The Effects of Fossil Fuel Consumption-Related CO₂ on Health Outcomes in South Africa

Abstract

The consumption of fossil fuel significantly contributes to the growth of South Africa’s economy but produces carbon dioxide (CO₂), which is detrimental to environmental sustainability with overall effects on health outcomes. This study sought to (i) examine the impacts of fossil energy consumption-related CO₂ emissions on the under-five mortality and infant mortality rates in South Africa and (ii) analyse the causal relationship between fossil energy consumption, CO₂ emissions, and mortality rates in South Africa. Linear and nonlinear ARDL bounds and the Toda–Yamamoto causality test were used to establish the equilibrium property in the long run and the causal effects of the models’ variables. Health outcome data include the under-five mortality rate (MTR1) and infant mortality rate (MTR2). Other explanatory variables include fossil energy consumption (FOC), inflation (Inf), carbon dioxide emissions (CO₂), and government expenditure (GEH). It is evident from the results of linear ARDL that the first lag of the under-five mortality rate in the short run has a positive and significant impact on the under-five mortality rate in South Africa. Holding the other variables constant, the under-five mortality rate in South Africa would increase by 0.630% for every 1% increase in its lagged values. Fossil energy consumption has a positive and significant effect on the under-five mortality rate in South Africa. This significant relationship implies that a 1% increase in fossil energy consumption increases the under-five mortality rate per 1000 persons per year in South Africa by 0.418% in the short run, all things being equal. The results from the Toda–Yamamoto causality test revealed that there is no causality between the under-five mortality rate and both the consumption of fossil fuel and CO₂ emissions in South Africa. The results from nonlinear ARDL presented four separate scenarios. In the short run, during increasing levels of CO₂ in the initial period (lag of CO₂), a 1% increase in CO₂ would decrease the under-five mortality rate by 1.15%. During periods of decreasing levels of CO₂ in the short run, a 1% increase in CO₂ would increase the infant mortality rate by 0.66%. Again, during previous and current periods of decreasing levels of FEC, a 1% increase in FEC would increase the infant mortality rate by 0.45% and 0.32%, respectively. In the long run, during periods of increasing levels of CO₂, a 1% increase in CO₂ would decrease the infant mortality rate by 4.62% whereas during decreasing levels of CO₂, a 1% increase in CO₂ would increase the infant mortality rate by 2.3%. The risk posed by CO₂ emissions and their effects on humans can then be minimised through a government expansionary policy within health programmes.

1. Introduction

The growing importance of energy consumption can no longer be ignored in the modern world since most nations’ economic activities are driven by it. Continued energy consumption must be moderated to mitigate pressure and negative consequences on the environment. The environment needs to be sustained to encourage individuals to live in a manner that must not exert stress on human health and natural resources. The strive for environmental sustainability helps to attain ecological balance in the natural environment across the globe, which in turn protects future generations as it serves to benefit today’s ecosystems. This argument strongly supports the United Nations Brundtland Commission on the need to preserve the present environment without necessarily compromising the potential of future generations to meet their own health demands [1].

Energy has been code-named the oxygen of an economy. This could be true since without energy, countries will fail to run factories, houses, and all forms of transportation. All economic activities aim to produce goods and services to enhance quality of life, improve the standard of living, and promote health outcomes [2]. Thus, energy consumption is a significant indicator of economic productivity.

Fossil energy consumption has enormously contributed to the growth of the economy both in developing and developed countries. Direct and indirect channels are the two significant ways through which energy contributes to economic growth in any given country [3]. The consumption of energy is usually linked to productive activities that increase economic prosperity, create employment opportunities, and enhance value through the transformation, distribution, and extraction of energy products and services. Regarding the indirect channel, energy, with special reference to fossil energy (coal, oil, and gas), is mostly used as input for the various economic sectors. Most industries use fossil energy as input to produce petrochemical and plastic products. Similarly, certain sectors of the agricultural industry rely heavily on fossil energy for the management of crops, particularly for fertilisers and machinery production [4].

Moreover, energy is required to empower the transportation, manufacturing, construction, and service sectors. Clearly, the growth of the economy has a strong connection with the consumption of energy. Any attempt to boost economic growth in any nation will require energy consumption in certain sectors of the nation’s economy. The proliferation or increase in fossil energy consumption causes carbon dioxide (CO₂) emissions. This increases the amount of greenhouse gases responsible for climate change [5]. CO₂ emissions have a direct and indirect impact on people’s health outcomes. The direct impact of CO₂ emissions on South Africa’s health, specifically focusing on infant and under-five mortality rates, is this study’s major concern.

Climate change manifests in higher temperatures, water scarcity, rising sea levels, flooding, drought, human exposure to heat and ultraviolet radiation, and the recurrence of infectious and vector-borne diseases. Moreover, climate change may destabilise agricultural activities and cause uncertainty regarding food systems (World Health Organization) [6]. The CO₂ from various sectors of the economy may also exacerbate air pollution, which poses a significant danger to human health. This affects health outcome indicators, namely, infant mortality rate, years lost to disability, under-five mortality, and low birth weight [7]. The burning question here is: should energy consumption be increased to enhance economic growth at the expense of life quality, or should it be reduced to enhance quality of life? This question pervades the major discussions in the international communities, notably within the WHO.

South Africa is endowed with vast energy resources. It is one of the most resource-rich countries in Southern Africa. In 2014, fossil energy consumption, as a share of the total energy consumption in South Africa, was 87%. South Africa’s overall primary energy consumption was about 5.63 quadrillion British thermal units (BTUs) out of the 5.9 quadrillion BTU production. With South Africa’s total energy intensity of 9 MJ per dollar of GDP, the country’s import energy was −14.5%. Alternative nuclear energy was about 2.7% [8]. The foregoing figures strongly imply that fossil energy and greenhouse energy consumption in South Africa could result in emissions that could be harmful to human health. Furthermore, the increase in South Africa’s population, and the decline in access to stable electricity supply because of continuous load shedding, engendered an increased usage of alternative power supplies in homes, offices, and shops. This, in turn, increases energy consumption and CO₂ emissions in the country, with associated consequences.

The theoretical argument developed by Grossman [9] demonstrates the nexus between health inputs and outputs. Theoretically, it is believed that the environment could serve as an input for health outputs. Such functioning of the environment depends on ecological modernisation and sustainable development theories. These theories promote the utilisation of finite resources with respect to fossil energy consumption in a way that natural resources can be preserved and the needs of future generations are adequately sustained. The increasing use of fossil energy consumption is vital to the improvement of the population’s health status. This is linked to the manufacturing of final products for human consumption [10].

In 2016, fossil-related CO₂ emissions in South Africa amounted to 390,557,850 tons. This reflects a slight decline in CO₂ emissions by −0.49% compared to the 2015 level. South Africa ranked 14th as the emitter of greenhouse gases globally. The elevated CO₂ emissions recorded in South Africa predominantly emanated from heavy reliance on coal. The fact that the energy sector contributed 84% to the overall emissions in 2012 indicates that South Africa’s emission capacity is an average of 464 million metric tonnes. Greenhouse gas emissions grew by 44% from 1990 to 2012, with energy sector emissions increasing by 127 Mt CO₂e within the same time frame [11]. South Africa has witnessed an increase in both the consumption and production of fossil energy [12].

CO₂ emissions from the South African manufacturing and construction industries stood at 500 million tons in 2019. Due to its large consumption of coal, South Africa is one of the largest CO₂ emitters in the world [13]. This concentrated emissions results in harsh weather conditions, elevated greenhouse gas levels, and higher-temperature heat waves. All of these CO₂-emission-related effects have grave consequences for humanity. According to [14], higher-temperature heat waves affect cardiovascular and respiratory health by increasing ozone wave intensity, particularly in developing countries such as South Africa. Youths and children are the most affected by the harsh weather conditions engendered by CO₂ emissions [15].

Thus, this study differentiated the under-five mortality rate from the infant mortality rate in the study’s regression model. This was carried out to allow for the investigation of how each of the variables reacts to CO₂ emissions. While the former is known as the chances of dying between birth and the age of five, expressed per 1000 live births, the latter could be defined as the likelihood of dying between birth and the age of one, expressed per 1000 live births. The relationship between each of these variables and CO₂ is the contribution to the body of knowledge this study seeks to achieve. While fossil energy consumption, including natural gas, oil, coal, and non-renewable resources, is believed to have been formed as prehistoric animals and plants died and were buried by layers of rock, the consumption of such energy is the total of such energy needed for a given process for human use to drive engines and other home tools.

Therefore, the objectives of this study are to examine the impact of CO₂ emissions and fossil energy consumption on the under-five mortality rate in South Africa; analyse the causal relationship between fossil energy consumption, CO₂ emissions, and both infant mortality rate and under-five mortality rate in South Africa; and establish the trade-off relationship. The study covers the period from 1981 to 2022 because South Africa recorded more fossil energy consumption, environmental hazards, and a higher mortality rate in this period, despite efforts by the South African government and international organisations. It is expected that the outcome of this research work will offer policy suggestions that will help reduce the impact of environmental hazards, thereby aiding a sustainable environment both in the short and long run.

Section 2 of this paper presents the literature review comprising relevant concepts, theories, and empirical findings. Section 3 discusses the theoretical framework and research methodology. In Section 4, an analysis is conducted, and the findings are discussed in detail. Section 5 concludes the study and relevant policies are recommended.

2. Literature Review

Ref. [16] defined health as a state of complete physical fitness or the absence of diseases and injuries. For [17], health outcomes can be regarded as human capital servicing. It is a relevant input in the growth process of developed, emerging, and developing countries. Health outcomes have been measured in different ways. Some of the most common measures include life expectancy at birth, infant mortality, years lost to disability, anaemia, under-five mortality, and low birth weight [7]. Energy consumption refers to all of the energy mobilised from various energy sources as part of human efforts in all industrial and technological sectors. Energy consumption has profound implications for people’s health. Energy utilisation can be disaggregated into electricity, petroleum, natural gas, and coal consumption. CO₂ emissions emanate from solid, liquid, and gas consumption, fueled by households and different economic sectors, notably manufacturing and transportation. Rapid population growth and increasing industrial development have also exacerbated the concentration of CO₂ emissions in the atmosphere. Thus, CO₂ constitutes the most significant environmental concern of the 21st century.

Numerous theories in economics are widely applied to health, energy, and environmental issues. Notable among these theories are the Grossman model, the Gary production function, the Environmental Kuznets Curve (EKC), the conservative hypothesis, and the Ramsey–Cass–Koopmans infinitely-lived agent framework. The Gary production theory offers explanations of the implications and interacting effects of energy consumption on production for infant health-related risks. The theory differentiates between two types of health. While the first form of health enters the utility function as an output, the latter serves as an input in the production function [9,18,19]. The Grossman model views people as producers and consumers of health. The theory argues that investment in health will continue until the marginal benefit of health equals the marginal cost. At this point of equilibrium, Grossman argues that an individual’s longevity of life is thus determined endogenously.

Emerging literature on the relationship between CO₂ emissions, energy consumption, and health outcomes exists. Empirical studies are divided into country-specific and cross-country studies and investigations. For cross-sectional studies from 1990 to 2013, [20], for instance, engaged a quantile regression method to investigate the impact of environmental quality on an individual’s health in the Anglophone countries of West Africa (Sierra Leone, Gambia, Ghana, Liberia, and Nigeria). Their findings revealed that CO₂ emissions from liquid and gaseous fuel consumption, public and commercial services, residential buildings, solid fuel consumption, and transport impacts negatively on human health in the subregion.

The authors of some other works have also investigated energy consumption effects and environmental quality on health outcomes in developing, developed, and emerging economies around the world. For example, Ref. [21] analysed the effects household energy consumption have on respiratory disease prevalence in India. Data were obtained from 2012 to 2013 for the study from the District Level Household Survey. A total of 117,752 respondents were diagnosed with various forms of chronic diseases over the course of the study. The study’s outcome revealed that energy consumption strongly impacts respiratory disease prevalence in India. It was recommended that households that adopt solid fuels such as wood, coal, and biomass as sources of energy are more likely to suffer from respiratory-related diseases.

Ref. [16], using data from 1985 to 2016, investigated the nexus between greenhouse gas emissions and health outcomes in Nigeria. An auto-regressive distributive lag model was employed as the study’s estimating technique. The evidence from the study revealed that an increase in greenhouse gas emissions decreases life expectancy significantly. Again, the study further confirmed that greenhouse gas emissions result from fossil fuel combustion through human activities.

The results obtained from the reviewed studies confirmed that certain issues are yet to be resolved. For instance, Refs. [16,22] found that energy consumption has a positive effect on health quality, while [17] reported an insignificant relationship. Furthermore, mixed results have been reported on the causality between CO₂ emissions, energy consumption, and health outcomes [17,23,24]. While a bidirectional causal relationship exists between life expectancy and all of the other explanatory variables, in [23]’s study, CO₂ emissions from coal failed to exhibit a causal relationship. Nevertheless, the causal relationship between health expenditure and other variables is also bidirectional. This implies that the nexus between energy consumption, CO₂ emissions, and health outcomes is still controversial. This warrants further studies on the relationship among these variables.

Most of the existing studies [17,24], for example, only used one indicator of health outcomes. Consequently, they were unable to account for the peculiarity of the other types of health outcomes. Thus, the present study addresses this limitation of existing studies by using two different indicators of health outcomes in South Africa to produce more robust results. CO₂ emissions have been increasing with the rise in energy consumption in South Africa. Considering the worsening quality of life, which has led to an increase in mortality rates, the increasing CO₂ emissions from energy consumption need to be urgently addressed.

3. Model Specification and Methodology

The study’s theoretical framework is the foundational work on Health Production Function (HPF) developed by [25]. Ref. [25] built a model to investigate the impact of medical and non-medical inputs, as well as socio-economic factors and physical conditions, on health outcomes. The health promotion model has been commonly adopted in investigating the impact of medical and non-medical factors [17]. The baseline model of the health promotion function, according to [25], is specified as follows:

$$ht=f\left(md,e\right)$$

(1)

where md, ht, and e represent the medical inputs, health outcomes, and vector of non-medical indicators. The Heath Production Function predicts a positive nexus between health outcomes and medical inputs. This is relevant to the current study because as healthcare improves, it enhances people’s quality of life. However, this study considers the possibility of the law of diminishing returns to scale setting after a certain level of resources has been reached in supporting the healthcare system [25]. Regarding the effects of non-medical input on health outcomes, Ref. [25]’s results differ insofar as the direction of the relationship. This is particularly the case among the non-medical indicators. Thus, it is concluded that the effects are dissimilar. This is because non-medical variables comprise the socio-economic and physical environments. Each of these variables impacts health outcomes differently. Following [25]’s model specification in Equation (1), the present study’s model is estimated as follows:

$$HT{O}_t=f(FE{C}_t,{{CO}_2}_t,GE{H}_t,In{f}_t)$$

(2)

The general specification of the model, with the white noise term (u), was specified as follows:

$$HT{O}_t={\beta }_0+{\beta }_{1}FE{C}_t+{\beta }_2{{CO}_2}_t+{\beta }_{3}GE{H}_t+{\beta }_{4}In{f}_t+\mu$$

(3)

where $HTO$ represents the health outcomes as measured by the under-five mortality rate (MTR) and the infant mortality rate (MTR2). The specifications of the explanatory variables are as follows: fossil energy consumption was proxied by FEC, carbon dioxide emissions was proxied by CO₂, inflation rate was proxied by Inf, and medical input was proxied by GEH, which also stands for government expenditure on health. Non-medical indicators include FEC, CO₂, and Inf, which denote fossil energy consumption, carbon dioxide emissions, and inflation rate, respectively. While FEC and CO₂ are indicators of the physical environment, Inf stands for the socio-economic indicators. β₀ designates the intercept; β₁, β₂, β₃, and β₄ are parameter coefficients denoting fossil energy consumption (FEC), carbon dioxide emissions (CO₂), government expenditure on health (GEH), and inflation (Inf), and t denotes the time-series properties of the model.

3.1. Definition, Measurement, and Justification of Variables

Health Outcomes ($HTO$): Ref. [17] argued that health outcomes, viewed as human capital, are important input components in the growth process. The mortality rate has recently been widely used to measure the quality and length of life [17,22]. The under-five mortality rate expresses the probability per 1000 that newborn babies will not survive to the age of five; while the infant mortality rate describes the total number of infants dying before the age of one, per 1000 live births in a year. Although the adoption of the under-five mortality rate and the infant mortality rate are rare proxies for health outcomes in the literature, where available, these variables have generally been used separately. However, the adoption of these variables for health outcomes in the current study seems justified since other researchers have adopted similar variables to proxy health outcomes [24,26,27].

Fossil Energy Consumption (FEC): This is the consumption of energy from various sources such as coal, oil, petroleum, and natural gas products. Energy consumption may enhance the reduction in the infant mortality rate and the under-five mortality rate, provided that the revenue generated from the sales of solid fuel is utilised to improve human welfare. Conversely, energy consumption may increase the under-five and infant mortality rates if the pollution linked to its consumption negatively affects the physical environment. Energy consumption, as denoted by FEC, is measured in terms of fossil fuel energy consumption (% of total). Refs. [24,26] justified this variable in their studies.

Carbon Dioxide Emissions (CO₂): These emissions are the outcomes of fossil fuels burning in hydrocarbons and industries such as cement manufacturing. They also include emissions from the flaring of gases that could be in partly solid–liquid form and the gaseous consumption of fuel [23]. The emission of CO₂ is expected to have a direct effect on health outcomes. This implies that CO₂ emissions increase the under-five mortality rate. This could be because CO₂ emissions cause greenhouse effects. This, in turn, results in higher heat levels that pose a significant danger to humanity. CO₂ is measured in metric tons per capita. A comprehensive justification of this variable was provided by [26].

Government Expenditure on Health: This is the amount of money expended on healthcare services by the government. Health expenditure is the total of all expenditures incurred in delivering health services to the people. This includes nutrition and family planning activities, and emergency aid [8]. By implication, increasing health expenditure should improve health outcomes. However, empirical evidence from the extant studies disproves this conclusion. Research [25] revealed that reports from various authors on the nexus between health education and health outcomes are either contradictory or weak. Thus, the present study adopted the government’s recurrent health expenditure as a percentage of the total recurrent expenditure to proxy the health expenditure. This is because data on the total expenditure on health are unavailable. This is not the first study to use government expenditure as a control variable in the fossil energy consumption mortality rate nexus. Refs. [24,26] justified the use of the same variable in their studies.

Inflation Rate: The inflation rate is the persistent increase in the general price level. [28] argues that an increase in food prices on international markets translates to higher domestic prices, which affects people’s purchasing power. The rise in the general price level worsens people’s standard of living and negatively influences their health conditions. Furthermore, inflation-induced pressure can cause mental health issues due to socio-economic frustration [29]. Hence, inflation is expected to worsen the mortality rates of the under-fives and infants. The consumer price index serves as the proxy for inflation in this study. [25] provided a comprehensive justification of the inflation rate and its relationship with mortality rates.

3.2. Data Sources

This study uses time-series data from 1981 to 2022 to analyse the impacts of energy consumption and CO₂ emissions on health outcomes in South Africa. Data on health outcomes, including infant mortality rate, under-five mortality rate, energy consumption, inflation, and carbon dioxide emissions, were sourced from the World Bank Development Indicators (WBDIs) [13].

3.3. Estimation Methods

For the preliminary investigation of the time-series data, descriptive statistics along with a correlation matrix were employed. The study adopted the Phillip–Peron tests and the augmented Dickey–Fuller (ADF) test to conduct a stationarity assessment. Furthermore, in this study, we used the linear and nonlinear ARDL bounds test to ascertain asymmetry effects, and the equilibrium properties of the study’s model. The study investigated the nexus and nature of the nexus between fossil fuel consumption-related CO₂ and health outcomes in South Africa by adopting the ARDL estimation built by [30]. Although the ARDL can handle various models with I(1), or a mixture of order one and zero-order, the estimating technique is suitable for analysing the adjustment speed from the short-run disequilibrium to equilibrium in the long run.

3.4. The Panel ARDL with Symmetric Properties

The symmetric properties of the panel ARDL could be applied as the representations of the augmented ARDL(a,b,c,d,e) model for Equation (3), given below:

$$\begin{array}{c}HT{O}_t={\beta }_0+{\displaystyle \sum _{i=0}^a}{\beta }_{1}FE{C}_{t-i}+{\displaystyle \sum _{i=0}^b}{\beta }_2{{CO}_2}_{t-i}+{\displaystyle \sum _{i=0}^c}{\beta }_{3}GE{H}_{t-i}\\ +{\displaystyle \sum _{i=0}^d}{\beta }_{4}In{f}_{t-i}+{\displaystyle \sum _{i=1}^e}{\beta }_{5}HT{O}_{t-i}+{\epsilon }_t\end{array}$$

(4)

Equation (4) represents the model when a stable long-run nexus is strongly supported by the Wald test. ${\beta }_0$ is the intercept, ${\epsilon }_t$ is the white noise at time t, and ${\beta }_1-{\beta }_5$ are the parameters for the estimations of the long-run coefficients. For the analysis of the short-run relationship, this paper specifies the ECM as follows:

$$\begin{array}{c}\Delta HT{O}_t={\alpha }_0+{\alpha }_{1}HT{O}_{t-1}+{\displaystyle \sum _{i=0}^a}{\alpha }_3\Delta FE{C}_{t-i}+{\displaystyle \sum _{i=0}^b}{\alpha }_4\Delta {{CO}_2}_{t-i}+{\displaystyle \sum _{i=0}^c}{\alpha }_5\Delta GE{H}_{t-i}\\ +{\displaystyle \sum _{i=0}^d}{\alpha }_6\Delta In{f}_{t-i}+{\displaystyle \sum _{i=1}^e}{\alpha }_7\Delta HT{O}_{t-i}+\phi ec{m}_{t-1}+{\epsilon }_t\end{array}$$

(5)

where $ec{m}_{t-1}$ is the lagged one time error correction term, ${\alpha }_1$ and ${\alpha }_5$ are the short-run parameters in the estimates, $\phi$ denotes the adjustment speed running from short-run disequilibrium to long-run equilibrium, and ${\epsilon }_t$ is the error term. While the pervious empirical works made use of the traditional Granger causality test to measure the causality direction, the current research adopted the approach offered by Toda and Yamamoto (1995) for causality estimation with the modified WALD statistic (χ² distribution). This method has been used extensively in the existing literature. [31,32] checked the direction of causality when the variables are stationary after first difference.

3.5. The Asymmetric ARDL

To generate the nonlinear ARDL model, this study augmented the model developed by [33]. Consequently, Equation (6) below reflects the asymmetric relationship between health outcomes and the factors that determine them.

$$\begin{array}{c}\Delta HT{O}_t ={\alpha }_0+{\alpha }_{1}HT{O}_{t-1}+{\alpha }_2^{+}lag1 FE{C}_{t-1}^{+}+{\alpha }_2^{-}lag1 FE{C}_{t-1}^{-}+\\ {\alpha }_3^{+}lag1 GE{H}_{t-1}^{+}+{\alpha }_3^{-}lag1 GE{H}_{t-1}^{-}+{\alpha }_4^{+}lag1 In{f}_{t-1}^{+}+\\ {\alpha }_4^{-}lag1 In{f}_{t-1}^{-} {\alpha }_5^{+}lag1 {{CO}_2}_{t-i}^{+}+{\alpha }_5^{-}lag1 {{CO}_2}_{t-i}^{-}{\sum }_{i=0}^a{\alpha }_3\Delta +\\ {\sum }_{i=0}^b{\beta }_1^{+}\Delta lag1 FE{C}_{t-1}^{+}+{\beta }_1^{-}\Delta lag1 FE{C}_{t-1}^{-})+{\sum }_{i=0}^c({\gamma }_1^{+}\Delta lag1 GE{H}_{t-1}^{+}+\\ {\gamma }_1^{-}\Delta lag1 GE{H}_{t-1}^{-})+{\sum }_{i=0}^d({\mu }_1^{+}\Delta lag1 In{f}_{t-1}^{+}+{\mu }_1^{-}\Delta lag1 In{f}_{t-1}^{-})+\\ ({\mu }_1^{+}\Delta lag1 Co{{CO}_2}_{t-i}^{+}+{\mu }_1^{-}\Delta lag1 {{CO}_2}_{t-1}^{-}){\sum }_{i=1}^e{\alpha }_6\Delta HT{O}_{t-i}+\phi ec{m}_{t-1}+{\epsilon }_t\end{array}$$

(6)

where $FE{C}_t^{+}$ estimates the fossil energy consumption that increases health outcomes, $FE{C}_t^{-}$ estimates the fossil energy consumption that decreases health outcomes; ${{CO}_2}_t^{+}$ indicates the level of carbon dioxide that positively impacts health outcomes; and ${{CO}_2}_t^{-},$ indicates the level of carbon dioxide that could cause a risk to health outcomes.

3.6. Limitations of the Study

This study adopted the government’s health expenditure to proxy the health-outcome variables. This is because data on the government’s recurrent health expenditure, which are the most preferred data, are not readily available and could not be accessed during data gathering for the countries under investigation. This may exaggerate or inflate the results; however, since data availability is recognised as one of the major challenges of developing economies, we have been encouraged to continue despite this limitation.

We measured CO₂ emissions and fossil energy consumption—the variables that need to be measured to really get at how climate impact might impact infant mortality. For example, am increase in income from economic productivity needs to be viewed by income level. And climate change impacts need to be viewed from the perspective of the actual health threat, for example, infectious disease rates, dangerous weather incidents, food security, water quality, and availability.

3.7. Justification of the Adopted Estimation Technique

The auto regressive distributive lag (ARDL) was formally adopted by [30]. The present study engaged with this estimation technique to investigate the co-integrating relationship among the regressors in this study. This is because it can cope with both linear and nonlinear regression, depending on whether the asymmetric relationship is a concern of the study. The adoption of the time-series ARDL became necessary because of the peculiarities inherent in the ARDL time-series model. It is highly flexible, as it is less restrictive of the possibility of integrating variables of the same order. This model is applicable to variables integrated at both order I(1) and I(0). Ref. [30] suggested that ARDL predicts consistent and dependable analysis for long-run coefficients, provided that such variables are asymptotically normally distributed, irrespective of their order of integration, I(1) or I(0). Despite its endogeneity, the ARDL model provides reliable coefficients. This is because it uses the lags of the findings and independent variables [30]. The preliminary problems commonly experienced in the conventional co-integration estimation are controlled when the ARDL is adopted [34].

3.8. Limitation of the ARDL Approach

Ref. [30] argued that the method is an approach that is fundamentally of a single equation. He further stated that another major limitation of the ARDL is that the approach seems to only restrict one level nexus among the variables under investigation and fails to allow for a greater number of long-run relationships.

4. Empirical Results

4.1. The Study’s Descriptive Statistics

The under-five mortality rate (MTR), infant mortality rate (MTR2), fossil energy consumption (FEC), carbon dioxide emissions (CO₂), and inflation rate (Inf) descriptive statistics are presented in

Table 1. Summary statistics.

	MTR	MTR2	FEC	CO2	GEH	Inf
Mean	61.79000	43.50250	86.96291	8.927220	42.14262	8.771061
Maximum	92.20000	67.00000	90.50629	9.979458	54.97161	18.65493
Minimum	34.50000	27.50000	84.24343	7.727642	33.43943	−0.692030
Std. Dev.	15.90247	10.29783	1.653438	0.639679	7.730395	4.652915
Obs.	40	40	35	37	39	40

Source: Authors’ computation, 2024.

below. A close examination of the table reveals that the mean values of MTR and MTR2 are closer to the minimum than the maximum, suggesting that the infant mortality and under-five mortality rate in South Africa are low. On average, the value of health outcomes is far from the maximum. In other words, the mean value is closer to the minimum observation. This implies that the unit of fossil energy consumed by households, industries, and other economic actors is low over the period under investigation. This could be attributed to the imposition of an eco-tax or carbon tax, which is expected to discourage the consumption of fossil energy. Conversely, the mean values of CO₂ and Inf are closer to the maximum than the minimum. The implication is that those CO₂ emission and inflation rates are on the high side, concerning health outcomes. This could increase the risk posed to the under-five mortality rate and infant mortality rate in South Africa.

For GEH, the table shows that the average value draws closer to the minimum than the maximum, implying that government expenditure on healthcare is low in South Africa. Furthermore, the value of the standard deviation suggests that the volatility exhibited by FEC, CO₂, GEH, and Inf is low compared to MTR and MTR2. As can be seen in

Table 2. Correlation matrix.

MRT: Under-5 Mortality Rate
	MTR	FEC	CO2	GEH	Inf
MTR	1
FEC	0.360	1
CO2	0.069	0.547	1
GEH2	−0.529	0.131	0.268	1
INFR	0.130	0.393	0.166	−0.476	1
MRT2: Infant Mortality Rate
	MTR2	FEC	CO2	GEH	Inf
MTR2	1
FEC	0.476	1
CO2	0.061	0.547	1
GEH2	−0.702	0.131	0.268	1
INFR	0.490	0.393	0.166	−0.476	1

Source: Authors’ computation, 2024. Note: Health outcome data include under-five mortality rate (MTR1) and infant mortality rate (MTR2). Other explanatory variables include fossil energy consumption (FOC), inflation (Inf), carbon dioxide emissions (CO₂), and government expenditure (GEH).

, which reports the strength and direction of the relationship among the variables, the coefficients are below 0.7, which is the rule of thumb for multicollinearity. Thus, it is reasonable to conclude that no perfect relationship exists among the independent variables.

4.2. Unit Root Test

This study utilised the augmented Dickey–Fuller (ADF) and Phillip–Perron tests to investigate the stationary property of the dependent and independent variables used for the analysis. The need for the adoption of two different methods of testing is to authenticate the consistency of stationarity among variables. The results of the test are presented in

Table 3. Augmented Dickey–Fuller and Phillp–Peron unit root tests.

ADF				Phillp–Peron
Variables	t-Stat	Sig. Level	Order of Integration	t-Stat	Sig. Level	Order of Integration
CO2	−2.555377	0.0153	I(0)	−2.822499	0.0651	I(0)
FEC	−6.48842	0.0000	I(1)	−6.440005	0.0000	I(1)
GEH	−5.526667	0.0000	I(1)	−5.620110	0.0000	I(1)
Inf	−5.363541	0.0001	I(1)	−9.676244	0.0000	I(1)
Lex	−3.826002	0.0064	I(0)	−3.917764	0.0005	I(0
Mtr1	−1.753107	0.0756	I(0)	8.834571	0.0000	I(0)
Mtr2	10.27119	0.0000	I(0)	−2.208084	0.0335	I(0)

Source: Authors’ computation, 2024.

. The probability values of the ADF and PP indicate that both dependent variables (the under-five mortality rate and the infant mortality rate), Lex and CO₂, are stationary at this level. All other variables, namely inflation (Inf), government health expenditure (GEH), and fossil energy consumption, are non-stationary at this level, indicating that other variables are stationary after the first difference, with this being the I(1) series. Conversely, the fossil energy consumption, government health expenditure, and inflation rate become stationary at 1% after the first difference.

Figure 1 reflects the result of the CUSUM test of stability. The overall stability nature of all the variables used in our model was tested under this model. The CUSUM test result from the recursive analysis above clearly indicates that the model is stable at 5%. The conclusion was premised on the stability theory, which stipulates that for the stability of a model, the blue line must lie in between the two red boundaries (lines). Consequently, this model is very stable and reliable.

Figure 2 explains the measurement of lag strength. The study further used the automatic lag selection process to analyse the strength of the chosen AIC model over all of the other criteria such as HQ and SIC. A criteria graph was employed to describe the top twenty (20) ARDL models. The lower the AIC, the better the model; the model (4,4,4,4,1,4) is considered the best because it offers the lowest positive AIC value [35]).

Table 4. The result of the test for heteroskedasticity.

Heteroskedasticity Test by Breusch–Pagan–Godfrey
F-statistic	0.728612	Prob. F(26,9)	0.7495
Obs R-squared	24.40535	Prob. Chi-Square(26)	0.5528
Scaled explained SS	2.351869	Prob. Chi-Square(26)	1.0000

Source: Authors’ computation, 2024.

explains the result of heteroskedasticity. From the Breusch–Godfrey heteroskedasticity LM test conducted, the null hypothesis cannot be rejected because the probability is greater than 5%. Consequently, no heteroskedasticity exists in the model estimated.

5. Empirical Results

Table 5. Result of the linear ARDL analysis.

	Model 1: Under-5 Mortality Rate (MRT) Selected Model: ARDL(4,4,4,4,1,4)			Model 2: Infant Mortality Rate (MRT2) Selected Model: ARDL(4,4,4,4,1,4)
Short-Run Results
Variables	Coef.	t-Stat	Prob.	Coef.	t-Stat	Prob.
D(MRT(−1))	0.630	8.448	0.000 ***	-	-	-
D(MTR2(−1))	-	-	-	1.169	10.778	0.000 ***
D(MTR2(−2))	-	-	-	−0.567	−5.287	0.000 ***
D(FEC)	0.418	2.431	0.024 **	0.137	1.539	0.140
D(FEC(−1))	-	-	-	−0.113	−1.360	0.190
D(CO2)	−0.442	−1.590	0.126	−0.087	−0.594	0.560
D(GEH)	−0.042	−0.525	0.605	0.033	0.909	0.375
D(GEH(−1))	-	-	-	−0.116	−3.184	0.005 ***
D(INFR)	−0.109	−2.011	0.057	−0.012	−0.458	0.652
D(INFR)	0.169	2.820	0.010 ***	-	-	-
CointEq(−1)	−0.172	−8.210	0.000 ***	−0.083	−2.887	0.009 ***
Long-Run Results
Variables	Coef.	t-Stat	Prob.	Coef.	t-Stat	Prob.
FEC	4.104	5.911	0.000 ***	1.364	1.327	0.201
CO2	−2.575	−1.462	0.158	−1.053	−0.570	0.575
GEH	−2.208	−10.853	0.000 ***	−1.369	−7.518	0.000 ***
INFR	−2.087	−6.886	0.000 ***	−0.834	−2.541	0.012 ***
C	−161.561	−2.852	0.009 ***	−4.917	−0.056	0.956
Post Estimation Results
Jarque–Bera Stat		0.490 (0.783)		2.633 (0.268)
Breusch–Godfrey Serial Correlation LM Test		3.486 (0.050)		7.208 (0.005)

10%, 5%, and 1% significant levels are denoted by *, **, and ***, respectively. The probability values of the F statistic are presented in parentheses. Source: Authors’ computation, 2024. Note: Health outcome data include the under-five mortality rate (MTR1) and infant mortality rate (MTR2). Other explanatory variables include fossil energy consumption (FOC), inflation (Inf), carbon dioxide emissions (CO₂), and government expenditure (GEH).

reports the results of the first and second models. The short-run and long-run results of the first model are presented in the second column, while the short-run and long-run results of the second model are presented in the third column. Looking closely at the results of the first model, it is evident that the first lag of the under-five mortality rate in the short run has a positive and significant impact on the under-five mortality rate in South Africa. Holding other variables constant, the under-five mortality rate in South Africa increases by 0.630% for every 1% increase in its lagged values. Similarly, fossil energy consumption has a positive and significant effect on the under-five mortality rate in South Africa. This significant relationship implies that a 1% increase in fossil energy consumption increases the under-five mortality rate per 1000 persons per year in South Africa by 0.418% in the short run, all things being equal. Similarly, the coefficient of the inflation rate is positive and significantly influences the under-five mortality rate in South Africa. This result suggests that the mortality rate of children below the age of five per 1000 persons per year increases by 0.169% for every 1% increase in the inflation rate, given that all the other variables are constant.

In the long run, all of the independent variables in the first model are significant, except for CO₂ emissions. For instance, the coefficient of fossil energy consumption is positive and significant at 1%, suggesting that the under-five mortality rate increases by 4.104% for every 1% increase in fossil energy consumption, all things being equal. Government health expenditure has a significant adverse effect on the under-five mortality rate in South Africa. In other words, for every 1% increase in the South African government’s health expenditure, the under-five mortality rate declines by 2.208%, provided that the values of all of the other variables remain the same. Likewise, the inflation rate has a significant negative impact on the under-five mortality rate in South Africa. Thus, holding the effect of fossil energy consumption, CO₂ emissions, and government expenditure on health constant, a 1% increase in the inflation rate reduces the South African under-five mortality rate by 2.087% in the long run. It also reduces South Africa’s CO₂ emissions by 2.575%. These two results seem counterintuitive and unexpected. However, following the principles of the Philip curve, an increase in the inflation rate could induce more employment in the system. Since this could result in more welfare for households, we expect a reduction in the South African under-five mortality rate by 2.087% in the long run. Similarly, industrial CO₂ emissions may decrease when industries become more labour intensive than capital intensive.

As for the second model, the empirical results reveal a significant positive relationship between the infant mortality rate and its first lag and a significant negative relationship between the infant mortality rate and its second lag. Put differently, the infant mortality rate increases by 1.169% for every 1% increase in its first lag and decreases by 0.567% for every 1% increase in its second lag in the short run, all things being equal. While the government’s expenditure on health is positive and insignificant, the first lag of government expenditure on health is significant at a 1% level of significance, though with a negative impact on the infant mortality rate in the short run. Given that all of the other variables are constant, this result implies that a 1% increase in the first lag of the government’s health expenditure reduces the infant mortality rate by 0.116% in the short run in South Africa.

The coefficients of the co-integration equations of the first and second models are negative, below one, and significant. This is congruent with the a priori expectation and justifies the results of the ARDL Bound, which revealed the existence of a long-run relationship among the variables in the first and second models. Furthermore, the values of the first and second models’ co-integration equations indicate that for any temporary shock, the first and second models adjust back to the equilibrium at a speed of 0.172% and 0.083%, respectively. Conclusively, the results of the post-estimation tests indicate that the adopted models are suitable for the analysis.

After the appropriate lag had been selected, the ARDL Bound was employed to investigate the existence of a long-run relationship among the variables in the two models. As shown in

Table 6. ARDL bound test results.

MRT: Under-5 Mortality Rate					MRT2: Infant Mortality Rate
F-Bounds Test		Null Hypothesis: No Levels of Relationship			F-Bounds Test	Null Hypothesis: No Levels of Relationship
T-Statistic	Value	Sign.	I(0)	I(1)	Value	Sign.	I(0)	I(1)
F-statistic	15.94627	10%	2.45	3.52	6.569643	10%	2.45	3.52
K	4	5%	2.86	4.01	4	5%	2.86	4.01
		2.5%	3.25	4.49		2.5%	3.25	4.49
		1%	3.74	5.06		1%	3.74	5.06

Source: Authors’ computation, 2024.

, the values of the ARDL bounds’ F-statistic of the two models are higher than the respective values of the lower I(0) and upper bounds I(1). These results imply that the null hypothesis of no levels relationship can be rejected for the two models, and it can be concluded that the variables are co-integrated.

6. Granger Causality Results

Table 7. The results of the LA-VAR Granger causality test.

Case	Null Hypothesis	Chi-Sq	Prob.	Decision
Under-5 Mortality Rate Model
1	MTR does not Granger Cause FEC FEC does not Granger Cause MTR	2.630 1.833	0.269 0.400	No Causality
2	MTR does not Granger Cause CO2 CO2 does not Granger cause MTR	0.536 3.850	0.765 0.146	No Causality
3	MTR does not Granger Cause GEH GEH does not Granger Cause MTR	31.936 1.972	0.000 *** 0.373	Unidirectional Causality
4	MTR does not Granger Cause INF INFR does not Granger Cause MTR	10.936 0.704	0.004 *** 0.703	Unidirectional Causality
Infant Mortality Rate Model
1	MTR2 does not Granger Cause FEC FEC does not Granger Cause MTR2	1.747 1.889	0.623 0.598	No Causality
2	MTR2 does not Granger Cause CO2 CO2 does not Granger Cause MTR2	3.505 2.212	0.320 0.530	No Causality
3	MTR2 does not Granger Cause GEH GEH does not Granger Cause MTR2	14.120 7.391	0.003 *** 0.060 *	Bi-directional Causality
4	MTR2 does not Granger Cause INFR INFR does not Granger Cause MTR2	2.829 6.143	0.419 0.105 *	Unidirectional Causality

1% and 10% significance levels are denoted by *** and *, respectively. Source: Authors’ computation, 2024. Note: Health outcome data include under-five mortality (MTR1) and infant mortality rate (MTR2). Other explanatory variables include fossil energy consumption (FOC), inflation (Inf), carbon dioxide emissions (CO₂), and government expenditure (GEH).

explains the result of the LA-VAR Granger causality test. The Toda–Yamamoto causality test was employed to investigate the direction of the causality among the variables. The results are presented in Table 7. The results reveal the absence of causality between the under-five mortality rate, fossil energy consumption, and CO₂ emissions in South Africa. In other words, no directional causality exists between the under-five mortality rate, fossil energy consumption, CO₂ emissions, and the under-five mortality rate in South Africa. However, a unidirectional causality running from the under-five mortality rate to the government’s health expenditure and the inflation rate was recorded in South Africa between 1980 and 2019. This implies that an increased mortality rate prompts the government to increase its expenditure on health. This boosts the ability of health facilities and infrastructure to cater to the needs of the people and enhance South Africans’ health quality. Similarly, a unidirectional causality flowing from the under-five mortality rate to the inflation rate was established. This may be attributed to the high prices emanating from an increase in the South African government’s health expenditure.

The results of the second model mirror those of the first model, except for the feedback relationship between the infant mortality rate and the government’s health expenditure, as well as the unidirectional causality from the inflation rate to the infant mortality rate. The bi-directional causality reflects the government’s efforts to lower the infant mortality rate by channelling more funds to healthcare centres and boosting health facilities and infrastructure. Inflation may also worsen health quality and increase the infant mortality rate. This is because higher food prices increase economic hardship, causing pain for individuals and increasing the mortality rate.

Table 8. Nonlinearity test using Teräsvirta sequential smooth threshold linearity tests.

Test for Nonlinearity Using MTR(−3) as the Threshold Variable
Taylor Series Alternatives: b0 + b1 × s [+ b2 × s^2 + b3 × s^3 + b4 × s^4]
Linearity Tests
Null Hypothesis	F-statistic	d.f.	p-value
H04: b1 = b2 = b3 = b4 = 0	5.529617	(5, 22)	0.0019
H03: b1 = b2 = b3 = 0	5.529617	(5, 22)	0.0019
H02: b1 = b2 = 0	5.529617	(5, 22)	0.0019
H01: b1 = 0	5.529617	(5, 22)	0.0019
The H0i test uses the i-th order Taylor expansion (bj = 0 for all j > i).
Teräsvirta Sequential Tests
Null Hypothesis	F-statistic	d.f.	p-value
H3: b3 = 0	NA	(0, 22)	NA
H2: b2 = 0|b3 = 0	NA	(0, 22)	NA
H1: b1 = 0|b2 = b3 = 0	5.529617	(5, 22)	0.0019
All tests are based on the third-order Taylor expansion (b4 = 0).
The linear model is rejected at the 5% level using H03.
Recommended model: first-order logistic
Pr(H3) < = Pr(H2) or Pr(H1) < = Pr(H2)
Escribano–Jorda Tests
Null Hypothesis	F-statistic	d.f.	p-value
H0L: b2 = b4 = 0	2.949831	(1, 21)	0.1006
H0E: b1 = b3 = 0	0.520373	(5, 21)	0.7580
All tests are based on the fourth-order Taylor expansion.
The linear model is rejected at the 5% level using H04.

Source: Authors’ computation, 2024.

explains the nonlinearity test through the use of Teräsvirta sequential smooth threshold linearity tests. The Teräsvirta sequential smooth threshold linearity tests were employed to test whether the dependent variable has a linear or a nonlinear relationship with the independent variables in the model. If the model analysed is a nonlinear model, the values would be significant, as indicated in the case of H1 above. The results of the test presented in Table 5 indicate that the linear model is rejected at the 5% level using H03, as the probability values of the sunder-five and infant mortality rates are significant. The implication of these results is that the models are nonlinear. This further justified the use of nonlinear ARDL for the analysis.

Table 9. Nonlinear ARDL cointegration and short-run form estimates for the under-five mortality rate.

Dependent Variable: MTR
Selected Model: ARDL(2, 2, 1, 0, 2, 0, 0, 2, 0)
Variable	Coefficient	Std. Error	t-Statistic	Prob.
D(MTR(−1))	0.549127	0.069940	7.851345	0.0000
D(CO2_POS)	−0.662872	0.746586	−0.887872	0.3857
D(CO2_POS(−1))	−1.154347	0.563362	−2.049034	0.0545
D(CO2_NEG)	0.549669	0.529105	1.038867	0.3119
D(FEC_POS)	0.382327	0.384257	0.994977	0.3323
D(FEC_NEG)	0.137507	0.270160	0.508983	0.6166
D(FEC_NEG(−1))	0.975217	0.336541	2.897768	0.0092
D(GEH2_POS)	0.150449	0.082144	1.831520	0.0827
D(GEH2_NEG)	−0.443526	0.208144	−2.130861	0.0464
D(INF_POS)	−0.197226	0.122877	−1.605066	0.1250
D(INF_POS(−1))	0.571024	0.120170	4.751791	0.0001
D(INF_NEG)	−0.460640	0.115009	−4.005237	0.0008
CointEq(−1)	−0.209677	0.027029	−7.757572	0.0000

Source: Authors’ computation, 2024. Note: Health outcome data include the under-five mortality rate (MTR1) and infant mortality rate (MTR2). Other explanatory variables include fossil energy consumption (FOC), inflation (Inf), carbon dioxide emissions (CO₂), and government expenditure (GEH).

showcases the short-run results of the asymmetry effects of the under-five mortality rate during an increasing and decreasing scenario. During an increasing level of CO₂ in the initial period (lag of CO₂), a 1% increase in CO₂ would decrease the under-five mortality rate by 1.15%, all things being equal. However, during a decreasing level of CO₂, no impact of CO₂ and FEC is felt on the under-five mortality rate. Hence, they are not statistically significant.

Table 10. Nonlinear ARDL cointegration and long-run form estimates for the under-five mortality rate long-run coefficients.

Variable	Coefficient	Std. Error	t-Statistic	Prob.
CO2_POS	−5.682400	3.298812	−1.722559	0.1012
CO2_NEG	6.038897	2.206958	2.736298	0.0131
FEC_POS	1.823414	1.729215	1.054475	0.3049
FEC_NEG	−4.211236	1.372107	−3.069176	0.0063
GEH2_POS	0.717527	0.399511	1.796015	0.0884
GEH2_NEG	−2.115285	0.976855	−2.165404	0.0433
INF_POS	−4.759683	0.431638	−11.027026	0.0000
INF_NEG	−2.196906	0.586598	−3.745166	0.0014
C	81.429859	4.372275	18.624140	0.0000

Source: Authors’ computation, 2024. Note: Health outcome data include the under-five mortality rate (MTR1) and infant mortality rate (MTR2). Other explanatory variables include fossil energy consumption (FOC), inflation (Inf), carbon dioxide emissions (CO₂), and government expenditure (GEH).

reflects the long-run findings of the asymmetry effects of the under-five mortality rate during an increasing and decreasing scenario. During periods of decreasing levels of CO₂ and FEC, a 1% increase in CO₂ would increase the under-five mortality rate by 6.04%, all things being equal. However, FEC would decrease the under-five mortality rate by 4.21%. During increasing levels of CO₂ and FEC, no impact of CO₂ and FEC would be felt on the under-five mortality rate. Hence, they are not statistically significant.

Table 11. Nonlinear cointegration and short-run form of ARDL estimates for infant mortality rate.

Dependent Variable: MTR2
Selected Model: ARDL(2, 1, 0, 2, 2, 0, 0, 2, 0)
Cointegrating Form
Variable	Coefficient	Std. Error	t-Statistic	Prob.
D(MTR2(−1))	0.437050	0.069343	6.302702	0.0000
D(CO2_POS)	−0.699606	0.435603	−1.606063	0.1248
D(CO2_NEG)	0.659153	0.272359	2.420160	0.0257
D(FEC_POS)	−0.019753	0.189553	−0.104207	0.9181
D(FEC_POS(−1))	−0.343358	0.234436	−1.464612	0.1594
D(FEC_NEG)	0.322777	0.160885	2.006266	0.0593
D(FEC_NEG(−1))	0.445536	0.192483	2.314674	0.0320
D(GEH2_POS)	0.104211	0.044113	2.362398	0.0290
D(GEH2_NEG)	−0.389433	0.118315	−3.291487	0.0038
D(INF_POS)	−0.091121	0.070118	−1.299551	0.2093
D(INF_POS(−1))	0.303518	0.068039	4.460961	0.0003
D(INF_NEG)	−0.300177	0.063664	−4.715001	0.0002
CointEq(−1)	−0.282903	0.035302	−8.013809	0.0000

Source: Authors’ computation, 2024. Note: Health outcome data include the under-five mortality rate (MTR1) and infant mortality rate (MTR2). Other explanatory variables include fossil energy consumption (FOC), inflation (Inf), carbon dioxide emissions (CO₂), and government expenditure (GEH).

reflects the short-run findings of the asymmetry effects of the infant mortality rate during an increasing and decreasing scenario. During periods of decreasing levels of CO₂, a 1% increase in CO₂ would increase the infant mortality rate by 0.66%.

Again, during previous and current periods of decreasing levels of FEC, a 1% increase in FEC would increase the infant mortality rate by 0.45% and 0.32%, all things being equal. However, no impact of CO₂ and FEC was felt on the infant mortality rate for any increasing scenario of CO₂ and FEC, both in the present and previous periods. Hence, they are not statistically significant.

Table 12. Nonlinear cointegration and long-Run form ARDL estimates of infant mortality rate long-run coefficients.

Variable	Coefficient	Std. Error	t-Statistic	Prob.
CO2_POS	−4.620310	1.897508	−2.434936	0.0249
CO2_NEG	2.329958	0.871931	2.672181	0.0151
FEC_POS	0.351869	1.004603	0.350257	0.7300
FEC_NEG	−1.135061	0.525116	−2.161543	0.0436
GEH2_POS	0.368364	0.158335	2.326488	0.0312
GEH2_NEG	−1.376557	0.379723	−3.625161	0.0018
INF_POS	−1.962164	0.171275	−11.456232	0.0000
INF_NEG	−1.061059	0.247890	−4.280358	0.0004
C	61.854048	2.032172	30.437411	0.0000

Source: Authors’ computation, 2024. Note: Health outcome data include under-five mortality (MTR1) and infant mortality rate (MTR2). Other explanatory variables include fossil energy consumption (FOC), inflation (Inf), carbon dioxide emissions (CO₂), and government expenditure (GEH).

reflects the long-run findings of the asymmetry effects of infant mortality rate during an increasing and decreasing scenario. During periods of increasing levels of CO₂, a 1% increase in CO₂, decreases the infant mortality rate by 4.62% whereas, during decreasing levels of CO₂, a 1% increase in CO₂ would increase the infant mortality rate by 2.3%. However, there is no significant impact of FEC on the infant mortality rate during a high FEC scenario. During periods of decreasing levels of FEC, a 1% increase in FEC decreases the infant mortality rate by 1.1%.

7. Discussion of Results

The study’s key variables are the infant mortality rate and under-five mortality rate, (health outcome variables), and CO₂ emissions. Two separate results emerged from the analysis for nonlinear and linear ARDL.

Empirical results of linear ARDL.

It is evident from the empirical results of the linear ARDL presented in Table 6 that the first lag of the under-five mortality rate in the short run is significant with a direct impact on the under-five mortality rate in South Africa. Holding the other variables constant, the under-five mortality rate in South Africa would increase by 0.630% for every 1% increase in its lagged values. The implication is that factors that are responsible for the rise in the under-five mortality rate could be endogenously determined in both the current and immediate past periods. Consequently, these factors that were left uncorrected in the economy caused a repeat of the under-five mortality rate in the current period. While this result is in concord with the Grossman hypothesis, which views an individual as a producer and consumer of health, the longevity of such individuals could be endogenously determined. This result is in tandem with the work of [20] who noted that CO₂ emissions negatively impact human health in the region under investigation.

Similarly, fossil energy consumption has a direct and significant impact on the under-five mortality rate in South Africa. This significant relationship implies that a 1% increase in fossil energy consumption increases the under-five mortality rate per 1000 persons per year in South Africa by 0.418% in the short run, all things being equal. This implies that for the CO₂ emissions from fossil energy consumption, many factors could be attributed to this result. For instance, authors in the literature argue that the environmental concerns directly linked to energy consumption and production include climate change, air pollution, water pollution, solid waste disposal, and thermal pollution. The air pollution from the emission of fossil fuel combustion is known as the main factor responsible for urban air pollution [1] and increase the under-five mortality rate per 1000 persons per year in South Africa by 0.418% in the short run, all things being equal. Apart from the fact that the result supports the Gary production theory, which clarifies possible implications of the interaction between energy consumption and production on health, the results support the results of [16,22], who found that energy consumption has an impact on health. Specifically, ref. [21] found that energy consumption has a positive effect on the prevalence of respiratory diseases in the country under investigation.

However, in the short and long run, no significant relationship exists between the infant mortality rate, under-five mortality rate, and CO₂ emissions in South Africa. This result is highly germane and informative to the current study. This is because this outcome is a true reflection of the South African economy. The result shows that South Africa, as the most technologically developed country on the African continent, has implemented some moderating policies, in compliance with the international environmentally friendly laws on pollution. Consequently, CO₂ emissions could not impact the infant mortality and under-five mortality rate and in South Africa. This finding confirms those of [17]. However, the study’s authors noticed that in the short and long run, the consumption of fossil energy significantly contributes to the decline in the under-five mortality rate. This is congruent with the theoretical prediction and a substantial part of the existing studies. The industrial consumption of fossil energy engenders an increase in this sector’s production capacity. This results in a rise in aggregate output, which enables the government to prioritise funds to boost the population’s health quality. Moreover, a rise in the aggregate output causes a rise in the per capita income. This provides individuals with the opportunity to increase their calorie intake and improve their health quality. Expectedly, an increment in health quality lowers the under-five mortality rate in South Africa in the short and long run.

Government health expenditure has a significant and inverse effect on the mortality rate in the short and long run. This significant negative relationship reflects the extent of government spending on health in the country. Being a highly developed country on the African continent, South Africa recognises the essential role of the health sector in its development process. The realisation of the importance of the health sector motivates the inflow of funds to this sector to enhance health infrastructure, healthcare facilities, and health workers’ productive potential. Similarly, the inflation rate decreases the infant mortality rate and the under-five mortality rate in the long run, through a negative and significant relationship recorded between the under-five mortality rate, infant mortality rate, and inflation in the long run. This implies that, although higher prices increase socio-economic frustrations and increase the under-five mortality rate in the short run, the inflation rate moves in the opposite direction in the long run. This is certainly not out of place because people adapt to the higher prices and adjust their consumption patterns in the long run.

Empirical results of nonlinear ARDL.

Short run result under nonlinear ARDL.

The results from nonlinear ARDL presented four separate scenarios. In the short run, during increasing levels of CO₂ in the initial period (lag of CO₂), a 1% increase in CO₂ would decrease the under-five mortality rate by 1.15% whereas during decreasing levels of FEC lag_, a 1% increase in FEC would increase the under-five mortality rate by 0.96%. The result shows that there is a link between the lag of FEC and CO₂ emissions. Such cases occur in a high level of industrial activities, where production increases in the economy result in more exhaust emissions from machines that are not environmentally friendly. However, this improved production will increase the economic gain and make more cash available for tax payment to the government, and, consequently, more money in the treasury will improve the welfare of both adults and children. Both scenarios of increasing and decreasing government expenditure significantly impact on under-five infant mortality rates. Scenarios of increasing government expenditure tend to increase the under-five infant mortality rates by 0.15%. This occurs in cases where government-expended projects failed to put in place or incorporate environmental conservation policies into the execution of certain projects whose by-products tend to be detrimental to health. Scenarios of decreasing government expenditure tend to decrease the under-five infant mortality rate by 0.44%. This holds true as no emission capable of impacting negative effects on health during such scenarios.

Long-run result under nonlinear ARDL

Table 10 reflects the long-run result. During periods of decreasing levels of CO₂ and FEC, a 1% increase in CO₂ would increase the under-five mortality rate by 6.04%. However, FEC would decrease the under-five mortality rate by 4.21%. This occurs when the economic performance leading to more production in the short run cannot be sustained. The decline in economic gain will decline welfare in the economy since less money through taxation is available. This will worsen the health of people and consequently result in an increase in the under-five mortality rate by 6.04%. This case is similar to the result acquired with infant mortality in both the short and long run. During periods of decreasing levels of CO₂ in the short run, a 1% increase in CO₂ will increase the infant mortality rate by 0.66%. Again, during previous and current periods of decreasing levels of FEC, a 1% increase in FEC would increase the infant mortality rate by 0.45% and 0.32% since it is an indication of poorer economic performance in the economy. In the long run, during periods of increasing levels of CO₂, a 1% increase in CO₂ decreases the infant mortality rate by 4.62% whereas during decreasing levels of CO₂, a 1% increase in CO₂ will increase the infant mortality rate by 2.3%.

8. Conclusions and Recommendations

This study aimed to determine the implications of the poor management of emissions on infants in South Africa. The study’s investigations show that both fossil energy consumption and CO₂ emissions could impact infant and under-five mortality rates. The findings show the diverse effects of emissions on the under-five mortality rate based on linear ARDL results and the four categories of scenarios as identified in the nonlinear ARDL results for South Africa. For instance, holding the other variables constant, the under-five mortality rate in South Africa would increase by 0.630% for every 1% increase in its lagged values. The increased level of the under-five mortality rate per 1000 persons could be due to an increase in fossil energy consumption resulting from environmental pollution. This pollution could lead to the depletion of the ozone layer, leading to a rise in environmental temperature. Again, the emitted carbon-related compounds potentially react with human haemoglobin which appears to be dangerous both to adults and the under-five age group. Therefore, the under-five age group in these cases will be unable to cope with the CO₂ emissions resulting from fossil energy consumption. In addition, exposure to high levels of air pollution could result in a variety of adverse health outcomes in under-fives and infants as this can increase the risk of heart disease, respiratory infections, and lung cancer, ultimately leading to a high mortality rate among infants, and, consequently, showing the positive and significant effects of emissions on the under-five mortality rate in South Africa. Government policies should include the building of housing estates at distances away from where emissions could pose a threat to health. People could also be moved to government apartments with affordable costs to the poor to decrease the risk posed by CO₂ emissions.

Results from the nonlinear ARDL confirmed that in the short run, during periods of decreasing levels of CO₂ and FEC, a 1% increase in CO₂ would increase the under-five mortality rate by 6.04%, all things being equal. However, FEC would decrease the under-five mortality rate by 4.21%. As discussed earlier, this occurs when the economic performance leading to more production in the short run cannot be sustained. It is therefore recommended that economic activities that could lead to a decline in CO₂ and FEC consumption should not be encouraged in the economy. However, technology and sanitary measures that could mitigate its effects on people (particularly infants) as spelt out by the World Health Organisation should be maintained as the government vigorously engages with more energy-driven activities.

Furthermore, the nonlinear ARDL results revealed that in the long run, during periods of increasing levels of CO₂, a 1% increase in CO₂ decreases the infant mortality rate by 4.62% whereas during decreasing levels of CO₂, a 1% increase in CO₂ would increase the infant mortality rate by 2.3%. There are several policy recommendations derived from the long-run results for sustainability goals. The infant mortality rate decreases with increasing levels of CO₂ emissions, and this shows that the possibility of the improvement of energy consumption to be environmentally friendly is expected in the near future. The government should allocate more expenditure to research and development to build machines to achieve environmentally friendly energy consumption. Outdated equipment should be removed to give way to more modern machines. The government could reduce the carbon footprint of the population by enacting policies that encourage changes in the way people travel, switching to clean energy sources, and reducing people’s consumption patterns through reuse and recycling as this would enhance environmental sustainability.

The coefficient of the inflation rate is positive and significantly influences the under-five mortality rate in South Africa. This suggests that the mortality rate of children below the age of five per 1000 persons per year increases by 16.9% for every 1% increase in the inflation rate. This could occur when inflation reduces the purchasing power of the population, particularly when the inflation rate violates the Phillip curve assumption of the trade-off nexus, such that an increase in the inflation rate fails to create more employment.

The results suggest that the mortality rate of children below the age of five per 1000 persons per year increases by 16.9% for every 1% increase in the inflation rate. Therefore, it is recommended that South Africa’s monetary regulatory bodies consider the state of children’s health while enacting policies targeting inflation. Again, consideration of the four economic scenarios for policy implications on the economy would help reduce under-five and infant mortality rates. For instance, during a decline in economic performance, the policy of the government should be flexible enough to enhance environmental sustainability and protect under-fives and infants within the economy.

The study’s hypothesis, which states that fossil energy consumption and CO₂ emissions impact the under-five mortality rate and the infant mortality rate in South Africa, has been proven. The results are both empirically and theoretically supported by studies conducted by other researchers in various nations across other regions of the world. However, the present study contributes to the body of knowledge and the literature in two important directions. In the first direction, the study adopted two different indicators of the under-five mortality rate and the mortality rate of children below the age of five per 1000 persons as health outcome variables in South Africa, thereby yielding more robust outcomes. In the second direction, a more advanced, novel method was used to determine the causal relationship in this study. Unlike other studies that used the traditional Granger causality test to measure the direction of causality, the current research adopted the approach proposed by [36] for causality estimation involving a modified WALD statistic (χ² distribution).

The main constraint of this study was the limited availability of data, which hindered the broadening of the scope of the health outcomes. Consequently, the study was constrained to the under-five mortality rate and infant mortality rate.

Room for further research on this concept exists since the study was unable to identify variables such as life expectancy. More importantly, further investigations are required to establish why certain variables were not able to significantly impact the dependent variables.

The study has implications for South Africa in terms of economic policy. Since fossil energy consumption has a positive and significant effect on the under-five mortality rate, the results indicate that if other forms of energy are used as substitutes for fossil energy consumption, the under-five mortality rate in South Africa can be significantly reduced.