Exploring the interrelationship among health status, CO₂ emissions, and energy use in the top 20 highest emitting economies: based on the CS-DL and CS-ARDL approaches

Abstract

Carbon dioxide emissions (CO₂e) which is caused by energy use contributes to the global average surface temperature increase by 1.5 °C as compared to the mid-1800s which is causing a certain change in the climate and becoming an adverse effect on health and economy. The relationship between health status, CO₂e, and energy use has yet to be thoroughly investigated in the top 20 highest emitting economies. The data from 2000 to 2019 is analyzed by using advanced techniques of cross-sectional augmented distributed lag (CS-DL) and cross-sectional augmented autoregressive distributed lag (CS-ARDL) which take into consideration crucial elements of panel data, namely dynamics, heterogeneity, and cross-sectional dependence. Moreover, cross-sectional augmented error correction method (CS-ECM) and the common dynamic process of the augmented mean group (AMG) are applied for robustness checks. The empirical findings revealed that (i) CO₂e weakens the health status only in the short-run, whereas health expenditure improves the health status in the both short- and long-runs, while economic growth is not contributing to the health status in the both short- and long-runs; (ii) health expenditure and economic growth only help to mitigate CO₂e in the long-run, whereas energy use causes CO₂e in the both short- and long-runs; (iii) energy use causes high economic growth in the both short- and long-runs, whereas CO₂e aids economic growth in the short-run but is extremely damaging to economic growth in the long-run, while in the both short- and long-runs health expenditure is not aiding the economic growth. This study provides policy recommendations on improving human health by advocating massive health expenditures, CO₂e easing, promoting renewable energy use or low-emission energy, and steering the economy toward green economic growth.

Graphical abstract

Introduction

The United Nations member states aim to achieve Sustainable Development Goals (SDGs) by 2030. A principal goal of the SDGs is to improve healthy living and well-being (Goal-3) and ensure easy access to affordable, clean, and sustainable energy, and improve people’s health worldwide (Goal-7) (UN, 2015; WHO, 2018). Healthy living is a fundamental human right. However, recently, the COVID-19 pandemics¹ have emphasized the need to rethink many health-related problems globally. The problems of healthy living and the increasing challenge of global warming are the result of deteriorating environmental quality around the world (Balan, 2016). The world’s population is undergoing a significant impact on its health status because of a lack of consideration and ignorance of many health-related factors such as CO₂ emissions (CO₂e), energy use, health expenditures (both public and private), environmental quality, and economic development. Energy usage and economic activities all have a major environmental impact, which contributes to an increasing worldwide burden of disease, lost work and school days, and harm to trade, transportation, agriculture, and other businesses (Jacobo & Viviana, 2013; Clark & Hammond, 2019). Therefore, the threat of global environmental change to human health and the economy is becoming more generally acknowledged, indicating the potential interaction among those variables.

Over the past few decades, the topic of environmental degradation and CO₂e has been receiving considerable attention from both researchers and policymakers worldwide (Adebayo and Kirikkaleli 2021; Adeneye et al., 2021). From the 19th to the 20th centuries, global CO₂e increased by 30% (Erdoğan et al., 2022). The problem of environmental deterioration and CO₂e are among the major challenges facing the global economies, and has a significant negative impact on human health, income, society, environmental quality, economic growth, and the natural world (Muntean et al., 2018). Greenhouse Gases (GHGs) emissions have been universally acknowledged as contributing to climate change and global warming (Kirikkaleli et al., 2022). CO₂e contributes approximately 75% of total worldwide GHGs emissions (Diffenbaugh, 2020); as a result, the global average surface temperature increased by 1.5 °C² as compared to the mid-1800s, causing a certain change in the climate and becoming an adverse effect on human lives, ecosystems, and long-term stability of the top 20 highest emitting economies (T20HEE)³ (Gavurova et al., 2021; NASA, 2020). A prior study on the top emitting countries concluded that climate change causes additional 400 000 deaths each year, and by 2030, it will be responsible for 700 000 annual deaths (McKinnon, 2012). As CO₂ temperatures rise, global climate change and extreme weather events, including floods, droughts, storms, heat waves, ice sheets melting, sea-level growth, and water system disturbances, are becoming more common (IPCC, 2021). The previous study on the “global climate risk index 2016” reported that over 15, 000 severe weather incidents occurred in the 10 developing countries between 1995 and 2014, resulting in over 525, 000 human deaths along with financial losses of more than 2.97 trillion U.S. dollars (Purchasing Power Parity) (Kreft et al., 2016). Japan was devastated by massive disasters in 2018, such as floods, severe weather events, the typhoon in Jebi, resulting in 1282 deaths and a 0.64 percent decline in gross domestic product (GDP). In Australia, a wildfire destroyed more than 8.4 million hectares, 1300 houses, 27 lives, and a record one billion livestock in 2019. The World Health Organization (2019) reported that approximately 1.6 million people suffer from lung cancer, 65 million from a "chronic obstructive pulmonary disorder," and around 334 million suffer from asthma, and these statistics are rapidly increasing. In this regard, problems regarding limiting environmental degradation are a global challenge that impacts all the countries. The global community, such as “the Kyoto protocol” which was signed in 1997 and implemented in 2005, “the Paris Climate Agreement (PCA)” in 2015, the COP26⁴ in Glasgow, Scotland in 2021, held many climate agreements aimed at limiting global warming average temperature below 1.5 °C and protect human health worldwide, and the most recent “United Nations Climate Change Conference (COP27)⁵ in Sharm El-Sheikh, Egypt” in 2022 which is important and timely for two reasons. First, the protracted war in Ukraine and the COVID-19 catastrophe compelled a number of countries to recommence coal-fired energy production, resulting in record-high CO₂e. Second, the global crisis revealed our sensitivity to energy shortages and the necessity to develop alternate energy sources. Nonetheless, despite all of these climate agreements and policymaking, global temperatures continue to rise, and CO₂e increased by 2.7% in 2018, prompting environmentalists, international organizations, and policy makers to focus more to identifying the critical factors and sources of CO₂ influencing human health (Diffenbaugh, 2020).

The current study of casual relationships between health status, health expenditures, CO₂e, energy use, and economic growth has recently received considerable attention from both developing and developed economies, particularly in the T20HEE. The T20HEE represents two-thirds of the world's population, which accounted for about (80%) of the world’s total CO₂e (Mott et al. 2021). Countries such as China, the USA, India, Russia and Japan are the world highest emitters; therefore, global environmental success largely depends on them (Fig. 1). Meanwhile, these top-economies account for (77%) of global energy consumption and (81%) of global renewable energy, and their rapid economic growth contributes to (85%) of global GDP (World Bank, 2020). With such a great economic development, the top-emitting economies are among the world’s fastest-growing nations. According to World Bank statistics, global GDP has doubled between 1992 and 2019, with an average annual growth rate of 2.9% (World Bank, 2020). However, due to their large populations and rapid economic growth, these top economies have experienced a massive increase in energy demand as a whole, resulting in high amount of GHGs that are adversely impact on the environmental quality (Munir and Riaz, 2019; Zou and Zhang 2020). Thus, higher GHGs emissions are as a result of increased economic development and growth (Wang et al. 2022). As the economic development continue to increase in the top-emitting countries, environmental quality deteriorates due to the adverse effects of CO₂e and polluting fossil fuels (Sinha et al., 2019), thus, posing a serious threat to the health condition of many countries (Li et al., 2022; Taghizadeh-Hesary et al., 2020). Globally, CO₂ emissions are majorly caused by rising industrial activities and high energy use (Callan et al. 2009). According to the International Energy Agency (IEA), the top emitting economies' energy demand is expected to rise by 80% between 2013 and 2035 (IEA, 2017). The World Health Organization (WHO) reported that approximately (41%) of the world's population (about 2.8 billion people) are using fuel energy sources to fulfill their daily energy demands (WHO, 2015). The IEA (2017) estimated that the energy sector accounts for 69% of global CO₂e, fuel energy (coal, gas, and petrol) generates about 44%, and the intensive usage of industrialization, manufacturing, transportation, and refrigeration has led to an excessive increase in other GHGs and is affecting environmental quality (Verma et al., 2021). Each country fulfills its energy demands using various types of energy sources based on several factors, including climate, culture, economic and social activities, population, war, and wealth (Gallo Cassarino et al., 2018). To alleviate environmental challenge, these top-emitting economies not only focus on reducing fossil fuels consumption, but also upsurge energy efficiency. Therefore, in order to achieve the SDGs, these selected countries needed significant improvements in their economies, particularly in their energy sectors. Due to the high population, rapid economic growth, excessive CO₂e, and energy use, these top-economies are mostly prone to the threat of environmental degradation and climate change, therefore, it is crucial to understand the association and the factors of CO₂e in these economies.

Fig. 1

CO₂ emissions (%) distribution based on top emitters in 2020

The deterioration of the environment and air quality increases the demand for healthcare expenditures in order to keep a healthy living (Alimi et al. 2019). Air pollution alone causes 9 million premature deaths globally each year, if urgent legislative and regulatory actions are not taken, deaths rate will increase in the coming future (Vohra et al., 2021). Furthermore, the World Health Organization (2019) declares that air pollution is expected to increase 250.000 additional deaths per year from 2030 and 2050. The increase in environmental pollution due to anthropogenic emissions such as CO₂e significantly influence health expenditures (Ahmad et al., 2021). Health expenditure is continuously rising because governments are financing more in healthcare infrastructure and other healthcare facilities such as public health insurance. The demand for health insurance and healthcare infrastructure is increasing due to industrial development, infrastructure development, rise in energy consumption, urbanization, and rural-to-urban migration (Apergis et al., 2020). Furthermore, it is widely recognized that healthcare facilities play a vital role in reducing the adverse effects of climate change. The WHO (2011) reported that healthcare facilities represent between 3% to 8% of climate change footprints in developed countries, such as (8%) in the USA and (3%) in the United Kingdom. The total health expenditures were approximately 7.12% of GDP, among which the public health expenditures comprised (59.47%), and private health expenditures accounted for (40.28%), respectively, in the T20HEE (World Bank, 2020). Moreover, as life expectancy increases, the influence of health expenditures on CO₂e and economic growth should not be neglected. According to the World Health Statistics (2019), the average age of the T20HEE was 77.6 years old. The complete detail of the average life expectancy of the T20HEE are provided in Fig. 2. Therefore, careful consideration is required of the issues of global CO₂e, energy use, health expenditures (both public and private), and economic growth, promoting researchers and policymakers to conduct in-depth empirical research and provide critical analysis for the betterment of human health and life in the T20HEE.

Fig. 2

The average life expectancy of top emitters in 2019

The previous studies have extensively examined the influence of CO₂e on health expenditures and economic growth. At present, however, the empirical research on the link between health expenditure and CO₂e is limited (Bilgili et al., 2021; Ghorashi and Alavi 2017). To this end, the first branch of studies focused on the impact of health expenditures on CO₂e (Gündüz, 2020; Apergis et al., 2018). In addition, those studies neglected to consider the association between health expenditures and CO₂e. Therefore, this study adds to the limited research to investigate the association between health expenditures and CO₂e in the T20HEE. Furthermore, earlier conducted studies on health status have had mixed results and are often inconclusive, making it challenging to formulate consistent strategies in the health sector (Rahman and Alam 2021). These inconsistent results may be due to the different countries, methodologies, data, and variables, which create an ambiguous relationship among the related variables. Moreover, the challenges of environmental pollution have attracted worldwide attention due to its severe long-term effects on human health and economy. In spite of world attention, there are still some policy uncertainties in the current strategy due to the deficit of comprehensive and wide-ranging study, thus providing the opportunity for further research into the formulation and execution of appropriate policy initiatives for improving people’s health and environmental policy framework, to accomplish the SDGs. In this regard, this study is intended to examine the interaction of health status, CO₂e, and energy use among the top 20 highest emitting economies.

The main contributions of the present study are the following: First, this study emphasizes T20HEE, which accounted for the large amount of the world population and significant contributors to global economic development and contribute considerably to CO₂e. Second, to the best of our knowledge, this is the only study that investigates the relationship between health status, health expenditures (both public and private), CO₂e, energy use, and economic growth using a dataset of the T20HEE from 2000 to 2019. Third, this study fills the gaps in previous studies by employing advanced econometric techniques, including the Cross-Sectional Augmented Distributed Lag (CS-DL) and Cross-sectional Augmented Autoregressive Distributed Lag (CS-ARDL) techniques by Chudik et al. (2016) to estimate the both short-run and long-run coefficients, where the two methods are complementary in nature. The combination of some of these countries consists of developing and developed countries as well as regions and continents; thus, in such an analysis, the CS-DL and CS-ARDL econometric procedures take into consideration three crucial elements of panel data, namely dynamics, heterogeneity, and cross-sectional dependence. Unlike earlier research studies, the current study applied econometrics analysis of second-generation panel methods such as the Pesaran unit root test, the Westerlund cointegration test, and the Dumitrescu-Hurlin Granger causality test was employed to determine the causal relationships between the variables. In addition, second-generation panel econometrics methods are regarded to be superior, advanced, and reliable to first-generation panel methods in terms of providing significant empirical results for guiding future policy decisions. This study, therefore, produces more effective and comprehensive estimates than previous studies in the context of T20HEE. Fourth, the Augmented Mean Group (AMG) and the Cross-sectionally Augmented Error Correction Method (CS-ECM) were employed to check the robustness. In addition, because slope coefficients change between nations and cross-country averages (and their lags) are proxies for unobserved shared characteristics, these models allow for common factors to influence countries. Eventually, the significance of the study is expected to help in achieving SDGs, particularly, SDG-3 and SDG-7, and these would serve as guiding principles for policymakers to use the outcomes of this study and improve human health by advocating massive health expenditures (both public and private), CO₂e mitigation, increased use of renewable energy or low-emission energy, and steering the economies toward green economic growth.

Though the broad objective of this paper is to study the interaction of health status, CO₂ emissions, and energy use among the top 20 highest emitting economies, it has the following specific objectives:

• i. To investigate whether health status is affected by CO₂ emissions of the top 20 highest emitting countries;

• ii. To explore whether health expenditure is helping in mitigating CO₂ emissions of the top 20 highest emitting countries;

• iii. To assess how energy use in the top 20 highest emitting countries is related to their production of CO₂ emissions; and

• iv. To examine how energy use and CO₂ emissions are related to the economic growth of the top 20 highest emitting countries.

The remaining sections of this study are distributed as follows: Section 2 will follow a literature review comprising both theoretical and empirical reviews; Section 3 describes research methodology, including data description, model specification, and estimation techniques; Sections 4 and 5 present results and discussion of the results, respectively; and the final Section 6 provides conclusion and recommendation for policy.

Literature Review

The topic setting of this study has theoretical justification from several well-known theories, including the Nightingale's Environmental Theory, Life Course Health Development (LCHD) framework, the Carbon Dioxide Theory, Health Expenditure and Carbon Emissions Systematic Linkage Model, and Environmental Kuznets Curve (EKC) Theory. On the other hand, the empirical review of the literature on the relationship between health status, CO₂ emissions (CO₂e), energy use and economic growth has been divided into three categories: (i) health status and CO₂ emissions; (ii) energy use and CO₂e; and (iii) energy use, CO₂e and economic growth.

Theoretical Review

From a theoretical perspective, several theorists have shown a relationship between environmental problems and health consequences throughout the time. Some of these theories or postulations include: (1) The Nightingale's Environmental Theory, which was first proposed by Florence Nightingale in 1859 in England, which remains relevant today in health discourse, ties the individual's health is determined by his environment. In this theory, Florence Nightingale addresses that clean water, efficient drainage system, cleanliness/sanitation, pure air, sufficient food, and sunlight are crucial component of an environment that will assure good health outcomes or status of population. The Environmental Theory has led to improved human health. A healthy environment is important for healing. She believed that “nature alone cures.” (2) The Life Course Health Development (LCHD) framework, which was published by Halfon and Hochstein's in 2002, which relates variations in health of an individual to a number of determinants operating in a nested genetic, biological, and environmental network. Furthermore, based on to the LCHD framework, health development is an adaptive process that is influenced by biological, economic, environmental, genetic, and social circumstances. As a result of these developing contributing factors, health outcomes or status are also invariably changing. Considering comprehensive empirical assessment of the association between these variables (biocapacity, environmental quality, human capacity, and health status or outcomes), the research has been deficient. (3) The Carbon Dioxide Theory, as the amount of CO₂ in the atmosphere increases, the atmosphere becomes opaque over a longer frequency interval. As a result, emitted radiation is trapped more efficiently close to the surface of the earth, which causes the temperature to rise. According to the most recent figures, if the amount of CO₂ in the atmosphere were to be halved, the surface temperature would drop by 3.8 degrees Celsius, while if the amount were to be doubled, it would cause a rise in temperature of 3.6 degrees Celsius. These results come from calculations that used the most recent data available. The carbon dioxide theory, which was first proposed by John Tyndall in 1861. The first significant computations were necessarily done using very approximative techniques. Because of carbon dioxide, there are hundreds of spectral lines which are caused by absorption, and each of these lines occurs in a sophisticated pattern with fluctuations in the intensity and the width of the spectral lines. Furthermore, the pattern does not appear the same even at all heights in the atmosphere due to the fact that the breadth and strength of the spectral lines change depending on the temperature and pressure of the environment. A relatively accurate solution to the problem of the impact of CO₂ on surface temperature has only recently been possible due to accurate infrared measurements, availability of a high-speed electronic computer, and the theoretical advancements. (4) Health Expenditure and Carbon Emissions Systematic Linkage Model postulated that spending money on healthcare equipment that is efficient in its energy consumption is a powerful lever for reducing carbon emissions and addressing rising operational energy costs for care providers. Specifically, the model stated that this would be the case if the expenditure was made on energy-efficient healthcare equipment. According to the most recent report published by the World Economic Forum in 2022 on how the healthcare industry can reduce its carbon footprint, hospital buildings have the highest energy intensity of all publicly funded buildings. As a result, hospital buildings emit 2.5 times more greenhouse gases than commercial buildings. Moreover, the amount of energy that is used by medical equipment is yet another area that has the potential to have a significant effect on large-scale, energy-intensive activities such as hospitals. Therefore, allocating funds for healthcare toward the purchase of energy-efficient medical equipment may have a significant bearing on the amount of carbon footprint reduced. (5) The Environmental Kuznets Curve (EKC) hypothesis was proposed by Kuznets (1955) regarding the income inequality and economic development and was named after it. The EKC examine the relationship between several indicators representing environmental pollution and per capita income. The EKC hypothesis was put forward by Grossman and Krueger (1995), which refers to an inverted U-shaped relationship between environmental degradation and per capita income. The term EKC argues that environmental pollution rises in the early steps of economic growth, when a certain level of income per capita are reached, the economic growth contributes to environmental improvement (Bilgili et al., 2021). For example, advancing production and consumption technologies for manufacturing processes, raising environmental awareness, increasing the usage of environmental-friendly technologies, establishing environmental protection laws, etc. Therefore, Ulucak and Bilgili (2018) stated in the prior study that the EKC hypothesis found an inverse U-shape association between per capita income and environmental deterioration. Numerous studies have documented EKC’s existence, while others have found no evidence of its existence, and still others have discovered EKCs in various shapes, thus leaving the conclusion ambiguous (Ahmad et al. 2020). Chen et al. (2020) confirmed conventional EKC while using China's city-level data.

Empirical Review

CO₂ Emissions, Health Expenditure, and Health Status Nexus

The first section of the study looked at the association between health and CO₂ emissions (CO₂e) by employing a number of methodologies, data sets, and locations. The important issue of the influence of CO₂e on health status is clearly revealed by many recent studies in the context of developing and developed economies. In this regard, Yang and Liu (2018) examine the impact of environmental deterioration on human health and health inequality from the 2014 national survey of “China Family Panel Studies (CFPS)”. The 2014 survey includes approximately 16000 households from 25 regions. They revealed that environmental pollution has a significantly negative impact on human health, and air pollution contributes to increasing health inequality. Another group of studies reported the same findings: Ahmad et al. (2020) for China, Balan (2016) for 25 European Union (EU) countries, Osabohien et al. (2021) for Nigeria, Rahman and Alam (2021) for SAARC-BIMSTEC countries, and Taghizadeh-Hesary et al. (2020) for middle-income Asian countries. Igbinedion (2019) identified environmental pollution as having a significant impact on neonatal life expectancy, and both enhanced sanitary facilities and public health expenditure had a significant and positive impact on health status and increased longevity in Nigeria from 1990 to 2016.

Furthermore, several empirical studies have explored the link between CO₂e and health expenditures. In this regard, several recent studies employing different estimation techniques revealed that the increase in healthcare expenditures would significantly increase CO₂e in various countries around the world. For example, Ahmad et al. (2021) in China; Alimi et al. (2019) for 15 ECOWAS countries; Blázquez-Fernández et al. (2019) for 29 OECD countries; and Khan et al. (2020) for 58 Belt and Road Initiative countries. Bilgili et al. (2021) applying the quantile regression analysis, they ascertained that both public and private health expenditures could help to decrease the adverse effects of CO₂e. Ganda (2021) revealed that there is a negative linkage between health expenditures and CO₂e. When the effect is disaggregated, public-health expenditure is positively related to CO₂e, whereas private-health expenditure is adversely related to CO₂e in the BRICS countries during 2000-2017 by using Fully Modified Ordinary Least Squares (FMOLS), Vector Error Correction Model (VECM), and the Dumitrescu-Hurlin causality tests. A recent study by Apergis et al. (2020) examined the long-run association between environmental pollution and healthcare expenditures among four global income groups in 178 countries during the period 1995–2017. According to the findings, they noted that a 1% increase in CO₂e will increase health expenditures by about 2.5%. Apergis et al. (2018) revealed that both usage of renewable energy and health expenditures caused a reduction in CO₂e, while real GDP leads to increase the CO₂e levels in African countries during the period 1995-2011. Zeeshan et al. (2021) found that CO₂e and environmental pollution positively increase the amount of household health expenditures in both the short- and long-runs. Also, the results demonstrate that there is a bidirectional link between household health spending, CO₂e, and environmental pollution, indicating that CO₂e significantly increases health expenditures in China from 1990 to 2019 by using the Nonlinear Autoregressive Distributed Lag (NARDL) model.

Energy Use and CO₂ Emissions Nexus

The second section of the study focuses on the association between different types of energy usage and CO₂e. Scholars have used a number of methodologies and data sets to conduct a study in various countries, regions, and continents. In this regard, Ummalla (2019) applying the Autoregressive Distributed Lag (ARDL) and Panel Quantile Regression (PQR) model to examine the effects of hydroelectric energy consumption on CO₂e in the BRICS countries during the period 1990-2016. They discovered that using hydroelectric electricity results in lower CO2e emissions. Another recent study by Ashin and Muhammed (2020) examines CO₂e and economic growth in India using Artificial Neutral Networks (ANN) and machine learning methodologies. They revealed that transitioning from fossil fuels (coal, gas, and petrol) to renewable energy can significantly minimize CO₂e without affecting economic growth. Zou and Zhang (2020) investigate the linkage between CO₂e, energy consumption, and economic growth in 30 Chinese provinces from the period 2000–2017. Employing the Spatial Durbin Model (SDM), they observed a two-way association between CO₂e and energy use, implying that CO₂e and energy use are related to a negative spatial spillover effect on CO₂e in neighboring provinces and cities. Adeneye et al. (2021) determined the relationship between CO₂e and energy use was highly significant in Asia during 2000–2014 by applying panel data models with dynamic common correlated effects. Litavcová and Chovancová (2021) employed the ARDL Model and found that energy usage is a significant long-run influence on CO₂e across 14 Danube countries from 1990 to 2019. Salari (2021) identified a long-run relationship between several kinds of energy use and CO₂e in the U.S. In addition to this, non-renewable energy and household energy are positively associated with CO₂e, whereas renewable energy (RE) has a negative association with CO₂ in the U.S. from 1997 to 2016. Murshed et al. (2021) noted that the usage of RE contributes to the sustainability of the environment in Bangladesh. Similarly, several prior studies confirmed that RE contributes negatively to environmental deterioration by Sharif et al. (2020) in Turkey, Destek and Sinha (2020) in the Organization for Economic Cooperation and Development (OECD) countries and Kayani et al. (2020) in the top 10 highest emitter countries. Bekun et al. (2021b) investigate the applicability of the conventional Environmental Kuznets Curve (EKC) with an extension for the case of emerging industrialized economies (E7) of the world, including China, India, Brazil, Mexico, Russia, Indonesia, and Turkey during the period 1995-2016. Applying second-generational panel econometrics strategies such as the Augmented Mean Group estimator, the Common Correlated Effects Mean Group estimator, the Driscoll-Kraay and Dumitrescu and Hurlin Causality analysis. The EKC phenomenon is supported by the declining impact of fossil fuel-based energy use in the bloc. However, renewable acts as a panacea for reducing pollution emissions, because RE has a negative and statistical association with CO₂e. Bekun et al. (2021a) investigate the long-term and causal relationship between tourism development and the real income level in EKC framework for emerging industrialized economies (E7). Employing second-generation panel estimation techniques, in conjunction with the Driscoll-Kraay robust estimator, the findings support the notion that GDP growth and non-renewable energy degrades environmental quality. Adebayo et al. (2023a) applied the CS-ARDL technique to analyze the complex relationship among technological innovation, use of renewable energy sources, natural resources, and CO₂e in the context of BRICS countries during the period 1990 to 2019. The findings discovered that natural resources, renewable energy consumption, and technological innovation increases environmental sustainability by reducing CO₂e both in the short- and long-runs.

Energy Use, CO₂ Emissions, and Economic Growth Nexus

The third section includes many studies that aim to investigate the link between environmental consequences and economic growth, which is known as the "Environmental Kuznets Curve (EKC)." A number of studies have been conducted to investigate the relationship between CO₂e, energy consumption, and economic growth by employing a number of data sets, locations, and methodologies in the both developing and developed countries around the world. In this regard, Pao and Chen (2019) revealed that a long-run balanced relationship exists between CO₂e, energy use, and economic growth for the developed countries by employing the panel cointegration test. Adebayo and Akinsola (2021) and Cai et al. (2018) applied different estimation techniques and discovered that changes in economic growth contributed to changes in CO₂e. They also found that both short- and long-runs associations exist between economic growth and CO₂e in Thailand and the G7 countries. Azam et al. (2016) employ the FMOLS model to examine the impact of CO₂e on economic development in particular higher CO2 emitting countries from 1971 to 2013. They revealed that CO₂e have a significantly positive influence on economic growth for China, Japan, and the USA, while significantly negative for India. Mensah et al. (2018) employing the STIRPAT model to evaluate the influence of innovation on CO₂e at the individual level in 28 OECD countries from 1990-2014. The economic-EKC model was found to be inapplicable for most economies because rising GDP per capita increases CO₂e in most OECD countries while reducing emissions in only a few OECD countries. Ummalla (2019) uses the ARDL model and PQR estimations to examine the effects of hydroelectric energy consumption on economic growth in the context of BRICS countries (Brazil, Russia, India, China, and South Africa) from 1990 to 2016. The findings support the idea that hydropower consumption is related to economic growth. Using dynamic simultaneous equation models, Ghorashi and Alavi (2017) examine the link between CO₂e, health spending, and economic growth in Iran during 1972–2012. According to the empirical findings, a bidirectional causal relationship exists between CO₂e and economic growth, whereas health expenditures and economic growth exhibited a one-way relationship. Litavcová and Chovancová (2021) identified that economic growth has only a long-term impact on CO₂e by employing the ARDL model in 14 Danube Region countries during the period 1990–2019. Employing multi-spatial Convergent Cross Mapping (CCM), Liu et al. (2019) have discovered various causal relationships in the G7 countries, including unidirectional, bidirectional, and neutral nexus; whereas a bidirectional nexus exists between energy use and CO₂e has been demonstrated in China and India from 1965 to 2017. Ssali et al. (2019) revealed a bidirectional causation exists between energy use and CO₂ in the short-run, but in the long-run, there is one-way causation from energy use to CO₂ in six Sub-Saharan African countries from 1980 to 2014 using the ARDL model. They found a rise in economic growth increases CO₂, while an increase in economic growth squared results in a decrease in CO₂. As a result, the EKC theory is supported by the positive and negative effects of economic development, as well as their square. Irfan et al. (2023) demonstrate the dynamic impact of the tourism industry on economic output, energy consumption, and environmental quality in China from 2001Q1 to 2019Q4. Using the ARDL approach and the Gradual shift causality method, they revealed that all tourism industry sub-sectors have a significant and positive impact on economic growth and energy consumption, while tourism-related travel consumes more energy. Similarly, entertainment and shopping boost the economic development more than other sub-sectors. Interestingly, globalization has had a negative impact on air pollution and GHGs emissions in the long-run. Adedoyina et al. (2021) revealed that rising energy usage, real GDP per capita, and tourism across the four EU macro regions contribute to increase in CO₂e by employing the one-step system Generalized Method of moment (GMM) model in the 26 EU member states during the period 1995-2018. Caglar et al. (2021) using second-generation panel methodologies to investigate the factors that influence CO₂e in the BRICS economies, including the contribution of partnerships' investments in energy and economic complexity from 1990-2018. The results demonstrate that improvements in economic complexity and trade openness promote environmental quality. On the other hand, environmental deterioration is a result of economic expansion, use of natural resources, and public-private partnerships. Bekun (2022) investigates the impact of renewable energy, non-renewable energy, economic growth, and investment in the energy sector on CO₂e in the Indian economy. Using Canonical Cointegration Regression (CCR), Fully Modified Least Squares (FMOLS), and Dynamic Least Squares (DOLS), they demonstrated that, despite the economic growth trajectory, renewable energy acts as a panacea for sustainable development. However, there is positive association exists between CO₂e and both non-renewable and GDP growth. Adebayo et al. (2023c) employ Wavelet local multiple computation (WLMC) to analyze the time- and frequency-varying effects of hydroelectricity energy consumption, natural gas energy consumption, and economic growth on environmental sustainability in the United States of America from 1965/Q1 to 2020/Q4. The findings provide new insights into how CO₂e are caused by economic growth, which is demonstrated by a positive linkage exists between both factors at short-term and long-term time-frequencies. Adebayo et al. (2023b) employ CS-ARDL method, common correlated effects mean group (CCEMG), and AMG to investigate the impact of country risks and renewable energy usage on environmental quality and economic growth in MINT countries (Mexico, Indonesia, Nigeria, and Turkey) from 1990-2018. The findings indicated that economic growth, political risk, trade openness, and urbanization, all contribute to an upsurge in ecological footprint; nevertheless, environmental quality is positively influence by renewable energy use as well as economic and financial risks.

Literature Gap

According to the exploration of related literature, it is evident that relatively few studies have investigated the combined relationships among health status, both public and private health expenditures, CO₂ emissions, and energy use, particularly in the top emitting economies. In addition, previous studies result on health status were often inconclusive and not encouraging to formulate consistent strategies towards a better health and environmental policy framework (Rahman and Alam 2021). Furthermore, earlier studies have mainly focused on the influence of CO₂e on health expenditures and economic growth; however, minimal has studied the link between health expenditures and CO₂e (Ghorashi and Alavi 2017). Moreover, unlike previous studies, this study fills the gaps of methodological advancement by the use of Cross-Sectional Augmented Distributed Lag (CS-DL) and Cross-sectional Augmented Autoregressive Distributed Lag (CS-ARDL) advanced techniques robust against dynamics, heterogeneity, and cross-sectional dependence. Specifically, we examine the unit root test, panel cointegration analysis, and causal relationships among these variables. Therefore, the current study aims to fill the existing gaps in the prior studies and provide a prudent way of exploring the interaction of health status, health expenditures (both public-private), CO₂e, energy use, and economic growth in the T20HEE and the formulation of appropriate environmental policy implications in the selected countries.

Research methodology

Data description

The variables used in this study include health status, which following Rahman and Alam (2021) was proxied by life expectancy at birth (HS), public health expenditure (% of total health expenditure) (PBHEX), private health expenditure (% of total health expenditure) (PRHEX), total health expenditure (% of GDP) (HEX), CO₂ emissions (in kilotons) (CO₂), energy consumption (in quadrillion Btu) (ENERGY), and economic growth (GDP growth rate) (GDPGR). In addition to following the existing literature, the rationale for choosing the variables is that a combination of both public and private health expenditures ensures better health status; excessive CO₂e damage improves health outcomes; the green economy reduces energy consumption and provides adequate facilities for good health status, and economic growth contributes to improved amenities for ensuring health status. However, the data were sourced from the "World Development Indicators (2020)" and the "U.S. Energy Information Administration international energy statistics database (2020)". The data range is for the period 2000 to 2019 from the T20HEE, which comprises "China, the United States, India, Russia, Japan, Iran, Germany, Indonesia, the Republic of Korea, Saudi Arabia, Canada, South Africa, Brazil, Mexico, Australia, Turkey, the United Kingdom, Italy, France, and Poland." All the variables are in their natural values; thus, the paper did not take the log of any because none of the HS, CO₂, and ENERGY values are up to 100, while GDPGR, PBHEX, PRHEX, and HEX are all-in rates. As a result, there is no need to take either log.

Model specification

Following the objectives, three models will be estimated: to achieve the first objective, the first model specified in equation (1); to achieve the second and third objectives, the second model specified in equation (2); and to achieve the fourth objective, the last model specified in equation (3).

$$Health\ statu{s}_{it}=a+{\beta}_1C{O}_2 emission{s}_{it}+{\beta}_2 health\ expenditur{e}_{it}+{\beta}_3 economic\ growt{h}_{it}+{e}_{it}$$

(1)

$$C{O}_2 emission{s}_{it}=a+{\beta}_1 health\ expenditur{e}_{it}+{\beta}_2 energy\ use+{\beta}_3 economic\ growt{h}_{it}+{e}_{it}$$

(2)

$$Economic\ growt{h}_{it}=a+{\beta}_1 energy\ us{e}_{it}+{\beta}_2C{O}_2 emissions+{\beta}_3 health\ expenditur{e}_{it}+{e}_{it}$$

(3)

where for model one, the dependent variable is health status, while the independent variables are CO₂ emissions, health expenditures, and economic growth. For model two, the dependent variable is CO₂ emissions, while the independent variables are health expenditure, energy use, and economic growth. Lastly, for model three, the dependent variable is economic growth while the independent variables are energy use, CO₂ emissions, and health expenditure. However, a represents the constant term, β₁, β₂, and β₃ represent the respective coefficients of the parameters, e represents the error term, and i represents country i at time t.

Estimation techniques

The estimation techniques of this study will start with the graphical illustration of data to understand the nature of the data trend. The second method is the descriptive analysis for understanding the statistical characteristics of the data. The third method is the multicollinearity test. The fourth method is the heterogeneity test for checking cross-section dependence (CD) in the panels, such as Pesaran (2004) test, on which the majority of the research relies. The fifth method is the panel unit root tests for checking the stochastic properties of the panels, which were employed as none of the panels are free from cross-section dependence, such as the cross-sectional Augmented Pesaran and Shin (CIPS) panel unit root test of Pesaran (2007) that deals with the problem of cross-section dependence and heterogeneous slope coefficients. The sixth method is the panel cointegration test, which will be used to check for the long-run association between the variables, such as the Westerlund (2007) panel cointegration test, because the panels are cross-sectionally dependent and it does not necessitate previous knowledge of the order in which variables are integrated (Bhujabal et al., 2021), even though the significance or otherwise of the Error Correction Term (ECT) will be used for reaffirming the presence of the cointegration. The seventh method is the execution of the Cross-sectional Augmented Distributed Lag (CS-DL) and Cross-sectional Augmented Autoregressive Distributed Lag (CS-ARDL) techniques by Chudik et al. (2016) to estimate both short-run and long-run coefficients, where the two methods are complementary in nature. Nonetheless, this study will employ Eberhardt and Teal (2010) Augmented Mean Group (AMG) and the Cross-sectionally Augmented Error Correction Method (CS-ECM) for robustness checks, both of which produce better results with unobserved common factors, non-stationarity, cross-section dependence, and heterogeneous slope (Wahab et al., 2021). Both approaches are suitable for nonlinear or nonstationary variables in complex common factor structures where variables may be influenced by other factors. The reason for choosing the AMG is that it employs explicit estimates for unobserved common factors (f_t) that exhibit a common dynamic process as a meaningful construct, whereas the CS-ARDL and CS-DL estimators employ implicit estimates for f_t using cross-sectional averages of regressors. While the reason for choosing the CS-ECM estimator is that it is a second approach to the CS-DL technique that also considers multifactor error structure, Lastly, the Dumitrescu-Hurlin (2012) Granger causality test can be used to check the direction of causality between pairs of variables. Fig. 3 presents the flowchart of the empirical analysis.

Fig. 3

Flowchart of the empirical analysis

Cross-Sectional Augmented Autoregressive Distributed Lag (CS-ARDL) Model

The classic ARDL model is based on cross-sectionally independent error factors and homogenous units (countries). Error terms are likely to be cross-dependent because the models incorporate unobserved shared components. As a result, the fundamental ARDL model has been adjusted to account for both slope variation and error cross-sectional dependency. Therefore, Chudik et al. (2016) developed the CS-ARDL technique, a recent and advanced ARDL type estimator that can handle cross-sectional dependence in the presence of I(0) or I(1) order of integration and generate pooled long-run type estimates. The estimator also takes into consideration omitted variable bias. Apart from cointegration among variables, the only reason for using this type of estimator is that the model has a dynamic specification, which allows for weak exogeneity among the regressors to be taken into account and residuals to be no longer linked. The CS-ARDL approach, on the other hand, can be stated as follows:

$${y}_{it}=\sum_{j=1}^r{\beta}_{it}{Y}_{i,t-j}+\sum_{j=0}^p{\theta}_{ij}^{\prime }{X}_{i,t-j}+{\forall}_i+{e}_{it}$$

(4)

$${e}_{it}={\tau}_i^{\prime }\ {f}_{t,}+{u}_{it}$$

(5)

when f is an n × 1 vector of unobserved common factors; ${\tau}_i^{\prime }$ is the corresponding factor loading and e_it are the idiosyncratic errors. According to the both studies by Chudik et al. (2011) and Chudik et al. (2017), these idiosyncratic errors could be serially, cross-correlated, or uncorrelated with the common components. Unobserved common factors affect variables differently across nations, leading in erroneous cross-section dependence.

$${X}_{i,t-1}={\pi}_i^{\prime }{f}_{t,}+{v}_{it}$$

(6)

where X_{i, t − 1} are the vector of the explanatory variables. The assumption is that the explanatory variables could be influenced by the unobserved common factors 𝑓_t, ${\pi}_i^{\prime }$ is the n × m matrix of factor loadings, and 𝑣_it are the idiosyncratic errors associated with X_{i, t − 1} which are believed to be uncorrelated with 𝜀_it.

According to Chudik et al. (2011), the common factor structure can be expressed as a combination of a finite number (mfs) of strong factors (${f}_{it}^s$) that can be correlated with the regressors of the original model, and a finite number (mfw) of weak, semi-weak, and semi-strong factors ${f}_{it}^w$ that may affect a subset of the sample countries. Furthermore, according to Chudik et al. (2017), strong factors or shocks include structural changes, such as changes in the commodity price shocks, global financial cycle, and interest rates.

The weak, semi-weak, and semi-strong variables include spillovers from agricultural production, climate factor, domestic consumption, domestic industrial activity, research and development, geographical proximity, and natural resources (Eberhardt and Vollrath, 2016; Eberhardt et al., 2013; Holly et al. 2010).

Cross-Sectional Augmented Distributed Lag (CS-DL) Model

Despite the fact that the CS-ARDL technique can address cross-sectional dependence in the presence of I(0) or I(1) order of integration and obtain pooled long-run type estimates while accounting for omitted variable bias and having a dynamic specification, because the CS-ARDL technique calculates the short-run coefficients first and the residuals are no longer calculated (Pesaran 2015). However, there is a problem when the rate of convergence to the long-run estimate is slow and the time dimension is insufficient (Chakraborty & Mazzanti, 2021). Furthermore, another problem may develop if the sampling uncertainty is considerable and the short-run coefficient is subject to a slight T bias (see Pesaran 2015, page 782). As a result, Chudik and Pesaran (2015) developed a new estimation method in which the long-run coefficients are estimated without first calculating the short-run coefficients. The following is the CS-DL methodology, which is based on the ARDL method:

$${y}_{i t}={c}_{yi}+{\theta}_i^{\prime }\ {x}_{i t}+\sum_{\iota =0}^{p-1}{\delta}_{i\iota}{x}_{i,t-\iota }+\sum_{\iota =0}^{p\overline{y}}{w}_{y_{i\iota}}{\overline{y}}_{t-\iota }+\sum_{\iota =0}^{p\overline{x}}{\textrm{w}}_{{\textrm{x}}_{\textrm{i}\upiota}}^{\prime }{\overline{x}}_{t-\iota }+{e}_{i t}$$

(7)

where ${\overline{x}}_t={N}^{-1}\sum\limits_{i=1}^N{x}_{it}$ and ${\overline{y}}_t={N}^{-1}\sum\limits_{i=1}^N{y}_{it}$ and${p}_{\overline{x}}$is set equal to integer of T^2/3, $p={p}_{\overline{x}}$ and ${p}_{\overline{y}}$ is set to zero. Therefore, the mean group estimator for cross-sectional augmented distributed lag (CS-DL) can be stated as:

$${\overset{\frown }{\theta}}_{MG}=\frac{1}{N}\sum_{i=1}^N{\overset{\frown }{\theta}}_i$$

(8)

and

$${\overset{\frown }{\theta}}_i={\left({\overset{\sim }{X}}_i^{\prime }{M}_{qi}{\overset{\sim }{X}}_i^{\prime}\right)}^{-1}{\overset{\sim }{X}}_i^{\prime }{M}_{qi}{\overline{y}}_i$$

(9)

The CS-DL pooled estimator of the long-run coefficients can be written as:

$${\overset{\frown }{\theta}}_p={\left(\sum_{i=1}^N{w}_i{\overset{\sim }{X}}_i^{\prime }\ {M}_{qi}{\overset{\sim }{X}}_i^{\prime }\ \right)}^{-1}\sum_{i=1}^N{w}_i{\overset{\sim }{X}}_i^{\prime }\ {M}_{qi}{\tilde{y}}_i$$

(10)

Pesaran (2006) uses a heterogeneous number of regressors in his CCE estimator, but in the CS-DL estimator, θ_MG and θ_P incorporate ρ_T lags of ∆x_it and its cross-sectional averages. The length of ρ_T increases with T, and the optimal lag length can be determined using Chudik et al. (2016), Chudik and Pesaran (2015), and Pesaran (2006).

Nonetheless, the CS-DL and CS-ARDL econometric procedures take into consideration three crucial elements of panel data, namely dynamics, heterogeneity, and cross-sectional dependence.

Results and Discussions

Fig. 4 is the graphical representation of health status (HS), CO₂ emissions (CO₂), energy use (ENERGY), public health expenditure (PBHEX), private health expenditure (PRHEX), health expenditure (HEX), and GDP growth rate (GDPGR) of the top 20 highest emitting economies. According to the figure, HS, Energy, PBHEX, and HEX displayed an upward trend, though the trend of energy declined in the year 2009 but picked up by the year 2010, while PRHEX portrayed a downward trend. CO₂ increased until 2008, and then slowed until 2010, when it peaked in 2011, after which it plateaued until 2014, when it began to decline. GDPGR fluctuated throughout the period, but there was a huge decline in the year 2009. As a result, the trends of energy, PBHEX, and HEX moved in the same direction as that of HS, whereas PRHEX moved in the opposite direction, but for CO₂ and GDPGR, it is a fluctuating direction that cannot be directly compared. Looking at the trends in recent years, it can be seen that HS, ENERGY, PBHEX, PRHEX, and HEX are increasing, but CO₂ and GDPGR are decreasing.

Fig. 4

Graphical representations. Source: Author’s Computation. Note: HS stands for health status, CO₂ stands for CO₂ emissions, ENERGY stands for energy use, PBHEX stands for public health expenditure, PRHEX stands for private health expenditure, HEX stands for health expenditure, and GDPGR stands for GDP growth rate.

Table 1 displays the statistical characteristics of the variables. According to the table, the means and standard deviations of HS, CO₂, ENERGY, PBHEX, PRHEX, HEX, and GDPGR for the countries are 77, 8.9, 16.6, 62.7, 37.2, 7.4, 3.2, and 4.9, 5.1, 21.5, 16.5, 16.2, 3.1, 3.2, respectively. Therefore, for HS, CO₂, PBHEX, PRHEX, HEX, and GDPGR the mean value is greater than the standard deviation; thus, it can be stated that the data were not widely dispersed in the countries, but for ENERGY it is. The countries' maximum and minimum statistics of HS, CO₂, ENERGY, PBHEX, PRHEX, HEX, and GDPGR are 84, 20.5, 101.2, 84.3, 79.8, 17, 14.2 and 62.1, 0.9, 2.9, 18, 15.7, 3, -7.79, respectively. The skewness shows that the elements of CO₂, ENERGY, PRHEX, HEX, and GDPGR of the countries are positively skewed, but that of HS and PBHEX is negatively skewed, so none is 0; also, none of the kurtosis is 3. This highlights that the distribution of the individual series is not likely to be symmetrical and passively explosive. Yet, according to the p-value of the Jarque-Bera statistic, the distribution of all the variables seems to be volatile, as the p-value is significant at the 1% level.

Table 1 Descriptive statistics

Table 2 displays the correlation analysis of the variables for each of the three models. In model I, HS is moderate negative and positive correlated with CO₂ and HEX, respectively, while GDPGR is weakly but positively correlated with HS. For model II, CO₂ is moderate positive correlated with ENERGY, while HEX, and GDPGR both are moderate negative related to CO₂. For model III, there is a high positive correlation between GDPGR and ENERGY but a moderate negative correlation between GDPGR and CO₂ whereas there is a very low negative correlation between GDPGR and HEX.

Table 2 Correlation analysis

Table 3 reported the multicollinearity test results for each of the three models. The results indicated that all the variables evidenced that there is a serious multicollinearity problem between PBHEX and PRHEX variables as their variance inflation factor (VIF) is up to 10 and beyond. This depicts that between the two major remedies to the multicollinearity issue, such as dropping one of the variables or merging the two variables, the appropriate measure, in this case, is to merge the two variables, such as public health expenditure (PBHEX) and private health expenditure (PRHEX), which means total health expenditure (HEX). Therefore, total health expenditure will be used henceforth.

Table 3 Multicollinearity test results

Table 4 presents the Pesaran (2004) test for cross-sectional dependence (CD) of panel variables such as health status, CO₂ emissions, energy use, health expenditure, and economic growth. The results indicated that all the variables rejected the null hypothesis of no cross-sectional dependence in the data. Thus, it is concluded that there is cross-sectional dependence. Therefore, the Pesaran (2007) panel unit root test is used because none of the variables are free of cross-section dependence, which is indifferent, or rather, not sensitive to cross-section dependence, and will be used in place of the panel conventional unit root test.

Table 4 Cross-sectional dependence (CD) test results (Pesaran, 2004)

Table 5 shows the results of the Pesaran (2007) CIPS panel unit root test. The CIPS results show that the variables HS and GDPGR reject the null hypothesis of homogeneous non-stationary at 1% level, whereas CO2, ENERGY, and HEX do not reject the null hypothesis of homogeneous non-stationary at 1% level. The results of the panel unit root test indicate that all the variables are found to be stationary at the first difference at the 1% level. This demonstrates that the variables' order of integration is a mixture of I(0) and I(1).

Table 5 CIPS Panel Unit root test results (Pesaran, 2007)

Table 6 displays the panel cointegration test results obtained applying the method by Westerlund (2007) for each of the three models. According to the results, the test rejected the null hypothesis of "no cointegration" among the variables at 1%, 10%, and 1% for the models, respectively. Therefore, this indicates that the variables in each of the models are linked in a long-term cointegration relationship among the selected variables. This enabled the use of the CS-DL and CS-ARDL models.

Table 6 Panel Cointegration test results (Westerlund 2007)

Table 7 displays the results of the CS-ARDL, CS-DL, CS-ECM, and AMG techniques for both the short- and long-runs coefficients, except for the AMG technique, which has no short-run coefficients. According to the CS-ARDL and CS-DL results, the coefficient of CO₂ is negatively related to HS in the short-run, and the CS-ECM technique has also confirmed that it is negatively related to HS. The impact is significant at the 5% level, but the CS-DL technique's impact is higher, showing that a unit increase of CO₂ kilotons will deteriorate HS by 0.8. This means that in the short-run, an increase in the production of CO₂e is weakening the health status of people. The coefficient of HEX is positively and negatively related to HS, according to the CS-ARDL and CS-DL techniques, respectively, and both are significant at 10% and 5%, respectively. Since, both are significant in the opposite direction, the decision will be based on economic theory, indicating the relationship should be positive. Hence, a 1% increase in HEX will induce a 0.2 rise in HS. This shows that in the short-run, an increase in HEX improves the HS of people in the T20HEE. However, none of the techniques found the coefficient of GDPGR to be significant, though it was positively related to HS. This implies that economic growth is negligible in the short run, improving the countries' HS. The coefficient of ECM_t-1 is negative and significant at a 1% level, as shown by CS-ARDL and reconfirmed by CS-ECM techniques. The significance level confirms the earlier cointegration test result that there is a long-term association between the variables, and the model will automatically adjust itself by 61% (0.61) every year if there is a short-term shock.

Table 7 CS-ARDL, CS-DL, CS-ECM, and AMG test results

In the long run, according to the estimates of the CS-ARDL and CS-DL, the coefficient of CO₂ is negatively related to HS, which is also confirmed by the AMG technique, but is still significant as shown by all the techniques. This means, unlike in the short-run, an increase in the production of CO₂e is negligibly deteriorating the HS of people in the long-run. The coefficient of HEX is positively related to HS according to the techniques and also confirmed by the CS-ECM and AMG techniques, both of which are significant at the 5% level, which is also confirmed by the AMG technique. But the impact of CS-DL is higher, where a 1% increase in HEX will induce a 1.0 rise in HS. This implies that an increase in HEX improves individuals' HS in both the short and long run. However, in the short-run, none of the methodologies found the GDPGR coefficient to be significant in the long-run, implying that in the long-run, the GDPGR has no impact on the health of the people living in those countries. In all four techniques, the p-value of the heteroscedasticity test statistic of the estimated technique residuals is insignificant, which tells us that the residuals are homoscedastic, indicating healthy estimates of the techniques.

Table 8 shows the results of the CS-ARDL, CS-DL, CS-ECM, and AMG techniques for both the short- and long-runs coefficients, except for the AMG technique, which has no short-run coefficients. The short-run results show that the estimates of CS-ARDL and CS-ECM are the same but completely opposite in sign, except for the coefficient of ECM_t-1. Based on economic theory or following economic assertion, the results of CS-ARDL are consistent, whereas those of CS-ECM are diametrically opposed. Nonetheless, they are exactly the same in the long-run. Hence, the attention in the short-run will be on the CS-ARDL and CS-DL techniques, while in the long-run, it will be on all four techniques in the analysis. According to the estimates of the CS-ARDL and CS-DL, the short-run results indicate that the coefficient of Energy is positively related to CO₂, but only the impact of the CS-ARDL technique is significant at the 1% level, where a 1 unit increase in Quadrillion Btu of Energy will raise the production of CO₂ by 0.7. This means that in the short-run, an increase in energy use is causing the release of more CO₂e in these countries. Furthermore, the coefficient of HEX is negatively and positively related to CO₂ according to the CS-ARDL and CS-DL techniques, respectively, but neither is statistically significant. Hence, the decision will be based on economic theory, and accordingly, the relationship should be negative. This suggests that in the short-run, an increase in HEX is negligibly reducing the production of CO₂e in the T20HEE. However, none of the techniques found the coefficient of GDPGR to be significant, though negatively related to CO₂. This means that, in the short run, economic growth is moving in the direction of countries reducing CO2 emissions. The coefficient of ECM_t-1 is negative and statistically significant at a 1% level, as shown by CS-ARDL and reconfirmed by CS-ECM techniques. The significance re-endorses the earlier cointegration test result that there is a long-run association between the variables and that in the event of a shock in the short-run, the model will automatically adjust itself by 89% (0.89) per annum.

Table 8: CS-ARDL, CS-DL, CS-ECM, and AMG test results

In the long-run results, according to all techniques, the coefficient of Energy is positively associated with CO₂ and statistically significant at least at the 5% level, with the exception of the CS-DL technique, which is in the opposite sign and insignificant, thus, contradicting economic theory. However, the coefficient of the CS-ARDL technique is higher, where a 1 unit increase in Quadrillion Btu of Energy will raise the production of CO₂ by 0.9. This indicates that an increase in energy use causes the release of more CO2 in the countries in both the short and long run. From all the techniques, the coefficient of HEX is negatively associated with CO₂ and statistically significant at 10%, 10%, 10%, and 5% levels, respectively, but that of the CS-DL technique is higher where a 1% increase in HEX will reduce CO₂ by 0.7. This suggests that, unlike in the short-run, an increase in HEX reduces CO₂e in the long-run. However, none of the techniques found the GDPGR coefficient to be significant in both the short and long runs, with the exception of the AMG technique, which shows that GDPGR is negatively related to CO2 and the impact is statistically significant at the 5% level, with a 1% increase in GDPGR reducing CO2 by 0.03. This implies that in the long-run, the economic growth of the T20HEE countries is assisting in the reduction of CO₂e in the countries. In all the four techniques, the p-value of the heteroscedasticity test statistic of the estimated techniques residuals is insignificant, indicating that the residuals are homoscedastic; therefore, the techniques' estimates are healthy.

Table 9 presents the results of the CS-ARDL, CS-DL, CS-ECM, and AMG techniques for both the short- and long-runs coefficients, except for the AMG technique, which has no short-run coefficients. In the short-run, according to the estimates of the CS-ARDL and CS-DL, the coefficient of Energy is positively related to GDPGR and is significant at 10% and 1% levels, respectively, also confirmed by the CS-ECM estimate at 5% level, but the coefficient of the CS-DL technique is higher, where a 1 unit increase in Quadrillion Btu of Energy will raise GDPGR by 18.6%. This suggests that in the short-run, an increase in energy use causes high economic growth in those countries. The coefficient of CO₂ is positively related to GDPGR using CS-ARDL and CS-ECM techniques but negatively related to GDPGR using CS-DL and CS-ECM techniques; however, only the coefficient of the CS-DL technique is significant at the 10% level, where a 1 unit increase in kilotons of CO₂ will raise GDPGR by 1.3%. This indicates that, in the short-run, CO₂e is related to the increase in economic activities, such that the greater the CO₂e, the higher the economic growth of the T20HEE. However, none of the techniques found the coefficient of HEX to be significant, though positively related to GDPGR. This means that in the short-run, the HEX is negligibly aiding countries' economic growth. As indicated by CS-ARDL and validated by CS-ECM methods, the coefficient of ECM_t-1 is negative and significant at the 1% level. The significance confirms the earlier cointegration test, which concluded that there is a long-run association between the variables and that, in the event of a shock in the short-run, the model will automatically adjust itself by 120% (1.2) each year, resulting in a speedy convergence to the equilibrium path.

Table 9 CS-ARDL, CS-DL, CS-ECM, and AMG test results

In the long-run, according to the estimates of all the techniques, the coefficient of Energy is negatively related to GDPGR using CS-ARDL and CS-ECM techniques and negatively related to GDPGR when applying CS-DL and AMG techniques; however, only that of the CS-DL is statistically significant at the 5% level, where a 1 unit increase in Quadrillion Btu of energy induces a 34.5% increase in GDPGR. This means that an increase in energy consumption causes rapid economic growth in the countries in both the short and long run. The coefficient of CO₂ is positively related to GDPGR by CS-ARDL and CS-ECM techniques but negatively related with GDPGR by CS-DL and AMG techniques; however, only the coefficient of the CS-DL technique is statistically significant at the 10% level, where a 1 unit increase in kilotons of CO₂ will decline GDPGR by 17.8%. This means that unlike in the short-run, where it is aiding the economy, in the long-run, the production of CO₂e is extremely damaging the economic growth of the T20HEE. However, all the AMG techniques found the impact of HEX to be significant in relation to GDPGR at the 1% level, and the impact is negative, where a 1% increase in HEX will retard GDPGR by 1.4%. This implies that health expenditures have no impact on economic growth in the long-run. In all four techniques, the p-value of the heteroscedasticity test statistic of the estimated technique residuals is insignificant, and this indicates that the residuals are homoscedastic, which shows healthy estimates of the techniques.

Table 10 illustrates the Pairwise Dumitrescu-Hurlin (2012) panel Granger causality test results. According to the findings of the study, we observed that there is a bidirectional causality between HS and CO₂, HS and HEX, HS and GDPGR, HS and Energy, HEX and CO₂, CO₂ and GDP, Energy and CO₂, and Energy and HEX, but a unidirectional causality running from Energy to GDPGR. However, there is no causality between HEX and GDPGR. The outcomes of the Granger causality are graphical presented in Fig. 5.

Table 10 Pairwise Dumitrescu-Hurlin (2012) Granger causality test results

Fig. 5

Graphical representation of Granger causality test results

Discussion of the Results

This paper explored the interaction of health status, CO₂ emissions, and energy use among the T20HEE. The first model is to achieve the first objective, which is to examine whether health status (HS) is affected by CO₂ emissions (CO₂) of the T20HEE, where variables – health status, CO₂ emissions, health expenditure, and economic growth were used. The results revealed that in the short-run, CO₂ emissions are weakening the health status, but in both short- and long-runs, health expenditure improves the health status, while in both the short- and long-runs, economic growth is not contributing to the health status of the countries. This finding is in line with that of Igbinedion (2019) in Nigeria on CO₂, Taghizadeh-Hesary et al. (2020) in low- and middle-income Asian countries on CO₂, Wang et al. (2020) in China on CO₂, Osabohien et al. (2021) in Nigeria on CO₂, and Zeeshan et al. (2021) in China on CO₂, and environmental pollution shocks increase health expenditure both in the short- and long-runs. Therefore, health status is affected by CO₂e of the T20HEE.

The second model is to achieve the second and third objectives, which are to explore whether health expenditure is helping in mitigating CO₂ emissions and to assess how energy use is related to their production of CO₂ emissions in the T20HEE, where variables – CO₂ emissions, energy use, health expenditure, and economic growth were used. The results revealed that in the long-run health expenditure and economic growth are helping to mitigate CO₂e, while in both the short and long runs; energy use is increasing CO₂e in the selected countries. This is consistent with the findings of Keho (2017) in 59 countries on energy use, Apergis et al. (2018) in sub-Saharan African countries on health expenditure, Gündüz (2020) in the USA on health expenditure, Bilgili et al. (2021) in 36 Asian countries on health expenditure, Ganda (2021) in BRICS countries on health expenditure, Liu et al. (2019) in China, India, and G7 countries on energy use, Ssali et al. (2019) in six sub-Saharan African countries on energy use, Zou and Zhang (2020) in 30 Chinese provinces on energy use, Litavcová and Chovancová (2021) in 14 Danube-area nations on energy use, Salari (2021) in the United States on energy use, and Wu et al. (2022) in the Nordic countries on energy use. However, it is in contrast with the finding of Mensah et al. (2018) in 28 OCED nations on economic growth. Therefore, health expenditure is helping to mitigate CO₂e, and energy use is raising the level of CO₂e in the selected countries.

The third model is to achieve the fourth objective, which is to examine how energy use and CO₂ emissions are related to the economic growth of the T20HEE, where variables for economic growth, energy use, CO₂ emissions, and health expenditure were used. The results revealed that both in the short and long runs, energy use contributes to high economic growth in the selected countries, whereas CO₂e is aiding economic growth in the short run but extremely damaging it in the long run, while both in the short and long runs, health expenditure is not aiding the economic growth of the countries. This finding is in line with that of Azam et al. (2016) in selected higher CO₂ emitting countries on CO₂e, Ghorashi and Alavi (2017) in Iran on CO₂e, Ummalla (2019) in the BRICS countries on energy use, Wang et al. (2020) in China on CO₂e, Radmehr et al. (2021) in the European Union on CO₂e, Adebayo (2022) in China on CO₂e, and Adebayo et al. (2023d) in BRICS countries on CO₂e. Therefore, energy use is influencing the economic growth of the countries, while CO₂e is retarding it.

Conclusion and Policy Recommendations

The current study examined the interrelationship of human health status, both public-private health expenditures, CO₂ emissions, energy use, and economic growth, which could be pretty interesting in the context of the top-emitting economies from 2000 to 2019. To achieve this goal, we employ advanced techniques of Cross-Sectional Augmented Distributed Lag (CS-DL) and Cross-sectional Augmented Autoregressive Distributed Lag (CS-ARDL), which take into consideration crucial elements of panel data, namely dynamics, heterogeneity, and cross-sectional dependence. The study adopted advanced econometrics analysis of second-generation panel methods such as Pesaran unit root test, Westerlund cointegration test, and the Dumitrescu-Hurlin Granger causality test was employed to determine the causal relationships between the variables. The study also employs the Cross-sectional Augmented Error Correction Method (CS-ECM) and the common dynamic process of the Augmented Mean Group (AMG) for robustness testing. The findings revealed that CO₂e weakens the health status only in the short-run, whereas health expenditure improves the health status in the both short- and long-runs, while economic growth is not contributing to the health status in the both short- and long-runs. Furthermore, health expenditure and economic growth only help to mitigate CO₂e in the long-run, whereas energy use causes CO₂e in the both short- and long-runs. Moreover, energy use causes high economic growth in the both short- and long-runs, whereas CO₂e aids economic growth in the short-run but is extremely damaging to economic growth in the long-run, while in the both short- and long-runs health expenditure is not aiding the economic growth. Additionally, the Pairwise Dumitrescu Hurlin Panel Causality Test revealed that a bidirectional causality exists between health status and CO₂ emissions, health status and health expenditure, health status and economic growth, health status and energy, health expenditure and CO₂ emissions, energy use and CO₂ emissions, CO₂ emissions and economic growth, and energy use and health expenditure, while a unidirectional causality exists between energy use and economic growth. On the other hand, there is no causal relationship between health expenditures and economic growth.

Therefore, CO₂ emissions are harming the health status, health expenditure is tackling the CO₂ emissions, energy use is fueling the rate of the CO₂ emissions and is stimulating economic growth, while CO₂ emissions are retarding economic growth. Hence, the current study offers the following policy recommendations for the T20HEE and global policymakers to improving people’s health and environmental policy framework which will lead to accomplish the SDGs, particularly, SDG-3 and SDG-7, as follows:

First, the governments of the T20HEE should increase both public-private health expenditures by adopting a cost-effective service and energy-efficient health management approach in order to achieve sustainable healthcare services and create a sustainable environment. As a result, the proceeds of economic growth should be prudently channeled by increasing the portion of healthcare expenditure, thus improving health status and addressing the negative effects of CO₂e, thereby, creating a healthy society with a green and clean environment for long-term development for the people. In addition, the health and well-being of patients can be improved by helping healthcare workers communicate more effectively. Improving health also necessitates strategies to ensure that medical professionals are aware of suggested services and treatment protocols.

Second, as CO₂e has a significant negative impact on health by generating various health-related diseases, such as heart problems, lung problems, and recently, flu related diseases such as COVID-19. In this regard, policymakers should adopt green and clean environment policy to protect the environment quality and minimize the incidence of health diseases. It is also essential to develop and implement suitable environmental laws, particularly in the T20HEE to avoid the adverse impact of environmental pollution on human health.

Third, the governments and policymakers of the top-emitting economies should made additional efforts to promote technical innovation and cleaner production for manufacturing purposes in order to reduce consumption-based CO₂e to secure a healthy environment for the people. In addition, the policymakers should be imposed a heavy carbon tax on emissions-intensive products and imported resources for both production and consumption purposes. Because the production process of these top-emitting countries emits high CO₂ and resulting serious pollution in the environment. Therefore, policymakers should be implemented strict policy rules that assure the usage of modern environmental-friendly technologies and promote CO₂e reduction technologies to improve environment quality. Moreover, governments of these economies should encourage locals to use less energy, recycle, and place a higher value on environmental products and services. It could be helpful to reduce the number of cars on the road and the amount of fossil fuels usage, and to encourage people to use public transportation and bikes instead of their cars.

Fourth, the governments of the top-economies must increase necessary investments in the field of technological innovation, especially in the energy sector, which will enhance energy efficiency infrastructure and boosting the usage of pollutant-free alternative energy sources and increase the supply of green energy. Additionally, public procurement reform can have a significant influence. Governments may reduce their own carbon footprint by using greener goods and services, while also motivating other players to enhance their own sustainable consumption and production habits by establishing strong green contract criteria.

Fifth, the T20HEE could not achieve environmental sustainability if they continue to increase their dependence on fuel energy sources (coal, gas, and petrol) to fulfill their domestic demands in the energy sector. As renewable (clean) energy consumption is essential for reducing CO₂e, the governments of these economies should take considerable steps to develop supportive policies which increase the investment in clean energy sources and reduce the reliance on fossil fuel energy. These top-emitting countries consume extensive energy, therefore policymakers should carefully implement energy-saving policies with caution, as changes in energy could obstruct their economic growth rate, and also, they must strike a balance between energy production and economic growth. The industries which contribute to fuels energy sources should develop alternative energy such as clean energy sources in the manufacturing and transportation processes for sustainable economic growth. Therefore, policymakers should facilitate industries and organizations to promote green energy technologies and increase the investment in clean energy sources in order to achieve the SDGs-7.

Finally, subsidies and tax breaks are further strategies to increase the demand for eco-friendly goods and services, including electric vehicles, solar energy, and renewable energy. To encourage innovation and the development of game-changing technologies like renewable energy, carbon capture, waste management, and energy efficiency, governments must provide subsidies and grant funding to research organizations, academic institutions, and private R&D enterprises.

However, the current study has some limitations that should be addressed in future research studies. The large amount of the world population and high level of development and production of the top-emitting economies, the findings of this study may not be applicable to less emitting or other group of countries. However, further studies could conduct similar research on less developed economies or countries with less than one percent of the total world emission and examine the nexus among health status, CO₂ emissions, and energy use across the economies. Nonetheless, such study could be conducted among the top industrial economies as well as among the energy transition economies. Furthermore, since CO₂e are major contributor of environmental pollution, it is important to consider the negative impact of other GHGs emissions on human health, such as methane (CH₄), nitrous oxide (N₂O), and sulfur dioxide (SO₂). Moreover, potential limitations of this study could not be extended to the current time period due to the unavailability of relevant data. Lastly, this study explores the relationship between health status, CO₂e, and energy use in the T20HEE. Further studies can not only investigate this relationship in other economies but also need to include other important economic variables, such as financial development, technological innovation, globalization, etc.