Repercussions of Hydroelectricity use on Carbon Emissions in Bangladesh: Evidence using Novel Fourier-Bootstrapped ARDL and Fourier-Gradual Shift Causality analyses

Introduction

In the modern day, pursuing economic activities without simultaneously controlling the negative environmental externalities is often discouraged. Consequently, isolating environmental degradation from economic growth, especially in the form of mitigating Carbon dioxide (CO2) emissions, has gained tremendous attention worldwide (Alam, 2022; Murshed, Mahmood, et al., 2022). As per the preconceived notion, the consumption of energy resources is the leading contributor to CO2 emissions (Alam et al., 2022; Manigandan et al., 2022); consequently, the global countries need to strategize relevant CO2 emission-inhibiting plans (IEA, 2019). Moreover, committing to control greenhouse gas emissions, most of the world economies have pitched to execute climate change tackling strategies by ratifying the Paris Accord (Khan, Murshed, et al., 2021; Doğan et al., 2020). As a result, from the perspective of complying with the commitments, the objective of minimizing the emissions generated from the consumption of unclean energy resources needs to be embedded within the economic development policies. In the same line, the United Nations had also laid out the 2030 Sustainable Development Goal (SDG) agenda to ensure simultaneity between environmental and economic well-being (Shahbaz et al., 2021; Guang-Wen et al., 2022). However, ensuring a harmonious relationship between economic growth and environmental improvement is a major challenge given the fact that most of the world economies still rely heavily on non-renewable energy supplies for meeting the persistently surging global energy demand. According to statistical estimates provided by the World Bank (2022), non-renewable energy on average accounts for more than 80% of the global final energy consumption level. This monotonic non-renewable energy dependency can be assumed to catalyze the global discharge of CO2 emissions and, ultimately, induce chronic climate adversities worldwide. Accordingly, for combating the environmental hardships, lessening reliance on unclean energy resources is assumed to be of prime importance (Hamid et al., 2021; Doğan et al., 2021).

Besides, it is also imperative to know that the power sector generates almost half of the global CO2 emission figures (World Bank, 2022). Hence, it is relevant to explore the channels through which CO2 emissions generated within the power sector can be contained. But firstly, before initiating an appropriate energy policy reform, it must be understood why the global power sector is such a high CO2-emitting sector. Notably, statistics portray that coal accounts for the highest share of primary energy inputs employed within the power sectors worldwide; close to 40% of the global electricity output is generated from coal while natural gas comprises an additional 20% of it (World Bank, 2022). Moreover, statistics also reveal that non-renewable energy use collectively generates over three-quarters of the aggregate electricity output produced worldwide (World Bank, 2022). Under such predominant non-renewable energy dependency across the world economies, the power sector has evolved as the major contributor to greenhouse gas emission-led atmospheric pollution. As a result, greening the global energy systems is of paramount importance so that the associated discharges of CO2 emissions can be contained. Among the different greening channels, the related literature often recommends substantial employment of renewable energy resources as primary inputs within the power sectors worldwide (Sinha et al., 2022; Nathaniel et al., 2021).

Against this background, this study attempts to evaluate the repercussions of renewable electricity (hydroelectric power) consumption on fossil fuel-based CO2 emissions in Bangladesh’s context using annual data spanning from 1972 to 2019. Besides, the analysis controls for other key macroeconomic variables such as economic globalization, financial development, urbanization, and economic growth. The decision of choosing Bangladesh as a case study is influenced by the fact that this major South Asian economy has always turned to both locally sourced and imported primary fossil fuels for producing electricity (Murshed & Alam, 2021). Hence, it can be said that fossil fuel-based electricity production and consumption have played significant roles in expanding the Bangladesh economy; however, the corresponding environmental impacts have largely been undesirable (Hasan et al., 2021). Although Bangladesh is regarded as one of the lowest CO2-emitting global economies, the robust economic growth trends of the nation, especially over the last couple of decades, suggest that the nation should proactively think of adopting policies to decouple environmental deterioration from its robust economic growth performances. Besides, the nation has ratified the Paris Accord and has agreed to mitigate its greenhouse gas emissions to collectively tackle the global climate change adversities. Moreover, at the recently concluded 26^th Conference of Parties (COP26), which was held in 2021 in Glasgow, Bangladesh promised to reduce its CO2 emission figures by 22% by 2030 (The Business Standard, 2021). Furthermore, it is to be noted that the power sector of Bangladesh accounts for more than half of the nation’s total volume of CO2 emissions. Thus, mitigating energy consumption-based CO2 emissions is critical for this rapidly emerging South Asian country. Therefore, considering the national objective of curbing CO2 emissions by more than one-fifth by 2030 and the international commitments of the Paris Agreement, it is pertinent for Bangladesh to adopt policies that can effectively inhibit its energy use-based CO2 emission figures.

Several contributions to the literature can be associated with this current study. Firstly, this is the only study that estimates the determinants of disaggregated levels of energy consumption-related CO2 emissions, especially those resulting from the combustion of gas, coal, and oil. In contrast, the existing studies have mostly considered the aggregate level of CO2 emissions in Bangladesh (Murshed, Rashid, et al., 2022; Ali et al., 2020). Analyzing the determinants of disaggregated CO2 emissions is important for formulating comprehensive policies for energy diversification, fossil fuel dependency reduction, and environmental well-being enhancement from Bangladesh’s perspective. Besides, the Bangladesh government plans to reduce its CO2 emission levels particularly by integrating renewable energy into the national energy mix, improving energy efficiency levels, and avoiding depletion of the domestic fossil fuel reserves (The Business Standard 2021). Thus, it is imperative to know whether or not greater use of hydroelectricity and globalization exert heterogeneous impacts on CO2 emissions generated from the combustion of different fossil fuel sources in Bangladesh. Secondly, this study assesses the effects of economic globalization on Bangladesh’s disaggregated energy use-related CO2 emission figures while the previous studies have mostly evaluated the globalization-CO2 emissions nexus using the aggregate globalization index of Bangladesh (Islam et al., 2021). However, since globalization embodies several forms, it can be expected that different components of globalization exert heterogeneous environmental impacts. In this regard, as economic globalization considers both trade openness and Foreign Direct Investment (FDI) inflows, investigating the relationship between economic globalization and CO2 emissions would help policymakers understand the efficacies of the existing international trade and financial globalization policies in safeguarding the nation’s environmental attributes. Lastly, apart from evaluating the independent environmental effects of renewable energy use and economic globalization, this study also aims to unearth the interactive effects of these variables on disaggregated CO2 emission figures, as well. None of the previous studies has emphasized evaluating these joint impacts for environmental quality assessment purposes in the context of Bangladesh. However, relevant to the objectives of this study, it is pertinent to predict the possible joint effects since economic globalization in the form of intra-regional trade among the South Asian nations has been recognized as a means for Bangladesh to import hydroelectricity from the South Asian neighbors (Timilsina, 2018; Rahman and Alam 2021).

The next section provides an overview of the historical trends in energy consumption and CO2 emissions in Bangladesh. The review of relevant literature is presented in the subsequent section followed by the presentation of the empirical model, estimation strategy, findings, discussion of the results, and conclusion with potential policy recommendations.

Trends in Energy Consumption and CO2 Emissions in Bangladesh

It is known that Bangladesh has traditionally traded-off low environmental quality for stimulating the rapid growth of its economy. This phenomenon can be clearly understood from the corresponding trends in the country’s national income and energy consumption-based CO2 emission figures shown in Figure 1. It is evident that between 1972 and 2019, the annual per capita GDP figures of Bangladesh have almost quadrupled whereby the nation graduated from a low-income to a lower-middle-income status. Now, considering the corresponding rising trends in both the aggregate and disaggregated energy consumption-related CO2 emission figures, it can be asserted that the robust economic growth performances of Bangladesh were achieved at the expense of poor environmental quality. The aggregate annual CO2 emission per capita levels are observed to have risen by more than 10-fold over the 1972–2019 period. Correspondingly, the annual disaggregated CO2 emission per capita figures generated from the combustion of gas, coal, and oil have multiplied by around 35%, 15%, and 3%, respectively. Therefore, compared to coal and oil consumption-based CO2 emissions, the growth in the annual gas consumption-based CO2 emission per capita levels have been significantly higher; especially, from early 1980 onwards, the gas consumption-based CO2 emission figures took off particularly due to the government’s decision to gradually replace imported petroleum oils with local natural gas for electricity generation purposes (Murshed, 2020). Furthermore, the increasing aggregate and disaggregated energy-related CO2 emission trends portray the possible rise in Bangladesh’s fossil fuel dependency.OPEN IN VIEWER

Figure 2 illustrates the trends in the shares of renewable and non-renewable electricity in the total electricity outputs of Bangladesh. Two distinct scenarios can be identified from this figure. First, the electricity sector of Bangladesh has always depended on primary fossil fuel sources whereby the share of non-renewable electricity output has traditionally been substantially larger than the corresponding renewable electricity output shares. Second, not only has the electricity sector of Bangladesh been fossil fuel-dependent, the dependency has risen over the last four decades or so. This claim is certified by the upward-sloping graph depicting the difference between non-renewable and renewable electricity shares in the total electricity outputs of Bangladesh. This is an alarming situation given the fact that the current government of Bangladesh is trying to diversify the national energy mix by replacing fossil fuels with renewable alternatives.OPEN IN VIEWER

As far as the different primary energy sources employed to generate electricity in Bangladesh is concerned, Figure 3 shows a couple of concerning energy-related facts. First, it can be seen that Bangladesh has not been able to diversify its national energy mix before the 2000s. In the 1970s, 1980s, and 1990s, electricity in Bangladesh was generated using oils, natural gas, and renewables (especially hydroelectricity) while from the 2000s onwards coal made its way into the national energy mix. Therefore, despite managing to somewhat diversify the energy mix, the diversification was not done in an environmentally sustainable manner since coal is regarded as an unclean fossil fuel that releases hefty volumes of CO2 emissions when combusted, particularly, for producing electricity. Second, natural gas-based electricity is observed to account for a lion’s share in the total electricity output of Bangladesh from the 1980s onwards. Before that, oil-based electricity also held a significant share of Bangladesh’s aggregate electricity output. However, following the world oil price hikes during the early 1980s, the government turned to domestic natural gas supplies to replace oil for meeting the local energy demand. This justifies the relatively larger surge in Bangladesh’s gas-based CO2 emission figures as shown in Figure 2. Third, it can also be observed that rather than boosting renewable energy use, the share of renewable electricity outputs has rather declined. During the 1970s, the average share of renewables in the total electricity output was close to 23% (Murshed, Rashid, et al., 2022) which steadily dropped to reach a little more than 1.5% between the 2011–2015 period. Hence, it can be assumed that Bangladesh is not technologically and financially advanced enough to stimulate renewable energy transition within its electricity sector.OPEN IN VIEWER

Review of Literature

In this section, we first provide the theoretical underpinnings relevant to our study followed by the summarization of the corresponding empirical findings documented in the literature.

Empirical Model and Estimation Strategy

Empirical Model

In line with the assumptions of the EKC hypothesis, and keeping into consideration the theoretical underpinnings discussed in the previous section, our baseline empirical model is expressed as:

Model 1

ln F C O 2 k , t = φ 0 + φ 1 l n H E L E C T t + φ 2 E G I t + φ 3 l n R G D P t + φ 4 F D t + φ 6 U R B 5 + ε t

(1)

where the subscript t refers to the period of analysis (1972–2019), φ 0 is the intercept parameter, φ m ( m = 1 , 2 , … , 5 ) are the elasticity parameters, and ε t is the error term. The dependent variable FCO2_k stands for a set of three (k=3) types of fossil fuel-based CO2 emissions. Precisely, in this study, we consider CO2 emissions generated from the combustion of gas, oil, and coal in Bangladesh.

Among the explanatory variables, HELECT refers to the annual level of hydroelectricity consumption in Bangladesh. Hydroelectricity consumption is used as a proxy for renewable electricity use in the country. On the other hand, the variable EGI refers to the economic globalization index which is constructed using the levels of trade openness and FDI inflows, in particular (Gygli et al., 2019). A higher value of this index is synonymous with a rise in the degree of Bangladesh’s economic globalization with the other world economies. The predicted sign of the corresponding elasticity parameter ( φ 2 ) can be used to verify the PHH. In contrast, the negative sign of this elasticity parameter ( φ 2 < 0 ) would invalidate the PHH and support the pollution halo effect in Bangladesh. The variable RGDP refers to the real GDP level of Bangladesh. This variable is used as a proxy for the nation’s affluence or economic growth level. The variable FD abbreviates for financial development which is measured by the share of domestic credit extended to the private sector by banks in Bangladesh’s GDP (Murshed, Rashid et al., 2022). In this regard, a higher share portrays a more developed financial sector and vice versa. Lastly, the variable URB stands for the rate of urbanization which is measured as a percentage of the urban population in the total population of Argentina (Murshed, Rashid et al., 2022). The variables FCO2_k, HELECT, and RGDP are converted into natural logarithms (as denoted by the prefix ln) to predict the elasticity parameters ( φ m ).

Further, keeping into consideration the prospects of Bangladesh to import hydroelectricity from the neighboring South Asian nations like Bhutan and Nepal, in particular, we also evaluate the possible joint impacts of hydroelectricity consumption and economic globalization on disaggregated fossil fuel consumption-based CO2 emissions. Hence, we augment an interaction term in Model 1 which can be shown as:

Model 2

ln F C O 2 k , t = φ 0 + φ 1 l n H E L E C T t + φ 2 E G I t + β ( l n H E L E C T t ∗ E G I t ) + φ 3 l n R G D P t + φ 4 F D t + φ 6 U R B 5 + ε t

(2)

where the variable HELECT*EGI is the interaction term between hydroelectricity consumption and economic globalization in Bangladesh. Hence, the predicted sign of the elasticity parameter β would indicate the joint impacts of these variables on Bangladesh’s disaggregated fossil fuel consumption-based CO2 emission figures. Both models 1 and 2 are separately estimated using disaggregated figures of CO2 emissions stemming from gas, oil, and coal consumption, respectively. For checking the robustness of the long-run ARDL estimates, the DOLS estimator of Stock and Watson (1993) is utilized to re-estimate Models 1 and 2. Annual frequency data from 1972 to 2019 is utilized in this study while the missing data issue is handled using the linear interpolation technique used in the preceding studies by Menegaki and Tiwari (2018); Nasir et al. (2020). The selection of this study period is solely based on data availability. Besides, since Bangladesh achieved independence in 1971, it is not relevant to consider data from the pre-independence era. The definitions, measuring units, and data sources of the selected variables are shown in Table 1. ¹

OPEN IN VIEWER

Table 1.

The Definitions, Units of Measurement, and Data Sources of the Variables.

Symbol	Definition	Unit	Source
FCO2	Fossil fuel consumption-based CO2 emissions (generated from gas/oil/coal)	Kilograms per capita	Global Carbon Atlas (Andrew & Peters, 2021)
HELECT	Hydroelectricity consumption	Kilowatt hours per capita	British Petroleum Statistical Review of World Energy
EGI	Economic globalization (it contains shares of portfolio investments, trade, FDI, and income payments to foreign officials in the GDP)	Index	Gygli et al. (2019)
RGDP	Real per capita GDP	Constant 2015 US$ per capita	World Bank (2022)
FD	Financial development (domestic credit provided to the private sector by banks)	Percentage of GDP	World Bank (2022)
URB	Urbanization rate (urban population)	Percentage of the total population	World Bank (2022)

Estimation Strategy

The estimation process commences with the unit root analysis which is important since the choice of regression techniques is conditional on the order of stationarity of the variables in the respective model. The traditional unit root estimators such as the Dickey–Fuller Generalized Least Squares (DF-GLS), Augmented Dickey–Fuller (ADF), Phillips–Perron (PP), and Kwiatkowski–Phillips–Schmidt–Shin (KPSS) are incapable of controlling for the potential structural break problems in the data (Qahtan et al., 2021). Therefore, these methods are likely to yield inconsistent unit root outcomes. Therefore, considering the limitations of these techniques, we utilize the recently developed Residual Augmented Least Squares-Lagrange Multiplier (RALS-LM) unit root estimation technique of Meng et al. (2017). This technique considers trend breaks and non-normal errors and predicts the unit root properties in the presence of up to two structural breaks in both the intercept and trend (Krištić et al., 2019). For robustness check, the Lee–Strazicich one break and the Narayan–Popp two break unit root estimation methods of Lee and Strazicich (2013) and Narayan and Popp (2013), respectively, are also utilized in this study. All these three methods predict test statistics to verify the alternative hypothesis of stationarity of the series at the level [I (0)] or the first difference [I (1)].

Following the conclusion of the unit root analysis, and given that the variables have a maximum order of integration at the first difference, the next step involves the application of the novel Fourier-Bootstrapped Autoregressive Distributed Lag (BARDL) bounds testing technique introduced by Solarin (2019). This method for predicting cointegration (long-run associations) among the variables is an extended and advanced variant of the conventionally used ARDL bounds testing approach introduced by Pesaran et al. (2001) and the advanced BARDL method of McNown et al. (2018). According to the method of Pesaran et al. (2001), our baseline model (i.e., Model 1) can be represented as an Error-Correction Model (ECM) as follows

∆ ln F C O 2 k , t = ϑ 0 + ∑ i = 1 p − 1 ϑ 1 ∆ l n F C O 2 k , t − i + ∑ i = 1 p − 1 ϑ 2 ∆ l n H E L E C T t − i + ∑ i = 1 p − 1 ϑ 3 ∆ E G I t − i + ∑ i = 1 p − 1 ϑ 4 ∆ l n R G D P t − i + ∑ i = 1 p − 1 ϑ 5 ∆ F D t − i + ∑ i = 1 p − 1 ϑ 6 ∆ U R B t − i + ϑ 7 l n F C O 2 k , t + ϑ 8 l n H E L E C T t + ϑ 9 E G I t + ϑ 10 l n R G D P t + ϑ 11 F D t + ϑ 12 U R B t + ϑ 13 T R E N D + ϑ 14 S B D U M 1 + ϑ 15 S B D U M 2 + ρ t

(3)

where the optimal lag order, which is allowed to vary for each variable, is denoted by p; SBDUM1 and SBDUM2 are the structural break dummy variables that are constructed using information regarding the locations of the structural breaks for the dependent variable generated from the RALS-LM unit root analysis. Under the traditional ARDL bounds test of Pesaran et al. (2001), an F-overall statistic and a t-dependent statistic are predicted considering the null hypothesis of no long-run associations between the variables of concern. The null hypotheses can be shown as

N u l l H y p o t h e s i s ( F − o v e r a l l s t a t i s t i c ) : ϑ 7 = ϑ 8 = ϑ 9 = ϑ 10 = ϑ 11 = ϑ 12 = 0

(4)

N u l l H y p o t h e s i s ( t − d e p e n d e n t ) : ϑ 7 = 0

(5)

The conventional F-overall statistic considers lagged levels for both the dependent and independent variables within the ECM while the t-dependent statistic considers only the lagged level of the dependent variable (Alam et al., 2022). Under such circumstances, cointegration among the variables is affirmed if any of these test statistics are statistically significant. However, McNown et al. (2018) extended this cointegration analysis by presenting an additional F-statistic (i.e., the F-independent) which considers the lagged levels of only the independent variables within the ECM model. The corresponding null hypothesis of no cointegration can be shown as

N u l l H y p o t h e s i s ( F − i n d e p e n d e n t s t a t i s t i c ) : ϑ 8 = ϑ 9 = ϑ 10 = ϑ 11 = ϑ 12 = 0

(6)

For cointegration to exist, all these three test statistics (F-overall, t-dependent, and F-independent) have to be statistically significant; in simple terms, the absolute values of these statistics have to be greater than the corresponding bootstrapped critical values (Pata & Aydin, 2020). However, despite being superior to the conventional ARDL bounds test, the BARDL method has a limitation in terms of not being efficient in accounting for the smooth and sharp shifts/transitions in the cointegrating relationships among the variables (Solarin, 2019). Hence, Solarin (2019) extended the BARDL method of McNown et al. (2018) by introducing Fouriers into the ECM model. The inclusion of the Fourier is believed to capture the possible smooth transitions in cointegrating relationships (Pata & Aydin, 2020). Under the Fourier BARDL approach, the inclusion of Fouriers in equation (3) can be shown as

∆ ln F C O 2 k , t = ϑ 0 + ∑ i = 1 p − 1 ϑ 1 ∆ l n F C O 2 k , t − i + ∑ i = 1 p − 1 ϑ 2 ∆ l n H E L E C T t − i + ∑ i = 1 p − 1 ϑ 3 ∆ E G I t − i + ∑ i = 1 p − 1 ϑ 4 ∆ l n R G D P t − i + ∑ i = 1 p − 1 ϑ 5 ∆ F D t − i + ∑ i = 1 p − 1 ϑ 6 ∆ U R B t − i + ϑ 7 l n F C O 2 k , t + ϑ 8 l n H E L E C T t + ϑ 9 E G I t + ϑ 10 l n R G D P t + ϑ 11 F D t + ϑ 12 U R B t + ϑ 13 T R E N D + ϑ 14 S B D U M 1 + ϑ 15 S B D U M 2 + ϑ 16 sin ( 2 π m t T ) + ϑ 17 cos ( 2 π m t T ) + ρ t

(7)

where the frequency of the Fourier is represented by m. Given that cointegration exists among the variables, the short- and long-run marginal effects of the explanatory variables (hydroelectricity consumption, economic globalization, financial development, urbanization, and economic growth) on the outcome variable (fossil fuel consumption-based CO2 emissions) are predicted using Fourier BARDL regression analysis. Furthermore, for evaluating the robustness of the regression outcomes an alternative estimator is also utilized to predict the long-run marginal effects. In this study, the Dynamic Ordinary Least Squares (DOLS) technique of Stock and Watson (1993) is employed.

In the last stage of the analysis, the causal associations among the variables are ascertained. Conventionally, the Toda–Yamamoto causality estimation technique proposed by Toda and Yamamoto (1995) has been widely used in the literature. This technique predicts causal properties using a Vector Autoregressive (VAR) model. However, predicting causality using a VAR model may lead to inaccurate outcomes since the structural changes are not accounted for in this traditional approach (Enders & Jones, 2016). Therefore, to address this limitation, Nazlioglu et al. (2016) recently introduced the Fourier approximation within the Toda–Yamamoto approach. Consequently, the Fourier Toda–Yamamoto (FTY) causality estimator of Nazlioglu et al. (2016) emerged for conducting Granger causality analysis with (a) single Fourier frequency and (b) cumulative Fourier frequency (Pata & Aydin, 2020). In this study, we utilize the Fourier frequency version of the FTY method since its application is appropriate in the context of the number of observations being between 50 and 100. Since this study utilizes 48 years of data, the use of the single Fourier frequency version of the FTY is chosen. For comparison purposes, the non-Fourier Toda and Yamamoto (1995) method is also utilized to predict the causal properties of the variables.

Findings and Discussions

Firstly, the results from the unit root analysis are reported in Table 2. As mentioned earlier, the statistical significance of the predicted test statistics, for all three unit root estimation techniques, verifies the alternative hypothesis to affirm the stationarity of the variable of concern. Hence, the estimates suggest that almost all the variables are stationary at the first difference apart from EGI and lnRGDP. As a result, the findings reveal two major outcomes concerning the unit root properties of the variables. First, a mixed order of integration among the variables is evidenced. Second, the variables are integrated at a maximum order of I (1) [i.e., the first difference]. Both these outcomes fulfill the pre-requirement of performing the Fourier BARDL analysis. Moreover, the locations of the structural breaks are also reported in Table 2. It is interesting to see that for all three dependent variables, the first structural break is located in the early 1980s. This is a justified finding since the government of Bangladesh made an important energy policy reform during this period whereby it was decided to replace imported oils with local natural gas for electricity generation purposes. This major policy intervention by the government ensured that the share of natural gas-used electricity in Bangladesh’s total electricity output went up while the share of oil-based electricity gradually declined. The identified structural break locations for the three dependent variables are utilized to construct dummies and then inserted into the respective models to account for structural break problems within the cointegration and regression analyses to follow.

OPEN IN VIEWER

Table 2.

The Outcomes From the Unit Root Analysis.

Estimator	RALS-LM test (two breaks)		Lee–Strazicich test (one break)		Narayan–Popp test (two breaks)
Variable	Test statistic	Break dates	Test statistic	Break date	Test statistic	Break dates
lnFCO2 (gas)	−1.980	1983, 2001	−2.980	2000	−1.601	1983, 2000
lnFCO2 (oil)	−1.770	1984, 2000	−3.145	2000	−1.982	1983, 2001
lnFCO2 (coal)	−2.123	1980, 2000	−3.080	2001	−1.980	1981, 2000
lnHELECT	−2.201	1988, 2007	−2.898	2005	−1.140	1988, 2008
EGI	−2.850**	1985, 1998	−4.105***	2000	−6.789***	1985, 1997
lnRGDP	−2.950**	1990, 2010	−4.630***	2010	−5.930***	1988, 2011
FD	−2.010	1998, 2008	−2.225	2006	−2.665	1998, 2008
URB	−1.308	1995, 2005	−2.002	2006	−2.380	1995, 2005
∆lnFCO2 (gas)	−4.505***	1984, 2000	−7.600***	2000	−4.879**	1983, 2000
∆lnFCO2 (oil)	−3.945***	1984, 2001	−8.213***	2000	−4.925**	1984, 2001
∆lnFCO2 (coal)	−5.200***	1982, 2000	−8.100***	2000	−6.620***	1982, 2000
∆lnHELECT	−4.910***	1988, 2010	−6.930***	2010	−6.540***	1987, 2009
∆EGI	−3.100**	1990, 2005	−5.213***	2005	−7.140***	1990, 2006
∆lnRGDP	−4.213***	1985, 2010	−5.550***	2010	−7.750***	1985, 2010
∆FD	−4.804***	2000, 2009	−5.910***	2008	−6.889***	1998, 2008
∆URB	−4.201***	1995, 2005	−6.200***	2005	−5.280**	1995, 2005

Next, we move on to the Fourier BARDL analysis which firstly evaluates the possible cointegrating relationships among the variables included in each model. The cointegration outcomes, as shown in Table 3, reveal that for all models the predicted test statistics (F-overall, t-dependent, and F-independent) are statistically significant whereby the alternative hypothesis of cointegration among the variables can be verified. Therefore, it can be claimed that CO2 emissions generated from gas, oil, and coal combustion have long-run relationships with hydroelectricity consumption, economic globalization, economic growth, financial development, and urbanization in the context of Bangladesh. The validation of these cointegrating associations allows us to conduct the regression analysis to predict the short- and long-run marginal environmental impacts associated with changes in the explanatory variables.

OPEN IN VIEWER

Table 3.

The Results From the Fourier BARDL Cointegration Analysis.

Model	Specification	F-overall	t-dep.	F-indep.
1	lnFCO2 (gas) = f {HELECT, EGI, lnRGDP, FD, URB, SBDUM1, SBDUM2}	6.85***	−3.85**	8.34***
1	lnFCO2 (oil) = f {HELECT, EGI, lnRGDP, FD, URB, SBDUM1, SBDUM2}	5.20**	−3.62**	7.15**
1	lnFCO2 (coal) = f {HELECT, EGI, lnRGDP, FD, URB, SBDUM1, SBDUM2}	5.75***	−3.77**	7.23**
2	lnFCO2 (gas) = f {HELECT, EGI, (lnHELECT*EGI), lnRGDP, FD, URB, SBDUM1, SBDUM2}	6.99***	−4.13**	8.45**
2	lnFCO2 (oil) = f {HELECT, EGI, (lnHELECT*EGI), lnRGDP, FD, URB, SBDUM1, SBDUM2}	4.90**	−2.95**	5.95**
2	lnFCO2 (coal) = f {HELECT, EGI, (lnHELECT*EGI), lnRGDP, FD, URB, SBDUM1, SBDUM2}	5.35**	−4.44**	8.20***

The short- and long-run elasticity estimates from the Fourier BARDL analysis in the context of Model 1 are presented in Table 4. The results show that hydroelectricity consumption is effective for curbing the disaggregated fossil fuel consumption-based CO2 emission figures. An increase in the per capita level of hydroelectricity consumption is predicted to reduce per capita CO2 emissions related to gas, oil, and coal consumption respectively by 1.634%, 1.320%, and 0.240% in the short run and by 1.875%, 1.551%, and 0.287% in the long run, on average, ceteris paribus. Similar findings, but for the aggregate CO2 emission level, were also reported in the study by Bello et al. (2018) for the case of Malaysia. Therefore, these results highlight that CO2 emissions generated from the combustion of gas and oil are elastic to a marginal increase in the hydroelectricity consumption level, whereas CO2 emissions stemming from coal consumption are evidenced to be quite inelastic. This is quite understandable since natural gas and petroleum oils account for around 96% of Bangladesh’s total electricity output (World Bank, 2022). Hence, switching to hydroelectricity is more likely to reduce the utilization of these primary fuels for generating electricity, especially natural gas. It is to be noted that Bangladesh is currently facing natural gas supply shortages whereby energy diversification within the power sector is likely to significantly curb the share of natural gas-based electricity in the total electricity output level of Bangladesh. In contrast, since coal comprises a very low share of the aggregate electricity output, enhancing hydroelectricity use may not induce a large reduction in the use of coal for electricity generation purposes. Moreover, it is also evident that the long-run CO2 emission-inhibiting effects are relatively larger than the corresponding short-run effects. Hence, it can be said that gradually increasing the share of hydroelectricity in the total electricity output would be an effective means of curbing fossil consumption-based CO2 emissions in Bangladesh. Therefore, the pertinence of initiating renewable electricity transition within the power sector for achieving environmental sustainability cannot be denied.

OPEN IN VIEWER

Table 4.

The Fourier BARDL Short- and Long-Run Results for Model 1.

Dep. Var.	FCO2 (gas)		FCO2 (oil)		FCO2 (coal)
Regressors	Short run	Long run	Short run	Long run	Short run	Long run
lnHELECT	−1.634***	−1.875***	−1.320***	−1.551***	−0.240***	−0.287***
	(0.499)	(0.380)	(0.469)	(0.562)	(0.036)	(0.040)
EGI	0.631***	0.703***	0.539***	0.696***	0.425***	0.510***
	(0.122)	(0.156)	(0.163)	(0.242)	(0.096)	(0.116)
lnRGDP	2.336***	2.129***	2.771***	2.225***	1.951***	1.390**
	(0.804)	(0.551)	(0.810)	(0.784)	(0.563)	(0.690)
FD	0.102**	0.146**	0.147	0.245	0.048	0.039
	(0.048)	(0.069)	(0.106)	(0.295)	(0.041)	(0.032)
URB	0.125	0.185***	0.098	0.160***	0.988	0.156***
	(0.089)	(0.035)	(0.079)	(0.047)	(0.699)	(0.015)
SBDUM1	−0.143***	−0.713***	−2.286***	−2.165***	−2.463***	−2.762***
	(0.008)	(0.256)	(0.072)	(0.340)	(0.715)	(1.015)
SBDUM2	−0.105***	0.394***	−0.192**	1.195***	−0.985**	1.800**
	(0.030)	(0.120)	(0.078)	(0.322)	(0.495)	(0.891)
ECTt-1	−0.698***		−0.700***		−0.825***
	(0.115)		(0.136)		(0.256)
Constant		5.569***		−3.375		−18.198***
		(1.854)		(2.885)		(4.166)
Obvs.		45		45		45
Adj. R2		0.777		0.785		0.810
Diagnostic tests	Test stat.	p -value	Test stat.	p -value	Test stat.	p -value
LM	1.750	0.205	1.566	0.350	1.180	0.650
Heteroscedasticity	0.650	0.770	0.950	0.440	0.335	0.800
Ramsey-Reset	1.879	0.156	1.711	0.325	1.300	0.490
Jarque–Bera	3.130	0.112	3.553	0.215	3.301	0.165
CUSUM	Stable		Stable		Stable
CUSUMSQ	Stable		Stable		Stable

On the other hand, the results also indicate that the trade and financial globalization policies of Bangladesh are not favorable for the quality of the environment. This is because it can be witnessed that a 1% rise in the economic globalization index is associated with a rise in the levels of per capita CO2 emissions generated from combusting gas, oil, and coal on average by 0.631%, 0.539%, and 0.425%, respectively, in the short-run while reducing these emissions further by 0.703%, 0.696%, and 0.510%, respectively, in the long run. Hence, it can be clearly understood that the environmental adversities linked to higher degrees of economic globalization persistently go up with time. In the context of Bangladesh, this finding of economic globalization exerting negative environmental impacts is justified since it has been recognized in the literature that the environmental protection laws in Bangladesh are not strong enough to restrict environmental degradation (Rahman and Kashem 2017). In line with this notion, it can be expected that participation in international trade and FDI inflows foster the development of the unclean industries in Bangladesh; consequently, the expansion of these industries is likely to scale up the fossil fuel demand and, thereby, boost the CO2 emission levels as well. The finding of the positive correlation between the economic globalization index and disaggregated CO2 emission levels also validates the PHH in Bangladesh since a higher degree of economic globalization is synonymous with a rise in the influx of FDI. It is important to note that Islam et al. (2021) concluded that globalization is effective in curbing the total CO2 emission figures in Bangladesh. In contrast, the disaggregated fossil fuel consumption-based CO2 emission-stimulating effect of economic globalization found in this study implies that although globalization as a whole may be effective in improving Bangladesh’s environmental quality, the economic dimension of globalization does not facilitate the environmental development goals of the nation. Thus, the decision to evaluate the globalization-CO2 emissions nexus using disaggregated figures can be claimed to be justified from the perspective of designing comprehensive globalization policies in Bangladesh.

Among the other key findings reported in Table 4, the results certify the economic growth-environmental degradation trade-off scenario in Bangladesh. The short- and long-run elasticity estimates reveal that economic growth triggers higher emissions of fossil fuel consumption-based CO2 emissions. A rise in the per capita level of real GDP by 1% is evidenced to boost per capita CO2 emissions generated from combusting gas, oil, and coal on average by 2.336%, 2.771%, and 1.951%, respectively in the short run while reducing these emissions in the long run by 2.129%, 2.225%, and 1.390%. Therefore, considering the relatively smaller CO2 emission-boosting impacts in the long run, it can be said that the detrimental environmental impacts associated with economic growth are likely to decline over time. This trend is in line with the principles of the EKC hypothesis based on which this hypothesis can be claimed to hold in the context of Bangladesh. The EKC hypothesis, concerning total CO2 emissions, was also confirmed in the previous studies conducted by Murshed et al. (2021) for Bangladesh, Ahmed and Long (2013) for Pakistan, and Aydoğan and Vardar (2020) for the E7 countries. On the other hand, the results also indicate that financial development boosts CO2 emissions generated from the combustion of gas both in the short- and long run; however, financial development was evidenced to be ineffective in influencing the per capita levels of CO2 emissions stemming from oil and coal consumption. These findings indicate that as the private sector credit access is enhanced, it might trigger the electricity demand in Bangladesh. Since natural gas holds a lion’s share in the total electricity output of the country, the surge in electricity demand is likely to boost the gas consumption-based CO2 emissions as well. Pata (2018) also documented evidence of financial development being responsible for surging the total CO2 emission levels in Turkey. However, the findings of Khan, Hossain, and Chen (2021) can be linked with the results found in this paper regarding the ineffectiveness of financial development to affect CO2 emissions generated from using oil and coal.

Moreover, it can also be observed from the elasticity estimates reported in Table 4 that urbanization is hurting Bangladesh’s long-term environmental well-being. The corresponding elasticity outcomes show that a rise in the share of urban residents in Bangladesh’s total population does not generate any short-run impact on the nation’s disaggregated fossil fuel consumption-based CO2 emission figures. However, in the long run, urbanization boosts these emissions. These imply that unplanned urbanization within the Bangladesh economy is likely to substantially surge the electricity demand which, in turn, can be held responsible for the emission of higher CO2 emissions in the long run. Once again, the environmental adversities of urbanization can be associated with the fossil fuel dependency within Bangladesh’s energy sector whereby urbanization-driven energy demand is most likely to simultaneously trigger higher fossil fuel consumption-related CO2 emissions. Ali et al. (2019) and Mahmood et al. (2020) also found evidence of urbanization boosting the total CO2 emission figures of Pakistan and Saudi Arabia, respectively. It is to be noted that Bangladesh, Pakistan, and Saudi Arabia have all traditionally remained largely reliant on fossil fuels to meet their respective energy demand. As a result, the similarity of the findings is justified. Lastly, the Fourier BARDL results shown in Table 5 reveal that in most of the cases the predicted coefficients attached to the structural break dummies are statistically significant which rationalizes their inclusion in the respective model. Notably, the negative sign of the coefficient estimates in the context of the first structural break dummy for all models, which was identified to be around the early 1980s, implies that compared to other years, the disaggregated CO2 emission figures of Bangladesh are more likely to be lower. This is believable since during this period the surge in crude oil prices in the world market motivated the Bangladesh government to replace imported petroleum oils with locally sourced natural gas for electricity generation purposes. Since natural gas is a relatively cleaner fossil fuel, this energy switch within the power sector can be expected to be associated with lower fossil fuel consumption-based CO2 emissions in Bangladesh.

OPEN IN VIEWER

Table 5.

The Fourier BARDL Short- and Long-Run Results for Model 2.

Dep. Var.	FCO2 (gas)		FCO2 (oil)		FCO2 (coal)
Regressors	Short-run	Long-run	Short-run	Long-run	Short-run	Long-run
lnHELECT	−1.745***	−2.122**	−1.365***	−1.617***	−0.345***	−0.383***
	(0.306)	(0.910)	(0.455)	(0.439)	(0.110)	(0.139)
EGI	0.721***	0.738***	0.627***	0.704***	0.519***	0.559***
	(0.197)	(0.164)	(0.172)	(0.165)	(0.136)	(0.185)
lnHELECT*EGI	−0.088**	−0.092**	−0.016***	−0.027***	−0.028**	−0.035***
	(0.045)	(0.046)	(0.004)	(0.004)	(0.015)	(0.011)
lnRGDP	2.115**	2.010***	2.800***	2.645***	1.519***	1.350***
	(0.851)	(0.585)	(0.640)	(0.646)	(0.345)	(0.445)
FD	0.128***	0.165***	0.129	0.215	0.057	0.061
	(0.041)	(0.055)	(0.101)	(0.192)	(0.059)	(0.072)
URB	0.135	0.195***	0.121	0.174***	0.888	0.258***
	(0.089)	(0.036)	(0.088)	(0.050)	(0.705)	(0.045)
SBDUM1	−0.139***	−0.248***	−2.268***	−2.457***	−2.606***	−2.685***
	(0.029)	(0.032)	(0.038)	(0.510)	(0.755)	(0.975)
SBDUM2	−0.102	0.303	−0.173**	1.134***	−0.820***	1.625***
	(0.082)	(0.235)	(0.078)	(0.334)	(0.125)	(0.220)
ECTt-1	−0.558***		−0.735***		−0.815***
	(0.151)		(0.200)		(0.225)
Constant		5.165**		−3.522***		−19.790***
		(2.445)		(0.955)		(4.285)
Obvs.		45		45		45
Adj. R2		0.652		0.665		0.757
Diagnostic tests	Test stat.	p-value	Test stat.	p-value	Test stat.	p-value
LM	1.560	0.295	1.660	0.290	1.024	0.771
Heteroscedasticity	0.510	0.780	1.155	0.229	0.413	0.650
Ramsey-Reset	1.913	0.114	1.860	0.212	1.120	0.510
Jarque–Bera	2.920	0.133	3.121	0.152	2.991	0.190
CUSUM	Stable		Stable		Stable
CUSUMSQ	Stable		Stable		Stable

Table 4 also shows that the error-correction terms predicted for all cases are negative and statistically significant at the 1% level of significance. These imply that these models converge to a long-run equilibrium level at a speed of around 70%–82.5%. Besides, the predicted adjusted R-squared values are seen to be quite high which indicated that around 78%–81% of total variations in Bangladesh’s per capita fossil fuel consumption-based CO2 emission figures can be explained by positive shocks to the nation’s levels of per capita hydroelectricity consumption, economic globalization index, per capita real national income, financial development, and urbanization rate. As far as the diagnostic tests are concerned, the corresponding results certify that the Fourier BARDL models are free from serial correlation, model misspecification, heteroscedasticity, and non-normality problems. Furthermore, the stability of the parametric estimations in these models is affirmed by the findings from the CUSUM and CUSUMSQ tests. ²

Further, to assess the interactive effects of hydroelectricity consumption and economic globalization, we augmented Model 1 with the interaction term between these variables. Table 5 reports the Fourier BARDL model outcomes for Model 2. Overall, the predicted signs of the parameters presented in both Tables 5 and 6 are evidenced to be identical. However, in the context of Model 2, it is important to note that hydroelectricity consumption and economic globalization jointly help to reduce the disaggregated fossil fuel consumption-based CO2 emission figures of Bangladesh. Hence, it can be said that hydroelectricity consumption and economic globalization not only directly affect the emission levels but also exert indirect impacts by interacting with one another. This finding supports the hypothesis that Bangladesh can look forward to liberalizing the trade barriers to facilitate the process of importing hydroelectricity from neighboring nations like Bhutan and Pakistan. Besides, it is also possible that Bangladesh attracts clean FDI to develop the technologies to scale up hydroelectricity and other renewable electricity outputs. As a result, such forms of economic globalization would not only enhance renewable electricity production and consumption in Bangladesh but would also neutralize the adverse environmental impacts associated with traditional economic globalization activities. Besides, the diagnostic test results reported in Table 5 once again affirm that the Fourier BARDL models in the context of Model 2 do not have serial correlation, model misspecification, heteroscedasticity, and non-normality issues. Furthermore, the stability of the parameter estimates for Model 2 is confirmed by the outcomes from the CUSUM and CUSUMSQ tests. ³

OPEN IN VIEWER

Table 6.

The Results From the DOLS Method (Robustness Analysis).

Dep. Var.	Model 1			Model 2
	lnFCO2 (gas)	lnFCO2 (oil)	lnFCO2 (coal)	lnFCO2 (gas)	lnFCO2 (oil)	lnFCO2 (coal)
Regressors
lnHELECT	−2.555***	−1.657***	−0.395***	−2.240***	−1.797***	−0.423***
	(0.200)	(0.290)	(0.111)	(0.395)	(0.386)	(0.112)
EGI	0.860***	0.915***	0.566**	0.824***	0.862***	0.917***
	(0.145)	(0.217)	(0.285)	(0.106)	(0.113)	(0.227)
lnHELECT*EGI	—	—	—	−0.087***	−0.038***	−0.056***
				(0.005)	(0.009)	(0.016)
lnRGDP	2.860***	3.225***	1.990***	2.552***	2.600***	1.518***
	(0.500)	(0.680)	(0.657)	(0.501)	(0.399)	(0.555)
FD	0.203***	0.326	0.092	0.173**	0.221	0.088
	(0.023)	(0.278)	(0.085)	(0.030)	(0.182)	(0.078)
URB	0.210***	0.176***	0.157***	0.235***	0.216***	0.296***
	(0.025)	(0.055)	(0.037)	(0.016)	(0.022)	(0.082)
SBDUM1	−0.742***	0.419***	−0.305	−0.469***	−2.825***	−2.905***
	(0.155)	(0.0973)	(0.260)	(0.043)	(0.835)	(0.512)
SBDUM2	0.520***	0.487***	1.318***	0.424	1.702***	1.822***
	(0.135)	(0.0853)	(0.228)	(0.431)	(0.675)	(0.331)
Constant	12.612***	−4.380***	−33.740***	11.238***	−4.156***	−29.578***
	(2.555)	(1.288)	(5.165)	(0.635)	(1.211)	(7.422)

The robustness of the long-run findings across alternative regression methods is assessed using the DOLS estimator. The DOLS elasticity estimates, for both Models 1 and 2, are reported in Table 6. It can be observed that DOLS elasticity estimates have similar signs as the signs of the corresponding Fourier BARDL elasticity estimates. However, the sizes of the DOLS elasticity estimates are relatively larger. Hence, upon comparing the relative magnitudes, it can be said that the Fourier BARDL estimator corrects for the overestimate bias of the DOLS estimator. The capacity of the Fourier BARDL method to incorporate variables of mixed order of integration, whereas the DOLS technique requires all variables to be commonly integrated at the first difference, could be the reason behind the differences in the sizes of the elasticity estimates across these two alternative methods. In addition, since the Fourier BARDL technique can assign different lag lengths to the variables, which is not the case for the DOLS technique, it can account for the overestimate bias of the DOLS analysis outcomes. Therefore, the robustness of the long-run findings across alternative regression techniques could not be established. Nevertheless, considering the relative advantages of the Fourier BARDL approach, the elasticity estimates predicted using this technique can be assumed to be more valid.

Lastly, to support the regression outcomes and also identify scopes for comprehensive policy-making purposes, the causality analysis is conducted. Table 7 reports the findings from both the Fourier Toda–Yamamoto and Toda–Yamamoto causality exercises. Overall, it can be seen that the Fourier Toda–Yamamoto approach generates better causality outcomes which could be due to the superiority of this technique over the conventional one by Toda and Yamamoto (1995). Nevertheless, it can be observed that unidirectional causalities are extending from hydroelectricity consumption to all three fossil fuel-based CO2 consumption levels in the context of Bangladesh. Hence, these findings support the corresponding regression findings to further emphasize the pertinence of consuming cleaner energy resources for neutralizing the associated environmental quality-inhibiting effects. Besides, unidirectional causations from economic globalization to these emissions are also ascertained whereby, keeping the related regression outcomes into consideration, it is clear that the trade and external financing policies of Bangladesh are hurting the nation’s environmental well-being. Moreover, we find evidence of bidirectional causations between economic growth and fossil fuel consumption-based CO2 emissions. This is a key finding which implies that not only the current environmental consequences are conditional on the nature of the economic activities but also the future economic performances are likely to be affected by the quality of the environment. This makes sense since in the contemporary era sustaining economic progression is doubted without simultaneously facilitating environmental welfare. Lastly, although no causal association between financial development and fossil fuel-based CO2 emissions could be established, we find unidirectional causality running from urbanization to all three forms of fossil fuel consumption-based CO2 emissions. Therefore, in support of the corresponding regression outcomes, it can be claimed that urbanization within Bangladesh is not complementing the nation’s environmental development objectives.

OPEN IN VIEWER

Table 7.

The Results From the Causality Analysis.

Dep. Variable	Null hypothesis	Fourier Toda–Yamamoto test		Toda–Yamamoto test
		Test Stat.	Prob.	Test Stat.	Prob.
lnFCO2 (gas)	lnHELECT ≥ lnFCO2 (gas)	13.235***	0.007	6.145	0.125
	lnFCO2 (gas) ≥ lnHELECT	2.342	0.350	1.750	0.450
	EGI ≥ lnFCO2 (gas)	16.580***	0.002	7.245*	0.085
	lnFCO2 (gas) ≥ EGI	3.234	0.290	1.910	0.410
	lnRGDP ≥ lnFCO2 (gas)	11.245***	0.010	6.950*	0.091
	lnFCO2 (gas) ≥ lnRGDP	8.010**	0.045	3.135	0.150
	FD ≥ lnFCO2 (gas)	4.120	0.225	1.801	0.435
	lnFCO2 (gas) ≥ FD	3.133	0.350	1.922	0.411
	URB ≥ lnFCO2 (gas)	18.235***	0.001	6.501	0.119
	lnFCO2 (gas) ≥ URB	2.500	0.340	0.980	0.650
lnFCO2 (oil)	lnHELECT ≥ lnFCO2 (oil)	14.990***	0.005	7.224*	0.091
	lnFCO2 (oil) ≥ lnHELECT	3.225	0.310	2.035	0.300
	EGI ≥ lnFCO2 (oil)	12.400***	0.009	5.133	0.119
	lnFCO2 (oil) ≥ EGI	2.660	0.475	3.260	0.233
	lnRGDP ≥ lnFCO2 (oil)	12.434***	0.010	8.234**	0.042
	lnFCO2 (oil) ≥ lnRGDP	13.223***	0.008	5.113	0.231
	FD ≥ lnFCO2 (oil)	5.121	0.174	3.224	0.165
	lnFCO2 (oil) ≥ FD	5.400	0.171	2.119	0.230
	URB ≥ lnFCO2 (oil)	21.224***	0.000	7.500	0.075
	lnFCO2 (oil) ≥ URB	5.134	0.139	1.980	0.510
lnFCO2 (coal)	lnHELECT ≥ lnFCO2 (coal)	8.085**	0.045	5.800	0.154
	lnFCO2 (coal) ≥ lnHELECT	1.560	0.665	1.600	0.520
	EGI ≥ lnFCO2 (coal)	16.150***	0.009	1.350	0.660
	lnFCO2 (coal) ≥ EGI	1.450	0.302	0.075	0.790
	lnRGDP ≥ lnFCO2 (coal)	9.750**	0.031	5.850	0.113
	lnFCO2 (coal) ≥ lnRGDP	8.850**	0.042	3.850	0.200
	FD ≥ lnFCO2 (coal)	4.450	0.240	1.149	0.574
	lnFCO2 (coal) ≥ FD	3.301	0.282	2.334	0.600
	URB ≥ lnFCO2 (coal)	19.132***	0.001	6.904*	0.055
	lnFCO2 (coal) ≥ URB	4.700	0.214	2.113	0.453

Conclusion and Policy Recommendation

Bangladesh has traditionally pursued economic growth-enhancing policies without paying much heed to the environmental adversities that have accompanied the growth achievements. This is because much of Bangladesh’s economic expansion has been attained by utilizing both imported and locally sourced fossil fuels. However, in the contemporary era, Bangladesh is concerned regarding the aggravation of its environmental quality whereby the government has recently pledged at the COP26 to reduce CO2 emissions by more than one-fifth by the end of 2030. Thus, decarbonizing the economy has become imperative for the Bangladesh government especially by lessening the nation’s traditional fossil fuel dependency within the electricity sector, in particular, and greening its future globalization strategies. Against this backdrop, this study evaluated the effects of hydroelectricity consumption and economic globalization, controlling for economic growth, financial development, and urbanization, on disaggregated fossil fuel consumption-based CO2 emissions in Bangladesh over the 1972–2019 period. The results from the empirical analysis revealed long-run cointegrating relationships among CO2 emissions, generated specifically from combusting gas, oil, and coal, and the other macroeconomic variables of concern. Besides, hydroelectricity consumption was found to be associated with higher CO2 emissions while economic globalization was evidenced to induce CO2 emission-stimulating effects. However, both hydroelectricity consumption and economic globalization were found to jointly curb the disaggregated fossil fuel-based CO2 emission figures of Bangladesh. In addition, the findings validate the EKC hypothesis in the context of Bangladesh. Furthermore, financial development led was seen to boost CO2 emissions generated from the consumption of gaseous fuels. Lastly, the results indicated that urbanization boost all forms of fossil fuel consumption-based CO2 emissions only in the long run. Therefore, in line with these major findings, this study recommends several decarbonization policies for the Bangladesh government to consider.

Firstly, it is imperative for the nation to get over its fossil fuel dependency within its electricity sector by substantially enhancing the share of hydroelectricity and other types of renewable electricity in the total electricity output of Bangladesh. Achieving a renewable electricity transition would not only benefit the nation by curbing the fossil fuel consumption-related CO2 emissions but would also reduce the acute natural gas supply shortages faced by the nation to ensure energy security in Bangladesh. Therefore, overcoming the technological and financial constraints impeding renewable electricity transition in Bangladesh must be made a prioritized agenda of the government. This leads us to the second recommendation put forward in this study. Secondly, the traditional economic globalization policies pursued by Bangladesh need to be restructured. More precisely, Bangladesh should emphasize boosting international trade of relatively less pollution-intensive commodities and inhibit the influx of dirty FDI. In this regard, to undergo a renewable electricity transition, Bangladesh may think of boosting its intra-regional import of hydroelectricity from the neighboring hydroelectricity-surplus nations. Besides, liberalizing trade barriers imposed on primary inputs needed for generating solar power can also be an additional policy initiative. On the other hand, Bangladesh should also consider attracting clean FDI, especially to develop the latest technologies required for producing electricity using renewable energy sources. Accordingly, the fossil fuel consumption-based CO2 emissions associated with economic globalization can be expected to be controlled.

Thirdly, the nation must focus on adopting environmentally sustainable production and consumption policies to eliminate the traditional economic growth-environmental deterioration trade-off. This can be particularly achieved by greening the domestic financial sector of Bangladesh. In this regard, issuing green bonds can prove to be a mechanism for neutralizing the environmental adversities associated with financial development in Bangladesh. Besides, subsidizing the interest charged against credit extended to the private sectors, especially for financing green projects can further reduce the financial development-driven CO2 emission problems. Finally, aligning the urbanization policies with the decarbonization objectives of the nation is imperative to cut down the CO2 emission problems linked with unplanned urbanization in Bangladesh. In this regard, meeting the urban energy demand with renewable electricity is critically important which further justifies the importance of undergoing a renewable electricity transition in Bangladesh. Successful adoption and execution of these key environmental development policies can be expected to assist Bangladesh in significantly decarbonizing its economy.

The key limitation faced in this study was the lack of availability of sectoral CO2 emission figures for Bangladesh whereby a sector-specific analysis could not be done. Consequently, the findings and policy recommendations are more relevant for all sectors collectively but may not be homogeneously valid for each of the sectors of Bangladesh. Future studies can look to conduct extensive work in this area so that the findings from this study can be compared for further robustness checks.