference

No recycling

0% HH1 -5%

HH2

HH3

HH4

HH5

HH6

Payroll tax decrease Green cheque

-10% -15%

In the end all income groups—including the highest—are better-off compared to the payroll tax option, which may not be intuitive at first (Figure 14). Not surprisingly, the payroll tax scheme is the best towards employment and alleviates the carbon tax burden to reach an employment rate very close to the reference. In addition, the payroll tax scheme has a small distributive impact among income groups compared to the green cheque. This is linked to the steady share of labour income in total income across income groups in the model (Figure 15). Eventually the green cheque option has by far the highest social performance in terms of inequality reductions and poverty alleviation. 141

Figure 15. Sources of income at base year (2005) for the 6 income groups 100% 90% Other income

80% 70%

Public

60%

Private

50%

Financial income

40%

Land rent

30% 20%

Capital Income

10%

Labour income

0% HH1

HH2

HH3

HH4

HH5

HH6

As a conclusion, these two studies suggest that maintaining low emission levels in the medium term in Brazil seems feasible at low economic cost in a context of global climate action. Furthermore it seems there is also room for an inclusive low-carbon development pathway if carefully designed fiscal reforms include transfers towards the lowest income groups.

4. Conclusion: some insights from the comparison of the Brazilian and South African exercises The exercises on Brazil outlined above have been conceived and conducted independently from the South African study, for different research and policy analysis purposes. The Brazilian studies are also in large part ongoing. This limits the information that can be drawn from the comparison of the two. Still, some methodological and substantive remarks can be drawn from pulling together the two sets of studies, since they both share the same objective, namely exploring how environmental— and more specifically mitigation—objectives can be articulated with other key development goals, and since they both share the same methodology, namely the use of a prospective model derived from the IMACLIM family to explore plausible images of the future in a prospective framework. The first remark that can be made is that the key challenges at the energy–environment–development nexus in Brazil and in South Africa are strikingly different. In the latter, the key issue is to reduce high reliance on coal while maintaining economic growth, reducing unemployment and lessening inequalities. In the former, the key issue is to limit growth in emissions—from a comparatively very low base—, notably in the industry and energy sectors while developing domestic oil and gas production, lessening inequalities—from a comparatively lower level—, and ensuring steady growth per capita. Comparing the two exercises thus serves as a reminder that the articulation between decarbonisation and development objectives is extremely country-specific, even within the same group (here, the BRICS). This is an obvious statement, but with widespread implications for global mitigation 142

objectives, since the participation of a large number of countries—including notably the BRICS—is required to reach significant mitigation objectives at the global level, and since the participation of each country will crucially depend on the extent to which decarbonisation can be articulated with key domestic development objectives.[74] A corollary from the first remark is that the tools suited to analysing the articulation between environment and development objectives in each country also differ. For example, the importance of the agriculture and agro-industry sectors in Brazil requires not only considering those explicitly in the sector disaggregation of the CGE, but also to include land as a productive factor along with labour and capital. Including land did not appear to be necessary in first approximation in South Africa. Similarly, the importance of skill shortage in South Africa requires a more in-depth representation of the labour market. Though quite an obvious statement again, the need for country-specific tools is a challenge for the research and policy analysis community, in two ways. First, building relevant country-specific tools requires time, while the demand for policy analysis is growing fast.[75] Second, the need for cross-country comparison requires a careful balance between country specificities and general principles. It follows from the previous discussion that the policy packages able to jointly address environmental and development challenges are also likely to differ markedly from one country to the next. Precisely, if carbon pricing (typically in the form of a carbon tax) is on the research agenda both in South Africa (since it is envisioned in the White Paper on Climate Change) and in Brazil (as a possible outcome of the PNMC), carbon tax recycling modalities will differ. Similarly, (non-price) policies and measures in various sectors are considered (education in the South Africa study, land use in the Brazilian case). This has two major methodological consequences. First, economy-wide tools of analysis (such as CGEs) are necessary to capture both the diversity of these policy packages and the complexity of their economic implications. On the other hand, economy-wide tools are unlikely to be able to capture the sectoral details and implications of each measure within the policy packages. Hence the need to dialogue with detailed sectoral knowledge (be it in the form of expert knowledge, other models, etc.). The coupling of IMACLIM-BR with sectoral models in the first study is an illustration. Finally, the comparison between the Brazilian and South African cases also sheds light on an important limitation of both analyses, namely the lack of articulation between national studies and a global analysis of the interactions between national plans and objectives. If global models can be (and are) used to provide “limit conditions” to national exercises, such as trends in international energy prices, national decisions are likely to retrofit at the global scale. For example, the amount of oil and gas that Brazil ultimately releases on the international markets is likely to affect the international oil and gas prices, thereby modifying the conditions under which all other countries will undertake their mitigation efforts. The articulation between national and global scales (and even local scales if one considers, for example, the diversity of mitigation actions taken by cities) is a major challenge for future policy analysis and for the research on the global mitigation policies. [74] The “paradigm shift” at COP16 in Cancun towards “equitable access to sustainable development”, and the subsequent emphasis put on national mitigation proposals consistent with national circumstances (through the Nationally Appropriate Mitigation Action and now the Intended Nationally Determined Contributions) underlines the necessity that each country embed decarbonisation within its broader development agenda. [75] And is likely to grow even more rapidly if COP21, as expected, builds on the INDCs—with a signficant need, country by country, for ex ante and on-course assessments, and for the comparison of outcomes across countries.

143

Appendix references Empresa de Pesquisa Energética (EPE) 2013. Plano decenal de expansão de energia 2022. Empresa de Pesquisa Energética (EPE) 2014. Cenário econômico 2050. Empresa de Pesquisa Energética (EPE) 2014. Demanda de energie 2050. Grottera, C., 2013. Impactos de políticas de redução de emissões de gases do efeito estufa sobre a desigualdade de renda no Brasil, Rio de Janeiro: UFRJ/COPPE. Heaps, C., 2013. Long-range Energy Alternatives Planning (LEAP) system. [Software version 2012.0049]. Stockholm Environment Institute. Somerville, MA, USA. IAEA, 2007. MESSAGE. Model for Energy Supply Strategy Alternatives and their General Environmental Impacts. User Manual. International Atomic Energy Agency. June 2007. ICONE, 2014. Land-use model for Brazilian Agriculture – BLUM. Instituto de Estudos do Comércio e Negociações Internacionais – ICONE. IEA, 2013. World Energy Outlook 2013. IEA: Paris. Johansen, L., 1960. A multi-sectoral study of economic growth, North Holland, Amsterdam. La Rovere, E. L., C. Burle Dubeux, A. O. Pereira Jr, and W. Wills, 2013. Brazil beyond 2020: from deforestation to the energy challenge. Climate Policy 13:70–86. La Rovere, E. L., W. Wills, A. O. Pereira Jr, S. H. F. da Cunha, and C. Burle Dubeux, 2015. Mitigation scenarios for Brazil – 2030, preliminary version. COPPE: Rio de Janeiro. SAE (Secretariat for Strategic Affairs), 2013. The Middle Class in Numbers, SAE, Brasilia. Thissen, M., 1998. A classification of empirical CGE modelling. SOM Research Report 99C01. University of Groningen: Groningen.

145

List of abbreviations AFD

Agence Française de Développement

CIRED

Centre International de Recherche sur l’Environnement et le Développement

CTL

Conversion of coal to liquid fuels like gasoline, diesel and LPG

DoE (/RSA) Department of Energy (of the Republic of South Africa) ERC

Energy Research Centre of the University of Cape Town

ESKOM

South Africa’s main electricity producing and distributing utility company

Final Consumption in the national accounting sense

FCC

Fixed Capital Consumption

GFCF

Gross Fixed Capital Formation

GTL

Transformation of natural gas to liquid fuels like gasoline, diesel and LPG

Intermediate consumption in the national accounting sense

I-O

Input-Output table (see also TES)

LNG

Liquefied natural gas

Mtoe

Million tonnes of oil equivalent

NERSA

National Energy Regulator South Africa

Quarterly Bulletin

South Africa

SAM

Social Accounting Matrix

SAPIA

South African Petroleum Industry Association

SARB

South African Reserve Bank

SATIM

South Africa TIMES energy model

SNA

System of National Accounts

SU tables

Final Supply and Use Tables 2005 as presented by StatsSA

TEE

Current, financial, and capital accounts (Tableau Économique d’Ensemble)

TES

Input-Output table (Tableau Entrées-Sorties)

TIPP

National Tax on Petroleum Products (Taxe Interieure Sur Les Produits Petroliers)

VA Value-Added VAT

Value-Added Tax 147

List of tables Table 1. A parallel between 2011 South Africa and France . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 2. Characteristics of sectors according to sectoral disaggregation criteria . . . . . . . . . . . .

Table 3. Characteristics of sectors according to sectoral disaggregation criteria (continued) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 4. Correspondence of IMACLIM-SA to national accounts sectors . . . . . . . . . . . . . . . . . . .

Table 5. 2035 savings rates assumptions, 5 household classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 6. Distribution of 2005 working-age population (15 to 64 years old) by educational attainment and labour market status, thousand individuals . . . . . . . . . . . . . . . . . .

Table 7. Classification of job types by skill level and corresponding educational attainment in the SAM 2005 and IMACLIM-SA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 8. Population by educational attainment and IMACLIM-SA skill, 2005 . . . . . . . . . . . . . . .

Table 9. Population by educational attainment in IMACLIM-SA’s Low Educational Progress (LEP) 2035 population scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 10. Active population by skill level in 2005 and 2035, CEA approach . . . . . . . . . . . . . . . . .

Table 11. Distribution of active population across household classes according to skill, 2005 base year and all 2035 projections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 12. Share of skills in the labour content of household classes’ consumption, 2005 (BY) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 13. Sectoral employment, base year (BY) and reference projection (RP) . . . . . . . . . . . . .

Table 14. Skill disaggregation in BY and RP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 15. Change in volumes of resources (Y, M) in RP (2035) as percentage of the respective resources and uses in BY (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 16. Uses variations in volumes from BY (2005) to RP (2035) . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 17. GDP shares of VA by sector in 2005 and 2035 RP. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 18. BY (2005) resource structure of sectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 19. RP (2035) resource structure of sectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 20. Change in physical intensities of production by sector from BY to RP . . . . . . . . . . .

Table 21. Population, total gross disposable income and per capita incomes per class and the composition of income for BY and RP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

53 149

Table 22. Per capita consumption of household classes, RP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 23. Government revenue and expenses in 2005 and in 2035 for the Reference Projection (RP) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 24. 2035 carbon tax levels explored in policy scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 25. Key outcomes for tax revenue recycling scenarios with carbon tax levels of ZAR2005 100/tCO2 and ZAR2005 300/tCO2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 26. Ex ante sectoral price impact of 100 ZAR2005/tCO2 per energy product . . . . . . . . . .

Table 27. Ex ante price impact of 300 ZAR2005/tCO2 per energy product by consuming sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 28. Employment by sector for BY, RP and 4 policy scenarios . . . . . . . . . . . . . . . . . . . . . . . . .

Table 29. Employment by skill level for BY, RP and 4 selected 2035 C tax scenarios . . . . . . .

Table 30. Changes of resource volumes versus BY for RP and 4 C tax scenarios . . . . . . . . . . .

Table 31. Sectoral shares in total Value-Added, BY, RP and 4 policy scenarios . . . . . . . . . . . . . .

Table 32. Changes in volumes of uses compared to BY of IC, Household FC and Exports for RP and four selected carbon tax policy scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 33. Composition of GDP by different uses for BY (2005), and for RP and four selected carbon tax policy scenarios in 2035 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 34. Changes in factor intensities vs. BY for RP and 4 selected policy scenarios . . . . . . .

Table 35. Gross disposable income (after taxes) per household class and per capita, for RP and four selected carbon tax scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 36. Sensitivity analysis of the Reference Projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 37. Sensitivity of R2+ results to impacts on productivity growth . . . . . . . . . . . . . . . . . . . . .

Table 38. Final hybrid I-O table for Base Year 2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

106

Table 39. Quantities of uses and resources at base year 2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

108

Table 40. Base Year agent specific income distribution (top) and expenditures (bottom) . .

109

Table 41. South Africa current and financial accounts 2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

110

Table 42. Data need for current and financial accounts for IMACLIM, availability from SAM 2005, and correspondence with I-O tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

111

Table 43. Demography and distribution of actives, employed and broad unemployed by skill level and household class for the Base Year* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

113

Table 44. RP demography and distribution of actives by skill level and household class for 2035* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

114

Table 45. Estimations of KLE elasticities of substitution by Van der Werf (2008) and Okagawa and Ban (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

115

Table 46. Nested-CES production function elasticities of IMACLIM-SA sectors. . . . . . . . . . . .

116

Table 47. Nested-CES household consumption elasticities of IMACLIM-SA . . . . . . . . . . . . . . . .

116

150

Table 48. Terms-of-trade elasticities of exports and imports of IMACLIM-SA . . . . . . . . . . . . .

117

Table 49. Growth in jobs of RP compared to BY by sector and skill . . . . . . . . . . . . . . . . . . . . . . . .

119

Table 50. Main settings and key outcomes for RP and all policy scenariosRuns . . . . . . . . . . . . .

120

Table 51. Parameters and variables of the SATIM runs used to derive technical coefficients for the ELC sector in IMACLIM-SA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

122

Table 52. Variables of selected IMACLIM-SA runs for comparison to parameters and variables from SATIM runs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

123

Table 53. Scenarios – Study 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

133

Table 54. Mitigation measures included in Scenarios M1 and M2 of Study 1 . . . . . . . . . . . . . . . .

134

Table 55. Macroeconomic results – Study 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

139

151

List of figures Figure 1. Schematics of the hybridisation procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 2. Nested production function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 3. Households’ consumption decision tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 4. Growth and unemployment projections under constant educational attainment (CEA) definition of skills . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 5. Main performance indicators, base year (BY) and reference projection (RP) . . . . . .

Figure 6. GDP shares of uses in BY and RP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 7. Principles of alignment of material balances and monetary flows . . . . . . . . . . . . . . . . .

103

Figure 8. Total GHG emissions of Brazil, 2015–2030 per sector in Scenario B . . . . . . . . . . . . . .

130

Figure 9. Energy-related GHG emissions in Brazil per sector, 2015–2030, in Scenario B . . . .

131

Figure 10. GHG Emissions (Mt CO2 eq.) - Study 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

138

Figure 11. Sectors output levels – Study 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

140

Figure 12. Net export levels of selected industry sectors – Study 1 . . . . . . . . . . . . . . . . . . . . . . . . .

140

Figure 13. Trade-offs between environmental, economic and social performance in Study 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

141

Figure 14. Real disposable income variations from reference for the 6 income groups – Study 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

141

Figure 15. Sources of income at base year (2005) for the 6 income groups . . . . . . . . . . . . . . . . .

142

152