The Impact of Forest Wood Product Exports on Environmental Performance in Asia

Abstract

The pressure on governments has been increased to improve their environmental performance across the globe. To ensure sustainability, every country is now responsible for educating its citizens on its environmental policies for decreasing pollution and managing natural resources. Therefore, it is crucial to manage all elements that have a negative impact on a country’s environmental performance in order to ensure sustainability. This study’s main goal was to investigate how factors such as forest wood product exports, energy use, urbanization, and gross domestic product (GDP) per capita in Asia affect environmental performance. In order to investigate the empirical links, this study analyzed annual data for 31 Asian countries from 2001 to 2020. Various econometric methodologies were used, including the generalized method of movement (GMM) approach and the Hausman test for fixed and random effects. According to the results of these econometric methodologies, forest wood product exports, urbanization, energy consumption, and GDP are all significantly contributing to Asia’s increasing ${\mathrm{CO}}_2$ emissions and deteriorating environmental performance. Based on the findings of this article, the selected Asian countries should curtail forest products in their trade basket in order to enhance environmental performance in the region. Furthermore, the alternative use of forest products and their exports could be increased to meet requirements.

1. Introduction

Governments have been under increasing pressure to improve their environmental performance around the world. According to the 2030 Agenda for Sustainable Development, nations now have a duty to teach their citizens about their environmental policies for reducing pollution and managing natural resources in order to uphold the sustainability of their own countries. The Environmental Performance Index (EPI) measures a country’s efforts to combat environmental concerns. Analyzing the EPI’s behavior and affecting factors will, thus, give a solid foundation for sound policy-making. This can aid in determining the factors that influence environmental progress and in maximizing the rate of return on investment in environmental management in order to achieve long-term environmental sustainability [1].

In the environment, opposite effects occur when forests are burned, degraded, or cleared. Along with other greenhouse gases, a large amount of carbon dioxide (${\mathrm{CO}}_2$) is emitted. It is commonly believed that the release of such gases into the atmosphere changes the radioactive balance, which results in more absorption and the trapping of heat from the sun inside the earth’s atmosphere, resulting in global warming. Carbon dioxide, with an 85% contribution to greenhouse gases, is known to be the largest anthropogenic contributor into the atmosphere. According to Butler [2], the clearing and burning of tropical forests and peatlands causes 3.7 billion tons of ${\mathrm{CO}}_2$ emissions each year.

By 2050, the World Bank predicts that three regions—Latin America, sub-Saharan Africa, and Southeast Asia—will produce 143 million more climate migrants. In 2017, more people were forcibly displaced (68.5 million) than at any other time in human history. Although it is difficult to calculate, it is believed that between 22.54 and 245 million people were displaced due to "sudden-onset” meteorological catastrophes, such as flooding, forest fires following droughts, and stronger storms. The remaining two-thirds of displaced people are the outcome of several humanitarian crises. It is, therefore, becoming evident that climate change is causing what are known as slow-onset catastrophes, such as desertification, sea-level rise, ocean acidification, air pollution, altered rainfall patterns, and biodiversity loss. Despite the fact that climate change is rarely the only cause of migration, it is commonly acknowledged as a contributing and aggravating element in both migration and conflict [3].

People who are displaced as a result of environmental deterioration, such as deforestation, sea-level rise, growing deserts, and catastrophic weather occurrences, are known as "environmental refugees." According to a Red Cross study, environmental disasters are now displacing more people than conflict [2]. The manufacture of internationally traded goods accounts for about 30% of worldwide ${\mathrm{CO}}_2$ emissions [4]. As a result, while the increased consumption of carbon-intensive commodities in one nation may not result in increased emissions in that country, it will contribute to increased emissions in other countries that are carbon-intensive product suppliers. The majority of carbon-intensive trade movements go from developing to developed nations [5].

Even though it has been crucial to the exchange of goods and for the growth of economies, international trade also has unintended negative effects on the environment. A nation can lower its CO₂ emissions by importing carbon-intensive goods from the rest of the world. Developed countries continue to engage in this activity, which results in consumption-based carbon emissions in developing nations, despite the fact that it may have a negative impact on efforts to reduce global emissions. While neglecting the fact that one country can potentially raise carbon emissions through international trade, most strategies have concentrated on reducing CO₂ emissions where they are produced [6].

The trade in forestry causes huge amount of CO₂ emission per year. Between 2010 and 2014, 2.6 billion tons CO₂ per year was caused by deforestation to meet the demands of internal and external markets. This was considered to be 6.5% of the total global CO₂ emissions. The remaining 29% is concerned with the production of goods that are sold to satisfy demand in global markets, with 71% of production taking place to satisfy demand in domestic markets. Deforestation in other parts of the world is caused by numerous wealthy countries. High-income countries are the main importers of deforestation products, accounting for 40% of total deforestation. This means they are responsible for 12% of all deforestation on the planet. As a result, it is a proven truth that affluent countries cause deforestation in poorer (low-to-middle) income countries through importing forestry goods [7].

The distribution of emissions through international supply chains is presented in Figure 1 below. Regions where emissions are produced are represented on the left, while on the right are the regions or countries where these products are consumed. The paths between these boxes indicate where emissions are produced and being traded, measured as the annual average from 2011 to 2014. Latin America exports 23% of its total emissions, while the remainder are generated by products that are consumed within domestic markets. In the Asia–Pacific area, Malaysia and Indonesia export 44% of all emissions to overseas markets. Deforestation in Africa is mainly caused by local residents and markets, i.e., only 9% of the total emissions are exported.

There is no place on earth that is immune to the effects of environmental change. Asia is the world’s largest continent in terms of land and population, with the most carbon-emitting countries and the most polluted cities in the world. The intricate problems that Asia is dealing with include biodiversity loss, declining agricultural yields, the availability of clean and safe drinking water, and deforestation, all of which have a significant detrimental influence on the environment’s performance and welfare. The 2018 EPI report showed that Asian countries are the worst performers in the world based on their environmental performance. Although the economy of this region is expanding quickly, poor environmental governance, which has a very negative impact on environmental performance [1].

Six Asian countries, including China, India, and Japan, emit 60% of the global ${\mathrm{CO}}_2$ to the atmosphere. According to a research by the World Meteorological Organization, weather, climate, and water risks were to blame for 50% of all disasters, 45% of all reported fatalities, and 74% of all reported economic losses between 1970 and 2019. These changes are often linked to natural causes and associated with direct and indirect human activities [8]. Abnormalities in temperature, changing patterns of precipitation, frequent river floods, rises in sea-level, melting ice caps, and other extreme weather conditions are implied and considered due to changes in climate, and all of these have impacts on what and how the world trades. It is shown by recent estimates that out of 32 billion tons, 8 billion tons of ${\mathrm{CO}}_2$ emissions are caused by the production and distribution of goods and services that are traded around the world [9].

Assessing how exports of forest wood products affect environmental performance in Asia will be a unique and uncommon case. Over time, this article will be remembered as useful in the future when establishing long-term plans for environmental sustainability and evaluating environmental performance. Furthermore, to improve environmental performance in future, this article will not only be helpful for governments to formulate effective environmental strategies and policies, but it will also be helpful to measure the outcomes of previous efforts. The research gap will be filled by this study in two ways. First, the outcomes of the research will attract the interest and attention of the general public. Second, the government may consider the study’s implications in their agenda for environmental sustainability. Today’s world is experiencing a new era of data-based policy making; therefore, it helps to identify problems and formulate solutions.

The rest of the paper is structured as follows. The second section discusses the review of the literature. The third section presents the study’s methodology, findings, and discussion. The conclusions and recommendations are presented in the fourth and fifth sections, respectively.

2. Literature Review

Environmental degradation is the result of many factors, with continuous deforestation being a major factor that has been given the least attention around the world, specifically in Asian countries. Several analyses and studies have been conducted by researchers regarding environmental deterioration and further mitigating the overall impact on human life. In this section, we review the earlier empirical research on the causes of environmental degradation.

Recently, Alola, Bekun, and Sarkodie [10] concluded that the major factor that devalue the environmental layers are trade between two uneven economic entities, continuous industrialization, and the production of finished goods from raw materials imported from the developing world. Hasanov, Liddle, and Mikayilov [11] classified trade into exports and imports and propounded that finished goods in contemporary times require high-powered technologies that use non-renewable energy sources including ${\mathrm{CO}}_2,$ which exacerbate the environment at a drastic rate. The authors argued and compared the production at the cardinal scale with the greater ${\mathrm{CO}}_2$ emitted by the high-powered technologies. The same ${\mathrm{CO}}_2$ further impacts O3, which causes different diseases on earth. The equilibrium between the demands for finished goods and the ${\mathrm{CO}}_2$ emitted can be mitigated through the natural process of forestation. Kurniawan and Managi [12] connected coal (i.e., ${\mathrm{CO}}_2$) consumption with urbanization to fix the rate of emission and the devaluation of atmospheric layers. In the same way, Nathaniel [13] introduced modalities of energy consumption in the urban area and recommended green energy uses for achieving overall stable environment. Other studies [12,13] noticed a linear relationship between the continuous urban demands and the emission of ${\mathrm{CO}}_2$. Kurniawan and Managi [12] claimed that trade escalates the consumption of non-renewable energy sources by the populations that demand finished goods to meet their livelihood requirements. The same relationship was found by Gozgor and Can [14], who discovered parallel trends between trade and ${\mathrm{CO}}_2$ emissions. Sushmita et al. [15] recommended that as more resources are imperative for domestic as well as commercial usages in the process of globalization, the opening of high-volume trade shall further exacerbate the environment’s deterioration. Khan et al. [16] separated the trade volume into exports and imports and further estimated its impact on emissions of ${\mathrm{CO}}_2$ into the atmosphere. Regarding the top 10 emitter nations’ environmental performance and international trade, Ali et al. [17] also examined the function of eco-innovation and the use of renewable energy sources. Their findings showed a long-term correlation between these variables.

The researchers’ main conclusion was that environmental degradation is caused by many factors, such as trade, continuous increases in the population rate in these regions, the advancement of high-power technologies that emit ${\mathrm{CO}}_2$, deforestation and soil erosion, the failure to switch to green energy sources, and the usage of rare trees at rapid speed for decorations. From this perspective, we have not found any studies for Asian countries. Therefore, this article will cover the gaps by investigating how forest wood product exports affect environmental performance in Asia.

3. Research Methodology

This section includes information on the data specifics, empirical model, and methods of estimation. The goal is to prove a connection between the exports of forest wood products and environmental performance by utilizing yearly balanced panel data from 2001 to 2020. The study considers 31 Asian countries from the region, including Armenia, Azerbaijan, Bangladesh, Bahrain, Brunei, Cambodia, China, Cyprus, Georgia, India, Indonesia, Iran, Israel, Japan, Jordan, Kazakhstan, Kuwait, Kyrgyzstan, Lebanon, Magnolia, Malaysia, Oman, Pakistan, Philippines, Russia, Saudi Arabia, Singapore, Sri Lanka, Thailand, Turkey, and Vietnam. However, the rest of the Asian countries are not included in the article due to the unavailability of data for some variables used in the model. The required information was gathered from several sources, including World Development Indicators (WDI) [18] and World Integrated Trade Solution (WITS) data [19]. Everyone can access the data for free on the websites of the source organizations. In this study, we use the model below to determine the relationship between forest wood product exports and environmental performance:

$${\mathrm{CO}}_2=\mathrm{f} (\mathrm{FE}, \mathrm{UR}, \mathrm{EC}, \mathrm{YP})$$

(1)

where ${\mathrm{CO}}_2$ emissions are used as a proxy for environmental performance. About 75% of GHG (greenhouse gas) emissions are caused by the emission of carbon, a significant pollutant [20]. FE in the model represents the total value of forest wood product exports. Forest wood products are used to generate economic activity in a country and the forest exports variable captures foreign cash. When a country’s trade basket grows more reliant on forest wood product exports, then the current forest resources may be exhausted faster than the forests regrow. With their annual removal of 2.4 Pg of carbon from the atmosphere, forests play a significant part in the carbon cycle. The amount of wood harvested globally in 2011 was 3 billion m3, or 0.6% of the expanding supply. This roughly corresponds to 8 Gt CO₂, of which about half was industrial round wood and half was wood used for energy [21]. In the model, "UR" stands for the annual total urban population, while "EC" stands for energy consumption. Urban regions now account for a large portion of the energy consumption and CO₂ emissions in Asian countries due to rapid urbanization. The effect of urbanization on fossil fuel energy use has disrupted and elevated the carbon levels in the atmosphere, producing warming. Climate change and global warming are the results of this process [22]. YP is the per person GDP. Based on the idea that higher production equates to higher pollution, economic expansion is frequently blamed for environmental problems [23].

Table 1. Description of the study variables.

Variable	Elaboration	Sources
Carbon Dioxide Emission	${\mathrm{CO}}_2$ emissions in kilo tons (kt)	World Bank, 2021
Forest Wood Product exports	Total value of forestry exported goods measured in 1000 US$	WITS, 2021
Urbanization	Total number of peoples living in urban areas	World Bank, 2021
Energy Consumption	Usage of energy (kilograms of oil equivalent to energy use per capita)	World Bank, 2021
Gross Domestic Product	GDP per person in constant 2015 US$	World Bank, 2021

Source: The author’s compilation.

below contains a list of variables, their measurements, and their data sources.

The methodology of this article is three-fold. First, several preliminary tests will be performed to validate the properties of the variables. These firstly consist of the variance inflation factor (VIF), a measurement that identifies the variables in the data that have a high degree of correlation with one another. From 1 to infinity, the VIF value can be found.

Table 2. Properties of VIF.

VIF-Score	Correlation
VIF = 1	No
1 ˂ VIF ˂ 10	Moderate
VIF > 10	High

shows the properties of VIF. A VIF value of 1 indicates that there is no correlation between the two variables. When the VIF value is in the range of 1 to 10, there is a moderate connection. A value greater than 10 indicates that the variable is extremely well explained by other variables or that other variables in the model have captured the variation that the variable explained very well [24]. The VIF’s mathematical form is:

$$\mathrm{VIF}=\frac{1}{1-{\mathrm{R}}^2}$$

(2)

Testing the data series for cross-sectional dependence and homogeneity is a common procedure when choosing an acceptable methodology or econometric tool. Researchers are able to choose acceptable econometric techniques, such as unit root and cointegration tests, by determining whether or not there is homogeneity among the variables. The choice of a suitable test aids the researcher in obtaining trustworthy and unambiguous results.

In a panel data set, there are many cross-sectional units. Therefore, there is a possibility that these nations may be cross-sectionally dependent on one another or that these cross-sectional entities are connected politically, economically, or culturally. Additionally, the first generation panel unit root test will not be reliable if the data exhibit cross-sectional dependence. Therefore, when the cross-sectionals are dependent, we must perform the panel unit root test for the second generation. Consequently, this is essentially the first stage in a panel regression model before moving on to additional estimation stages.

A panel data model’s residual terms are probably going to be highly cross-sectionally dependent, which might be caused by the existence of common shocks and unobserved components that eventually contribute to the error term [25]. According to De Hoyos and Sarafidis [26], common shocks, externalities, overlooked shared effects, unobserved components, and spatial effects are the causes of cross-sectional dependence. These frequent shocks make coefficient estimates unstable and inefficient.

To ascertain cross-sectional dependence in the data, the Breusch–Pagan LM test [27], the Pesaran scaled LM test [28], and the Pesaran CD test [29] will all be utilized in this study. The data are heterogeneous or the panel-units are independent according to the null hypothesis, while the alternative hypothesis states that the cross-sectional units are interdependent. Either the null hypothesis or the alternative hypothesis will be accepted depending on the corresponding probability value. If the p value is ˂ 0.05, the alternative hypothesis that the cross sectional units are dependent on one another will be accepted.

Additionally, it is required to check each variable for the presence of a unit root in order to establish the order of integration and prevent erroneous findings. As a result, the unit root tests will be used to assess whether or not the data are stationary. The values of a variable’s mean, variance, and covariance remain constant if it is stationary. For a regression to produce correct findings, the data must be stationary, otherwise the results will not be reliable [30].

There are two generations of unit root tests for panel data. The first-generation unit root test is employed when cross-sectional dependence is not present. However, when cross-sectional dependence is evident, the second-generation unit root test is used. The second-generation panel unit root test corrects the first-generation test’s inadequacy regarding cross-sectional dependence [31]. According to Hurlin and Mignon [32], the differences between the unit root tests for panel data and time series data are issues of heterogeneity and cross-sectional dependence. The first-generation unit root test imposes the restriction of cross-sectional independence in the data [33,34,35,36,37].

In the presence of cross-sectional dependence, various tests have been proposed that allow correlations among different cross-sectional units called “second-generation tests”. However, to check the order of integration and to make the data stationary, in this article we apply the second-generation panel unit root cross-sectionally augmented Dickey–Fuller (CADF) and cross-sectionally augmented IPS (CIPS) tests developed by Pesaran [29]. The null hypothesis states that the data do not have a unit root, and vice versa. H0 will be accepted if the absolute CIPS and CADF statistics values exceed their critical values. The series is free of the unit root problem if H0 is accepted rather than H1.

Second, a variety of cointegration tests are employed to ascertain the variables’ long-term cointegration. Panel cointegration tests can be utilized even if some of the variables are integrated for the first difference and are stationary at that level [38]. Using cointegration tests, Pesaran [39] also permits the possibility that the series’ sequences of integration may vary. Therefore, this study will apply Pedroni [40,41], Kao [42], and Johansen–Fisher panel cointegration tests to capture cointegration among the variables. The null hypothesis of a panel cointegration test shows that the variables are not cointegrated, and vice versa.

Third, the Hausman test will be utilized to decide between fixed and random effects models before regressing the system’s GMM. The best linear unbiased estimates are provided by the random effects model (BLUE). They are reliable, effective, and impartial. However, if there is a relationship between the independent variables and the random effects model’s error term, the fixed effects model would be preferred to the random effects model. As a result, the estimates from the random effects model would be incoherent. Properties of the random and fixed effects models are presented in

Table 3. Properties of the random and fixed effects models.

Hypothesis	Random Effects Model Used	Fixed Effects Model Used
H0: $cov \alpha i, xit=0$ exogeneity	Consistent Efficient	Consistent Inefficient
H1: $cov \alpha i, xit \ne 0$ endogeneity	Inconsistent	Consistent Possibly Efficient

.

If there are omitted variables to which the fixed effect model is robust, the individual-specific component α may be associated with the independent variables in the random effects model. When compared to the estimates from the random effects model, the fixed effects model estimates are inefficient but they are always consistent. These characteristics of the panel data models’ estimates direct the researcher to employ the Hausman test. Below is a description of the Hausman test’s null and alternate hypotheses:

H0: Random effects is the best model. In the panel data model, there is no relationship between the error term and the independent variables, $cov \alpha i, xit=0$. H1: A fixed effects model should be used. The panel data model’s error term and independent variables have a statistically significant association, $cov \alpha i, xit \ne 0$. The alternative hypothesis will be chosen if p < 0.05.

The most important aspect of any form of analysis is the choice of a suitable technique. Endogeneity is the phenomenon where independent variables ($xit$) and residuals (${\mathrm{e}}_{\mathrm{it}}$) exhibit a correlation. Various circumstances can lead to this, but three instances are frequently seen during research: (1) when significant variables are left out of the model; (2) due to measurement error; (3) due to simultaneity or a bi-directional relationship, whereby two or more variables on the left side of the equation are functions of one another. Utilizing OLS is more productive if there is no endogeneity. It is advisable to use a two-stage least squares instrumental variable (2SLS IV) when endogeneity is present because the OLS is inconsistent in this situation. Therefore, to overcome the endogeneity issue, in this paper we use the GMM approach to investigate the connection between the exports of forest wood products and environmental performance in the Asian nations selected by Blundell and Bond [43] and Arellano and Bover [44]. The GMM is a common analytical strategy for handling endogeneity. In a dynamic Panel model, it assesses how explained and explanatory factors interact. Additionally, it contains the controls for ${\mathrm{e}}_{\mathrm{it}}$, the unobserved heterogeneity. Adopting the GMM strategy also ensures that the final model is homoscedastic and devoid of autocorrelation issues. Additionally, it offers reliable and more accurate estimations for hetroskedasticity issues [45]. Arellano and Bover [44] were the GMM pioneers for dynamic panel data, while Blundell and Bond [43] employed the GMM model to address the endogeneity issue. The GMM strategy employed in this investigation is expressed by the following equation:

$${\mathrm{CO}}_{2_{\mathrm{it}}}{=\mathsf{\beta }}_0{+\mathsf{\beta }}_1{\mathrm{CO}}_{2_{\mathrm{it}-1}}{+\mathsf{\beta }}_2{\mathrm{FE}}_{\mathrm{it}}{+\mathsf{\beta }}_3{\mathrm{UR}}_{\mathrm{it}}{+\mathsf{\beta }}_4{\mathrm{EC}}_{\mathrm{it}}{+\mathsf{\beta }}_5{\mathrm{YP}}_{\mathrm{it}}{+\mathrm{Z}}_{\mathrm{i}}{+\mathrm{e}}_{\mathrm{it}}$$

(3)

Where the cross-section (country) and year are denoted by the subscripts i and t, respectively. Zi is a country-specific effect and stands for the heterogeneity of specific countries, which is independent and identically distributed. Carbon dioxide (${\mathrm{CO}}_2$) is used as proxy for environmental performance. Here, ${\mathsf{\beta }}_1{\mathrm{CO}}_{2_{\mathrm{it}-1}}$ is the lagged dependent variable, FE is used for forest wood product exports, UR is the urban population, EC is the energy consumption, and YP denotes the countries’ actual GDP per person. Moreover, ${\mathsf{\beta }}_0$, ${\mathsf{\beta }}_1$, ${\mathsf{\beta }}_2$, ${\mathsf{\beta }}_3$, ${\mathsf{\beta }}_4,$ and ${\mathsf{\beta }}_5$ are the parameters to be computed, while the residual term (${\mathrm{u}}_{\mathrm{it}}$) is assumed to have a normal distribution with a mean of zero and constant variance.

The Hansen test [46] and the Sargan test [47] for over-identifying constraints and the autocorrelation correlation test for the residuals serve as the foundation for the GMM’s consistency. The homoscedasticity constraint and lack of serial correlation (in levels) of the idiosyncratic residuals are the limitations of the Sargan test. The Hansen test, which evaluates the general validity of instruments that are not permitted to be linked with residuals, is used to determine whether or not the model has been properly chosen. The instruments might be considered valid and uncorrelated with residuals if the Hansen test’s null hypothesis is not rejected. This presupposes that the model has been properly specified. The null hypothesis, or AR (1), which states that there is no first-order serial correlation, must be rejected in the serial correlation test. The null hypothesis, or AR (2), which states that there is no second-order serial correlation in the residuals, should not be disproved.

Comparing the system GMM technique to other panel approaches, there are some benefits. First, because other panel data models have lagged dependent variables or have serious endogeneity problems cause by explanatory factors, therefore, they produce inaccurate and inconsistent parameter estimations [48]. Additionally, estimates using the systemized GMM are accurate and reliable even when independent components are not strictly exogenous, as well as when hetroskedasticity and autocorrelation are present. Problems with endogeneity are also addressed by the systemized GMM approach [49]. In addition, even if the panel’s data are unbalanced, the systemized GMM is chosen over other GMMs [50].

4. Empirical Results

This section comprises the empirical results and a discussion. The descriptive statistics for the data are presented in the first section.

Table 4. Descriptive statistics.

	${\mathbf{CO}}_2$	FE	EC	UR	YP
Mean	502,846.60	192,615.80	2805.79	58,632,084	11,743.29
Median	66,850	50,210.81	1549.11	8,036,227	5245.41
Maximum	10,313,460	4,244,151	11,738.13	861,289,359	61,173.90
Minimum	2150	0.009	153.56	243,643	514.86
Std. Dev.	1,479,799	449,644.40	2888.44	134,961,434	13,063.39
Observations	620	620	620	620	620

Source: Author’s calculation.

gives some basic details about the data set, such as the mean, median, maximum value, minimum value, and standard deviation for each variable utilized in the model. The mean and median values of the series are displayed as the average and middle values, respectively. The highest and smallest values of each variable utilized in the data series are provided by the maximum and minimum values, respectively. The standard deviation, which is a measure of dispersion, reveals the variation between all of the observations and their mean values. The values are calculated by using twenty years of data form 2001 to 2020 for selected Asian countries.

Secondly, in this study we apply the variance inflation factor (VIF) test. The multicollinearity issue in the data can be found using the VIF test. It assesses how much one independent variable’s variance in the regression model is explained by other independent variables. Higher VIF values imply that the data have a significant degree of multicollinearity. The findings presented in

Table 5. VIF statistics.

	Co-Linearity Statistics
Variables	LFE	LUR	LEC	LYP	Mean VIF
VIF value	1.29	1.34	4.47	4.64	2.93

Notes: The symbol (L) stands for all of the variables in log form.

validate the absence of multicollinearity, i.e., the VIF value for each variable, such as forest wood product exports, urbanization, energy consumption, and GDP, is less than 10. If the value of the VIF is less than 10, this means the series is free from the problem of multicollinearity [24].

Third, the existence of cross-sectional dependence in the error terms is examined using the Breusch–Pagan LM test [27], the Pesaran scaled LM test [28], and the Pesaran CD test [29]. Cross-sectional dependency forces researchers to use the right methods, such as second-generation panel unit root tests. The null hypothesis here indicates that the cross-sectional units are independent. The alternative hypothesis, however, indicates that these cross-sectional units are interdependent. Alternative hypotheses will be accepted if the corresponding probability value is less than 0.05. The results obtained from these tests are presented in

Table 6. Findings of the cross-sectional dependence tests.

Tests	Statistics	d.f.	p–Value	Hypothesis
Breusch–Pagan LM	3425.23	465	0.0000	Accept H1
Pesaran scaled LM	97.07	465	0.0000	Accept H1
Pesaran CD	1.51	465	0.1313	Accept Ho

below. The probability values of the Breusch–Pagan LM and Pesaran scaled LM tests are less than at 5% significance level. Therefore, the alternative hypothesis will be accepted, which indicates that the cross-sectional units are interdependent. As a result, second-generation unit root tests will be used. On the other hand, the Pesaran CD test does not support the findings of Breusch–Pagan LM and Pesaran scaled LM tests by rejecting the null hypothesis that the cross-sectional units are independent.

Fourth, second-generation unit root tests such as the cross-section ADF (CADF) and cross-sectionally augmented IPS (CIPS) tests are used because the data are cross-sectionally dependent. Both examinations were created by Pesaran [29]. The results of these tests are shown in

Table 7. Panel unit root tests.

Variables	CIPS		CADF
	I(0)	I(1)	I(0)	I(1)
$\mathrm{L}{\mathrm{CO}}_2$	−1.92	−3.80 *	−2.13 **	-
LFE	−2.57 *	-	−1.95	−3.41 *
LUR	−2.75 *	-	−3.23 *	-
LEC	−2.15	−4.09 *	−2.37 *	-
LYP	−1.64	−2.50 *	−1.87	−2.08 ***

Note: *, **, and *** represent 1%, 5%, and 10% significance levels, respectively.

. The findings are given at the level of the first difference for each variable. The outcomes show that some variables are stationary at level I (0), while others are stationary at the first difference level, or I (1). Based on the CIPS test, LFE and LUR are stationary at level I(0) and LCO2, LEC, and LYP become stationary at the first difference I(1). The CADF findings presented in Table 7 below do not support CIPS decisions. Based on the CADF results, except for ${\mathrm{LCO}}_2$, LUR, and LEC, all other variables are stationary at this level.

Fifth, to ascertain the long run cointegration between the variables, this study employs various panel cointegration tests.

Table 8. Pedroni cointegration test.

	Statistic	Prob.
Within-dimension (homogeneous)
Panel v. Statistic	−0.43	0.67
Panel rho. Statistic	1.67	0.96
Panel PP. Statistic	−3.67 ***	0.0001
Panel ADF. Statistic	−3.05 ***	0.0012
Between-dimension (heterogeneous)
Group rho. Statistic	4.71	1.0000
Group PP. Statistic	−6.48 ***	0.0000
Group ADF. Statistic	−3.54 ***	0.0002

The level of significance at 1% is indicated by the symbol “***”.

displays the Pedroni cointegration test results. The two-within and two-between dimensions, represented by four of the seven statistics, provide evidence that the model exhibits high cointegration. The study concludes that the variables are in long-term equilibrium and rejects the null hypothesis of no cointegration. The findings of the Kao cointegration test are displayed in

Table 9. Kao (1999) Cointegration test.

	t-Statistic	Prob.
ADF	2.76 ***	0.0028
Residual variance	0.000521
HAC variance	0.000585

The level of significance at 1% is indicated by the symbol "***".

, which also suggests that the model is in long-term equilibrium and validates the conclusion of Pedroni cointegration.

Table 10. Fisher–Johansen test.

Hypothesized	Fisher Stat. *		Fisher Stat. *
No. of CE(s)	(from Trace Test)	Prob.	(from Max-Eigen Test)	Prob.
None	1193. ***	˂0.01	853.5 ***	˂0.01
At most 1	903.4 ***	˂0.01	597.8 ***	˂0.01
At most 2	483.5 ***	˂0.01	313.7 ***	˂0.01
At most 3	262.6 ***	˂0.01	198.2 ***	˂0.01
At most 4	184.6 ***	˂0.01	184.6 ***	˂0.01

Note: “***” indicates the level of significance at 1%.

shows the results of the Johansen–Fisher cointegration tests. At most, five cointegration equations can be found according to the reported trace and maximum eigenvalue statistics. This further supports the notion that the panel variables we have chosen show long-term cointegration. Hence, the findings of all these techniques proved that there is strong and long-term cointegration between the variables.

The results from the estimated random effects specification and a corresponding fixed effects specification were compared using the Hausman test in the study’s final section. The difference in coefficient is not systematic according to the null hypothesis in this case. The use of fixed effects instead of random effects is more acceptable because the Hausman chi-square probability value is smaller than 0.05. Finally, this study uses the systemized GMM technique to overcome the issue of endogeneity and to achieve reliable results, as indicated by Blundell and Bond [43] and Arellnao and Bover [44]. The use of the difference GMM estimator, according to Blundell and Bond [43], results in estimates that are both inefficient and biased if the dependent variable in an equation is persistent and close to being random. Different GMMs in such cases use poor instruments. Therefore, to redress this, we propose the use of a systemized GMM estimator.

The results obtained from the fixed effects, random effects, and GMM models are presented in

Table 11. Fixed effects and random effects models vs. the GMM.

Variables	FE (1)	RE (2)	Sys-GMM (3)
$\mathrm{L}{\mathrm{CO}}_2$(−1)	- -	- -	0.70 ** (0.04)
LFE	0.022 * (0.011)	0.248 * (0.011)	0.02 ** (0.004)
LEC	0.632 ** (0.042)	0.593 ** (0.038)	0.26 ** (0.07)
LUR	0.808 ** (0.037)	0.969 ** (0.022)	0.16 ** (0.04)
LYP	0.221 ** (0.031)	0.163 ** (0.029)	0.06 ** (0.02)
Hausman chi2	19.78	-	-
p-value	(0.0006)	-	-
Observations	620	620	620
Cross-sectional	31	31	31
	-	Hansen test’s p-value	0.43
	-	p-value for AR(1)	0.0000
	-	p-value for AR(2)	0.27
	-	Instrument rank	31

Note. *, ** show 5 and 1% significance levels, respectively.

. The findings indicate that the coefficients of all variables are positive and have significant impacts on the environmental performance in Asia. Both the models, i.e., the fixed effects and systemized GMM models, estimate different coefficients for each variable. However, the systemized GMM estimator is considered a powerful tool for estimating a dynamic panel model via an autoregressive process, where T is small compared to the cross-sectionals. Another benefit of the systemized GMM estimator, including the lagged dependent variable introduced here, is the endogeneity with respect to that particular variable, so the systemized GMM estimator can be used to resolve this endogeneity problem. Hence, the systemized GMM results are more reliable than for the fixed effects model. The findings for the systemized GMM show a positive and significant link between the dependent and independent variables. These findings show that a1% increases in forest wood product exports, urbanization, energy consumption, and GDP per capita contribute 0.02%, 0.26%, 0.16%, and 0.06%, respectively, of the ${\mathrm{CO}}_2$ emissions in Asia. Moreover, the coefficient for the log of ${\mathrm{CO}}_2$ is also positive and has a significant impact. Furthermore, the Hansen and Sargan tests confirm the validity of instruments and the second-order autocorrelation in the model, respectively.

5. Discussion

The findings from a variety of econometric methodologies provide credence to the economic theory that Asia’s environmental performance is impacted by exports of forest wood products. These findings are quite similar to the results found by Henders, Persson, and Kastner [51]. They measured the area of tropical deforestation and the carbon emissions from land-use change (LUC) brought on by the production and export of four commodities (beef, soybeans, palm oil, and wood products) in seven countries with significant deforestation rates (Argentina, Bolivia, Brazil, Paraguay, Indonesia, Malaysia, and Papua New Guinea). Between 2000 and 2011, these four items caused an average deforestation area of 3.8 Mha and LUC emissions of 1.6 ${\mathrm{GTCO}}_2$. Similar research was conducted by Nabuurs and Sikkema [52], who found a positive association between the trade in wood products and the carbon cycle for their respective countries, i.e., Gabon, Sweden, and the Netherlands. Likewise, numerous researchers from throughout the world have discovered similar outcomes. For instance, Peng, Ning, and Yang [6] focused on the exchange of CO2 emissions in three harvested wood product (HWP) sectors between China and its main international trading partners. China produced a significant amount of HWP sector emissions while satisfying the international demand for consumption.

6. Conclusions and Recommendations

In this study, which covered the period from 2001 to 2020, we made an empirical attempt to identify the main factors that influence carbon dioxide emissions in Asian countries. We looked at how exports of forest wood products, energy use, urbanization, and GDP per person affect ${\mathrm{CO}}_2$ emissions. This study used ${\mathrm{CO}}_2$ emissions as a proxy for measuring environmental performance, similar to the majority of earlier studies. Second-generation unit root tests were used in the study, which showed that all variables are stationary at either the first level or the first difference. This study also employed the second-generation cointegration tests after ensuring that each variable in the current study had a value of I (0) and I (1). These cointegration test results demonstrated that there is significant empirical support for a long-term equilibrium relationship between the target variables. The Hausman test was additionally utilized in this study to identify the best model between fixed and random effects. The findings of the Hausman test revealed that the fixed effects model is more appropriate than the random effects model. In addition, to obtain reliable results and to address the issue of endogeneity, this empirical study also employed the systemized GMM estimation technique. The elasticity results from the GMM showed that all independent variables, including the lagged dependent variable, have significant impacts on the environmental performance in the long run. Therefore, an increase in forest wood product exports will enhance the ${\mathrm{CO}}_2$ emissions in the long run, which will adversely affect the environmental performance in Asia.

In line with the aforementioned findings, this study urges decision-makers to combine trade and ${\mathrm{CO}}_2$ emissions control to integrate the regulation of ${\mathrm{CO}}_2$ emissions with trade policies. In fact, as long as forest wood product exports are positively correlated with carbon dioxide emissions, then it would be logical that these countries are improving foreign exchange earnings while incurring more costs in the form of environmental deterioration. Based on the findings of this article, the selected Asian countries should curtail forest wood products in their trade basket in order to enhance the environmental performance in the region. Furthermore, the alternative use of forest products and their exports could be increased to meet these requirements.

7. Study’s Limitations and Future Direction

Due to the availability of data, this study examined data from 2001 to 2020. Thirty-one Asian nations were taken into consideration for the study’s analysis, including Armenia, Azerbaijan, Bangladesh, Bahrain, Brunei, Cambodia, China, Cyprus, Georgia, India, Indonesia, Iran, Israel, Japan, Jordan, Kazakhstan, Kuwait, Kyrgyzstan, Lebanon, Magnolia, Malaysia, Oman, Pakistan, Philippine, Russia, Saudi Arabia, Singapore, Sri Lanka, Thailand, Turkey, and Vietnam. However, due to the lack of data for some of the model’s variables, Afghanistan, Bhutan, Iraq, Laos, Maldives, Myanmar, Nepal, North Korea, Palestine, Qatar, South Korea, Syria, Taiwan, Tajikistan, Timor-Leste, Turkmenistan, United Arab Emirates (UAE), Uzbekistan, and Yemen were excluded from the study. The data are freely available to everyone and can be accessed from the websites of the source organizations. Researchers will have a better understanding of the issue if they use a larger sample that covers the entire Asian continent. The scope of the future research can be expanded to examine the net forest change and how it affects the environmental performance in the area. The total of all forest gains (forest growth) and losses (deforestation) in a given period is known as the forest area net change.