Investigating the subsistence of Environmental Kuznets Curve in the midst of economic development, population, and energy consumption in Bangladesh: imminent of ARDL model.

The main purpose of this study is to analyze the existence of an environmental Kuznets curve (EKC) considering the midst of energy consumption, population and economic development. The main objective is to investigate the impact of energy consumption, population and economic development on CO2 emissions. This study has taken data from 1971 to 2020 to see the existence of an EKC in the country of Bangladesh. Besides population growth, energy consumption and economic development are also taken into consideration. An autoregressive distributed lag (ARDL) model was used to scrutinize cointegration based on selected variables and their respective I (0) and I (1) values. This study has confirmed the long-term existence of the EKC in the environment. The environmental Kuznets curve was also tested using economic performance coefficients on emissions. In the long run, EKC explains why per capita carbon output decreases with population expansion but turns down after a certain threshold level is achieved because of this inverted U-shaped pattern. For decades, increased energy consumption has been linked to worsening environmental conditions, according to this study. According to the findings, there are a wide variety of approaches to advancing Bangladesh's economy and improving its environmental quality. In the long run, the population has no positive impact on CO2 secretion. The use of fossil fuels such as gas and oil can have a detrimental environmental impact. As a result, if we want to conserve the environment, we need to use renewable energy sources like solar and biodiesel instead of traditional, nonrenewable fuels.

Introduction

International organizations and governments widely acknowledge the importance of environmental quality to a country's economic well-being. The concepts of “green growth” and “sustainable development” are derived from this notion. As it moves toward economic development and improvement, achieving and maintaining a high level of productivity is the ultimate goal. In addition, every country has its own unique set of obstacles when it comes to economic growth and advancement. Climate change and environmental contamination are two of the most serious consequences. Petroleum derivatives provide a significant portion of the world's energy needs, but the supply is rapidly depleting as a result of the enormous amount of energy being consumed (Rahman et al., 2020, 2021; Khan et al., 2022). Toxic carbon dioxide emissions harm both the environment and interior air quality. Environmental phenomena such as droughts, floods, tornadoes, and increasing ocean levels are all a result of climate change. For both poor and developed countries, an increase in the Earth's average temperature and climate change are among the most pressing ecological concerns of recent years (Wu et al., 2018). Concern has been expressed by natural scientists, sociologists, and others over the degrading of the environment caused by anthropogenic pollution. Other than human beings, wildlife and plants are also adversely affected by climate change in several different ways (Stern, 2008; Saud et al., 2019). As a result, any country must understand the rapport between economic progress and environmental damage. The EKC hypothesis is critical to resolving this problem. According to Kuznets (1955), inequality grows first and then declines as an economy develops, giving rise to the inverted U-shaped connection known as the Kuznets curve. This notion was first proposed in the 1950s, and it has since gained popularity. Moreover, the Environmental Kuznets Curve (EKC) hypothesis was developed based on Kuznets (1955) study of nexus between economic growth and income inequality. In his study he found that income inequality may initially increase as economic growth increases but that it would eventually tend to decline once economic growth reaches a certain level. An inverted U-shaped curve can depict this relationship. Based on Kuznets' study, Grossman and Krueger (1991) conducted a study on NAFTA to assess the impact of economic growth on environmental quality. They found that the environmental quality of NAFTA countries will deteriorate before the income of these countries reaches a certain level during the development process. Hence, this study indicates an inverted “U” relationship between environmental quality and economic development. However, Perman and Stern (2003) have shown in their environmental book in chapter 6 that the economic benefit is a function of emission (i.e. ). If we consider income as a proxy for economic benefit, and this income is generated from the production of goods and services that ultimately increase greenhouse gas (GHG) emissions, therefore, emissions increase the income of a country at the cost of deteriorating its ambient environmental quality. In this context, the government as well as different environmental organizations put pressure on firms and industries to curb GHG emissions to improve the ambient environmental quality. To make this pressure credible to firms and industries, government authorities can use various command and control instruments (i.e. technology control, input controls over quantity or mix of inputs, output controls (i.e. output quotas or prohibitions), emission licenses, and location controls (i.e. zoning, planning controls, relocation)), economic incentive instruments (i.e. emissions taxes, natural resource taxes, product taxes, emissions abatement subsidies, marketable emissions permits, deposit-refund systems, and liability payments etc.), and codification of liability as well as development of social responsibility for reducing environmental damage. To escape these rules and regulations, if the income of the firms and industries increases sufficiently, they have tried to use environmentally friendly technology to control their emissions discharge at environmental receptors. As a result, environmental quality will gradually improve in the long run. Therefore, in this way, the EKC hypothesis is supported by a country if their income increases significantly in the long run. The EKC concept has been applied to a variety of disciplines, including the relationship between environmental degradation and development, where it is anticipated that as an economy develops, environmental pollution will first increase and then decrease after a certain level of income. This idea is known as the inverted U-shaped EKC curve, which was first mentioned by Grossman and Krueger (1993); Panayotou (1993a); Schindler (1996); and Rothman (1998). Environmental Kuznets Curve (EKC) research began in the 1990s. Since then, there has been a steady growth in the number of research papers that have looked into the relationship between environmental conditions and economic development. Like other EKC model applications, this research seeks to investigate whether there is a threshold level of environmental deterioration that can be achieved when a nation has attained a particular degree of socioeconomic development (Laverde-Rojas et al., 2021; Adedoyin et al., 2020).

However, the key objective of this study is to learn more about the liaison between economic growth, population growth, and energy use on emissions. The STIRPAT model in its extended form was used in this paper. The EKC in Bangladesh was studied using energy consumption as a technological variable and oil and gas consumption as two separate subcategories. Power consumption in Bangladesh is dominated by the use of oil and natural gas. This study potentially examines the effects of these two variables on EKC separately, and the ARDL bound test has been used in conjunction with financial development, population, and energy consumption. The ARDL model outperforms alternative cointegration econometric techniques, and it is commonly utilized for the validation of the EKC, as well as other cointegration studies. There are four sections: the first provides an overview of the literature, followed by a discussion of the applied technique, followed by an in-depth analysis and discussion of the results, followed by an explanation of the conclusion and policy implications.

Literature review

Urbanization in Singapore from 1970 to 2015 was studied by Ali et al. (2017) to see how it affected carbon secretion. In order to evaluate time series data, they employed the ARDL approach. The findings of this study revealed that urbanization has a negative and statistically significant influence on CO2 emissions in the Singapore metropolitan area. The Consequences further noted that economic development has a progressive and large influence on carbon emissions. In Nigeria, Sulaiman and Abdul-Rahim (2018) looked at the connection between population increase and emissions. In this regard, they used the autoregressive distributed lag model based on three different periods. Emissions were not influenced by long-term population expansion, and population growth could have an impact on CO2 in the short term. When it came to environmental degradation in Nigeria, Maijama and Musa (2020) looked at how the country's urbanization and the country's crude oil price interacted using Kwiatkowski Philip Schmidt Shin and Augmented Dickey-Fuller tests in conjunction with the ARDL Model. Data from 1981 to 2016 was used in the study. According to the research, environmental damage was found to be inversely related to the price of crude oil and foreign direct investment (FDI) in both the long and short terms. However, it was discovered that urbanization was associated with environmental pollution in a statistically significant and positive way, showing that urbanization was the principal source of environmental pollution in the country. With the help of time series data from Lithuania from 1989 to 2018, Habib-Ur-Rahman et al. (2020) attempted to determine whether there was an association between EKC, financial expansion, and economic progress. They employed the ARDL bounds testing approach to conduct their empirical study, and they discovered that there was no correlation between the two variables evaluated. According to the findings, this association was found to be applicable in the short and long term, with a reversed U-shape between CO2 and economic growth. Rahman and Majumder (2020) investigated an energy and environmental issue in Bangladesh, where they discovered that energy was the cause of increased CO2 emissions which findings is silimar with Rahman et al. (2020). According to Ozden and Bese (2021), ARDL and nonlinear ARDL models were used to scrutinize whether an increase in economic growth was associated with a swell in greenhouse gas emissions over the period from 1960 to 1994. The EKC hypothesis was used to determine the relationship between gross domestic product, energy consumption, and CO2 emissions over this period. The research team wanted to find out if a rise in economic growth is connected to enhanced greenhouse gas emissions. They were successful. According to the findings, economic development and CO2 emissions did not appear to be connected in a symmetrical or asymmetrical way, as previously thought. In addition, the findings revealed that the EKC hypothesis had not been proven in the Australian setting, which was surprising. Sharri et al. (2020) conducted a 20-year study of the effects of gas and oil utilization on CO2 emissions in OIC member countries from 1990 to 2017. Dynamic heterogeneous panel approaches, including dynamic fixed effects (DFE), pooled mean groups (PMG), and mean groups (MG), found that national production correlates to higher environmental deprivation over the long term. However, even though national output has little impact on CO2 emissions in the short run, Also, the results indicated that the population may reduce emissions of CO2 in the short-term, but there was no long-term impact.

However, Villanthenkodath and Arakkal (2020) used data from 1970 to 2018 to investigate the prevalence of the environmental Kuznets curve in New Zealand's economic expansion, foreign direct investment, and openness to trade. The ARDL model has been used to check cointegration; empirical results verified the presence of EKC in New Zealand. This study's findings also demonstrated that financial development, FDI, and trade openness all contributed to an increase in environmental eminence, while the scatter plot revealed a reversed U–shaped relationship between CO2 and economic development. During the period from 1976 to 2013, Boufateh (2019) investigated the relationship between real GDP, oil prices, and emissions in China and the United States. Additionally, he was unable to locate EKC in either China or the United States. Rahman and Ahmad (2019) discovered that, between 1980 and 2016, there was an asymmetry link between emissions and capital formation as well as EKC in the country of Pakistan. This was true for the period from 1980 to 2016, when the data was collected. A significant contribution to emissions was discovered by Rahman and Ahmad in the form of coal consumption (CNS) and oil consumption (OCN). Using the NARDL model, Phiri and Andrew (2019) inspected the relationship between the GDP and environmental deprivation in the country of Eswatini between 1970 and 2014. They discovered that there was no EKC for the country of Eswatini. Alam et al. (2016) did a study on Brazil, China, India, and Indonesia by applying the ARDL model in two independent models (i.e., linear and quadratic models) spanning the period 1970 to 2012. They found that the EKC hypothesis exists in Brazil, China, and Indonesia, except for India. They also found that except for China and Indonesia, in India and Brazil the population growth has a statistically considerable impact on CO2 emanation both in the short-run and long-run under the linear and quadratic model. In each of the four countries investigated, income and energy usage had a significantly positive impact on emissions of CO2. Pata and Caglar (2020) conducted a study on China to investigate the existence of the EKC hypothesis in terms of income, renewable energy consumption, human capital, and trade openness. They identified that the EKC does not exist in China. However, increasing human capital worsens the ecological footprint as well as income, globalization, and trade openness deteriorates environmental quality, whereas consumption of renewable energy does not affect environmental quality in the long run. Pata (2021) conducted a study on the United States using the EKC framework to assess the impact of globalization, consumption of renewable and nonrenewable energy, and economic complexity on emissions and the ecological footprint. He found an inverted 'U' shaped EKC between economic complexity and emissions. He also found that consumption of renewable energy and globalization significantly reduces emissions, whereas consumption of nonrenewable energy worsens the environment. Tenaw and Beyene (2021) conducted a study on 20 Sub-Saharan African (SSA) countries based on the environmental sustainability-oriented modified EKC framework. They found an inverted ‘U’ shaped EKC holds for resource incentive SSA countries, whereas it does not hold for non-resource incentive SSA countries. They also found the detrimental effects of energy consumption, foreign direct investment and trade openness on the environment in the long run, whereas financial development and livestock production improve environmental quality in the long run. Alola and Ozturk (2021) investigated the risk of investment-induced EKC for the U.S.A. They found that EKC holds for the U.S.A. and consumption of renewable energy significantly improves environmental sustainability, but risk to investment under the EKC framework (i.e. GDP2) has a positive but not statistically significant impact on environmental sustainability. However, without the EKC framework, the low risk of investment significantly improves environmental quality both in the short and long run. Pata and Samour (2022) conducted a study in France to investigate the impact of consumption of renewable and nuclear energy on CO2 emissions, ecological footprint, and load capacity. They did not find any ‘U’-shaped nexus between CO2 emissions and income, as well as the consumption of renewable energy has no statistically significant impact on environmental quality, but the EKC hypothesis holds for load capacity factor. However, they also identified that consumption of nuclear energy significantly improves environmental quality and signifies green sustainability. Whatever the case, little research has been undertaken on the association between economic growth and CO2 secretion to establish the elements that influence CO2 emissions as well as feasible solutions for improving environmental quality and the environment. Variables taken into account include energy consumption, population growth, gross domestic product (GDP), and the openness of the economy to trade, foreign direct investment, urbanization, productivity, and the adoption of new technology. Recently, there has been an increase in the number of studies being conducted on the impact of monetary development on the increase or decrease of CO2 emissions. According to researchers, the impact of monetary development on the increase or decrease of CO2 emissions development of financial development, which comprises economic institutions and financial markets, has been recommended as a crucial factor in reducing CO2 emissions. The association between industrialization and carbon emissions has been seen regularly all over the world, but only a small amount of research has been done on the nonlinear relationship between industrialization and CO2 in Bangladesh. Murshed (2022) tested the deforestation-induced EKC hypothesis in three different scenarios for Bangladesh. In his empirical analysis, he segregated the deforestation into forest area coverage, deforestation rate, and forest depletion. However, when these variables are regressed on GDP, GDP2, agricultural land use, population, and democracy, the situation becomes clear. He revealed that the estimated results demonstrate a causal relationship among the variables; the cointegration exists in the long run; and the EKC hypothesis is supported in all three distinct scenarios. He also identified that democracy reduces deforestation, but while the interactive effect of economic growth and democracy persistently increases deforestation. Miah et al. (2010) reviewed the existing literature on EKC based on CO2, SOx, and NOx emissions to better understand how the economic development processes degrade environmental quality, contribute to global climate change, and suggest policy implications for Bangladesh. They observed that a monotonous straight line EKC exists in most cases for CO2 emissions, while the actual shape of the EKC for SOx exists in most countries and the EKC for NOx exists in most developed countries at lower income turning points. They also observed that Bangladesh has a straight line EKC and has the right to pollute because it is a low-income country, but in order to control environmental damage and climate change hazards, it should change its energy use policies, market regulation policies, and use clean technology. However, in this study, the CO2 induced EKC has been tested for Bangladesh in terms of GDP, GDP2, population, and energy use. The purpose of this study to work with these variables is to demonstrate that Bangladesh reached lower middle income country from low income country in 2015 and is on track to reach the UN's developing nation list after the completion of its LDC graduation criterion within 2026. This study has been analyzed the following research questions which are clearly explain the research contribution. As the income of Bangladesh has increased substantially in the last two decades, has the capacity to reduce the CO2 emissions of Bangladesh increased or not? However, in terms of population, Bangladesh ranked as the 8th most populous country in the world with a 166.30 million population as of July 1, 2021. Deforestation keeps going up because a large population needs more and more basic things. In this situation, how does this ever growing population contribute to CO2 emissions? In addition, according to the seven-five-year plan of Bangladesh, the per capita energy consumption will be raised to 514 kW h from 371 kW h. To meet this growing demand for energy consumption, Bangladesh basically relies on oil and coal-based power plants. As a result, the concentration of CO2 emissions and other greenhouse gases increases day by day in Bangladesh. In this context, how does energy use contribute to CO2 emissions in Bangladesh? Moreover, this research fills in the gaps by investigating the EKC and how population, energy, and economic development affect the environment. Environmentalists, government officials, industrial researchers, politicians, and the general public have found this study useful in analyzing and comprehending how population, energy, and economic development influence CO2 emissions. The significance of this study is that this is the first study on Bangladesh to take into account asymmetric impacts. As a result, the structural change brought about by industrialization has a variety of effects on CO2.

Conceptual framework, methodology, and data

Data

In order to construct the annual time series dataset from 1971 to 2020, the World Development Indicators (WDI) was employed. In order to calculate annual CO2 emissions per person in metric tons, the Carbon Dioxide Information Analysis Center (CDIAC) uses the WDI. The annual real GDP per capita from WDI source is used to estimate income (constant 2010 US dollars). The WDI predicts yearly energy consumption per person in kilograms of oil equivalents. In addition, the amount of oil equivalent per person in kilograms was calculated. A wide range of data from the entire population was also taken into account. In this system, each variable is represented by using natural logarithms which is presented in Table 1.

Table 1

List of the variables.

Log Form	Details	Sources
LnCO2	CO2 emissions (metric tons per capita)	WDI
LnGDPPC	GDP per capita (current US$)	WDI
LnGDPPC2	Square value of GDP per capita.	WDI
LnOil	Energy use (kg of oil equivalent per capita)	WDI
LnGas	CO2 emissions from gaseous fuel consumption (kt)	WDI
LnPop	Population, total	WDI

Model construction

For the purposes of this study, the ARDL model was utilized to determine whether the EKC hypothesis is valid in Bangladesh. The correlation between CO2 emissions and characteristics such as affluence and population, as well as with oil and gas use, was examined in this empirical inquiry. As a result, Eq. (1) is based on the fundamental model shown below. For time series analysis, the ARDL model developed by Pesaran et al., 1999, Pesaran et al., 2001 is better suited than the Vector Autoregression (VAR) model since it incorporates both endogenous and exogenous variables. Many benefits can be gained by using the ADRL model for environmental connections. First, when the sample is small and finite, the ARDL cointegration test is preferred to the Johansen cointegration test. Second, the ARDL model does not necessitate that all variables be integrated at the same level in order to run. If the variables' integration order is set to either I (1), I (0), or mixed, the ARDL model can be run without any confusion. Third, the ARDL model can be implemented using a single regression equation rather than many regression equations. As a result, the ARDL model's estimation and interpretation are simpler and more straightforward than those of the VAR model. Additionally, this model yields ECM as well as short-and long-run coefficients at the same time. ARDL model has some limitations such as, the stochastic process has been considering in model estimation and sometimes model continue with structural break Geyikci et al. (2022). Energy use and CO2 emissions have been extensively studied using the Impact, Population, Affluence, and Technology (IPAT) model. The elements that have an impact on the environment are rather simple to comprehend in the model. A consequence of this is that the IPAT model is being utilized in this study to evaluate the effects of national output and energy consumption on CO2 emission levels. IPAT is an abbreviation that stands for environmental impact (I), population (P), affluence (A), and technological advancement (T).

The fundamental model is as follows:

The analytical and functional framework of EKC, the literature recommends the use of the following Eq. (1) to estimate the long-run connection involving environmental degradation and related variables.

CO2 refers to carbon dioxide emissions, and GDPPC refers to real per capita gross domestic product. In light of the fact that real GDPPC is used to calculate the square component, GDPPC2 is defined as the square term of real GDP per capita. The term was created using orthogonal transformation in order to avoid complete multicollinearity with real GDP per capita. The letter EC is used to signify the amount of energy consumed per person. P is an abbreviation for population. For the empirical analysis point of view, Eq. (2) is converted to the following form consisting of parameters in Eq. (3).

Energy consumption (EC) will be separated into two types based on Eq. (3), namely natural gas consumption (G) and oil consumption (O), with the following equation describing each type:

αi stand for parameters, t here is time periods, and εt error terms for usual assumptions.

Econometric methods

The ARDL model, the limits testing technique to cointegration first described by Pesaran and Shin (1999) and then refined by Pesaran et al. (2001), is used because L (Gas) is I (0) by all tests. L (Pop) is I (0) only in KSSUR test and the other variables are I (1).

We write our Eq. (4) into ARDL from at Eq. (5):Where, Δ denotes the differenced operator, as well as all of the other variables, defined earlier in the expression. When using the ARDL bounds approach, the first step is to estimate Eq. (5) using Ordinary Least Squares (OLS), and then to use an F-test for the joint significance of the coefficients of the lagged levels of the variables to test for the existence of a long run (cointegration) relationship, thereby rejecting the null hypothesis that there is no cointegration relationship between the variables, defined as . , n = 1,2,…5, is compared to the alternative hypothesis of the presence of a cointegrating link. According to the alternative hypothesis, none of the variables are equal to zero. Pesaran et al. (2001) determined that there are two crucial values for assuming that any variable will be stationary at I (0) and I (1) using the F-statistic they developed. Based on the lag length, the AIC and BIC may find the optimal lag length for variables. It was estimated by the ARDL approach that there were as many regressions as (p + 1) K, where p represents the number of lags and K represents the number of variables in the equation. The AIC nominates the greatest lag length connected to the variables, and the SBC selects the lag length that is the shortest possible concerning the variables. As a result of the lag length being chosen and the presence of long-run association being detected, the error correction model is represented as-Where, the term refers to the long-run equilibrium speed of adjustment following a short-term shock. This research also includes diagnostic tests to determine the goodness of the model, such as the functional form, serial correlation, and heteroscedasticity tests, which are all included in this study. These tests were also used in this investigation, including the CUSUM and CUSUMSQ stability tests. The regression model will be considered stable if the results of these tests fall inside the key boundary of the 5% significance level of the significance level.

Result analysis and discussion

Table 2 delineates the descriptive statistics data of the variables. Column 2 and 3 shows mean and standard deviation. Also, column 4,5, and 6 shows skewness, kurtosis, JB test results (see Table 2).

Table 2

Descriptive statistics.

VARIABLES	(1)	(2)	(3)	(4)	(5)	(5)	(6)
	N	mean	sd	Skewness	Kurtosis	Jerque-Bera	P-Value
L (CO2)	50	9.853	0.917	-0.128	2.894	11.542	0.021∗∗∗
L (GDPPC)	50	5.979	0.759	0.147	3.121	82.154	0.0869∗∗
L (Pop)	50	18.52	0.288	0.123	12.253	110.14	0.6214
L (Oil)	50	4.950	0.313	-0.863	5.435	14.162	0.0541∗∗
L (Gas)	50	9.252	1.272	0.672	4.989	10.842	0.1711
L (GDPPC2)	50	36.31	9.351	2.68	14.453	124.251	0.2847

Source: author's calculation.

In order to be called normal, the series must have Skewness 0 and Kurtosis 3. The Jarque-Bera test statistic is used to determine if a series is normally distributed. According to this null hypothesis, the skewness (S) is equal to zero, and the kurtosis equals three. Skewness and kurtosis measurements reveal that most values are not in a normal distribution. Probability of normality, on the other hand, indicates the significance of three variables.

Before doing the ARDL bound test, it is important to execute a unit root test on the variables in question to guarantee that none of the variables is integrated of order greater than I (1). The ADF, PP, and KSSUR techniques were utilized in this investigation to determine whether the underlying variables were stationary. The findings of the stationary test are reported in Table 3a, Table 3b, which can be found below. A non-stationary level value for the underlying variables is found by performing a unit root test; nevertheless, a stationery level value is found by performing an ARDL technique. ARDL constraints cointegration requires that the order of integration of all research variables be stationary at I (0) or I (1) to produce reliable results (1). To demonstrate that the data are stationary, the presence of a unit root is checked for in the following variables: CO2 emissions (CO2), GDP per capita, GDP per capita square, Population (Pop), Gas and Oil are tested for the presence of a unit root to show that the data is stationary. Three unit-root tests, namely the Kapetanios, Shin and Snell (KSSUR), the Augmented Dickey-Fuller (ADF), and the Phillips-Perron (PP) tests, are used to determine whether unit roots exist, and the findings are shown in Table 3a. Each variable in an ARDL model must be either I (0) or I in order to pass the boundaries test (1) (Elahi and Rahman, 2021). According to contemporary scholarship, however, this is not the case. At the level, three tests show that the population and gaseous fuel consumption are both stationary, while CO2, GDP per capita, gross domestic product squared, and oil consumption have unit roots at the level and become stationary once the first difference has been observed. ARDL bounds cointegration is included to the analysis when the order of integration of all variables is confirmed.

Table 3a

Results from unit root tests.

Variable	KSSUR Test		PP Test		ADF Test
	Level	1st Diff.	Level	1st Diff.	Level	1st Diff.
L (CO2)	-0.055	-6.522∗∗∗	-1.431	-6.993∗∗∗	-1.431	-6.993∗∗∗
L (GDPPC)	2.200	-3.631∗∗∗	0.855	-7.733∗∗∗	1.125	-7.733∗∗∗
L (GDPPC2)	2.455	-3.380∗∗∗	1.814	-7.319∗∗∗	1.814	-7.319∗∗∗
L (Pop)	-5.638∗∗∗		-1.685	-8.709∗∗∗	-4.465	-8.709∗∗∗
L (Oil)	-0.052	-6.512∗∗∗	-0.711	-6.919∗∗∗	-0.711	-6.919∗∗∗
L (Gas)	-3.684∗∗∗		-3.955∗∗∗		-3.955∗∗∗

Source: Authors Computations.

(a) The AIC and SIC have determined the ideal lag duration. (b) In all unit roots testing, an intercept and a trend term are included as well. (c) There is a ∗∗∗, ∗∗, and ∗ signifying statistical significance at 1%, 5%, and 10% significance levels for the computed score.

Table 3b

Unit root tests with structural breaks.

Zivot-Andrews test
Variables	ZA statistic	Break	1%	5%	10%	Decision
L(CO2)	-8.145∗∗∗	2011	-4.95	-4.45	-4.14	Break Exist
L(GDPPC)	-4.362∗	1994	-4.95	-4.45	-4.14
L(GDPPC2)	-6.524∗∗∗	1995	-4.95	-4.45	-4.14
L(Pop)	-4.4882∗∗	2010	-4.95	-4.45	-4.14
L(Oil)	-7.364∗∗∗	1992	-4.95	-4.45	-4.14
L(Gas)	-6.254∗∗∗	1993	-4.95	-4.45	-4.14

In Table 3b, the Zivot and Andrews (1992 were applied, henceforth ZA) test to allow for an endogenous structural break in order to account for a probable shift in regime in the unit root testing. As shown in Table 3b, all variables exhibit I (1) processes at the 5% significant level when run as a univariate ZA unit root test. In the series of L (CO2), L (GDPPC), L (Pop), L (Oil), and L (Gas), there are structural breaks in 2011, 1994, 1995, 2010, 1992, and 1993. The early 1990s are the most likely years for the majority of the breaks to have occurred. Globalization's effect on Bangladesh's external economy and political instability in Bangladesh may be to blame. Another possible explanation for the 2010 structural split is that big changes in Bangladesh's trade policy occurred in the year 2000. But the year 2008 will always be associated with the global financial crisis, which has ramifications for the Bangladeshi economy as well as those of other Asian countries.

On the other hand, in Table 3c, Lumsdaine and Papell (1997) or the LP test, expand the models proposed by Zivot and Andrews (1992) to take into account the probability of two endogenous structural fractures. The results imply that a unit root cannot be rejected for all L (GDPPC2) and L (OIL) variables. Because their respective t-statistics are statistically significant, the L (CO2), L (GDPPC), L (Pop), and L (Gas) variables are all considered to be stationary. When there are two or more structural breaks in the data, the findings of unit root tests that only use one break or no break will not be reliable. It is crucial to run higher-order tests that take into consideration two structural breaks to compare the results. To attain this goal, the LP unit root test that incorporates two structural breaks is carried out (Table 3c). The results of the unit root tests with structural breaks conducted by ZA (1992) and LP (1997) are similar, and few breaks are contradictory. From both ZA and LP unit root tests, the early 1990s and early 2010s are the most likely years for the majority of the breaks to have occurred. Globalization's effect on Bangladesh's external economy and political instability in Bangladesh may be to blame for the 90s breaks. Another possible explanation for the 2010 structural split is that big changes in Bangladesh's trade policy occurred in the year 2000. But the year 2008 will always be associated with the global financial crisis, which has ramifications for the Bangladeshi economy as well as those of other Asian countries. To be clear, all of the variables are movable in their initial different forms. Thus, we may conclude that I (1) is fully integrated into all of the level series of the variables. To conclude, the outcomes of both traditional and structurally broken unit root tests are compared. If structural breakdowns in long-run time series are disregarded, the ADF, KSSUR, and PP unit root tests (Table 3a) are biased in favor of accepting stationarity. Both the ZA and LP tests, which permit a single or double structural shift in the series, produce identical findings: that all variables are stable. This is the case regardless of which test is used.

Table 3c

Unit root tests with two structural breaks.

Lumsdaine and Papell (1997) double structural breaks test
Variables	t-statistic for α	Break 1	Break 2	P-Value
L (CO2)	-5.18∗	1993	2011	0.0912
L (GDPPC)	-4.51∗∗	1994	2009	0.0325
L (GDPPC2)	-4.67	1995	2014	0.3125
L (Pop)	-5.73∗∗	2010	2013	0.0512
L (Oil)	-6.44	1992	2012	0.3625
L (Gas)	-5.27∗∗	2002	2014	0.0245

There is a ∗∗∗, ∗∗, and ∗ signifying statistical significance at 1%, 5%, and 10% significance levels for the computed score.

In Table 4, the F-statistic is significant even at 1%, according to a close examination of the table. Furthermore, the F-value exceeds the critical upper constraint. We can conclude that our variables are in long-term equilibrium by looking at this graph. With this background in mind, we will now look at how GDP per capita, population, oil, and gas all impact to CO2 emissions in Bangladesh.

Table 4

ARDL bound test.

Test Statistic	Value	K
F- Statistic	10.518	5
Critical Value Bounds
Significance level	I (0)	I (1)
10%	3.68	4.59
5%	5.63	5.28
1%	6.14	6.28

Source: Authors Computations.

Using Table 5, we can see that the coefficients of ΔGDPPC and ΔGDPPC2 are both negative and positive, like a normal U-shaped curve, and both are statistically significant, which did not support the EKC hypothesis and GDP impacts on emissions. This study also shows that the EKC does not clasp in the short term because the GDP squared is positive, which is consistent with the findings of Saboori and Sulaiman (2013). Energy usage has an optimistic impact on emissions when looking at the short-term implications of climate change. The following is the outcome: an increase of one percent in the use of oil in Bangladesh will result in an increase of 1.75 percent in emissions. In Bangladesh, on the other hand, a one percent increase in gas consumption results in a 0.029 percent rise in emissions. These energy consumption figures are congruent with those obtained by Khan et al. (2019), Dogan and Seker (2016), Jebli et al. (2016) and Rahman et al. (2020). In addition, in short-run we found that increase of population has negative and impact on CO2 emissions. The estimated result exposed that CO2 emissions decreased by 3.22 percent if the population increased by 1 percent. Increasing population growth rate consecutively is not responsible for such alarming emissions in Bangladesh and this inference is reliable with Brazil for liner method which was conducted by Alam et al. (2016), Wang et al. (2017) for china, Hussain and Rehman (2021) for Pakistan.

Table 5

Short-run coefficient estimates.

Lag order	0	1	2	3	4	5
Variables
Δ L (CO2)		-731∗∗	-.034	-.445
ΔL (GDPPC)	-.456	-.654	-.400
ΔL (GDPPC2)	.168	-.054	-.032	-.123∗∗	-.063	-.016
ΔL (Pop)	-3.22∗	3.72	7.45	-10.35	9.25	2.63
ΔL (Oil)	1.759∗∗	1.726∗∗∗	.059	1.097
ΔL (Gas)	.029∗∗	.471	.200	-.417	-.754

Note: L(CO2) is the determinant variable. The calculated ARDL's appropriate lag time is (4,3,6, 6,4, 0). For the selection of lag length, we use the AIC selection criterion. In parenthesis are the standard errors. ∗, ∗∗, and ∗∗∗ indicate that the null hypothesis has been rejected at the 1%, 5%, and 10% levels of confidence.

In Table 6, the findings show that the L (GDPPC) has a statistically positive influence on CO2 discharge in Bangladesh, while the L (GDPPC2) has a statistically negative and significant influence on CO2 discharge in Bangladesh, which is an important requirement for the presence of the EKC hypothesis. According to Bangladesh's Intended Nationally Determined Contribution (INDC) to UNFCCC on 25th September 2015, these estimated results support the 15 percent greenhouse gas (GHG) emissions reduction target within 2030 (Ministry of Environment, Forest and Climate Change of Bangladesh (2020)). However, in the long run, the burning of oil and natural gas has a largely positive and statistically significant impact on CO2 emissions, implying that the attention paid to CO2 emissions will be increased as an outcome of the use of energy as a result the economic growth and development increases with ensuring sustainability. Furthermore, the population growth in Bangladesh has an impact on CO2 emissions but is not significant to this point. This estimation is validated by the analysis of Alam et al. (2016) for India, Indonesia, China, and Brazil in the quadratic model estimation, Hasan and Chongbo (2020) for Bangladesh.

Table 6

Long-run coefficient estimates.

	Constant	L (GDPPC)	L (GDPPC2)	L (Pop)	L (Oil)	L (Gas)
Coefficient	-1.588∗∗∗	2.433∗	-.197∗∗	-1.49	3.26∗∗∗	1.64∗∗

From Table 7, Ct-1 indicates the speed of adjustment and its coefficient shows how a shock or disequilibrium merges to long-run equilibrium. For example, the coefficient of Ct-1 for ARDL is -0.48∗∗∗, which indicates that roughly 48 percent of the disequilibrium in CO2 emission caused by the shock the previous year is restored to the long run in the current year.

Table 7

Dynamic adjustment.

LnCO2 L1.	Coef.	Std. Err.	T	P>|t|	[95% Conf. Interval]
	-.483∗∗∗	.393	-2.78	0.013	-1.49	.527

However, several diagnostic tests on estimated ARDL model were applied to ensure that our estimated model is stable and resilient and that it has produced unbiased estimates. The diagnostic test results of the ARDL models are presented in Table 8, which consist of Serial correlation, Homeskedaticity, Normality test that was testified by (i.e., Breusch-Pagan/Cook-Weisberg test for heteroskedasticity), ARCH, and Ramsey RESET test. According to the Breusch–Godfrey Lagrange Multiplier (LM) test, the econometric model for L (CO2) does not demonstrate serial autocorrelation. Here, we have seen that no serial correlation, no autoregressive conditional heteroscedasticity (ARCH) exists in the estimated model. Moreover, to check the heteroscedasticity of the model we conducted Breusch-Pagan/Cook-Weisberg test for heteroskedasticity and found that the calculated value does not exceed the critical value. Overall, White's test and Breusch–chi2 Pagan's statistics indicate that the model does not suffer from heteroscedasticity. On the other hand, the J B Normality test confirms that the model is normally distributed. However, the Ramsey RESET test is called the regression specification error test to check whether the model is not correctly specified or not. If the calculated F value is significant, we can conclude that model is not correctly specified. Here, we have seen that the calculated F value is not significant. Ramsey RESET suggests that the nonlinear amalgamations of independent variables are useful in characterizing the dependent variable, based on its results and the model has no omitted variables. Therefore, we terminate that the ARDL model is correctly specified (Gujarati et al., 2012).

Table 8

The ARDL Model's diagnostic test results.

Statistics test	Statistics	P-Value	Decision
Breusch-Godfrey LM test for autocorrelation	Chi2 .345	0.5218	H0: no serial correlation. We cannot reject the null.
			So there is no Serial correlation.
White test for homoscedasticity	Chi2 27.16	0.4009	H0: homoskedasticity
			H1: unrestricted heteroskedasticity
J B normality test	Chi2 2.58	0.4452	Normal
LM test for ARCH	Chi2 0.02	0.9638	H0: no ARCH affects vs. H1: ARCH(p) disturbance
			We are unable to reject null. As a result, there are no ARCH consequences.
			There is no heteroskedasticity.
Ramsey RESET test	F (3, 36) = 0.37	0.7345	Ho: model has no omitted variables.
			So, no misspecification
Adjusted R2	0.652

The finding of the Toda-Yamoto Granger causality test is provided in Table 9 of this paper. Results reveal that unidirectional causality exits from L (GDPPC) to L (CO2) as well as from L (Gas) to L (CO2) in the long run. However, there is no causal relationship exists between L (CO2) and L (Pop) in Bangladesh. Results of Granger causality test demonstrate that unidirectional causality exists from L (CO2) to L (Oil) which is similar with Saboori et al. (2017), Rahman and Majumder (2022). Finally, the stability test was implied to determine the goodness of fit of the envisaged parameters. For this reason, the study performs the cumulative sum (CUSUM) and the cumulative sum of the squares (CUSUMSQ) test which are provided in Figure 1. It is obvious from both figures that plot for CUSUM and CUSUMSQ test lie within the borders of 5 percent level of significance and it has assured that all parameters are stable.

Table 9

Toda-yamamoto causality test results.

Null Hypothesis	Wald Statistic	P-value	Decision
L (CO2) ≠> L (GDPPC)	.61125	0.264	L (CO2) ⇐ L (GDPPC)
L (GDPPC) ≠> L (CO2)	12.9547∗∗	0.0462
L (CO2) ≠> L (Pop)	2.24	0.235	L (CO2) ≠ L (Pop)
L (Pop) ≠> L (CO2)	1.4544	0.124
L (CO2) ≠> L (Oil)	11.548∗∗∗	0.001	L (CO2) ⇒ L (Oil)
L (Oil) ≠> L (CO2)	.26297	0.521
L (CO2) ≠> L (Gas)	.064	0.247	L (CO2) ⇐ L (Gas)
L (Gas) ≠> L (CO2)	14.801∗∗∗	0.002

Note: (i)∗∗∗∗, ∗∗, and ∗ are used to signify the statistical significant levels of the derived coefficients at one percent (1%), five percent (5%), and ten percent (10%) levels of significance for the coefficients.(ii) The SIC has determined the optimal lag period. (iii) It is possible to acquire information about the direction of causation among the remaining variables if the request is made. (iv) Notation ≠> on the table indicates that there is a hypothesis of no Granger causality association between two variables given in the table.

Figure 1

Cusum and cusum of square test. Source: Author's estimation.

Conclusion

The primary aim of this research was to evaluate if Bangladesh had an EKC between 1971 and 2020. The ARDL framework uses economic and financial development as environmental indicators to test the EKC hypothesis. Empirical analysis has established a long-term association between carbon emanation and economic growth, energy utilization and population. The presence of the EKC hypothesis is established by the presence of a positive sign associated with economic growth and a negative sign associated with its quadratic term. According to Granger's causality, there is a one-way association between economic development and carbon emissions, gas consumption and CO2 emissions, and CO2 emissions and oil consumption. We discovered that the model's predicted coefficients are stable using CUSUM and CUSUMSQ. The CO2 emissions from Bangladesh's population and energy use were also found to have substantial contributions in this empirical analysis. These findings make it quite evident that Bangladesh's policymakers should create regulations that encourage green energy investments to be financially, economically, and environmentally sustainable. Substantial evidence supports an optimistic view of the EKC link between CO2 emissions and economic development. This evidence clearly demonstrates that the national strategy for climate change management policy execution does not affect economic growth at the current level of GDP. Because the EKC relationship between CO2 and financial development is not precise, extra factors must be taken into account when establishing energy-efficient measures, notwithstanding this study's crucial conclusions. However, this study recommended that the energy sector should be managed through the consideration of the supply chain. Alternative energy purchases such as technological innovation and solar panels could be effective to enhance renewable energy production and distribution. The sectors which are most responsible for consuming oil and gas, such as commercial vehicles, industries, electricity production houses, and others, should be identified and a limit of emissions would be an efficient solution. Clean and affordable clean energy would be another effective solution to reduce emissions in Bangladesh. This study is able to contribute to academic as well as scientific and social science fields in terms of economic growth and environmental degradation. The key limitation of this study is that we have used some of the selected variables, but in the future, studying Bangladesh's urbanization, energy security, and other environmental variables could prove useful. At the same time, regional comparisons such as N-11, South Asian countries, MENA countries, BRICS, the EU, and others would provide a great opportunity to expand this study.

Declarations

Author contribution statement

Liton Chandra Voumik: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Wrote the paper.

Md. Hasanur Rahman: Conceived and designed the experiments; Analyzed and interpreted the data; Wrote the paper.

Md. Shaddam Hossain: Performed the experiments; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Funding statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability statement

No data was used for the research described in the article.

Declaration of interest's statement

The authors declare no conflict of interest.

Additional information

No additional information is available for this paper.