Effects of the Impact Factors on Transportation Sector’s CO₂-eq Emissions: Panel Evaluation on South Africa’s Major Economies

Abstract

The paper utilized a panel dataset to investigate the effects of the impact factors, namely, economic growth, energy intensity, energy consumption, urbanisation, passenger vehicles and transport infrastructure investments on carbon emissions (CO₂-eq). The transport sector of the main buoyant economic provinces of South Africa for a consecutive period of five years was investigated using STIRPAT models. Carbon emissions coefficient method, (CECM) is made applicable to determine the quantification of the forms of fossil fuels utilized in the transportation sector. The characterized parametric panel dataset for Gauteng, KwaZulu-Natal and Western Cape were assessed from 2016 to 2020 to make objective function determinations. The results indicate that GDP and passenger vehicles contributed largely to CO₂-eq emissions, however, the impact varies across the three provinces. Energy intensity and the approach to energy consumption are significant in mitigating carbon emissions, which is due to the possibilities of high fuel efficiency and pollution decrease. The outcome of the research advances climate change mitigation strategy and proffers the required attention to policy makers in South Africa and Africa as a continent.

1. Introduction

Most scholastic pieces of work produced by experts and researchers all agreed that greenhouse gases (GHG) are the main culprit for global warming and, as further supported by Molla et al. [1] and Chang et al. [2], carbon dioxide (CO₂) accounts for the vast proportion of the emissions. There is a consensus among energy researchers that climate change is being driven by increasing atmospheric greenhouse gases, most notably CO₂. The direct global GHG emissions from the transport sector were estimated to be 8.7 GtCO₂-eq and accounted for 23% of global energy-related CO₂ emissions.

The reports of Intergovernmental Panel on Climate Change [3] in 2019 found that 70% of the transport emissions are produced from road vehicles, while aviation, shipping and rail have a share of 12%, 11% and 1%, respectively [4,5]. Several gases, collectively labelled GHG, contribute to global warming processes. A decade of rapid development in South Africa’s transport sector has resulted in tremendous environmental pressures, which include energy consumption and consequently carbon emissions.

From the accounts of [3], contributions of the various gases to global warming are CO₂ (53%), chloro-fluro carbons (CFCs) (24%), methane (CH4) (15%), and nitrous oxide (N₂O) (5%) [6,7,8,9]. It is observed that the largest contribution to the greenhouse effect is produced by CO₂ emissions. The transportation activities and primarily the combustion of fossil fuels and the resultant carbon emissions are among the activities significantly causing warming. Statistical records in recent times have shown that the ratio of increase in transport sector’s emissions in the developing regions of the world is higher than in the developed regions, specifically, North America and Europe [10,11,12]. With the introduction of additional energy policies, the trend is projected to persist in the developed regions. South Africa is the second largest economy and the biggest energy user in Africa, having the most sophisticated transport system, and remains the largest GHG emitter in the continent [13].

The CO₂ emissions in South Africa’s transport sector in 2019 was estimated to be around 60 MtCO₂-eq. Considering the tremendous increase in the vehicle stocks with the volume of both passenger and freight transport, South Africa’s fossil fuel consumption in the transport sector has increased to around 500,000 barrels per day (tb/d) in 2022 and the import reliance will leap by 85% [14,15]. These are not only related to energy security and environmental pressures but some prevailing challenges as well. These challenges have drawn nationwide attention as all indicators point to addressing the concerns around CO₂-eq emissions in the most industrially established provinces in South Africa. The three main economic buoyant provinces, namely, Gauteng, KwaZulu-Natal and Western Cape, are addressed in the paper fitting with the country’s outline plan to reduce energy consumption per unit of gross domestic product (GDP) by 9% and CO₂ emissions per unit of GDP by 8%, in comparison to the available reports of 2015 level. To address the challenges, a series of recommendations have been made both on a national and global scale, as related to fuel consumption limits, energy consumption tax and vehicle emission standards [13,16,17,18,19].

However, many of the recommendations fundamentally dealt with energy efficiency and conversions that were deemed not to be associated with the emissions, considering varying parameters [20,21]. More factors are deduced to be responsible for South Africa’s transport sector’s emissions [13]. Hence, it becomes urgently important to develop models for the three major buoyant provinces of South Africa to evaluate and analyse the varying factors responsible for carbon emissions in the transport sector.

South Africa’s GHG and CO₂ emissions relating to environmental science and in the industrial sector have been broadly investigated using an integrated approach, ARDL and wavelet coherence techniques [22,23] and in the transport sector covering South Africa and Lesotho [24]. Except for Oladunni et al. [13], the previous available studies pay attention to other sectors or GHG emissions being addressed, as combined with no specificity, in most cases neglecting the transport sector and in no way addressing Gauteng, KwaZulu-Natal and Western Cape. The study by Tongwane et al. [24] examined only road transportation and found that GHG emissions in South Africa are closely related to the number of vehicles on the road and that it is possible to calculate their coefficient ratios. It is largely evident in the analytical methods and empirical findings of Bakker et al. [25], Capallan-Perez et al. [26], Alam et al. [27], Gately et al. [28] and Diakoilaki et al. [29] using an integrated approach, bottom-up sectorial approach, ARDL and/or wavelet to perform GHG emissions’ reduction estimates, however, not with the case of applying econometric models.

For the desired level of economic growth, although energy policies for reducing the usage of mobility systems can serve a purpose, they cannot be proven to be sufficient for sustainable development, as the era of COVID-19 made it increasingly evident that a growing need for clean transport systems can drive behavioural changes and support the implementation of new transport technology options. Thus, it becomes imperative to stress the fact that South Africa possesses a vastly developed transport system, having peculiar characteristics in the continent with the account of the three investigated provinces, namely, Gauteng, KwaZulu-Natal and Western Cape. Although, with the use of cutting-edge energy production techniques such as hydrogen-inputs for high performance of heavy-duty gas engines, the contributions of natural gas and renewable energies to having serene residential areas and clean urban cities can be cushioned [30,31,32,33,34,35]. However, reliable estimation undertakings are essential in drawing functional emissions-mitigation methodologies. This research investigates CO₂-eq emissions in the transportation sector by constructing a panel dataset of the three provinces for the period of five years, which runs from 2016 to 2020 as sourced from public, private, local and international organizations [36,37,38].

At first—Apart from the industrial sector, which consumes more energy globally than any other end-use sector (more than 50%), the transport sector accounts for about 23% of global energy use [39,40,41]. This burning of energy, particularly by combustion of fossils in the transportation sector of South Africa, has resulted to around 13% of CO₂ emissions calling for reduction. If addressed, this will contribute to lowering the global warming intensity.

By looking into several systematic studies on the sectorial emissions in several transportation modes, the current research extensively prioritizes CO₂-eq emissions with the various impacting forces in the transport sector. Okada [42] examined CO₂ emissions by factoring in population, pricing, network growth, and metropolitan metro demand. Tongwane et al. [24] conducted research on road transportation involving two boundary countries, namely South Africa and Lesotho, determining the quantity, kind and carbon content of fuel combusted using similar criteria of [3]. Cheze et al. [43] conducted a study on the GHG emissions of air travel and came to some pertinent conclusions about passenger and freight transport. In studies of the transport industry in Thailand and Japan that addressed energy efficiency, some significant mitigation solutions are being derived [44,45].

When researching energy with focus on GHG and CO₂ emissions, there are primarily four conventional methodologies that are frequently used. The first is the index decomposition analysis, which breaks down emissions from the transportation sector into four categories: energy intensity, population size, GDP per capita, and carbon intensity. Timilsina and Shrestha [46] investigated the rise of the transport sector’s CO₂ emissions in Italy and Asia by including fuel-mix and modal shift. Mazzarino [47] widened the scope of his research to include transport intensity and divided CO₂ emissions in Italy essentially into five transport-type categories. The bottom-up sector-based analysis is the second approach. Using traffic data from the Highway Performance Monitoring System (HPMS) and on-road emissions estimates from the Emissions Database for Global Atmospheric Research Protocols, Gately et al. [28] examined on-road CO₂ emissions in Massachusetts, the United States (EDGARP). Rio Gonzalez [48] used input and output techniques created based on bottom-up sector-based to study the structure of CO₂ emissions from land transport in several European Union (UN) nations for the investigation of new energy penetration. The third method, system optimization, has been used extensively to forecast GHG emissions and energy consumption [49,50,51,52]. Gonzalez et al. [53] analysed CO₂ emissions from passenger vehicles using panel data approach, as Georgatzi et al. [54] proceeded to examine the determinants of CO₂ emissions caused by the transport sector in 12 European countries.

The econometric methodology is the fourth method that is highly sensitive. Using panel dataset estimation and econometric methods, Oladunni et al. [13] investigated GHG emissions and their driving factors in the South African transportation industry. Gonzalez [55] evaluated the effects of dieselization and induced road transport in Spain as Rentziou [56] calculated the total number of vehicle miles traveled (VMT) for a specified period for 48 states in the continental United States. For OECD nations, the correlations between gasoline prices, car ownership, income, and gasoline consumption were studied using highly econometric methods [57,58]. In addition, the detection of GHG emissions from inland containers was done using econometrically derived time-series data by Dirka and Acciaro [59].

Due to the increasing pressure from international organizations, it is, therefore, necessary to limit global warming by 1.5 °C [59,60]. Shafique et al. [61] used a panel dataset from 1995 to 2007 accessible for the main Asian economies, including China, South Korea, Singapore, Japan, India, Indonesia, Malaysia, Thailand, and the Philippines, to examine the relationship between transportation, economic growth, and environmental degradation. Of a special case, using LMDI method Wang et al. [62] studied CO₂ emissions in China’s transport sector. Using a comprehensive decomposition method, Zhou et al. [63] and Merven et al. [64] examined the CO₂ emissions in the Chinese transport sector and concluded that the country’s economic expansion is the primary cause of these emissions. It was established that the approach to energy consumption maintains its superiority [64] in South Africa and [65] in China, employing bottom-up sector-based analysis on multiple occasions to calculate emissions in the transportation sector. Xu and Lin [66] optimized the variables affecting CO₂ emissions in the BRICS (Brazil, Russia, India, China, and South Africa) nations by using the system optimization method. Oladunni [67] applied econometric techniques of a robustly expanded STIRPAT model from an industrial research-based constructed panel dataset to determine the effects of the various factors influencing the GHG emissions in the South African transport sector. The research report covered all the nine provinces of South Africa.

The challenges surrounding energy consumption and GHG-CO₂ emissions with the identification of the driving forces continue to be of grave concern to South Africa and its continent, as more study is still needed to catch up with developed countries such as the US, the EU, and China. Previous research papers highlighted some significant flaws, so this work is putting out more improvements to the niche. It is imperative to conduct in-depth research into the three provinces that are the most economically developed, namely Gauteng, KwaZulu-Natal and Western Cape. Hence, to assess the responsiveness of the numerous factors having an impact on CO₂ emissions in the transportation sector for the provinces, it is necessary to employ econometric approaches. Thereby, we can determine which important factors have the most and least influence on carbon emissions in the transportation sector.

2. Methodology

Recent energy analyses have shown that fossil fuel burning is the primary source of carbon emissions [66,67,68,69]. With the rarity of statistical releases, the carbon emissions coefficient method (CECM) is found applicable [70] in quantifying the quantity of carbon emissions in the three provinces. The CO₂ emissions coefficients, the rate of carbon oxidation reported by the Intergovernmental Panel on Climate Change (IPCC) [3], as provided by [71], and the conversion factors from physical units to tons of oil equivalence are all required to calculate carbon emissions in South Africa for the years 2016 to 2020. GHG or carbon emissions from the burning of fossils fuel are generally taken in the standard of CO₂ emissions. Thus, calculated as presented below:

$${\mathrm{CO}}_2-e{q}_t=\frac{44}{12}\times {\displaystyle \sum }_{i=1}^{11}(E{C}_{it}\times c{f}_i\times e{f}_i\times 0_i)$$

(1)

The CO₂-eq_t represents the total carbon emissions in the standard of CO₂ emissions of the equivalence (10⁴) in year t. We have i to denote the eight types of fossil fuels, which are crude oil, motor gasoline, natural gasoline, aviation gasoline, diesel, kerosene, lubricating oil, coal, natural gas, jet fuel and coke. We refer to EC_it as consumption of fuel type of I (expressed in physical units’ form). cf_i represents the conversion factors from physical unit to the tons of oil equivalent of fuel type i. ef_i denotes the carbon emissions coefficients of fuel type i (tC/tce). The ratio of $\frac{44}{12}$ expresses the molecular weight of CO₂ to the atomic weight of carbon (C) and 0_i is the rate of carbon oxidation of fuel type of i, as shown in Equation (1).

Table 1. Tons of oil equivalents in conversion factors from their physical units.

Source	Conversion Factor in Unit	Coefficient of CO2 Emission	Carbon Oxidation Rate
Crude oil	1.018	0.5850	0.970
Motor gasoline	1.050	0.5538	0.980
Natural gasoline	1.073	0.5439	0.920
Diesel	1.015		0.9820
Aviation gasoline	1.050	0.524	0.9840
Kerosene	1.032	0.5714	0.9860
Lubricating oil	1.007	0.8075	0.950
Coal	0.714	0.7480	0.9130
Natural gas	1.079	0.4440	0.990
Jet fuel	1.043	0.5230	0.9850
Coke	0.9714	0.8550	0.9280

Extract Source: IPCC, (2007).

presents, in tons, the oil equivalent in their conversion factors from their respective physical units [3].

To evaluate the effects of various factors on the change in energy consumption over GHG emissions, the IPAT (I = PAT) identity has been functionally applied [66,67,68,69,70,71,72]. In more general terms, the IPAT is to perform an analysis of the effects of economic activities on the environment. As initially suggested by Ehrlich and Holdren [73], IPAT possessed a limitation of not permitting hypothesis tests, as the already known values of some specific terms determine the value of the missing term. Hence, to overcome the limitation of IPAT model, the STIRPAT model, known as stochastic impacts by regression on population, affluence, and technology, is established by Dietz and Rosa [74]. The model of STIRPAT equation is presented as follows:

$$I_i=a{P}_i^b{A}_i^c{T}_i^d{e}_i$$

(2)

As $I$ denotes the impacts on the environment as a result of human activities, $P$ represents the population under study. The symbol $A$ stands for affluence taken as per capita GDP. T represents the technological factor while their subscripts suggest that the quantities are differential across units. The constant term of $a$ is fitted to scale the model; b, c and d are the exponential determinants of the parameters $P$, $A$ and $T$, respectively where $e$ serves as an error term. The STIRPAT model is formed such that more factors can be incorporated, provided that they are conceptually fitted. Wherein $a=b=c=d=1.$ By its functionality, in empirical investigation, Equation (2) may be converted to logarithmic form:

$$ln I=ln a+b ln P+c ln A+d ln T+ln e$$

(3)

This model considers seven cogent variables to perform analysis in the transportation sector based on the three major economic provinces of South Africa. The carbon emissions are indicated as CO₂-eq (in the equivalence of CO₂), which is equivalent to the impacts on the environment, as stated earlier. Passenger vehicles (PV) and economic growth (per capita GDP) are intentionally included to measure social economic activity. Energy intensity (EI), which measures the technological component, is taken into consideration together with energy consumption (EC), urbanization (URB) and transportation infrastructure investments. The CO₂-eq is the variable that is parametrically dependent on the other assessed variables for the transport sector in this research paper. To foster evaluation and test of hypothesis, all the variables are used in logarithmic form, as follows:

$$\begin{array}{c}ln {\mathrm{CO}}_2{}_{it}-eq=a_i+{\beta }_p ln P_{it}+{\beta }_a ln A_{it}+{\beta }_{pv}+{\beta }_{pv} ln P{V}_{it}+{\beta }_{urb} ln URB\\ +{\beta }_{tii} ln TII+{\beta }_{ei} ln E{I}_{it}+EC+e_{it}\end{array}$$

(4)

Herein, $a_i$ reports individual effects for the model and β with variable-specific subscripts represent the coefficients to be estimated. As in Equation (2) $i$ and $t$ denote province and year observed, respectively.

Prior to the model’s estimation, some essential tests were performed. These are Robust Hausman test, panel cointegration test, autocorrelation test, group-wise heteroskedasticity and cross-sectional independence test. The panel cointegration test is used to assess long-run equilibrium relationship among variables, as in the case of Martinez-Zarzoso and Maruotti [75]. Robust Hausman test is applied to give allowance for group-wise heteroskedasticity, while the autocorrelation and cross-sectional tests exercise independence with the panel dataset, as experimented by Hoechle [76]. When the case applies, other tests determine estimation methods that are reliable when in violation of the classic hypotheses, as demonstrated by Wooldridge [77]. The insensitivity of the estimation results from the authenticity, as observed, often emanated from the estimation error of the model parameters and bias of the covariance matrix as stipulated by Kim and Sun [78].

3. Data Source and Description

3.1. Data Source

A panel dataset covering the five years from 2016 to 2020 is constructed for this investigation. The research focuses on Gauteng, KwaZulu-Natal, and the Western Cape, which are South Africa’s three largest economic provinces. Some resourceful estimates taken from Oladunni’s [67] research project were used as the data in this study. They are technically well sourced; the reports are available with the [36,37,38] private and public organizations, as there is no single readymade available source of such bulky information. The data are also widely sourced from various private and public establishments and then designed to function as a uniform South African panel dataset for the GHG emissions’ impacting factors of the transport sector, as the research niche in this region of the world is still underdeveloped, in comparison to the panel datasets made available in the cases of North America, Europe, and China [66,79,80].

The data on population was drawn from South Africa’s Department of Home Affairs, while the data on per capita GDP and urbanization were both taken from the Department of Treasury, Statistics South Africa and the World Bank. The data on energy intensity was sourced from South Africa’s Department of Energy and South Africa’s Reserve Bank, while the data on transport infrastructure investments were drawn from South Africa’s Department of Works and the Department of Transportation. In the case of the Intergovernmental Panel on Climate Change (IPCC) [3], the authors used the applied approaches of calculating CO₂ emissions from the energy consumption in the transport sector at the provincial level for the three investigated provinces. Additionally, to quicken the comparison, affluence, $A$, (the per capita GDP), was used to standardize the worth of local currency (Rand) to the constant value of the year 2020.

3.2. Data Description

According to the drawn data of GHG in CO₂-eq emissions that are estimated in this research paper, the amount of emissions for the South African transport sector from 2016 to 2020 over a 5 year period reached its all-time highest in 2016, to the sum of 4.8 million tons, and its lowest in the very last year of the data period in 2020, at around 4.2 million tons. Figure 1 gives a pictorial representation of carbon emissions (CO₂-eq), as the dependent factor over the exogenous impacting variables, namely, economic growth, energy intensity, energy consumption, urbanization, transport infrastructure investments and the passenger vehicles. Figure 1 further shows that in the transport sector, land and road, under which passenger vehicles fall, are more of a concern.

The three main provinces, which are Gauteng, KwaZulu-Natal and the Western Cape, in the year 2020 contributed approximately 13.3, 7.0 and 6.3 million tons of CO₂ emissions, respectively. For this research, the parameters of concerns over the CO₂-eq emissions include economic growth (GDP), energy intensity (EI), energy consumption (EC), passenger vehicles (PV), urbanization (URB) and transport infrastructure investments (TII). GHG is taken from the equivalence of CO₂ emissions measured in metric tons per capita; GDP is measured in constant 2020 SA Rand; EI is measured in MJ/SA Rand 2020 PPP GDP; EC is taken in tons of oil equivalent; URB is dimensionless, hence, measured in percentage (%); PV is taken in 10⁴ units; TII is taken in SA Rand 2020. The analytical models for this investigation maintain CO₂-eq (taken in metric tons per capita) as the dependent variable. The parameters GDP, EI, EC, PV, URB and TII are the exogenous variables. The data are retrieved from various energy consumptions and intensity sources; greenhouse gas, carbon dioxide, atmospheric and environmental sources; transport and freight; trade; infrastructural investments; and South Africa’s urbanization planning establishment. These data sources are grouped in two distinct categories as follows:

Direct local data sources: Statistics South Africa, South Africa Department of Energy, South Africa Department of Transportation, South Africa Department of Treasury, South Africa Department of Works, South Africa Department of Home Affairs, South Africa Department of Energy and National Association of Automobile Manufacturers of South Africa.

Direct international data sources: World Bank Platform, United Nations Platform, International Energy Agency and British Petroleum Statistical Review. Both variables on the two sides of the (Equation (4)), dependent variable and exogenous variables, are reducibly converted into natural logarithms, as required in the practical investigations.

Table 2. Definition of variables represented with their codes and units of measurement.

Variable	Code	Units of Measure	Definition	Source
Carbon dioxide emissions	CO2-eq	10,000 ton	Standardized equivalence for carbon-emitting compounds	SA DoE, [81]; IEA, [82]
Gross domestic product	GDP	Rand per capita	Per capita gross domestic product in SA Rand	SA DoTreasury, [83]; Stat. SA, [84]; W.Bank, [85]
Energy intensity	EI	Tce per ZAR 10,000	Energy intensity level of the primary energy use in South Africa’s transport sector	SA DoE, [81]; SA Reserve Bank, [85]
Energy consumption	EC	kg of oil equivalent per capita	Oil consumption in equivalence to the total energy consumption	SA DoE, [81]; SA DoT, [36]
Urbanization	URB	% to the total development	Percentile ratio to the total urban development	SA DoTreasury, [83]; World Bank, [85]
Transport infrastructure Investments	TII	Rand per capita	South Africa’s investments made on transport infrastructures provincial level	SA DoTreasury, [83]
Passenger vehicles	PV	By provincial fleet units	The number of passenger vehicles on the road contributes to carbon emissions	SA DoT, [37]; NAAMSA, [86]

presents the used variables with the provision of definitions, codes, units of measurements and their sources.

Table 3. Three provincial 5-year panel dataset constructions for South Africa’s transportation sector 2016–2020.

Province	Year	CO2-eq	GDP	EI	EC	URB	TII	PV
Gauteng	2016	17,124,262	101,093	166.711	795,378.2	0.982	24.66764B	4,524,810
	2017	14,998,320	107,675	170.681	814,321.9	0.984	24.77246B	4,613,986
	2018	14,373,850	111,171	166.662	795,147.7	0.982	26.34476B	4,698,524
	2019	13,808,028	120,233	170.383	812,896.9	0.987	25.99536B	4,803,490
	2020	13,323,126	137,463	169.282	807,646.5	0.991	31.68207B	4,885,748
KwaZulu-Natal	2016	9,003,252	59,048	84.524	403,264	0.484	11.32424B	1,575,967
	2017	7,885,516	62,681	88.397	421,745.3	0.485	11.37236B	1,600,930
	2018	7,557,195	66,254	90.065	429,699.4	0.487	12.09416B	1,627,244
	2019	7,259,709	71,132	87.526	417,589.9	0.489	11.93376B	1,660,720
	2020	7,004,767	88,194	89.338	426,231.4	0.502	14.51620B	1,695,075
Western Cape	2016	8,148,409	84,707	93.37	445,466.3	0.937	9.78516B	1,868,946
	2017	7,136,800	85,941	85.645	408,614.6	0.940	9.82674B	1,923,765
	2018	6,839,653	97,664	79.737	380,425.4	0.941	10.45044B	1,977,518
	2019	6,570,412	103,510	79.148	377,616.3	0.943	10.31184B	2,020,510
	2020	6,339,676	126,738	89.222	425,677.1	0.957	12.54330B	2,063,299

shows a 5-year dataset for the three major provinces that was utilized to investigate the effects of the impact factors on carbon emissions in the transport sector. The produced carbon emissions in the equivalence of CO₂ are also provided in the yearly proceedings. The values of each of the contributing exogenous variables are stated in the order of economic growth, energy intensity, energy consumption, urbanization, transport infrastructural investments and passenger vehicles, respectively, in their related standard of assessments.

Table 4. Description of the econometric statistics for the main variables.

Province Investigated	Variable	Observations	Average	Standard Deviation	Minimum	Maximum	Log Mean	Skewness	Kurtosis
Gauteng	CO2-eq	250	15,207,559	1,065,933	13,375,346	17,108,303	7	−0.21215	−1.2638
	GDP	250	119,271	10,515	101,132	137,382	5	−0.16404	−1.2131
	EI	250	169	2	167	171	2	0	−1.2609
	EC	250	803,628	5110	795,213	812,656	6	−0.09696	−1.1588
	URB	250	99	0	98	99	2	0	−2.0109
	TII	250	2,817,557,180,857	207,210,076,874	2,470,465,545,892	3,169,899,709,227	12	−0.05109	−1.1838
	PV	250	4,694,688	104,742	4,525,334	4,885,435	7	−0.03574	−1.1791
KwaZulu−Natal	CO2-eq	200	7,988,230	589,551	7,021,405	8,981,176	7	−0.02593	−1.2599
	GDP	200	73,973	8559	59,382	88,136	5	−0.21541	−1.1589
	EI	200	87	2	84	90	2	−0.04379	−1.3422
	EC	200	416,176	7916	403,299	429,601	6	−0.05995	−1.3607
	URB	200	48	1	47	50	2	−0.055	−1.2981
	TII	200	1,112,643,341,608	83,838,205,711	981,040,163,682	1,247,731,514,685	12	0	−1.074
	PV	200	1,635,542	33,139	1,576,454	1,694,667	6	0	−1.1425
Western Cape	CO2-eq	150	7,288,116	553,915	6,364,933	8,121,483	7	0	−1.1472
	GDP	150	105,398	12,525	84,959	126,562	5	0	−1.1539
	EI	150	86	4	79	93	2	0	−1.1913
	EC	150	411,116	18,887	377,697	445,203	6	−0.0258	−1.1513
	URB	150	95	1	93	96	2	0	−1.3766
	TII	150	1,110,035,785,123	79,777,242,362	983,482,231,043	1,249,078,375,142	12	−0.35548	−0.8870
	PV	150	1,961,986	58,836	1,869,120	2,063,231	6	−0.1592	−1.0583
All three provinces	CO2-eq	421	11,653,022	3,149,951	6,356,707	17,100,795	7	−0.44976	−1.0275
	GDP	421	98,196	23,047	59,660	137,320	5	−0.32826	−1.1286
	EI	421	125	26	79	170	2	−0.242	−1.0598
	EC	421	1,632,270	15,251	1,605,375	1,659,547	6	0	−1.1953
	URB	421	75	15	48	99	2	−0.27593	−1.0629
	TII	421	2,088,809,066,568	625,995,505,074	985,517,397,500	3,169,430,922,472	12	−0.4132	−1.0286
	PV	421	3,243,894	918,201	1,645,195	4,783,508	6	0	−1.0442

depicts the—essentially—first-stage of a statistical analysis that was conducted. A total of 421 running observations on the exogenous and endogenous variables were made in the combined three provinces of the national case. To their highest level of accuracies, the reliable values of their averages, standard deviations, minimum and maximum possibilities were ascertained, and hence drawn. As shown in Table 4, the respective logarithm bases were employed. To avoid any form of error by skewness and kurtosis, the bearable limits were observed, as shown in Table 4.

Table 5. Econometric comparison of variables characteristic within the three provinces.

Province	CO2-eq	GDP	EI	EC	URB	TII	PV
Gauteng	maximum	maximum	maximum	very high	very high	very high	very high
KwaZulu-Natal	intermediary	minimum	intermediary	medium	low	low	high
Western Cape	minimum	intermediary	minimum	medium	very high	low	high

provides smart, quick referencing points of all the effecting and resulting variables on their first stage energy-econometric analyses. In all the impacting factors and for the investigated period of 5 years, the Gauteng province remains very high in their values and the maximum of all. Table 5 shows that KwaZulu-Natal, in comparison with the Gauteng province variables, is within the boundaries of intermediary, minimum, intermediary, medium, low, low and high for the parametric variables of CO₂-eq, GDP, EI, EC, URB, TII and PV, respectively. Further explanations, as succinctly shown in Table 5 of the fourth row, illustrate that the province of Western Cape’s parametric variables are within minimum for CO₂-eq emissions, intermediary for economic growth, minimum for energy intensity, medium for energy consumption, very high for urbanization, low for transport infrastructure investments and high on passenger vehicles, respectively.

4. Empirical Results

4.1. Results of Panel Unit Root Tests

There exist some highly sensitive performing unit root tests that are applied in this research paper, as also conducted by Hadri [79] and Levin et al. [80]. To analyse the stationarity of both the endogenous and the exogenous series, the works of Dickey and Fuller [87] were utilized. The panel unit root tests, as shown in

Table 6. Results of panel unit root test.

Variable	Level		1st Difference
	Intercept	Intercept and Trend	Intercept	Intercept and Trend
ADF-Test
CO2-eq	−13.3226	−20.32162 **	−11.02377 ***	−11.01005 ***
GDP	−20.4583	−20.45426 ***	−10.64380 ***	−10.63048 ***
EI	−10.3921	−17.25520 ***	−8.584478 ***	−8.565831 ***
EC	−13.32263 ***	20.32162 ***	−11.02377 ***	−11.01005 ***
URB	20.302	20.28933 *	−11.36690 ***	−11.35446 ***
TII	10.66187	10.75451 **	−11.06483 ***	−11.05049 ***
PV	20.37188 **	20.38554 **	−12.08296 ***	−17.07012 ***
PP-Test
CO2-eq	−20.11868 *	−20.33029 *	182.7463 ***	185.4594 ***
GDP	−20.45950 **	−20.45535 ***	345.3184 ***	358.6211 ***
EI	−16.93254 **	−17.25499 **	104.9367 ***	118.2252 ***
EC	20.11868 **	20.33029 **	182.7463 ***	185.4594 ***
URB	20.30133 **	20.28974 **	166.8125 ***	166.5240 ***
TII	22.19060 **	22.27828 ***	220.8640 ***	220.0184 ***
PV	20.37188 **	20.38550 ***	182.2821 ***	190.8106
GLS-Test
CO2-eq	−1.909201 *	−3.932731 *	−0.475976 ***	−1.933990 ***
GDP	−8.192766 ***	−19.5842	−1.813580 ***	−3.021176 ***
EI	−2.833733 **	−2.956060 **	−0.545694 ***	−0.915999 ***
EC	−1.909201 *	−3.932731 *	−0.475976 ***	−1.933990 ***
URB	−5.23819	−8.562990 ***	−0.912300 ***	−2.304651 ***
TII	−5.651923 **	−10.26940 **	−1.022295 ***	−2.114796 ***
PV	−1.692029 **	−3.367025 **	−12.76321 ***	−12.77482 ***

(*) denotes level of significance at 10%, (**) significance at 5%, (***) significance level at 1%, (0) at not significant.

, are applied to energy-econometrics analysis to avoid spurious regression. The analysis of unit root, namely, ADF-Test, PP-Test and GLS-Test, were performed with their intercepts and then intercept and trend. However, it is observed that not all the parametric values were stationary at level. Nevertheless, at the first difference, all the variables were found to be stationary and time variant. Consequently, both the dependent and the exogenous variables were integrated in an order of one. As with the case in the work of Pradhan et al. [88] and Beyzatlar et al. [89], the null hypothesis of a panel has a unit root with absolute values higher than the critical values, and their R-squared was lower than the Durbin–Watson statistics, and thus the null hypothesis was rejected. Hence, the long-term relationship between CO₂-eq and the exogenous variables were optimally determined.

4.2. Panel Estimation for Long-Run between CO₂-eq and the Exogenous

The empirical outcomes of LS, ML ARCH and GLS [90,91,92] panel estimations for the national case by combining the three main industrial economies are illustrated in

Table 7. Panel estimation results: National case for three provinces of CO₂-eq emissions in the transport sector.

Variables	LS (M1)	ML ARCH (M2)	GLM (M3)
lnGDP	0.142 *** (1.065)	0.130 *** (0.980)	0.142 *** (1.065)
lnEI	−0.086 *** (−1.337)	−0.096 *** (−1.480)	−0.086 *** (−1.337)
lnEC	2.051 *** (1.373)	1.981 *** (1.338)	2.051 *** (1.373)
lnURB	−0.038 *** (−0.559)	−0.043 *** (−0.645)	−0.038 *** (−0.559)
lnTII	−0.039 *** (−0.881)	−0.042 *** (−0.938)	−0.039 *** (−0.880)
lnPV	−0.065 *** (−1.401)	−0.063 *** (−1.363)	−0.065 *** (−1.401)
Constant	−12.255 *** (−0.568)	0.031 *** (0.337)	−12.2255 *** (−0.568)
R2	0.018	0.018
F-Stat	1.254
Observations	421	421	421

(*) denotes level of significance at 10%, (**) significance at 5%, (***) significance level at 1%, (0) at Not significant.

; more emphasis is placed on the LS estimation method for unified clarity. Hausman and F test were conducted to settle on LS estimation method. As analysed in Table 7, most of the variables are found to be statistically significant and possessed the required signs. Energy consumption in the panel estimation models is of the largest value of impact factors with the elasticity of 2.051 for LS Method 1, 1.981 for ML ARCH Method 2 and 2.051 for the GLM Method 3. The economic growth (GDP) followed, with an impact factor in elasticity of 0.142 at LS Method 1. Energy intensity followed with an impact elasticity of 0.086 possessing a negative sign. Passenger vehicle has an impact elasticity value of 0.065 having a negative sign of impact. To run diagnostics, heteroskedasticity test was performed on the variables using first the White method and then Glejser and Harvey tests as shown in Table 7. The White method performed diagnostics in the multiplication of variable by itself and later the multiplication of a given variable with subsequently surfacing variable.

The exogenous variable of urbanization (URB), as shown in Table 7, underwent t-test to produce outcome 0.038 elasticity, this possesses negative impact on carbon emissions. The transport infrastructure investments report a quantitative coefficient of −0.039; this shows a gap to fill in meeting the demands for more of its investments in the transport sector, as it indicates negativity. The econometric coefficient of GDP, which represents economic growth, was found to impact carbon emissions by a level of positive statistics. The result confirms the existence of GDP possessing positive impact factors on carbon emissions in the transport sector. The implication of the research analysis connotes that the prior increase in carbon emissions by some percentile levels was produced as a result of the GDP in the given period. This is shown in the panel dataset, however, there is a decrease alongside the increasing economic growth. In other words, a 1% increase in economic growth will decrease the level of carbon emissions in the transport sector by 0.142%.

Energy consumption (EC) and energy intensity (EI) are significant with coefficients of 2.051 and −0.086, respectively. As the former poses a huge threat to the transport sector with a high ratio of effects on carbon emissions, this implies that the outcome of carbon emission pollutions is highly dependent on how energy is being consumed.

A 1% utility (or disutility) could result in an exponential case of 2.051% (alternatively, of negative consequences), even when considering that the mitigation strategies in the South African transport sector have started taking shape. A 1% decrease would lead to carbon emission reduction by energy intensity. The interpretation of EC is that, with a 1% carbon emissions reduction, energy consumption can be made more efficient to a very high degree of 2.051 ratio 1. Furthermore, the outcome of the research produces supporting evidence for the existence of EKC [93,94]. The essential interpretation of the conducted analysis, as demonstrated in Table 7, is that energy efficiency improvements are effective in the reduction in carbon emissions in the transport sector.

4.3. Models Diagnostics Results: Heteroskedasticity of the Transport Sector CO₂-eq Emissions

To avoid spurious estimation and variable inconsistencies, diagnostics tests were conducted, as stated in

Table 8. Diagnostics results: heteroskedasticity of the transport sector CO₂-eq emissions.

Variables and Interconnections	White	Glejser	Harvey
lnGDP2	5.160 *** (1.008)
lnGDP×InEI	−1.232 *** (−0.572)	0.247 *** (1.019)	0.714 *** (0.787)
lnGDP×InEC	−11.046 *** (−1.149)
lnGDP×InURB	2.105 *** (0.926)
lnGDP×InTII	−0.933 *** (−0.609)
lnGDP×InPV	3.474 *** (2.254)
lnEI2	1.567 *** (1.376)	−0.018 *** (−0.156)	−0.344 *** (−0.788)
lnEI×InEC	−0.482 *** (−0.020)
lnEI×InURB	−1.423 *** (−1.261)
lnEI×InTII	0.427 *** (0.590)
lnEI×InPV	−0.698 *** (−0.931)
lnEC2	5.336 *** (0.793)	−1.614 *** (−0.510)	1.427 *** (0.140)
lnEC×InURB	−4.376 *** (−1.586)
lnEC×InTII	1.794 *** (0.708)
lnEC×InPV	−3.122 *** (−1.471)
lnURB2	1.041 *** (0.845)	−0.039 *** (−0.312)	−0.238 *** (−0.508)
lnURB×InTII	1.414 *** (1.807)
lnURB×InPV	−0.399 ** (−0.487)
lnTII2	−0.456 ** (−0.850)	−0.029 *** (−0.363)	−0.084 *** (−0.279)
lnTII×InPV	−0.240 ** (0.453)
lnPV2	−0.027 ** (−0.048)
Constant	−4.871 *** (−0.006)	21.902 *** (0.564)	−23.539 *** (−0.16)
R2	0.051	0.043	0.004
F-Stat	0.964	0.295
Observations	421	421	421

(*) denotes level of significance at 10%, (**) significance at 5%, (***) significance level at 1%, (0) at not significant.

. Three active methods of White, Glejser and Harvey heteroskedasticity were employed [95,96,97]. The second column presents the White diagnostics results. The approach is for a single prevailing variable to exercise diagnostics in a square of itself, then every other variable in line follows by multiplication to the subsequent exogenous parameter, as illustrated in the second column of Table 8. By a robust performance, as depicted, the outcomes of the White diagnostics are forecasted to be consistent and reliable with the series in Table 8. The two other methods are employed in the diagnostics on the dependent variable (CO₂-eq emissions) over the impacting factors, producing singular outcomes for each of the exogenous variables, as shown in the third and fourth columns, respectively, in the analysis of Table 8.

4.4. Provincial Analysis of CO₂-eq Emissions

Most of the exogenous variables of the Gauteng province are econometrically significant, which is found by observing, in particular, the LS estimation method at the confidence levels of 1, 5 and, mostly, 10%, as illustrated in

Table 9. Panel estimation results: Gauteng CO₂-eq emissions in the transport sector.

Variables	LS (M1)	ML ARCH (M2)
lnGDP	−0.0180 *** (−0.337)	−0.008 *** (−0.148)
lnEI	0.549 *** (0.966)	−0.544 *** (1.015)
lnEC	0.043 *** (0.056)	0.042 *** (0.166)
lnURB	−1.901 *** (−2.067)	−1.907 *** (−24.471)
lnTII	−0.004 *** (−0.057)	−0.006 *** (−0.079)
lnPV	−0.478 *** (2.215)	−0.478 *** (14.893)
Constant	14.849 *** (1.241)	14.837 *** (175.766)
R2	0.04	0.039
F-Stat	1.646
Observations	250	250

(*) denotes level of significance at 10%, (**) significance at 5%, (***) significance level at 1%, (0) at not significant.

. Energy consumption and economic growth impact CO₂-eq emissions of the transport sector with the coefficients of 0.043 and 0.018, respectively. Energy intensity and passenger vehicles also impact CO₂-eq emissions with elasticities of 0.549 and 0.478, respectively. By comparison with the national case, the panel estimation appears to be more peculiar in their status, which is due to human and economic density in the province. The impact of the development in urbanization with high predictive values and transport infrastructure investments decreases carbon emissions with elasticities of −1.901 and −0.004, respectively. The low value impact suggests the need for more investments to TII variables, although a percentile decrease in on-road passenger vehicles would mitigate carbon emissions by 0.478 elasticity. Energy (Oil) consumption has a critical value to impact on carbon emissions with a coefficient of 0.043. The measure of energy efficiency in the models, being the energy intensity (EI), possesses a parametric value of 0.549 at t-stat 0.996 (very high confidence level). The analytical comparison of energy intensity and energy (oil) consumption of the combined provinces and the Gauteng province proffer that a robust fuel mix-shift and optimal efficiency will suit carbon emissions mitigation.

As illustrated in

Table 10. Panel estimation results: KZ-Natal CO₂-eq emissions in the transport sector.

Variables	LS (M1)	ML ARCH (M2)
lnGDP	−0.023 *** (−0.555)	−0.027 *** (−0.685)
lnEI	−0.076 *** (−0.336)	−0.132 *** (−0.545)
lnEC	0.055 *** (0.214)	0.054 *** (0.208)
lnURB	0.133 *** (0.597)	0.113 *** (0.510)
lnTII	−0.045 *** (−0.606)	−0.046 *** (−0.617)
lnPV	0.286 *** (1.205)	0.204 *** (14.893)
Constant	14.849 *** (1.241)	14.837 *** (0.899)
R2	0.015	0.014
F−Stat	0.475
Observations	200	200

(*) denotes level of significance at 10%, (**) significance at 5%, (***) significance level at 1%, (0) at Not significant.

below, the province of KwaZulu-Natal displays its distinctive characteristics from the national case combined analysis and the province of Gauteng. Most of the variables are shown to impact CO₂-eq emissions significantly. The frictional coefficient of energy intensity is bulkiest, this stipulates a strong relationship over CO₂-eq emissions, albeit negatively. The economic growth for the province of KwaZulu-Natal is econometrically, negatively significant; this implies the need for an expansive effort to limit carbon emissions in the sector. Energy (Oil) consumption in the sector has a robust coefficient and remains econometrically significant to the value of 0.055. Both urbanization and transport infrastructure investments are statistically responsive to the values of −0.133 and −0.045 with their t-stat of 0.597 and −0.606, respectively. Passenger vehicles on the roads of the KwaZulu-Natal province are less frequent, in comparison with all the investigated cases. It is positive (0.286; 1.205) in terms of its coefficient and t-test values, as it pushes towards being econometrically significant.

Table 11. Panel estimation results: Western Cape CO₂-eq emissions in the transport sector.

Variables	LS (M1)	ML ARCH (M2)
lnGDP	−0.053 *** (−1.010)	−0.053 *** (−0.977)
lnEI	−0.071 *** (−0.562)	0.071 *** (0.560)
lnEC	0.234 *** (1.793)	0.234 *** (364.231)
lnURB	−0.160 *** (−0.320)	−0.160 *** (−0.341)
lnTII	0.002 *** (0.022)	0.002 *** (0.023)
lnPV	−0.046 *** (−0.231)	0.204 *** (14.893)
Constant	14.411 *** (2.966)	14.412 *** (2867.158)
R2	0.03	0.03
F-Stat	0.725
Observations	150	150

(*) denotes level of significance at 10%, (**) significance at 5%, (***) significance level at 1%, (0) at not significant.

presents the Western Cape analyses of the impacting factors on carbon emissions. Except for transport infrastructure investments (TII) having parametric elasticity and t-stat values of 0.002 and 0.022, respectively, all the independent variables, strongly possess explanatory powers. Economic growth in the GDP of the Western Cape Province is shown to be statistically significant at an econometric value of −0.053 and passes the t-test at a higher value of 1.010. The developed negative characteristics connote the excessive impact of cash flow per capita, leading to high energy use. The excesses of energy intensity appear unusual; however, it is statistically significant with a coefficient of 0.071. The reliability of energy consumption with a significance value of 0.234 exists, passing the t-test by a very robust ratio of 1.793. The effects of urbanization and passenger vehicles are statistically significant and contribute to the carbon emissions in the transport sector of the province.

5. Discussion

The outcomes of the panel assessments reveal many important scenarios for the impacts of the influencing factors and those of carbon emissions. On the surface, Figure 2 shows a graphical depiction of the supposed normal distributions over the actual datasets for both the dependent variable (CO₂-eq) and the exogenous variables; their impactful properties, along the period from 2016 to 2020, across the series were determined. For the plot, with regard to CO₂-eq on the Y-coordinate, a 0.4 graphical unit represents the quantile normal distribution, while a 0.2 graphical unit on the X-coordinate represents 10,000 tons of carbon emissions. It is observed that some linear points for carbon emissions were determined across the series, prior to preceding and exceeding the graphical parametric values. Thus, for all ideal samples, equidistant linearity can be maintained. For EC on the Y-coordinate, a 0.4 graphical unit represents the quantile normal distribution, while a 0.2 graphical unit on the X-coordinate represents the kilogram of oil equivalent per capita. It is observed that some linear points for EC were determined across the series, prior to preceding and exceeding the graphical parametric values. Thus, for all ideal samples, equidistant linearity can be maintained. The case also applies for EI ideal samples’ determination. Except for the differences in the adopted units, which come with variances of units, the parametric values for GDP, PV and TII across the series were graphically determined. However, as the graph depicts for URB, the case of urbanization indicated something extraordinarily sensitive. Some linearity points were maintained where the red sloppy line crosses the parametric values of URB; however, the values, as shown therein, are not concentrated, being rather too slanty. Consequently, the carbon emissions and the impacting exogenous variables (in blue colours), along with their linearity (the red label) in the South African transportation industry, are determined. The red label, which can be taken as constant, as shown in Figure 2, is applied to improve the predictive ability of the models. The implications in Figure 2 indicate the ease of variables’ computation to establish a valid relationship among the parametric estimates of the initial regression and the transformed variables.

Furthermore, the blue inclining curves along the two coordinates of Y-X as shown in the graphs of Figure 2, connote the trends for the parametric values of a given exogenous variable over the determining series. The effects exercised by each of the impacting factors (namely, GDP, EI, EC, URB, TII and PV), and the outcome of CO₂-eq emissions on air and largely environmental quality are illustrated in Figure 2. The red linear slopes in the graphical presentations of Figure 2 are to depict conditional boundaries for which positive increase and or negative decrease could effect changes to the transportation environment. In addition, the proper acceleration of a determined variable to fitting the linear slope can be the required estimates at country’s developmental level across the series.

Figure 3a illustrates the relative changes of CO₂-eq emissions developed by economic growth. Figure 3b illustrates the relative changes of CO₂-eq emissions caused with the results of energy intensity as Figure 3c indicates the relative changes of CO₂-eq emissions from the outcome of energy (oil) consumption. The impacted relative differences of CO₂-eq emissions as forced by urbanization are graphically depicted in Figure 3d, so as well are the relative differential impacts of transport infrastructure investments on CO₂-eq emissions are illustrated in Figure 3e. The effects for the impact of passenger vehicles on carbon emissions in the transport sector of South Africa are also demonstrated in Figure 3f. The spatial differences depict the closeness value of one variable to the other.

The scattered circled-dotted plots (in blues) are presented as the CO₂-eq emitted compounds with their relative responses to the respective exogenous impact factors. The preferably Red-falling lines for CO₂-eq emissions of the transportation industry, or as the case connoted, the Red-rising lines as illustrated in the graphical presentations of Figure 3 are applied to indicate the directions at which the impact factors are leading. More than any of the investigated variables, the response of passenger vehicles to the estimation models appears to be critical at the three-combined provincial level and for each of the provincial cases, namely, Gauteng, KwaZulu-Natal and the Western Cape. This conforms to the findings of Wang et al. [98]. Secondly, we found out that the impact of passenger vehicles on carbon emissions decreases from the Gauteng to Western Cape and then to KwaZulu-Natal provinces within the 5-year study period. In the combined case of the three provinces, the results indicated that a 1% increase in passenger vehicles would increase the possibilities of carbon emissions by 0.065%. However, as proved by Habib et al. [99], the results vary for each of the provincial cases. The notable differences are due to living standard and geographical conditions, as confirmed by He et al. [100] and Yu and Huang [101].

As the three studied provinces are among the most urbanized, the impacts of urbanization on CO₂-eq emissions are not adversely significant because their level of urbanization reaches a high degree of development. Hence, the three provinces become situated places for service industry, as in the works of Li and Li [102], and even more so in the investigations of Sun et al. [103]. Urban residents usually have a high social awareness for the merits of low carbon burning, which results in relatively low carbon economic development, as shown by He et al. [100] and then by Yu and Huang [101]. As rightly debated, the panel estimation revealed that shifting towards a service economy impacts the overall decrease in the energy intensity and optimization of energy allocations, consequently mitigating carbon emissions in the transport sector of the most urban-wise developed provinces, which tallies to the works put forward by Mulder et al. [104], followed by Adebayo and Odugbesan [23].

For the case of the three combined provincial analysis, the GDP is shown to be negative and statistically significant, even at the confidence level of 1%. More so, in the subsamples, GDP was negative and statistically significant for the three separately investigated provinces. Therefore, we confirm the EKC relationship between economic growth and CO₂-eq emissions in the transport sector. The impact of economic growth on CO₂-eq emissions in the province of Gauteng is the lowest (−0.0180) among the three provincial cases. This implies that the effects of GDP on carbon emissions are very low in most developed provincial economies. It confirms that the economic growth model of the most developed province possesses the bulkiest environmental efficiency, as suggested in the works of Song and Wang [105].

The econometric coefficients of energy intensity (EI)—starting from the combined scenario followed by Gauteng, KwaZulu-Natal and Western Cape—are −0.086, 0.549, −0.076 and −0.071, respectively. This shows that energy intensity is one of the critically significant factors that affect CO₂-eq emissions. It also shows the reduction in energy usage from the less developed to most developed provinces. This conforms to the fact that energy efficiency can be positively correlated with economic buoyancy, as is proved by Pan et al. [106]. More so, the results of EI, as with CO₂-eq emissions, possess a direct proportionality constant. The coefficient of transport infrastructural investments for all the investigated cases (−0.038, −0.004, 0.045 and 0.002, respectively) indicate relatively low investments and possess negative impacts on carbon emissions; however, they are not shown to be critical.

6. Conclusions and Recommendations

6.1. Conclusions

With the use of a designed panel dataset for the three main industrial economies (Gauteng, KwaZulu-Natal and Western Cape) of South Africa, multivariate analyses were conducted for 5 years from 2016 to 2020. The paper examined the characteristics of the impact factors and the mitigation potentials of CO₂-eq emissions in the transport sector of South Africa. The study considered the provincial differences by modelling STIRPAT as the basis and expanded with passenger vehicles, transport infrastructural investments, energy intensity, energy (oil) consumption and urbanization to facilitate in-depth interpretations of their complex and interwoven relationships. The carbon emissions coefficient method (CECM) is made applicable to determine the quantification of the forms of fossil fuels utilized in the transportation sector. Furthermore, three estimation methods (LS, ML ARCH and GLM) were employed to obtain accurate estimates, firstly for the combined scenario, and then for the three provincial subsamples.

For the analysis of a combined three-provincial sample, an inverted Kuznets relationship was evident along the urbanization and CO₂-eq emissions axis. This indicates that South Africa’s transport sector is on the threshold and that the carbon emissions are on the verge of decreasing with incremental differences in urbanization. The findings conform to the theory of ecological modernization. There is proof of the distinct characteristics for the three provinces evaluated separately. For the province of Gauteng and the Western Cape, the elasticity of carbon emissions to urbanization was negative for low urbanization and positive for high urbanization. Despite this, high urbanization remains statistically significant. However, there is no correlation to confirm the inverted Kuznets relationship. The KwaZulu-Natal province tested positive at low and negative at high and maintains its elasticity to be econometrically significant. This confirmed an inverted Kuznets relationship.

6.2. Recommendation

Based on the estimated forecast for the conducted research analysis, the following concise recommendations are proffered for the reduction of carbon emissions in the transportation sector as the on-road unit of the sector becomes more of concerning issue:

• Issues around migration, especially emigration to already dense cities, impact urbanization, which consequently affects CO₂-eq emissions in the transport sector of South Africa. In the peculiar case of the Gauteng province, measures should be taken in broadening the responsibilities of agencies and government.
• For further development on clean and quality living, urbanization should be cushioned to abide by low carbon eco-cities regulations. In addition, it is imperative to establish a motivational mechanism for urban low-carbon consumption.
• Advocating comprehensive plans to guide low carbon transportation systems, such as rail and mass public transit, should be strengthened.
• Steps to take multiple measures for improving energy efficiencies and fuel modal shifts with fiscal instruments should be advocated.
• Across the country, better-connected transport infrastructural systems should be erected for better technical functions.