Forecasting the Economic Growth Impacts of Climate Change in South Africa in the 2030 and 2050 Horizons

Abstract

In this paper, we estimate the effects of climate change by means of the systems generalised method of moments (System GMM) using panel data across South African municipalities from 1993 to 2016. We adapt the estimates to the municipal economic structures to forecast losses at the municipal level for the 2030 and 2050 horizons. The projections show that, relative to the 1995–2000 levels, South Africa’s economy would lose about 1.82 billion United States dollars (USD) on average due to climate change following the Representative Concentration Pathway (RCP) of 4.5 Wm⁻² radiative forcing scenario, and USD 2.306 billion following the business-as-usual (BAU) scenario by 2030. By 2050, the losses will be USD 1.9 billion and USD 2.48 billion, respectively. The results vary across municipalities depending on geographic location and sectors. Natural resources and primary sectors are the most impacted, while the economic losses are more than the gains in almost all municipalities in South Africa. This has a significant bearing on sustainable poverty reduction in South Africa through pro-poor industrialisation. The implication of the findings is discussed in the paper’s conclusion.

1. Introduction

South Africa’s National Development Plan (NDP) set an ambitious target of reducing unemployment from 24.9% in 2012 to under 6% in 2030, requiring a real economic growth rate above 2% and a nominal rate of about 7%. However, climate change will likely add a significant dent to these prospects. South Africa’s socio-economic landscape is still marred by the triple challenge of unemployment, poverty and inequality. Since the NDP’s launch, South Africa’s economy has grown rather sluggishly, following a weak global economic climate due to recent financial crises. However, as the developed world and other emerging markets are prospecting for a somewhat positive economic outlook, South Africa’s outlook has remained bleak due to internal governance issues. Although investor confidence seems to be slowly returning following the early 2018 change in government, climate change is likely to affect South Africa’s projected economic performance.

South Africa is one of the major examples of climate change’s adverse impact. Most of the country has warm, sunny, cool nights, with rainfall occurring most often during the summer months (November to March). However, around the Cape of Good Hope in the southwest, rainfall occurs in winter (June to August), and temperatures in that region are influenced by variations in sea level elevation. At the Intergovernmental Panel on Climate Change’s (IPCC) request, the climate science community developed a set of new guidelines for climate scenario projections known as the Representative Concentration Pathways (RCPs). Van Vuuren et al. [1] also reviewed the main RCPs commonly used in climate science. Four RCPs define a specific emission trajectory and ultimate radiative forcing. Radiative forcing is the ability of a scenario to significantly offset the net balance of energy into and out of the Earth’s atmospheric system, measured in watts per square metre (Wm⁻²). The pathways respectively lead to radiative forcing levels of 8.5, 6, 4.5 and 2.6 Wm⁻² by 2100. The baseline trajectory is the RCP8.5, which assumes high population, low economic growth and high energy intensity with little technological advancement. The RCP2.6 is an ambitious policy scenario leading to low levels of forcing. At the current levels of emissions-cutting commitments, this scenario is implausible. The two intermediary trajectories are the RCP6 and RCP4.5, but the RCP4.5 is considered the modal scenario, with most mitigation policy projections leading to it. In this work, we adopted the baseline and the modal mitigation scenario 4.5.

South Africa’s average temperatures are expected to rise significantly over the rest of this century. The RCP8.5 series shows an above 5-degree increase by the end of the century relative to the 1990–2000 average temperatures. The mitigation scenario that yields RCP4.5 would result in a maximum 2-degree increase in average temperatures by 2100. Over the 2050 forecast horizon considered in this work, the RCP8.5 will result in a 2.54-degree increase compared with a 1.64-degree increase in the RCP4.5 scenario.

These projected changes are likely to affect South Africa’s economic output to the extent that output responds to temperature changes. One feature of the assessment of climate change’s impact on the economy is the complexity of the climate–economy relationship. Various sector-specific studies around the world and in South Africa suggest climate change has significant effects on agriculture [2], ocean fisheries, access to fresh water, migration, tourism and other factors [3]. However, there is less emphasis on climate change’s direct link to economic growth, especially productivity. Ultimately, temperature changes can affect human capital through health [4], crime [5] and conflict [6]. Extreme events can also erode physical infrastructure, all of which affect economic activities directly or indirectly. Most of South Africa’s productive sectors have significant exposure to climate risks.

South Africa is an emerging economy with high dependence on extractive resources such as coal, copper, gold and platinum. The services sector is well-developed, and financial services dominate the sector, with a stock exchange ranking among the top 20 in the world and the highest in Africa. The 2016 population was estimated at 55 million, with a gross domestic product (GDP) per capita estimated at USD 13.4 in 2017. However, unemployment and inequality in South Africa are among the highest in the world, with a Gini coefficient between 0.66 and 0.70. The unemployment rate is around 27% of the workforce, with a higher percentage among the black youth. The country’s poverty headcount in 2017, estimated using 2015 data, was above 50% [7].

In order of magnitude, four main sectors contribute to the South African value addition, as shown in

Table 1. Trend in South Africa’s economic structure by value addition.

	1993–1999	2004–2000	2005–2009	2010–2014
Agriculture	2.48%	2.36%	2.12%	2.19%
Forestry	0.46%	0.44%	0.42%	0.43%
Fishery	0.15%	0.12%	0.11%	0.13%
Mining	14.96%	12.96%	10.58%	8.74%
Manufacturing	15.90%	16.21%	15.88%	15.09%
Electricity and gas	2.60%	2.41%	2.37%	2.08%
Water	0.81%	0.68%	0.60%	0.64%
Services	62.65%	64.82%	67.91%	70.70%
Total value-added (USD Billions)	696.892	603.135	738.774	802.235

Source: Computed by authors using data from Quantec.

. The largest sector is the services sector, followed by the manufacturing, mining and agricultural sectors. All these sectors are likely to be affected in one way or another by climate change through its direct impact on agriculture, fishery, forestry and water but also indirectly through its impact on physical and human capital inputs to production.

Based on existing literature on climate change’s impact on economic growth, viewed from the country-level perspective, this study attempted to investigate climate change’s impact on economic growth in South Africa at the municipal and provincial levels. Contrary to the approach adopted by this study, Akram and Hamid [8] investigated the impact of climate change on Pakistan’s economic growth with temperature as a proxy for climate change. The authors found that climate change has a negative and significant relationship with GDP and productivity at the sectoral level, including agricultural, manufacturing and services, with strong devastating effects on the agriculture sector. A similar country-level study conducted in Brazil by Tebaldi and Beaudin [9] showed that climate events could enhance inequality, with significant adverse rainfall variations affecting the GDP growth rate of the poorest regions. In addition to this finding, Brown et al. [10] argued that heavy precipitation (drought) is the most significant climate event that impacts GDP growth in sub-Saharan African (SSA) countries, whereas temperature variability had significant effects in some regions. Henceforth, precipitation, as a proxy of climate change events, is likely to produce more significant results in SSA countries rather than temperature as a proxy. However, it is critical to include both climate variables for the robustness analysis. Moreover, Abidoye and Odusola [11] found evidence that an increase of one-degree Celsius in temperature reduces GDP growth by 0.67%, significantly negatively impacting economic growth in Africa. Furthermore, they argued that the long-run temperature variability affects long-run economic growth in this part of the world. In addition to Brown et al. [10] arguing that temperature variability only has a significant impact on some African regions, Van Vuuren et al. [1] emphasised that higher temperatures substantially reduce economic growth in the poorest countries, while having little effect in the richest countries. Furthermore, they found evidence that higher temperatures affect developing countries by reducing agricultural output, industrial output, aggregate investment, and increasing political instability. Since South Africa is well known as a country with high inequality indicators among some rural municipalities, this finding will be a key guideline for further study.

Studies have examined potential interactions in the global economy based on climate change’s impact on productivity. Findings have shown that industrial transformation barriers can lead to an optimal path without taking up less climate-sensitive technology, decarbonisation, and stagnant growth. They emphasised that lower set-up costs can lead to an efficient growth path that incorporates less vulnerable and low carbon technology. These results are founded on the theory that the greater the carbon intensity of existing production, the more damaging accumulated emissions are for the growth, and the lower the rate of utility discounting increase the present value cost of long-term losses due to climate change.

In this section, we further provide an updated review of the economics of the problem and appraise appropriate literature on the empirical relationship between climate change and economic growth. Climate change affects economic growth through different channels, including output and productivity [12]. In the case of output and productivity channels, climate change reduces yield and productive capacity, respectively. The empirical research can be divided into three categories: (i) output channel effects, (ii) productivity channel effects, and (iii) evaluating both effects together. Researchers who examined the effect of temperature on aggregate economic activity include Fankhauser and Tol [13], Raddatz [14], Abidoye and Odusola [11], Alagidede et al. [15], Kahsay and Hansen [16], Rezai et al. [17], Liu et al. [18], and Talib et al. [19], among others. For example, Fankhauser and Tol [13] used a simulation approach to examine the impact of climate change on economic growth. The results revealed temperature’s negative impact on economic growth when capital accumulation and savings were controlled for. Dell et al. [20] examined the income–temperature nexus by using cross-sectional country and sub-national data at the municipal level for 12 countries in North and South America. Their findings determined that temperature has a negative impact on income both within and between countries, with the latter having a significant impact. As a result, the cross-country link between income and temperature is not driven by nations’ characteristics.

Alagidede et al. [15] examined the effects of climate change on economic growth for SSA countries using a long- and short-run effects panel cointegration econometric approach. The findings of that study revealed a non-linear relationship between real GDP per capita and temperature. Moreover, Burke et al. [21] found global support for non-linear temperature’s impact on economic growth, but the non-linear temperature effect vanished when they interacted with the developing country dummy. However, Dell et al. [22] found no evidence of non-linear temperature effects on economic growth in Africa, which was further confirmed by Burke et al. [21].

Liu et al. [18] investigated urban climate change–economic growth in the case of China over the period 2000–2015, using the coupling coordination degree (CCD) model. Their findings revealed the significant spatio-temporal heterogeneity of the CCD. Elshennawy et al. [23] also explored the nexus between climate change and economic growth, using an intertemporal general equilibrium analysis for Egypt. The model was used to simulate the effects of climate change on aggregate consumption, investment and income up to 2050. Their simulation analysis suggested that, in the absence of policy-led adaptation investments, real GDP towards the middle of the century will be 6.5% lower than in a hypothetical baseline without climate change.

Sheng et al. [24] examined the effects of climate risks (temperature growth and its volatility) on economic activity among a panel of US states by considering the role of uncertainty. The panel data set-up covered a monthly period from March 1984 to December 2019, applying impulse response functions (IRFs) from a linear local projections (LPs) model. The study determined that climate risk has a similar negative impact on economic activity, regardless of whether changes in temperature growth or volatility cause them. More crucially, using a non-linear LPs model, the IRFs show that the negative impact of climate risks is dependent on the states’ economic and policy-related uncertainty regimes, with the impact being much greater under higher rather than lower levels of uncertainty. Zhang et al. [25] also found something similar to Sheng et al. [24] when examining extreme climate events and economic impacts in China. However, their study applied and embedded the new damage function into the Cobb–Douglas production function of a computable general equilibrium (CGE) model and simulated future climatic losses on industrial economic systems. Copiello and Grillenzoni [26] further investigated the causality between economic development and climate change. Their study applied ARIMAX-ARCH models and found evidence for bidirectional causation between human growth and climate change. In addition, Olper et al. [27] examined the nexus between weather, climate and economic outcomes for the case of Italy over the period 1980–2014 for a panel of 110 provinces. They determined the statistically significant effects of temperature on both the GDP and agricultural reaction function.

Climate change has far-reaching consequences, and focusing solely on economic growth understates the actual economic cost of climate change. As a result, other researchers looked into the impact of climate change on agricultural production. Barrios et al. [28], Kahsay and Hansen [16], Adom et al. [29], Adom and Adams [30] and Talib et al. [19] all used agricultural value-added to look at the output channel of climate change. Higher temperatures, according to these studies, diminish agricultural yield.

Another important area of inquiry is temperature/climate change’s impact on productivity. For example, Hsiang [31] examined climate change’s impact on economic growth and labour productivity. The author claims that fluctuations in temperature have a comparable effect on aggregate output and productivity. The findings revealed that, compared to losses in the agricultural sector, climate change causes greater output losses in the non-agricultural sector. As a result, ignoring climate change’s effect on non-output or non-agricultural sectors, according to the author, understates the full cost of climate change. In support, Gosling et al. [32] estimated climate change’s regional and global effects on labour productivity. They showed that climate change reduced labour productivity by 2 to 21%, with a higher degree of variation among countries. Mexico, Central America, the Caribbean and South Africa suffered the greatest losses, while the UK, Ireland and northern Europe suffered the least. Martínez-González et al. [33] explored climate change’s inclusion in general debates by focusing on one of the colder periods of the last 500 years, known as the Maunder Minimum. The results of that study suggest that climate change and the resulting adaptations may have influenced the start of the English Agricultural Revolution, the Energy Transition and the European Divergence. Adom and Amoani [12] also explored whether climate change’s effect on economic growth and political stability (measure for productivity growth) depends on climate adaptation readiness, using data from 44 African countries. Their findings revealed that increases in temperature exert significant negative effects on economic growth and productivity, but these effects critically depend on the level of adaptation readiness.

To summarise, the following observations can be made based on previous studies. To start with, studies forecasting climate change’s impact on economic growth in South Africa in the 2030 and 2050 horizons are nowhere to be found to the best of our knowledge. Secondly, previous studies have failed to consider the growth effect of climate change at the municipal, provincial and national levels for different economic sectors in South Africa. Thirdly, previous studies have failed to apply the econometric approaches used in this study for the topic under investigation. This study addresses this research gap by forecasting the economic growth impacts of climate change on South Africa in 2030 and 2050 using panel data from 1993 to 2016. South Africa has the leading economy in Africa and is among the top CO₂ emitters (1.09 of the world emissions)—the country is ranked 15th globally [34]. It is therefore important to forecast climate change’s growth impact for the purpose of avoiding future demerits associated with climate change through relevant policy recommendations.

This study assessed climate change’s effect on South Africa’s economic output, captured by long-term changes in average temperatures. The estimated effects were then used together with projected future temperature scenarios for South Africa to forecast possible economic losses due to climate change. We adapted provincial-level estimates to the municipal economic structure in order to forecast losses at the municipal level for the 2030 and 2050 horizons. On this basis, we sought to forecast the economic growth impacts of climate change in South Africa in the 2030 and 2050 horizons.

Many scientific agencies and a growing chorus of environmental groups and governments are advocating for drastic reductions in CO₂ emissions, claiming that future generations’ interests are at stake. This study is important because the South African government, climate and yield modellers, among others, can use this study’s forecast to control fossil fuel burning or prepare other responses that may help mitigate the negative effect of climate change. Economic growth forecasting is an integral part of supporting public and private entities’ decision making. Economic forecasting considerations are utilised to explain decisions and policy actions. In other words, economic forecasting can be considered a set of arguments favouring a particular economic growth trajectory and its various deviations. It allows one to verify the reliability of economic forecasts to formulate economic policy. It also helps policymakers check economic forecasts’ accuracy before making policy decisions. The forecasting result from this study may increase the warning time for climatic hazards and recommend how this information may be used most effectively to minimise risk. The emerging ability to make timely, skilful climate forecasts could reduce different sectors’ economic vulnerability to the impact of climate variability. To that end, improved decision making is promoted to either prepare for expected adverse conditions or take advantage of expected favourable conditions, hence the rationale for this study.

2. Methodology

There are various modelling approaches to explore the effects of climate change. These approaches can also be applied at various levels, in the micro-local area, within-country sub-regional levels and at national and cross-country levels. A choice must also be made regarding which economic sectors to focus on. Several studies have paid attention to the agricultural sector because of its direct exposure to climatic factors. However, because of the inter-sectoral linkages of the economy, climate change can have direct and indirect effects on the economy. Each approach has its strengths and weaknesses. For example, examining effects that encompass direct and indirect effects may require the use of the computable general equilibrium methodology (CGE). However, this usually relies on social accounting matrix data that are typically obsolete. The CGE method is also very data-intensive and may lead to implausible assumptions in the absence of required data.

In this study, we were interested in assessing effects at the municipal, provincial and national levels for different economic sectors in South Africa. We opted for an econometric and forecasting methodology that could optimise the available data, while catering for certain biases that may arise due to data issues. These are discussed below.

2.1. Econometric Modelling

Our econometrics method followed the panel form specified by Colacito et al. [35] for US states, later used by Tebaldi and Beaudin [9] for Brazil. We adapted the panel for South African municipalities from 1993 to 2016. Economic growth is captured by changes in the aggregate output or real GDP, denoted as Y. Inputs into the function are labour (L) and capital (K), with labour–capital augmenting technology (A). The model is specified for municipality $i$ at time $t$ according to Equation (1) below:

$$Y_{it}={{\displaystyle e}}^{{{\displaystyle ({\beta }_1{T}_{it}+{\beta }_2{R}_{it}+{\beta }_{3}SP{T}_{it}+{\beta }_{4}SP{R}_{it}}}^{})}{A}_{it}{K}_{it}^{\theta }{L}_{it}^{\tau }$$

(1)

where $\beta \mathrm{and} \left(\tau =1-\theta \right)$ are parameters.

Climate variables temperature (T), rainfall (R) and respective spatial effects terms (SPT and SPR) affect total factors productivity exponentially.

At equilibrium, Equation (2) holds:

$$\frac{\Delta A_{it}}{A_{it}}=g_{it}+{\delta }_1{T}_{it}+{\delta }_2{R}_{it}+{\delta }_{3}SP{T}_{it}+{\delta }_{4}SP{R}_{it}$$

(2)

where $\delta$ represent the steady-state coefficient that measured the long-term growth and $gi$ is the short-term growth. Where $g_{it}$ is the growth rate of per capita output, $\beta$ is the level effect of climate shocks on output, and $\delta$ is the growth effect of climate shocks. Expressing (1) in natural logs of per worker terms and taking the derivative with respect to time yields:

$$\Delta \ln y_{it}=\Delta \ln A_{it}+{\beta }_1\Delta T_{it}+{\beta }_2\Delta R_{it}+{\beta }_3\Delta SP{T}_{it}+{\beta }_4\Delta SP{R}_{it}+\delta \Delta \ln k_{it}$$

(3)

Equation (2) is an expression of change in technology, or total factor productivity. Replacing the term in (3) with its expression in (2) and rearranging it gives:

$$\begin{array}{l}{g}_{it}=g_i+({\beta }_1+{\delta }_1)T_{it}+({\beta }_2+{\delta }_2)R_{it}+({\beta }_3+{\delta }_3)SP{T}_{it}+({\beta }_4+{\delta }_4)SP{R}_{it}\\ \begin{array}{ccc}& & -({\beta }_1\end{array}{T}_{it-1}+{\beta }_2{R}_{it-1}+{\beta }_{3}SP{T}_{it-1}+{\beta }_{4}SP{R}_{it-1})+\alpha \Delta \ln k_{it}+{\epsilon }_{it}\end{array}$$

(4)

Or

$$\begin{gathered} g_{it}=g_i+{\gamma }_1{T}_{it}+{\gamma }_2{R}_{it}+{\gamma }_{3}SP{T}_{it}+{\gamma }_{4}SP{R}_{it}-\left({\beta }_1{T}_{it-1}+{\beta }_2{R}_{it-1}+ \\ {\beta }_{3}SP{T}_{it-1}+{\beta }_{4}SP{R}_{it-1}\right)+\alpha \Delta k_{it}+{\epsilon }_{it} \end{gathered}$$

(5)

where lowercase letters and natural logs of the corresponding variables and $\gamma =\beta +\delta$ is a combination of the short-run and long-run effects. In Equation (5), ${\beta }_1$ is the short-run effect of temperature changes on output growth, while $\gamma -\beta$ is the long-run coefficient. Equation (5) is estimated at the provincial level for different economic sectors—manufacturing, mining, agriculture, forestry, fishery, electricity and gas and water. Municipalities under each province share the same coefficient estimates for the different sectors but vary in terms of their shares of these sectors in the GDP.

2.2. Variables and Data

Despite relatively low carbon dioxide (CO₂) emissions (0.97% of the world’s total emissions), climate change has already cost South Africa quite a fortune due to various floods and droughts. South Africa is located at 22o–34o latitude and 16o–32o longitude. Administratively, it is divided into nine provinces: the Eastern Cape, the Free State, Gauteng, KwaZulu-Natal, Limpopo, Mpumalanga, the Northern Cape, the North West and the Western Cape. According to Benhin [36], the country is divided into four main climatic zones. The first zone is the desert, covering a good part of the Northern Cape and the northeastern part of the Western Cape. The arid zone spans Limpopo, Mpumalanga, North West, Free State, western KwaZulu-Natal and Eastern Cape and the northern areas of the Western Cape. The sub-tropical wet zone covers the coastal areas of Kwazulu-Natal and the Eastern Cape. Finally, the Mediterranean or winter rainfall area spreads over the southwestern coast of the Western Cape. South Africa’s climate is generally dry with sunny days and cool nights, with rainfall mostly occurring during the summer period and most often during winter. Temperatures are more influenced by variations in elevation, with the sea level being higher in Cape Town and Durban. Bloemfontein appears to be the coldest city, with winter temperatures below −3 °C. According to South African Weather Services [37] the average temperature in Cape Town was 17 °C in 2002, yet this average increased to 19.4 °C in 2014 [38]. Winter temperatures, on average, vary across the country between 5 °C and 12 °C, with lower temperatures occurring in the Eastern Cape. The highest maximum summer temperature is recorded in the Northern Cape and Mpumalanga (49 °C), whereas the coolest temperatures between 0° and −2 °C are reported in the Western Cape [39].

The successful estimation of the true coefficient of climate change’s impact on economic growth requires six variables adapted to fit the underlying econometric model. These are GDP or GDP growth rates, capital per worker, annual average temperatures, annual average rainfall and spatial dependence indicators for temperature and rainfall.

The main climate-related variables (temperature and rainfall) are obtained using a variable resolution global climate model (GCM) developed by the Commonwealth Scientific and Industrial Research organisation. Two of the six representation concentration pathways were adopted (RCP4.5 and RCP8.5), downscaled to 50km resolution. The simulations cover the period from 1960 to 2100, with RCP4.5 and RCP8.5 capturing high and low mitigation scenarios, respectively. The simulations were undertaken by a team of experts from the Centre for High-Performance Computing (CHPC) at the Meraka Institute of the Council for Scientific and Industrial Research (CSIR) in South Africa. The specific GCM’s ability to simulate present-day southern Africa’s climate has been well demonstrated in existing literature [1]. The simulations were necessary for the predictions of future climates for our forecast horizon. However, the data used to estimate the econometric model were based on historical figures of actual observed climate data supplied by the CSIR.

The economic growth indicator is computed using value-added for different economic sectors at the municipal level in South Africa. Capital per worker is taken as the ratio of gross fixed capital formation to employment for different economic sectors at the municipal level. Value-added figures, together with capital and labour, were sourced from Quantec. The Quantec EasyData service provides online access to South African and international economic data covering macroeconomic, regional, industry and trade data.

A great deal of econometric bias can arise when one ignores the effect of cross-municipality dependence. In order to effectively isolate the effect of climate change, it is necessary to control and isolate the impact of neighbouring municipalities. We generated spatial effect variables for temperature and rainfall by computing a weighted matrix of temperature and rainfall, where the weights were proportional to the geographic distance between two municipalities. The econometric model allows for the isolation of short-run and long-run effects. For this reason, we introduced one period for each of our explanatory variables to produce the autoregressive model in Equations (4) and (5).

2.3. Estimation Technique

The econometric model is estimated using a technique known as the System’s Generalised Method of Moments (System GMM). This technique was most appropriate for the kind of data we were using for two reasons. Firstly, it is able to correct for the bias introduced by the problem of double causation, known in econometrics as endogeneity. This often arises when a model is specified with some omitted relevant variables, in addition to the fact that the variables in the model might mutually cause one another in a statistical sense. Secondly, this technique can exploit municipality-specific information to improve the estimated coefficient of the effect of climate change.

3. Empirical Results and Discussion

3.1. Short-Run and Long-Run Results and the Forecasting Technique

From the estimated coefficients, the long-run (L-R) and short-run (S-R) effects were computed, and the values are reported in

Table 2. Long-run and short-run coefficients based on the regression output.

		EC	FS	GT	KZN	LIM	MP	NW	NC	WC
Manu.	L-R coef	−0.021	−0.009	−0.004	−0.031	0.006	−0.004	−0.004	−0.014	0.009
	S-R coef	0.000	−0.019	−0.015	−0.011	0.017	0.015	−0.017	0.000	0.054
Mining	L-R coef	−0.073	−0.081	−0.065	−0.089	−0.099	−0.076	−0.090	−0.025	0.052
	S-R coef	−0.032	−0.083	−0.068	−0.047	−0.081	−0.081	−0.082	−0.025	0.068
Serv.	L-R coef	−0.008	0.003	−0.006	−0.012	−0.004	0.003	0.008	0.000	−0.014
	S-R coef	0.007	0.011	0.008	−0.001	0.015	0.018	0.010	0.019	−0.003
Agric.	L-R coef	0.026	−0.007	0.021	0.055	−0.030	0.094	0.068	0.046	−0.055
	S-R coef	0.013	0.035	0.070	0.098	0.080	0.184	0.101	0.083	−0.069
Forest.	L-R coef	0.025	0.025	0.037	0.040	0.081	0.033	0.042	0.048	−0.042
	S-R coef	0.046	0.079	0.093	0.074	0.170	0.091	0.064	0.060	−0.029
Fishery	L-R coef	0.021	0.025	0.012	0.059	0.062	0.036	0.042	0.057	−0.054
	S-R coef	0.042	0.076	0.074	0.097	0.180	0.089	0.083	0.066	−0.057
Elec. and Gas	L-R coef	−0.035	−0.005	0.006	−0.035	0.003	0.022	−0.008	0.026	−0.025
	S-R coef	0.035	0.075	0.067	0.053	0.114	0.121	0.057	0.075	0.010
water	L-R coef	0.024	0.021	0.017	0.016	0.045	0.009	0.039	−0.058	−0.076
	S-R coef	0.017	−0.003	0.007	−0.027	0.011	0.006	0.029	0.069	0.115

Note: S-R and L-R stand for short-run and long-run, respectively. The different column headings abbreviate the names of provinces of South Africa: EC is Eastern Cape; FS, Free State; GT, Gauteng; KZN, KwaZulu-Natal; LIM, Limpopo; MP, Mpumalanga; NW, North West; NC, Northern Cape; and WC, Western Cape.

. We used long-run coefficients for forecasting due to our forecasting model’s long time horizon (2030 and 2050). For the purpose of forecasting, the model was estimated at the provincial level, where sub-samples were drawn for the municipalities of a specific province. The provincial-level coefficients for each economic sector were allocated to each municipality within a given province.

The forecasting technique shocks the fitted model with temperature changes in 2030 and 2050 relative to the 1995–2000 average levels. The forecast is based on the underlying production function from which Equation (2) is derived. The equation links output levels or real value-added to climate variables and other inputs. In order to calculate the future impact of climate change, we held the effects of other variables constant. As such, the projected impact of climate change is simply the long-run coefficients in Table 2, times the average real value-added for the reference base period of 1990–2000, and the change in temperature from the reference period to the projection horizon of 2050. It is important to note that the projection horizon is taken as a 10-year average from 2041 to 2050. Hence, the temperature change is the difference between the 2041–2050 mean and the 1995–2000 mean.

Figure 1 reports the evolution of mean deviations from mean temperatures of the period 1990–2000 based on the baseline RCP8.5 and the plausible mitigation scenario RCP4.5. It is evident from the graph that the temperature cost of inaction is significant. However, the gap between the two scenarios in the 2050 horizon is less pronounced than in the latter part of the century. In this respect, our projections of the economic costs of climate change are conservative, given the significant gap between RCP4.5 and RCP8.5 beyond 2050. The projected economic losses are nonetheless significant.

3.2. Robustness Check

Before discussing the forecast of climate effects, we first undertook a robustness check of the estimated model’s forecasting performance. To do this, we used the estimated coefficients to conduct a within-sample forecast for the span of the data. The forecast series were compared with the t-test and fitted scatter plots. The scatter plot (The plots are not reported here for the sake of space) points are all within the 95% confidence interval. The results of the t-test are reported in

Table 3. T-test on mean difference between the forecast and actual values.

		Mean	SE	T-Stat (p-Value)	DECISION
Overall	Actual	0.025	0.001	1.614 (0.107)	No statistical diff. in mean
	Forecast	0.024	0.001
Eastern Cape	Actual	0.021	0.001	0.227 (0.821)	No statistical diff. in mean
	Forecast	0.021	0.001
Free State	Actual	0.033	0.003	0.801 (0.423)	No statistical diff. in mean
	Forecast	0.031	0.003
Gauteng	Actual	0.015	0.003	−0.003 (0.998)	No statistical diff. in mean
	Forecast	0.015	0.201
KwaZulu-Natal	Actual	0.027	0.001	1.221 (0.223)	No statistical diff. in mean
	Forecast	0.025	0.002
Limpopo	Actual	0.025	0.002	−0.634 (0.526)	No statistical diff. in mean
	Forecast	0.027	0.002
Mpumalanga	Actual	0.022	0.002	0.663 (0.507)	No statistical diff. in mean
	Forecast	0.021	0.002
North-West	Actual	0.023	0.002 *	1.890 (0.060)	Significant mean diff. at 10%
	Forecast	0.018	0.003
Northern Cape	Actual	0.023	0.002	0.900 (0.369)	No statistical diff. in mean
	Forecast	0.021	0.003
Western Cape	Actual	0.033	0.002	1.092 (0.275)	No statistical diff. in mean
	Forecast	0.032	0.003

Note: * denotes statistical significance at 10%.

.

At the national level, the forecast approximates the actual series relatively more precisely. The municipal-level analyses are performed using coefficients generated at the provincial level. Hence, in addition to the overall forecast, we also undertook forecasts for each provincial panel. Overall, eight of the nine provinces performed very well, with their respective forecasts closely mimicking the actual panels. The hypothesis of zero mean difference between the forecast and the actual was rejected at a 10% significance level in the North-West province. However, we still used the estimated coefficients to perform out-of-sample forecasts for municipalities in the North-West province since the mean difference is only 0.005. We could not treat the province differently as the relatively weak forecast power stemmed from the fact that there were fewer observations for that province. We were still comforted by the fact that the mean difference for this province is significant only at the 10% level. However, we raised the possibility that the results of the forecasts will be weaker for the municipalities and towns in the North-West province. With this confidence in the model’s performance, we proceed with the discussion of the forecast effects below.

The results show that climate change will have a significant drag on economic expansion in South Africa by 2030 and 2050. There is a very high economic cost to doing nothing about climate change (RCP 8.5); even the best plausible mitigation scenario (RCP 4.5) still yields significant economic losses by 2030 and 2050. In what follows, we first present the national picture by various sectors, comparing both the plausible mitigation scenario (RCP 4.5) with the business-as-usual (BAU) scenario (RCP 8.5). Thereafter, the provincial results are discussed. We end the discussion with interesting local (municipality-level) findings.

Our projections show that South Africa’s economy will lose significant output in GDP relative to the 1995–2000 levels. This amounts to about 28.4 billion (These values have not taken discount rates into account. They will be a lot higher with appropriate discount rates applied) local currency on average due to climate change following the RCP 4.5 scenario, and 35.7 billion following the BAU scenario by 2030. By 2050, the losses will be 29.3 billion and 38.1 billion, respectively. The year-by-year national percentage losses are plotted in Figure 2.

Table 4. National average effects of economic losses of climate change.

	Absolute Changes in Millions				Percentage Changes
Sector	2030 Horizon		2050 Horizon		2030 Horizon		2050 Horizon
	RCP4.5	RCP8.5	RCP4.5	RCP8.5	RCP4.5	RCP8.5	RCP4.5	RCP8.5
Elec. and gas	−3408.38	−4390.52	−3408.38	−4390.52	−10.35%	−12.89%	−10.35%	−12.89%
Forestry	−370.53	−449.56	−370.53	−449.56	−6.79%	−8.47%	−6.79%	−8.47%
Agriculture	−1654.77	−2140.36	−1654.77	−2140.36	−5.70%	−7.14%	−5.70%	−7.14%
Fishery	−3.54	−5.6	−3.54	−5.6	−5.97%	−7.26%	−5.97%	−7.26%
Manufacturing	−3015.94	−3610.42	−3887.22	−6032.67	−2.25%	−2.77%	−2.89%	−4.16%
Mining	−84.89	−12.2	−84.89	−12.2	−1.50%	−1.94%	−1.50%	−1.94%
Services	−20,085.33	−25,358.62	−20,085.3	−25,358.6	−1.98%	−2.49%	−1.98%	−2.49%
Water	214.14	281.6	214.14	281.6	1.78%	2.29%	1.78%	2.29%
All	−28,409.24	−35,685.66	−29,280.5	−38,107.9	−4.10%	−5.08%	−4.11%	−5.19%

shows the average national economic losses/gains attributed to climate change in South Africa, comparing the 2030 and 2050 horizons. Apart from the manufacturing sector, there are no significant differences between the losses in 2030 compared to 2050. We followed the 2030 projection scenarios for the rest of this work.

Overall, there is a one percentage point difference between the plausible mitigation scenario and BAU. The main sectors with the highest percentage losses are electricity and gas, forestry, fishery and agriculture. South Africa would forfeit up to 12.9%, 8.5%, 7.3% and 7.1% of its value-added, respectively, to climate change if no mitigating action is taken. With a mitigation scenario that leads to the RCP 4.5, losses in the respective sectors are likely to drop to 10.4%, 6.8%, 6.0% and 5.7%, respectively. The manufacturing sector will lose more than 2% of its value-added to climate change. The services and mining sectors will also register losses, though small relative to the other sectors (2.5 and 1.9 respectively for the BAU scenario).

Value-added in the water sector is projected to expand due to climate change. The RCP 4.5 (RCP 8.5) scenario leads to an expansion of 1.8% (2.3%). It is plausible that heavy investments will flow to the water sector in the future, following significant water scarcity bound to ensue due to the warming climate.

However, the projected effects for the different economic sectors and provinces vary greatly from municipality to municipality, according to Figure 3 and Figure 4. A number of municipalities have a few positive extreme values in all economic sectors. Nevertheless, apart from the water sector, the net effect of all sectors is negative, resulting in average negative effects for the respective sectors and the whole economy. From a policy point of view, watershed municipalities are therefore likely to gain economically, while those relying on agriculture, forestry, fishery, electricity and gas are likely to suffer the greatest losses. Policy attention should therefore focus on adaptations in these sectors.

At the provincial level shown in

Table 5. Percentage losses/gains: Scenario 4.5 versus 8.5.

Province	Agri	Man	Min	Ser	For	Fis	ELG	Wat	All
Scenario 4.5
Eastern Cape	2%	−3%	−5%	−2%	−3%	−3%	−9%	1%	−3%
Free State	−6%	2%	1%	−1%	−11%	−7%	−12%	3%	−4%
Gauteng	−5%	1%	0%	−2%	−6%	−6%	−7%	1%	−3%
KwaZulu-Natal	−5%	−3%	−4%	−2%	−4%	−5%	−11%	5%	−3%
Limpopo	−22%	−2%	−4%	−3%	−22%	−26%	−18%	5%	−11%
Mpumalanga	−13%	−3%	8%	−2%	−10%	−7%	−14%	0%	−5%
North West	−4%	2%	−1%	0%	−4%	−5%	−8%	1%	−2%
Northern Cape	−6%	−2%	0%	−3%	−2%	−1%	−6%	−1%	−3%
Western Cape	2%	−8%	−2%	−2%	−2%	1%	−6%	−6%	−3%
National	−5.70%	−2.25%	−1.50%	−1.98%	−6.79%	−5.97%	−10.35%	1.78%	−4.10%
Scenario 4.5
Eastern Cape	2%	−4%	−6%	−3%	−4%	−4%	−11%	1%	−4%
Free State	−7%	2%	1%	−2%	−13%	−8%	−15%	4%	−4%
Gauteng	−7%	2%	1%	−3%	−8%	−9%	−9%	1%	−4%
KwaZulu-Natal	−7%	−3%	−5%	−2%	−6%	−6%	−14%	7%	−4%
Limpopo	−27%	−2%	−5%	−4%	−28%	−32%	−23%	6%	−13%
Mpumalanga	−15%	−4%	10%	−3%	−12%	−8%	−17%	1%	−6%
North West	−5%	2%	−2%	0%	−5%	−6%	−9%	1%	−3%
Northern Cape	−7%	−3%	0%	−4%	−3%	−2%	−8%	−2%	−4%
Western Cape	3%	−9%	−2%	−3%	−3%	1%	−7%	−7%	−3%
National	−7.14%	−2.77%	−1.94%	−2.49%	−8.47%	−7.26%	−12.89%	2.29%	−5.08%

, the greatest impact of climate change will be felt in Limpopo, where the RCP 4.5 scenario will lead to economic losses of 11%, while the RCP 8.5 will bring about losses of 13%. This finding is attributed to Limpopo’s high exposure to agriculture, forestry, fisheries and electricity and gas sectors, where climate change effects will be more severe in 2050. The next most significantly affected province is Mpumalanga, with main economic losses in agriculture (15%), forestry (12%), fishery (10%) and electricity and gas (17%) in the case of no mitigation assumptions. There is only a 1% gain overall in Mpumalanga for the RCP 4.5 mitigation scenario, relative to RCP 8.5. It is worth noting that in the Western Cape, the average economic loss is 3%, and there is no difference between the BAU and the RCP 4.5 mitigation scenario. Of greater concern for this province is the high relative losses that will be recorded in the manufacturing sector. Whereas all other provinces record positive effects in the water sector, except for Mpumalanga with 0%, the Western and Northern Cape provinces’ water sectors will be affected negatively, with losses at 7% and 2%, respectively, with RCP 8.5.

Apart from the water sector, a number of other sectors will be positively affected by climate change. For example, there will be an 8% (10%) gain in the mining sector in Mpumalanga due to climate change by 2030 for the RCP 4.5 (RCP 8.5) mitigation scenarios. The Eastern and Western Cape agricultural sectors would witness a 2% increase in climate change. In Gauteng, the Free State, and the North West, the manufacturing sector will each expand by 2%, respectively. The mining sector will also expand in the Free State and Gauteng by 1% each for the RCP 8.5 scenario. In the Western Cape, the fishery sector will expand by 1% due to climate change by 2030. The provincial dimensions again suggest that policy action for adaptation has to pay attention to most provinces’ agricultural, forestry, fishery, and electricity and gas sectors, with particular attention in Limpopo, Mpumalanga and the Free State.

A number of individual municipalities present extreme effects on the positive and the negative sides within specific sectors. The case of these municipalities is summarised in Figure 5. The greatest negative impact is in the forestry subsectors in the Blouberg and Ephraim Mogale municipalities in Limpopo, with 50% and 48% losses for the RCP 4.5 scenario, respectively, and 54% and 47% for the RCP 8.5 scenario. In order of decreasing losses, the most affected municipalities are Mokalakwena, in terms of fishery and agriculture, and Mutale faces effects in agriculture, forestry and electricity and gas. Thulamela faces losses in agriculture and the fishery sector. Other municipalities to earmark for particular attention in adaptation policies are Elias Motsoaledi, Polokwane, Makhudutahmaga, Greater Giyani and Bela-Bela.

Positive effects are projected for several municipalities, mainly in the water and mining sectors. In KwaZulu-Natal, Mfolozi has the highest gains in the water sector, followed by Ulundi, Mtubatuba and Hlabisa. A number of municipalities in Limpopo are also projected to experience positive effects in the water sector, namely, Ephraim Mogale and Mutale. In Mpumalanga, the most significant positive effects are all in the mining sector for Mskahlgwa, Albert Luthuli, Thembisile, Dr JS Moroka, Thaba Chweu, Mkhondo, Lekwa and Bushbuckrige.

4. Conclusions and Policy Implications

The climate change effects projections have painted a picture not only of significant economic losses but also of significantly altered local economic structures in South Africa. Some sectors are projected to expand in certain localities, such as mining in most of Mpumalanga and water sectors in most of the country. However, losses outweigh gains in almost all municipalities in South Africa. Our findings show that the temperature cost of inaction is significant at national, provincial, municipal and sub-municipal levels. Therefore, beginning at the national level, the South African government needs to pay serious attention to comprehensive plans to address both mitigation and adaptation in the short, medium and long term (up to 2050) by encouraging intergovernmental relation frameworks among government departments. The national government’s adaptation efforts to reduce future economic losses must include early warning and forecasting for disaster risk reduction; medium-term (decade-scale) climate forecasting to identify potential resource challenges well in advance; and long-term climate projections that define the range of future climate conditions. The national government should formulate, design, implement and evaluate policies that promote climate change response measurement and evaluation systems through the South African Department of Environmental Affairs. Since economic losses were projected in almost all the sectors at the national level due to climate change, adaptation strategies should be integrated into sectoral plans. These should include the national water resource strategy, as well as reconciliation strategies for particular catchments and water supply systems; the strategic plan for South African agriculture; the national biodiversity strategy and action plan, as well as provincial biodiversity sector plans and local bioregional plans; the department of health’s strategic plan; a comprehensive plan for the development of sustainable human settlements; and the national framework for disaster risk management. We recommend that potential adaptation responses in the agriculture sector should range from national-level strategies that include capacity building in key research areas, extension and consideration of water resource allocation all the way to the local level, where responses may be specific to agriculture production methods and local conditions. The national government should also enact policies that set emission reduction outcomes for each significant sector and sub-sector of the economy based on an in-depth assessment of the mitigation potential, best available mitigation options, and a full assessment of the costs and benefits using a ‘carbon budget’ approach.

Given that the results of this study reveal that municipalities relying on agriculture, forestry, fishery and electricity and gas are likely to suffer the greatest losses, policy attention should focus on adaptation in these sectors. The Ministry of Agriculture, Land Reform and Rural Development of South Africa (MALRRDSA) should introduce climate-smart agriculture approaches that will help integrate several agricultural-friendly practices into the agricultural cropping system. The climate-smart agriculture approach is important because it will increase agricultural productivity and income, build crops’ resilience to climate change and reduce carbon emission levels, especially farm and agricultural-induced flaring. The provincial and municipal government agencies that deal with environmental issues should jointly collaborate with the MALRRDSA by investing in agricultural technologies (Agro-Tech) such as machinery, robotics and energy-efficient planters that could be deployed to commercial farms for agricultural activities.

Coal contributes a significant proportion of the energy supply (as electricity) in South Africa and is a major fossil energy resource that significantly contributes to CO₂ emissions. In addition, losses outweigh gains in almost all municipalities in the mining sector, according to our results. Therefore, we recommend that mining companies in South Africa should collaborate with the Department of Mineral Resources and Energy of South Africa to respond now in order to become more resilient in the future. The typical response to climate change among mining companies should be one that enhances energy efficiency, secures water sources and restructures portfolios to exit commodities (most notably coal) that negatively impact the environment, according to literature [40,41]. The high relative losses that will be recorded in the manufacturing sector also call for the need to work towards decarbonising South Africa’s manufacturing sector by 2050. This will be a demanding task that will necessitate a concerted effort from a wide range of stakeholders, including governments, enterprises, international organisations, development financing institutions, investors, and civil society. We therefore recommend the following areas of action to be considered by the government and key stakeholders for the sector, even though they are not directly linked to our study: shift to a net-zero mindset and policy environment; unlock green financing; upgrade green infrastructure; and accelerate research and development. These will help in decarbonising South Africa’s manufacturing sector. Although issues about renewable energy sources fall outside the scope of this study, they have nevertheless been identified in the energy and environmental literature as a means by which countries can reduce the impact of climate change [42,43]. Therefore, the national, provincial and municipal governments should invest more in producing renewable energy and making its consumption affordable.

In conclusion, given the international reluctance in committing to mitigation scenarios that will significantly reduce radiation forcings, the most plausible forcing scenario will not be able to avert significant economic losses. Therefore, adaptation strategies must be implemented as soon as possible. The more policy measures wait, the more losses are incurred; as the picture shows, we are already in an era of losses. Based on the varied effects by sector and geography, adaptation strategies have to be well customised to suit specific conditions. Private economic agents may not be able to make socially optimal choices in this regard, hence the need for careful public policy coordination, with the state taking the lead. The anticipated changes in economic structure imply that several sectors may shed jobs in varying proportions, suggesting careful human capital and skills development planning that accounts for the structural effects of climate change. Each municipality or provincial government will have to develop such plans based on anticipated economic structural changes. Climate change adaptable policies have to consider inter-municipality and inter-provincial impacts. Our results also suggest significant spatial effects of the impacts of climate change. Consequently, we suggest that policy measures transcend cities to involve all tiers of government in seamless coordination in order to ensure effective adaptation policy implementation. The main population groups to consider in developing adaptation strategies should be those that depend on agriculture, forestry and fisheries in municipalities in almost all provinces, especially Limpopo, Mpumalanga, the Free State and North-West. Attention is also to be paid to the manufacturing sector in the Western and Eastern Capes and the water sector in the Western Cape.

A limitation of this study is that it could not extend the period to cover the COVID-19 pandemic due to data unavailability for the variables used. Therefore, future studies should account for the pandemic period to present a better understanding in terms of forecasting the economic growth impacts of climate change in South Africa. Future studies should also consider COVID-19′s impact on climate change interventions and economic activities in South Africa and how the government at the provincial, local and national levels could partner with the private sector to reduce the effect of climate change.