“Economic growth and environmental degradation paradox in ASEAN: A simultaneous equation model with dynamic panel data approach”

Economic variables are dynamic in nature. This paper uses a simultaneous equation model to assess the complexity of the link between economic expansion and environmental deterioration in ASEAN. The study examines how CO2 emissions, economic growth, public health initiatives, and control factors interact using dynamic panel data from 2011 to 2020. The population, the amount of forested land, the use of renewable energy, foreign investment, the inflation rate, the total amount of foreign exchange reserves, and government health policies are just a few examples. In order to provide a reliable and accurate assessment of the long-term relationship, this study employs the generalized approach of the Arellano-Bond moment method. The econometric technique deals with the issues of nonstationary, endogeneity, cross-error correlation, and heteroscedasticity. Additionally, the two stage least square (2SLS) method was used to assess the results’ robustness. According to the statistical results, there is a causal link between CO2 emissions and economic growth, and between CO2 emissions and energy consumption. Furthermore, according to the data, ASEAN CO2 emissions showed a monotonically growing relationship during the sample period. Policymakers may use these findings since they can aid in implementing economic measures to promote sustainable and ecologically friendly development.


INTRODUCTION
Environmental damage has become a global issue and a common concern for people worldwide.Industrial growth has contributed to widespread environmental damage, especially in recent decades, and has affected the world's health, ecology, and climate balance (Radmehr et al., 2021).With the rapid expansion of industry, there has been exploitation and depletion of the earth's minerals and resources, as well as environmental degradation in the form of an increase in critical land, water pollution, and air pollution.The expansion of essential land has had unfavorable effects, such as flooding throughout the wet season and drought throughout the dry season.Ten ASEAN nations were compelled to reach a regional agreement on smog pollution that crosses international borders due to the severe effects of air pollution (Mughal et al., 2021) from smog and forest fires (Nazeer & Furuoka, 2017;Thanh et al., 2019).Energy use has both immediate and long-term effects on environmental quality in various ASEAN nations (Haruna & Mahmood, 2018).
Environmental degradation increased due to the significant increase in the economic growth of numerous ASEAN nations (Haruna & Mahmood, 2018;Hu et al., 2021;Mughal et al., 2021;Thanh et al., 2019).Due to this phenomenon, numerous studies have provided empirical evidence that economic growth will affect environmental quality changes during the early stages of development until a specific limit is reached.After this point, the condition will result in improved environmental conditions.Using a yearly data set covering the years 1980-2014, Bakhsh et al. (2017) identified the factors that influence foreign direct investment (FDI) inflows into Pakistan as well as their direct effects on the environment's degradation and the growth of the economy.CO 2 emissions in the atmosphere are another factor that can be used to gauge the severity of air pollution.Rizk and Slimane (2018) analyzed the association between poverty and CO 2 emissions from 146 nations between 1996 and 2014 as a kind of environmental deterioration.The major finding is that poverty and CO 2 emissions have a nonlinear relationship that might cause poverty to increase and the environment to degrade.Therefore, the fundamental policy advice is that all nations should strengthen their institutional foundation to reduce poverty and environmental damage (Abdouli & Hammami, 2020).
In most nations worldwide, environmental deterioration is mainly caused by economic expansion and FDI (Bakhsh et al., 2017;Ren et al., 2021).Through technological transfer, higher productivity, and the introduction of new managerial techniques and procedures, the flow of foreign investment serves as a direct source of capital to promote economic growth.Additionally, FDI inflows aid in the financial development of the investing nation.This suggests that FDI increases the amount of money the financial system has access to.As a result, these funds support economic growth as well as the development of financial markets.It is also claimed that international businesses can use banking services to get loans, overdraft facilities, or to pay suppliers of semi-finished items.On the other hand, higher FDI inflows and economic expansion result in a reduction in the environment's quality.
It is impossible to ignore environmental degradation, including water and air pollution, which can jeopardize the viability of development.Engineering and compositional impacts show that more economic expansion results in higher pollution emissions (Bakhsh et al., 2017).The scale effect demonstrates that while pollution has a detrimental impact on growth, the stock of labor and capital benefits Pakistan's economy.Economic expansion and FDI have a favorable and considerable impact on capital stock in terms of the effect of capital accumulation.Even if economic expansion increases pollution, economic growth declines when pollution levels go beyond a certain point.As a result, solutions must be found to combat more severe air pollution.However, the majority of research is based on cross-sectional studies, and very few studies have used the panel data simultaneous equation technique (Ali et  A simultaneous equation system has the characteristic that it consists of several equations (Baltagi & Liu, 2009).In addition to mathematical and phenomenal, there is a relationship between these equations (Rose et al., 2020).The model in this simultaneous equation system has endogenous and explanatory variables in each equation, unlike the single equation system.Explanatory variables in one equation may also be endogenous variables in another.As a result, these variables can be correlated with other explanatory variables.In this case, the estimation of ordinary least squares (OLS) cannot be used because OLS in a simultaneous equation system will produce biased and inconsistent parameter estimates (Gujarati, 2022).One alternative for estimation is the two-stage least square (2SLS) method.This method will produce one estimator for one parameter and a standard error for each estimator (Gujarati, 2022).Several previous studies have shown that it is possible to use a simultaneous equation model in environmental economics because the variables tend to have a simultaneous relationship (Omri, 2013).Based on the phenomenon, there is a relationship between economic variables, so it cannot be modeled

Theoretical literature review
A simultaneous equation model has multiple equations connected.As a result, a variable in a simultaneous equation can play the dual roles of both independent and dependent variables.This is possible in the simultaneous equation model.There are endogenous variables and predetermined variables in a simultaneous equation model.The predetermine variable is a variable whose value can be determined in advance but is not directly decided by the equation.In contrast, the endogenous variable is the dependent variable determined in the simultaneous equation system (non-stochastic).Exogenous variables and endogenous lag variables are the two categories into which predetermined variables fall (Gujarati, 2022).
In the simultaneous equation model, there are two kinds of equations: structural and reduced form.The structural equation is the original equation that describes the behavior of the relationship between the existing variables.In contrast, the reduced equation is an equation with endogenous variables only influenced by the predetermined variable and the error component.
In panel data regression, the dependent variable is also affected by the dependent variable from the preceding period (lag 1), so the model is said to be a dynamic panel data model.The dynamic panel data regression model in economic research is more suitable for finding the relationship between economic variables.This dynamic panel data model can be seen from the lag of the dependent variable between the independent variables.The dynamic panel data model can be written as (Gujarati, 2022): In the early 1980s, the theoretical links between governmental policy and economic growth (as measured by FDI and GDP proxies) were examined.To encourage investment in cutting-edge technology, a structure for financial development is necessary yet insufficient.Furthermore, according to some studies, FDI inflows and financial developments are tightly associated.This implies that FDI can help countries with more developed financial markets benefit more from economic growth promoting FDI.

Analyzing empirical literature and formulating hypotheses
Economic growth is proxied by the variables of GDP (Mughal et  without carrying out further studies of changes in the large-scale transformation of the economic system so fast (Rizk & Slimane, 2018).
The impact of economic growth on the deterioration of the natural environment has been the subject of many studies.For example, according to Bakhsh et al. (2017), who studied the impact of economic growth on carbon dioxide emissions, waste, and foreign investment in Pakistan, increased economic growth is likely to increase the amount of emissions that contribute to pollution.
The impact of pollution has a detrimental effect on the country's environmental quality.In contrast, the investment and labor of foreign nationals have a positive impact on the expansion of Pakistan's economy.This also occurred in several other countries that are still developing ( H1: There are contributions of CO 2 emissions, government policies in the health sector, FDI, consumption of fossil fuels, population, inflation rates, and total foreign exchange reserves to economic growth in ASEAN countries. Over the past two decades, a number of empirical studies have examined the connection between FDI inflows, economic growth, the environment, and governmental policy.A similar quadratic association between income and CO 2 emissions is demonstrated by Thanh et al. (2019).This suggests that China has an inverted U curve.In line with that, Garza-Rodriguez (2018) tested the EKC hypothesis in Malaysia from 1980 to 2009 using the auto-regressive distributed lag (ARDL) methodology.Using the fixed-effect model and the generalized methods of moment (GMM) methodology, Haruna and Mahmood (2018) discovered a unidirectional relationship between GDP and carbon dioxide emissions.
In addition, it was found that Indonesia is one of the countries in which there is a link in the shape of an inverted U between the governmental policies and the amount of carbon emission (Nazeer & Furuoka, 2017).More specifically, they draw attention to the fact that as the financial sector matures, government initiatives initially result in a drop in CO 2 emissions.

Econometric model
This study model uses a production function approach to explain the impact of CO 2 emissions, fossil fuel energy consumption (FFEC), total foreign exchange reserves (TR), inflation (INF), the government budget for health (DGGH), foreign investment (FDI), and population (POP) to economic growth (GDP) in the form of a function model: .
In the case of economics, variables that have a twoway relationship are often encountered.This twoway relationship that influences each other can be summarized in a system of simultaneous equations.Almost all approaches in macroeconomics have a simultaneous nature.Identifying a simultaneous equation with the order conditions provides information that an equation is exactly identified or overidentified.Based on the order conditions, the model is said to be identified if the equations ( 2) and ( 3) meet the requirements if the model shows then it is called precisely identified, and if Kk < m -1, then the equation is unidentified.
Due to the correlation between endogenous variables and disturbances in simultaneous equations, the OLS estimator will result in a biased and unstable estimate.Therefore, an alternative estimation method is needed, which is called the 2SLS method.The 2SLS method is the application of OLS in two stages.First, a simultaneous test (Hausman test) is needed to test whether explanatory endogenous variables correlate with disturbance.In data processing using panel data, there are several stages of testing that aim to determine the best model to be used in a panel data study (Gujarati, 2010).The three models in panel data regression processing are the common effect model, the fixed effect model, and the random effect model.In addition, there are three stages of testing the model selection on the panel data: the Chow test, the Hausman test, and the LM test.The Chow test is useful for testing the model selection between the common effect model and the fixed effect model.The Hausman test is used to test the model selection between the fixed effect model and the random effect model.At the same time, the LM test is used to test the model selection between the random effect model and the common effect model.After knowing the best model to be used in the study, hypothesis testing, such as the coefficient of determination and partial and simultaneous significance tests, will be carried out.
Two basic specifications of the panel data simultaneous equation model are as follows (Figure 1).
The classical assumption test will be carried out in three stages, namely the classical assumption test of heteroscedasticity, autocorrelation, and multicollinearity.These three stages must be met so that the data used is tested for validity.Heteroscedasticity resulted in unbiased coefficient values, but the variance of the estimated regression coefficients was no longer minimal.The Harvey test can test the existence of heteroscedasticity.To prove the presence of heteroscedasticity with the white test, it can be done by comparing the value of n (amount of data) and R-square of the unadjusted R-square value in the auxiliary model.Autocorrelation shows the regression residual that is not independent from one observation to another.Autocorrelation can arise from an inappropriate specification of the relationship between endogenous variables and explanatory variables.The presence of autocorrelation can be detected through the Durbin-Watson Test.Multicollinearity arises when the independent variables are correlated with each other.Multicollinearity is the relationship between independent variables, which is a condition of a strong correlation between independent variables or vice versa.The existence of multicollinearity can be determined through a correlation matrix or regression between independent variables in the equation model.

Data
The data used are annual data for GDP per capita (constant 2015 USD), CO 2 gas emissions (metric tons per capita), inflation, total reserves (including gold, current USD), fossil fuel energy consumption (% of total), renewable energy consumption (% of total final energy consumption), population (total), foreign direct investment, net inflows (BoP, current USD), forest area (sq.km), and domestic general government health expenditure (% of general government expenditure).The data were collected for 2011-2020, sourced from the World Bank's World Development Indicators.To estimate the model, the paper divided the variable by the population to get the variable in per capita terms.The study covered 10 ASEAN countries that were selected based on data availability.These countries are Brunei Darussalam, Cambodia, Indonesia, Malaysia, Myanmar, Laos, Philippines, Singapore, Vietnam, and Thailand.Descriptive statistics of various variables for individuals and panels are presented in Table 1.

Model identification
Before estimating the parameters, it is necessary to identify the problem first to see if the equation model formed can be used using the two-stage least square approach.In identifying these problems, the paper uses order condition tests.The test results with order conditions are described in Table 2. Based on the results in Table 2, equation ( 4) and equation ( 5) models are over-identified equations, which means that the two models are correctly identified so that the two-stage least square approach can be used.

Panel unit root test
The panel unit root test must first be run to determine if the pertinent variables in a panel data analysis are stationary.Levin et al. (2002) used the LLC approach for the panel unit root test.In the two-unit root tests mentioned above, the null hypothesis is that there is a unit root (i.e., the variable is not stationary), and the alternative hypothesis is that the series has no unit root (that is, the variable is stationary).Table 3 displays the outcomes of the panel unit root test for the variable level.All variables at the level can be demonstrated to be statistically significant in the LLC test, proving that they are all integrated and stationary.

Hausman test
Hausman test aims to prove the existence of a simultaneous relationship between the two independent equations (Gujarati, 2022).The hypothesis for the Hausman specification test is H0, which means a simultaneous relationship between the GDP equation and CO 2 emissions, and H1, which means that there is no simultaneous relationship between the GDP equation and CO 2 emissions.The test statistic used is the t-test with the rejection area; if the p-value of the residual variable is less than 0.05, then H0 is rejected.Hausman test results are shown in Table 4.

Dependent variables Residual variables p-value
Table 4 shows that the p-value for the resCO2 and resGDP variables is greater than 0.05, indicatingthat the GDP equation has a simultaneous relationship with the CO2 equation.Therefore hypothesis 3 (H3) is supported.

Model estimation using the generalized method of Arellano-Bond moment
The equation estimation uses the GMM estimation that consists of equations ( 4) and (5).Based on the calculation results, the panel data simultaneous equation models for the GDP and CO 2 equations are obtained, as shown in Tables 5 and 6.It is indicted in Table 5 that the free variables of CO2 emissions, DGHH, log(FDI), FFEC, log(Pop), inflation and log(TR) have significance influence on economic growth.Thus, the first hypothesis (H1) is supported.
from Equation (5) and Table 5, the following simultaneous equation is formulated:  Based on study results presented in Table 6, it is concluded that all independent variables have a significant effect on CO2 emissions, therefore the second hypothesis (H2) is supported.
Referring to Equation ( 4) and Table 6, the simultaneous equation is obtained:

Significance of the parameters simultaneously
If there is a relationship between the variables in an equation model, it will be found via parameter significance testing.The parameter significance test also examines whether the independent variable impacts the dependent variable.Table 7 shows the results of the simultaneous parameter significance test.Based on Table 7, it can be seen that the p-value of the two equations, namely GDP and CO 2 , are less than 0.05, thus making the independent variables of the two models have a significant effect on the dependent variable (H0 is rejected).

Significance of the parameters partially
The paper performs the partial parameter significance test to determine whether the independent variable only has a partial impact on the dependent variable.Tables 5 and 6 show that all independent variables have p-values that are less than 0.05, so it is concluded that all independent variables have a significant effect on the equation ( 3) and ( 4) models.

Classic assumption test and normality test
The assumptions used in this study are the normality test, Arellano-Bond (AB) test, and Sargan test.The objective of the normality test in a regression analysis is to establish whether the model's independent and dependent variables have a normal distribution.The Jarque-Bera test is a statistical method that can determine whether the model is normally distributed (Gujarati, 2022).From Table 8, information is obtained that the residuals of the two models are normally distributed.

Arellano-Bond (AB) test
The Arellano-Bond test aims to test the consistency of the model.In the Arellano-Bond test, there are two tests with different functions, namely the ab(1) test, which serves to determine the influence of individual effects between observations, and the ab(2) test, which functions to determine whether or not there is a correlation between the first difference error in the i-th observation (Gujarati, 2022).
Based on Table 9, the p-value for ab(1) both models is 0.000, so the decision rejects H0, and it can be concluded that the GDP and CO 2 equation model does not have individual effects between variables.For the ab(2) test results, the P-values obtained are 0.774 and 0.9158, so H0 is not rejected.Thus, it is concluded that there is no lag effect of the dependent variable on the first difference error in the GDP and CO 2 models.Therefore, because the ab(1) and ab (2) tests are met, it can be concluded that the GDP and CO 2 equation models are consistent.

Sargan test
The Sargan test is used to evaluate the validity of using instrument variables that are more numerous than the estimated parameters (over-identified) (Gujarati, 2022).Table 10 shows the results of the Sargan test of the two-equation models.

DISCUSSION
Based on Table 5, it is shown that the coefficient of lag indicator of economic growth (GDPt-1) has a positive and statistically significant effect.Thus, every country in the ASEAN region can take appropriate macroeconomic policies with a backward look at achieving high and sustainable economic growth.For the panel results, FDI inflows per capita have a positive and significant effect on GDP per capita.Every 1% increase in FDI will increase GDP 0.219 times at the time of ceteris paribus.This shows that economic growth is elastic to FDI inflows.This implies that the technological changes brought about by FDI inflows promote economic development in the long run.This indicates that the influence of foreign investment has not been able to encourage economic growth for all countries in the ASEAN region.Although FDI is still only focused on a few countries, it positive- The consumer price index is an indicator used to describe price movement.Changes in public consumption patterns in the long term trigger an increase in aggregate demand to encourage an increase in the inflation rate.In the long term, an increase in a country's economic growth reflects people's income and consumption.Business actors followed up the increase in demand by increasing their production output.With the addition of output, the costs incurred for the production process become greater, causing an increase in the selling price of the product.If, for a relatively long time, most of the traders do the same thing, then the increase in the prices of consumer goods, in general, can encourage an increase in inflation.
The role of government spending in the health sector has a positive effect on boosting economic growth.Government spending on health has a positive effect on attracting foreign investment.This will indirectly increase economic growth so that welfare will increase.Considering the benefits for investors in addition to quality labor, the region will become a significant consumer due to the increasing welfare.An increase in government spending (DGGH) by one percent causes an increase in the economic growth of 0.01775 percent at the time of ceteris paribus.In the long term, an increase in government spending can increase foreign investment by 1.203 percent.Nazeer and Furuoka (2017) also showed that government spending has a positive and statistically significant effect on increasing economic growth, concluding that simultaneously government spending has a positive effect on investment development in the ASEAN region.
The empirical findings for the global panel are shown in Table 6, demonstrating that energy consumption significantly reduces CO 2 emission levels at the 1% level.This implies that increased energy use may result in increased CO 2 emissions.Economic growth is significantly and favorably influenced by panel estimates.A robust energy policy is required to promote sustained economic growth because energy is a critical component.These findings concur with those of Thanh et al. (2019).In terms of the pollutant variable, it was discovered that CO 2 emissions have a sizable impact on worldwide panel economic growth.This demonstrates that a 1% increase in CO 2 emissions causes a 0.005423% rise in economic growth.almost all nations, population increase is statistically significant and negatively influences economic growth at a 1% level.Population growth raises the level of government policies, including health, fuel, and other subsidies.Additionally, fuel consumption has a 1% favorable impact on CO 2 emissions.If there is a population of 1%, in the short term, it will cause a slowdown in the eco-nomic growth of 0.516775%.On the other hand, an increase in population by 1% will increase CO 2 emissions by 4.846382%.

CONCLUSION AND IMPLICATION
The primary findings suggest a bidirectional causal relationship between rising CO2 levels and a greater spectrum of economic activities.The concept is supported by both economic expansion and carbon dioxide emissions.The empirical evidence demonstrates a unidirectional causal connection between economic expansion and CO2 emissions.Therefore, nations must adopt rules to protect their citizens from carbon pollution.In ASEAN nations, economic growth and CO2 emissions have a two-way causal link.This means that economic growth is causing environmental damage.Currently, high economic growth contributes to environmental degradation, while decreasing economic growth produces unemployment, placing a significant strain on the economies of ASEAN member states.In order to increase their efforts to combat global warming, governments should reduce CO2 emissions without jeopardizing short-term or long-term growth.It will promote ecologically friendly, long-term economic growth in practice.In contrast, pollutant emissions have a negative impact on economic growth, indicating that environmental degradation is a causal factor in economic growth.In addition, as a result of its effects on human health, persistent environmental deterioration can have a negative externality on the economy and eventually reduce productivity.This is consistent with the Environmental Kuznets Curve theory, which states that environmental degradation results in a decreasing slope of economic growth.

Table 1 .
Description of research variable data per ASEAN countries

Table 2 .
Problem identification with order conditions

Table 3 .
Panel unit root test results for research variables

Table 5 .
Estimation results of the regression coefficient with the dependent variable log(GDP) VariablesCoefficient Std.error t-statistic Prob.

Table 6 .
Estimation results of the regression coefficient with the dependent variable CO 2 emissions

Table 7 .
Simultaneous significance test results

Table 8 .
Simultaneous significance test results

Table 10 .
Sargan test resultsBased on Table10, the results show that both models have a p-value of more than 0.05.The GDP equation model has a p-value of 0.2901.In contrast, the CO 2 equation model has a p-value of 0.2927, so the decision is to fail to reject H0.It can be concluded that the two models have no problem with the validity of the instrument variables.