Foreign Direct Investments, Renewable Electricity Output, and Ecological Footprints: Do Financial Globalization Facilitate Renewable Energy Transition and Environmental Welfare in Bangladesh?

In the context of developing nations, financial constraint is a major growth-inhibiting concern. At the same time, technological limitations and infrastructural backwardness limit the prospects of undergoing RET and safeguarding environmental sustainability in tandem. Under such circumstances, attracting FDI can be expected to lessen the financial constraints and simultaneously overcome the other barriers inhibiting RET and environmental development (Murshed et al., 2021b; Liu et al., 2016). Besides, mitigating CO₂ and other harmful emissions has become a global agenda following the ratification of the Paris Agreement (Chaudhry et al., 2020) by several of the world economies including Bangladesh. Thus, in this regard, is it important to reduce the use of fossil fuels and significantly enhance renewable energy use.

Renewable electricity generation is a relatively technology-intensive process. Thus, technological innovation is a pre-requisite to initiate the RET phenomenon (Ben Jebli & Kahia, 2020; Dagoumas & Koltsaklis, 2019; Eleftheriadis & Anagnostopoulou, 2015). However, the poor state of technology in the majority of the developing nations, including Bangladesh, often bottlenecks the prospects of phasing out the dependency on fossil fuel consumption to utilize renewable energy alternatives (Murshed et al., 2020b; Murshed, 2020a). Besides, the generation, transmission, and distribution of electricity generated from renewable resources require the energy infrastructure to be advanced which, once again, is a RET-inhibiting constraint faced by the developing nations. These limitations can be expected to be overcomed through a knowledge spillover impact associated with globalization, especially in the form of FDI inflows (Padhan et al., 2020; Hoang et al., 2021). In reference to the absorptive capacity hypothesis (Navas, 2019), FDI inflows can exert a positive spillover impact to promote renewable energy adoption within the host economies. Consequently, FDI inflows can induce renewable energy augmenting effect by discouraging the use of non-renewable fossil fuels in the host economies (Narayan and Doytch, 2016). At the same time, incoming FDI can also be expected to foster the development of the host economy’s renewable energy sector growth by upgrading the poor energy infrastructure (Ji & Zhang, 2019). Conversely, FDI inflows, especially the dirty FDI, can also exert dampening impacts on REO. This is because the relatively unclean FDI is often targeted at economies that are predominantly fossil fuel-intensive. Under such circumstances, such FDI inflows can boost non-renewable energy use to hinder the RET phenomenon across the host nations (Cole et al., 2006; Shahbaz et al., 2019). Hence, it is ideal for the developing nations to attract clean FDI, especially for the development of their respective renewable energy sector.

On the other hand, it is often recommended to pursue green economic and financial policies to ensure environmental welfare (Karyawati et al., 2020; Kovilage, 2020). In line with this notion, FDI inflows have been acknowledged to exert both positive and negative environmental impacts on the host nations. The adverse environmental effects associated with FDI inflows is explained using the theoretical underpinnings of the ‘pollution haven hypothesis’ (Nasil et al., 2019; Doytch, 2020). This hypothesis asserts that inflows of dirty FDI tend to boost the use of unclean energy resources which, in turn, stimulate the emission of greenhouse gases to degrade the overall quality of the environment in the host economies (Solarin et al., 2017). In contrast, the ‘pollution halo hypothesis’ postulates in favor of FDI inflows exerting positive environmental impacts within the host economies (Liu & Xu, 2021). In this regard, clean FDI is hypothesized to green the production and consumption processes within the host nations to improve the quality of the environment. Furthermore, since renewable energy use is believed to be inextricably linked to environmental development (Dogan et al., 2020; Nathaniel et al., 2020; Adedoyin et al., 2020; Sharma et al., 2021), a joint environmental impact of FDI inflows and renewable energy consumption can also be expected (Zhang & Zhou, 2016).

Although most of the preceding empirical studies have suggested the application of the conventional unit root tests, including the Augmented Dickey-Fuller Dickey and Fuller (1979) and Phillips-Perron (1988) unit root testing methods, these methods generate biased estimates of the stationarity properties in the case of potential Structural Break (SB) issues in the data (Perron, 1989). Therefore, to counter the limitation of these techniques, this study uses the recently developed unit root analyses proposed by Narayan and Popp (2010). The Narayan-Popp (NP) method accounts for up to two SB in the data and generates unbiased test statistics to evaluate the order of integration among the variables included in the respective models. The NP technique predicts the stationarity properties using two model specifications. In one of the models, the two SB are assumed to be in the level while in the other model the two SB are assumed to be in the level and slope of a trending series (Narayan & Popp, 2010). For robustness check, the Clemente-Monantes-Reyes (CMR) unit root test (Clemenete et al., 1998), which also accounts for two SB in the data, is also applied. The test statistics under both the NP and CMR approaches are estimated under the null hypothesis on non-stationarity against the alternative hypothesis of stationarity; thus, the statistical significance of the test statistic confirms the stationarity of the corresponding variable. The unit root investigations are followed by the cointegration exercises.

Cointegration exercises are performed to check the long-run associations between the variables included in the respective empirical model. The existence of long-run cointegrating relationships between the variables is a pre-requisite to estimating the long-run elasticities using appropriate regression techniques. This study primarily follows the Bayer-Hanck (BH) combined test for non-cointegration proposed by Bayer and Hanck (2013). The BH method prerequisites the variables to be commonly integrated in their respective first difference, I(1). This particular technique corrects for the limitations of the conventionally used Engle and Granger (1987) cointegration analysis due to being independent of certain nuisance parameters (Bayer & Hanck, 2013). Moreover, the BH approach provides efficient estimates of the cointegrating properties in the context of small samples. Besides, different cointegration tests are believed to provide dissimilar estimates of cointegration whereby generalizing the cointegration findings becomes a challenge. Thus, Bayer and Hanck (2013) came up with a solution to this problem by proposing a combined cointegration analysis which involves estimation of a Fisher statistic based on the following formula:where P refers to the probability values (p values) of individual cointegration tests; EG-J refers to the p values of the Fisher statistic for the combined Engle-Granger (EG) and Johansen (J) tests while EG-J-BO-BD refers to the p values of the Fisher statistic for the combined EG, J, Boswijik (1994) (B) and Banerjee, Dolado and Mestre (1998) (BDM) tests. The statistical significance of the Fisher statistic rejects the null hypothesis of no cointegration.

However, the BH method does not take into consideration the SB issues. Hence, this study also applies the Maki (2012) cointegration approach that can account for up to five SB in the data. There are four specific tests under the Maki cointegration approach: A (assumes the breaks in the intercept only), B (assumes the breaks in the intercepts and coefficients without the trend, C (assumes the break only in the intercept and coefficients but the model is assumed to have a trend) and D (assumes the breaks in the intercepts, coefficients and trend). The model specifications for these four tests can be shown as:

where y is the dependent variable and x is a vector of the independent variables. t denotes the time period, m denotes the highest number of breaks in the model which can take a maximum value of 5 and u denotes the error term. D_i,t is the dummy variable that is used to signal the presence of a break in the data. D_i,t takes a value of 1, denoting the existence of the SBs at specific break years T_bi (i = 1, …, m), if t_i are after the break years T_bi (i.e. t_i > T_bi) and a value of 0, denoting no SB at the specific break years T_bi (i = 1, …, m), if t_i are before the break years T_bi (i.e. if t_i < T_bi). Statistical significance of the test statistics in each of the four models (A, B, C, and D) rejects the null hypothesis of no cointegration to confirm the presence of cointegrating associations between the variables in the respective models. The identified SB from test A, within the Maki cointegration approach, is used to create SB dummy variables and included into the respective empirical models to control for time fixed effects. The regression analysis follows the cointegration exercises.

Following Hassan et al. (2019), the short and long-run elasticities are estimated using the Autoregressive Distributed Lag (ARDL) technique introduced by Pesaran et al. (2001). This regression model is chosen in the context of this study due to several of its beneficial features: (a) it can accommodate a mixed order of integration among the variables included in the respective model and generates unbiased elasticity estimates compared to the other estimation techniques that pre-requisite a common order of integration among the variables; (b) it can predict both the short- and long-run elasticities unlike the other conventionally used estimators which primarily focuses on the long-run effects only; (c) this technique is efficient in accounting for the possibility of endogeneity and serial correlation problems in the data; and (d) it is the most efficient estimator for handling data sets with short time dimensions. However, a limitation of this procedure is its inability to accommodate the SB in the data. Thus, the inclusion of the SB year dummies into the econometric models helps to resolve this problem. The ARDL approach is a two-step procedure. Firstly, it involves the estimation of an unrestricted Error-Correction Model (ECM). The ECM model in the context of model (1) can be specified as:where ∆ denotes first difference form of the corresponding variables. The short run dynamics for model (1) is observed from the model below:where ECT_t-1 represents the one-period lagged Error-Correction Term (ECT) which is estimated from the residuals of the long-run equation shown in Eq. (11). The ECT reflects the speed of adjustment of the variables in restoring long-run equilibrium following a deviation in the earlier period. The long-run elasticities are sourced from the model below:

Similarly, in the context of model (2) the ARDL model can be expressed as:

The short run dynamics for model (2) is observed from the model below:

The long-run elasticities for model (2) are sourced from the model below:

The stability of the ARDL estimates are checked using the Breusch-Godfrey Lagrange Multiplier test for Autocorrelation (BGodfrey) proposed by Breusch and Godfrey (1981), the Lagrange Multiplier test for the presence of Autoregressive Conditional Heteroskedasticity (ARCH) proposed by Engle (1982), and the Cumulative Sum (CUSUM) and Cumulative Sum of Squares (CUSUMSQ) methods proposed by Chow (1960) and Brown et al. (1975), respectively.

The causal associations between the variables are evaluated using the Hacker and Hatemi-J (HH) causality estimation technique proposed by Hacker and Hatemi-J (2012). This method is modified from the bootstrapped causality test proposed by Hacker and Hatemi-J (2006). According to Hacker and Hatemi-J (2006), bootstrapping the distribution reduces the concerns from small sample size distortions associated with the conventional Wald test introduced by Toda and Yamamoto (TY) (1995), irrespective of the presence of autoregressive conditional heteroscedasticity effects within the model. The modification of the modified Wald test statistic of Hacker and Hatemi-J (2006) is done by endogenizing the optimal lag selection criterion which tends to reduce the small sample size distortions further. The HH technique uses a Vector Autoregressive (VAR) model to calculate the modified Wald test statistics under the null hypothesis of no causality between the dependent and the independent variables against the alternative hypothesis of otherwise. The bootstrapping involves two stages: firstly, estimating the optimal lag structure and secondly predicting the Wald statistic for investing the Granger causality. The VAR model of order k can be specified as:where y_t, β₀ and u_t are vectors with dimension n × 1 and B_i (i > 0) is a parameter matric with a dimension of n × n. The error term u_t has no expected value and presumed to be independent and identically distributed with a non-singular covariance matrix (Hacker & Hatemi-J, 2012). In the first stage of the bootstrapping approach under the HH approach, Eq. (12) is estimated without imposing any restriction in terms of the non-causality null hypothesis. The predicted value, y*, can be given by:where \(\widehat{{{\mathbf{u}}*}}_{{\mathbf{t}}}\) is a vector of bootstrapped error terms and t (t = 1, …, T) is the time period. The set of T bootstrapped error term vectors is estimated by drawing randomly with replacement from the vectors of the modified residual to ensure that the mean value of the bootstrapped error term vectors is a zero vector. The modified residuals are the raw residuals modified via leverages within the HH approach. This modification is ideal to deal with heteroscedasticity issues within the model and also to account for the ARCH effects (Hacker and Hatemi, 2012). The bootstrapping mechanism is repeated M times to produce a Wald statistic each time which is based on the TY methodology. The resulting set of bootstrapped Wald statistics is termed as the bootstrapped Wald distribution which is then used to evaluate the causal properties of the pair of variables. In the case of the predicted Wald –statistic being greater than the corresponding bootstrapped critical value then causality stemming from the independent to the dependent variable is affirmed via rejection of the null hypothesis of non-causality. For comparison purposes, the TY causality technique is also applied.

The NP and CMR unit root test results are reported in Table 1. The results reveal that all the variables are non-stationary at their respective level forms but they become stationary at their first differences, irrespective of considering the SB in level or both level and slope. Besides, the identical results from the CMR unit root analysis confirm the robustness of the stationarity properties of the variables. Hence, the unit root results implicate a common order of integration among the variables, at first difference. Therefore, all the variables included in the empirical analyses are mean-reverting which nullifies the possibility of the regression analysis to follow being spurious.

Table 2 presents the results from cointegration analysis. In the context of the BH test, the statistical significance of the EG-J-BO-BDM test statistics, at 1% level, of the estimated test statistics confirms cointegrating associations between the variables of concern in the respective models. Besides, the statistical significance, at 1% and 5% levels, of the test statistics under the Maki (2012) approach also affirm the cointegrating relationships; thus, the results can be considered robust across different estimation techniques. In line with these results, it can be held that FDI inflows, REO shares, ecological footprints and the relevant control variables have long-run associations. These findings fulfill the pre-requisite to predicting the long-run elasticities. Moreover, the results from the Maki cointegration analysis also spit out the locations of five structural break points for each of the two empirical models. Accordingly, the identified break points, from test A under the Maki cointegration approach, are used to create break year dummies for inclusion into the respective models in order to accommodate these data issues within the regression analysis.

The ARDL short- and long-run elasticity estimates in the context of model (1) are reported in Table 3. The short-run elasticity estimates, accounting for the structural breaks in the data, reveal that incoming FDIs in Bangladesh exert a dampening impact on the nation’s REO shares. This can be perceived from the negative signs and statistical significance of the predicted elasticity parameters attached to current and lagged levels of FDI inflows. Hence, incoming FDI is found to be ineffective in facilitating the RET phenomenon in Bangladesh in the short-run. However, it is to be noted that the adverse impacts of incoming FDI on the REO shares tend to decline with time since the elasticity estimate at the current level of FDI inflows is relatively smaller than that of the corresponding elasticities at the lagged forms of FDI inflows. These imply that although FDI inflows initially depress the REO shares, it goes on to improve the quality of the environment as the volume of FDI inflows tend to persistently go up. Hence, it can be stated that the quality of the FDI flowing into Bangladesh possibly improves with time whereby the marginal negative impacts on the REO shares tend to diminish. Besides, a particular reason behind incoming FDI undermining the shares of renewable electricity in the aggregate electricity output figures of Bangladesh could be because of the fact that the FDI flowing into the country are predominantly directed at industries that are relatively more intensive in the use of non-renewable electricity; thus, marginalizing the overall REO shares. These short-run findings are parallel to the negative FDI inflow-REO nexus found by Lin and Li (2015) for China while contradicting the assertions made by Lin et al. (2016) where the authors failed to establish any statistically significant relationship between these variables in the case of China. Similarly, Kilicarslan (2019) also opined in favor of FDI inflows in the short-run being unable to explain the variations in the REO figures in Brazil, Russia, India, China South Africa, and Turkey.

Among the other short-run determinants of REO in Bangladesh, technological advancement is found to increase the REO shares. This is evident from the statistical significance of the negative elasticity parameter attached to the current level of energy use intensity. This negative correlation implies that as the intensity of energy use decreases, which is synonymous with technological advancement, the REO shares are likely to go up. Hence, it can be said that technological advancement is indeed a key determinant of higher REO in Bangladesh. A 1% decline in the energy intensity level is predicted to elevate the REO shares in the short-run by 0.21%, on average, ceteris paribus. Murshed (2021d) also found similar results and quoted that technological innovation-led energy efficiency improvements govern renewable electricity transition across South Asia.

Besides, economic growth is found to favor the use of electricity generated from the non-renewable energy resources, thus, reducing the REO shares. However, much like the case in the context of FDI inflows, economic growth is also evidenced to lower the REO shares. A percentage change in the one-period lagged level of the real per capita GDP of Bangladesh reduces the REO shares by 4.73% whereas a percentage change in the current level of real per capita GDP reduces the REO shares by 2.135%, on average, ceteris paribus. Hence, it can be asserted that higher levels of economic growth tend to empower the Bangladesh economy to gradually overcome the constraints that inhibit renewable electricity production. A similar negative, but statistically insignificant, short-run correlation between per capita real GDP and aggregate renewable energy use was reported by Fan and Hao (2020) in the context of 31 Chinese provinces. Likewise economic growth, the short-run REO-inhibiting impacts of domestic capital investments in the Bangladesh economy are witnessed. A percentage change in the current and one-period lagged levels of gross fixed capital formations in Bangladesh is found to depress the REO shares by 4.28% and 4.19%, on average, ceteris paribus. Furthermore, involvement in international trade is also found to suppress the REO shares which can be rationalized by the claim that the export sector of Bangladesh is predominantly dependent on the use of non-renewable electric power. Bangladesh relies heavily on its ready-made garments exports due to pursuing an export-led growth strategy (Shafiullah and Navatnam, 2016). However, the associated industries overwhelmingly intensive in the use of non-renewable electricity (Paul et al., 2017) whereby higher openness to trade justifiably dampens the REO shares. The short-run elasticity in this regard shows that a percentage increase in the trade openness index is associated with a decline in the REO shares by 0.92%, on average, ceteris paribus. This finding opposes the statistically insignificant short-run trade openness-REO nexus found by Khraief et al. (2018) in the context of Algeria.

In contrast, exogenous positive shocks to real crude oil prices in the international oil markets are predicted to stimulate a substitution effect that can plausibly be linked to lower use of imported oils for electricity generation purposes in Bangladesh. The positive signs of the statistically significant elasticity parameters attached to the current and lagged forms of the real crude oil price variable affirm this claim. Notably, the magnitude of the elasticity parameter attached to the current level of crude oil price is relatively higher than that attached to the lagged levels of crude oil prices. These imply that transitioning from the use of non-renewable to renewable resources for electricity generation is relatively difficult as the real oil price starts to rise. However, persistent rises in the oil prices with time gradually reduces the oil-dependencies which, in turn, can also be linked to increments in the REO shares, simultaneously. The initial detrimental impacts of such imported oil-dependency on the REO shares were also reported by Murshed and Tanha (2020) for four South Asian net oil-importing nations. The finding of the negative correlation between oil price and REO is in line with the findings put forward by Shahzad et al. (2021) from the understanding that the authors claimed that higher crude oil prices dampen the demand for overall energy in the newly industrialized fossil fuel-intensive countries. Consequently, as in the case of Bangladesh, it can be hypothesized that positive oil price shocks could possibly make way for greater renewable energy utilization in those newly industrialized countries as well. Finally, the short-run elasticity estimates reveal no correlation between CO₂ emissions and REO shares in Bangladesh. Furthermore, the negative sign and statistical significance of the estimated lagged ECT shows that any deviation from the long-run equilibrium is corrected at a rate of 78.1% in the next period.

As far as the long-run elasticity estimates are concerned, it can be seen that FDI inflows despite dampening the REO shares in the short-run, tend to marginally elevate the shares in the long-run. A percentage rise in the inflows of FDI is predicted to enhance the REO shares by 0.05%, on average, ceteris paribus. Therefore, it can be said that in the long-run foreign finance in the form of FDI mitigates the constraints to producing power from renewables on a mass scale and also inhibits non-renewable electricity generations in Bangladesh. This could be envisioned as the technological spillover effects of FDI on the energy sector of Bangladesh which could be effective in developing renewable electricity in the national energy mix. This finding matches the conclusions made by Azam et al. (2015) for Southeast Asian countries. These studies have asserted that incoming FDI promotes the use of clean energy resources which can be linked to higher REO in the long-run. In contrast, Lin et al. (2016) highlighted the adverse long-run impacts of inward FDI on the REO shares of China. Thus, in line with both the short and long-run estimates of the FDI inflow elasticities of REO shares, it is recommended that the government attracts clean FDI into the relatively greener industries in Bangladesh. More importantly, channeling FDI towards the energy sector with the uplifting the quality of the nation’s energy infrastructure could be expected to amplify the REO shares further.

On the other hand, technological advancement, as indicated by lower energy intensity levels, is also found to enhance the REO shares in the long-run. Hence, it can be asserted that the positive impacts of technological progress on the prospects of enhancing the REO shares in the short-run are sustained over the long-run as well. A percentage fall in the energy intensity level is predicted to account for 0.82% higher shares of renewable electricity in aggregate electricity outputs of Bangladesh. These results match the similar opinions put forward by Murshed (2021d) in the context of Bangladesh and five other South Asian nations. Moreover, the relatively higher magnitude of the long-run elasticity estimate, in comparison to that in the short-run, implicates that persistent advancement of the technological stock can be asserted to progressively enhance the REO shares as well. Thus, it is ideal to invest heavily in research and development purposes as a means of financing technological innovation in Bangladesh. Simultaneously, attracting FDI towards the energy sector can also be a potential mechanism to catalyze the rate of technological advancement in the economy; thus, intensifying the REO shares further.

Other results reveal that the REO-inhibiting impacts of economic growth in the short-run are reduced in the long-run which further certifies that economic empowerment, through increments in the national income levels, helps to gradually overcome the barriers that uphold mass-scale production of renewable electricity in Bangladesh. However, economic growth still does not ensure higher shares of REO in the country which is evident from the negative signing of the corresponding long-run elasticity estimate. A percentage rise in the real per capita GDP figures causes the long-run REO shares by 0.15%, on average, ceteris paribus. This finding opposes the findings by Murshed and Tanha (2020) in which the authors opined in favor of economic growth stimulating REO shares in the context of a panel of four South Asian net oil-importing nations. On the other hand, domestic capital investments unlike the case in the short-run are found to be incapable of explaining the variations in Bangladesh's long-run shares of REO. The statistical insignificance of the predicted elasticity parameter attached to GCF affirms this claim. Besides, the long-run elasticity estimates also show that international trade, in comparison to the short-run scenario, dampens the REO shares more in the long-run. A rise in the trade openness index by 1% is associated with a fall in the REO shares by 1.58%, on average, ceteris paribus. Hence, it is recommended that the Bangladesh government revisits its foreign trade policies and adopt appropriate policy measures to downsize the trade of goods and services that embody the use of electricity generated from conventional non-renewable energy resources. Simultaneously, the government is expected to incentivize the exporting industries, in particular, to independently generate a certain amount of electricity using renewable resources rather than solely being dependent on in-grid fossil fuel-fired electricity supplies. This finding is comparable to the remarks by Murshed (2020b) in the context of the lower-middle-income countries including Bangladesh; the author claimed that higher openness to international trade reduced the renewable energy consumption shares.

The long-run impacts of oil price shocks are similar to the corresponding short-run impacts. A 1% rise in the real crude oil prices are seen to enhance the REO shares by 4.3% on average which, in comparison to the comparatively lower impacts in the short-run scenario, implies that persistent positive shocks to oil prices in the international markets facilitate the replacement of fossil fuels by the renewable alternatives concerning electricity generation purposes in Bangladesh. This particular result condemns the findings put forward by Murshed and Tanha (2020) in which the authors, using panel data estimation methods, found that increments in crude oil prices monotonically dampened the REO shares in Bangladesh, India, Pakistan and Sri Lanka. In line with both the short and long-run estimates of the oil price elasticity of REO found in this study, it can be claimed that substantial hikes in the world crude oil prices would ultimately eliminate Bangladesh’s imported oil-dependency for electricity generation purposes; thus possibly elevating the nations REO shares in the future. This is a key finding in the sense that oil prices in the international markets have currently rock bottomed, and turned negative, courtesy of the global coronavirus (COVID-19) pandemic which, in turn, could trigger greater imports of crude oil at extremely low prices. Consequently, the nation's dismal REO shares could well be at stake of declining further in the post-pandemic period. Hence, the government has to keep this concerning issue into consideration and adopt appropriate to refrain from importing crude oils in bulk. Finally, the long-run elasticity estimates also reveal that CO2 emission, although not being able to enforce the transition from non-renewable to renewable electricity generation in Bangladesh, is capable of elevating the REO shares in the long-run. The corresponding statistically significant elasticity estimate suggests that a 1% rise in the per capita CO2 emissions, in the long-run, is associated with a 0.49% rise in the REO shares, on average, ceteris paribus. Hence, it can be said that apprehensions concerning the CO2 emissions-induced climate change adversities could spark urgency for the government to incentivize mass-scale production of electricity using the environmentally-friendly primary renewable energy inputs; hence, gradually uplift the nation’s REO shares in the aggregate electricity outputs.

The short and long-run elasticity estimates, from the ARDL approach, in the context of model (2) are reported in Table 4. The short-run elasticities indicate that inflow of FDI adversely impacts the environmental quality in Bangladesh; hence the results verify the authenticity of the pollution haven hypothesis. The statistical significance and positive signs of the predicted elasticity parameters attached to the current and lagged levels of FDI certify this claim. However, the negative environmental impacts are seen to cease with time since the magnitude of the elasticity estimates at the current level of FDI inflows is relatively smaller than that at the lagged levels of FDI inflows. Therefore, once again it can be said that, with time, the relatively cleaner FDI flow into the Bangladesh economy; thus, the marginal increments in the ecological footprints tend to decline simultaneously. A similar positive correlation between FDI inflow and ecological footprints was highlighted in the context of the United States by Zafar et al. (2019). Besides, the short-run elasticity estimates confirm the authenticity of the EKC hypothesis to validate the inverted-U-shaped association between economic growth and ecological footprints in Bangladesh. The positive and negative signs of the elasticity parameters attached to the current and lagged levels of per capita GDP and its squared term, respectively affirm this claim. These findings imply that at the initial stages of economic growth, there is a trade-off between economic and environmental welfare which seems to diminish at higher per capita GDP levels. The short-run validation of the EKC hypothesis was also reported in the study by Hassan et al. (2019) for Pakistan. Among the other short-run determinants of ecological footprints in Bangladesh, urbanization is found to dampen environmental quality in the short-run. A percentage increase in the urbanization rate at its current and one-period lagged levels boosts the ecological footprints figures by 0.34% and 0.60%, on average, ceteris paribus. Thus, these results provide statistical support to the unplanned urbanization-induced environmental woes of Bangladesh. Nathaniel et al. (2019) also found a similar short-run association between urbanization and ecological footprints in the context of South Africa. Besides, the negative sign and statistical significance of the estimated lagged ECT in the context of model (2) shows that any deviation from the long-run equilibrium is corrected at a rate of 65.9% in the next period.

On the other hand, the long-run elasticity estimates reported in Table 4 shows that the negative impacts of FDI inflows on the environment in Bangladesh are sustained in the long-run. The statistical significance and positive sign of the long-run elasticity parameter attached to FDI inflows affirm this claim. A rise in the volume of FDI inflows by 1% in the long-run is seen to increase the ecological footprints figures on average by 0.27%, ceteris paribus. Therefore, it can be said that the ‘pollution haven hypothesis’ is a long-term problem for the Bangladesh economy. However, it is to be noted that the magnitude of the long-run elasticity is relatively smaller than the corresponding short-run elasticities which, to some extent, implies that FDI inflows tend to have low technological spillover effects whereby the positive impacts on the environment are not so pronounced; rather the damages are slightly less. This is a concerning finding in the context of Bangladesh which warrants restructuring of the nation’s foreign financing and financial globalization policies. The nation is better-off attracting cleaner FDI and restricting the inflows of the relatively dirtier ones. This finding of the adverse environmental impact of FDI inflow in Bangladesh corroborates the conclusions made in the study by Doytch (2020) in which the author claimed the ‘pollution haven hypothesis’ usually hold for the developing countries since the dirty FDIs tend to flow into these countries to exploit their weak environmental laws; this scenario is pretty similar to the case of Bangladesh where little emphasis is put on enforcing strict environmental acts to safeguard the environmental attributes. The long-run validation of the pollution haven hypothesis was also put forward by Khan et al. (2019) for 54 BRI countries including Bangladesh. Besides, the long-run elasticity estimates also confirm the validity of the EKC hypothesis for Bangladesh. The corresponding elasticity estimates reveal that initially a 1% rise in the long-run per capita GDP figures is associated with a rise in the ecological footprints by 41.64%, but in the latter stages of growth, the marginal effect seems to reduce the ecological footprints by 3.05% on average, ceteris paribus. Hence, it can be said that economic growth is both the cause and the long-run solution to the environmental problems of Bangladesh. The results are parallel to the conclusions made by Altıntaş and Kassouri (2020) and Destek and Sarkodie (2019) for 14 European and 11 industrialized economies, respectively; while these results contradict the assertions made by Ozcan et al. (2018) in the case of Turkey.

More importantly, the long-run analysis shows that changes in the REO shares determine environmental well-being in Bangladesh. A percentage rise in the REO shares in the aggregate electricity outputs of Bangladesh is found to reduce the ecological footprints by 1.65%, on average, ceteris paribus. Hence it can be said that RET, as indicated by a rise in the share of renewables in total energy consumption volumes, is a long-run phenomenon that, although is ineffective in improving environmental quality in the short-run, safeguards the long-run environmental sustainability goals of Bangladesh. This finding is comparable to the claims made by Ulucak and Khan (2020) where the authors opined that renewable energy use, natural resource rent, and urbanization curb ecological footprints in Brazil, Russia, India, China, and South Africa.

Furthermore, exogenous positive shocks to world crude oil prices are found to reduce the ecological footprints of Bangladesh in the long-run. A percentage increase in the real prices of crude oil is found to curb the ecological footprints by 0.19%, on average, ceteris paribus. A plausible explanation to this finding could be put forward in the sense that rising crude oil prices are likely to induce the RET phenomena, also supported by the corresponding elasticity estimates found in the context of model (1), whereby replacing the conventionally consumed fuels by the renewable alternatives can mitigate environmental deterioration in Bangladesh. Similar conclusions were made by Murshed and Tanha (2020) in the context of four South Asian economies including Bangladesh. Finally, the long-run estimates also certify that the short-run adverse impacts of unplanned urbanization on the environmental quality of Bangladesh are sustained over the long-run as well. This is parallel to the finding by Shahzad et al. (2020) for the developing countries and by Nathaniel et al. (2019) for South Africa.

Therefore, the overall findings from the regression analysis are in line with the three hypotheses put forward in this study. As a result, it can be claimed that FDI inflows play a key role in initiating the RET phenomena in Bangladesh but do not guarantee environmental improvement as a whole. A possible explanation behind these inconsistent findings could be the fact that environmental degradation in Bangladesh is not merely confined in terms of air pollution. Rather multiple aspects collectively contribute to deteriorate the nation's environmental quality. Hence, these findings once again highlight the multidimensionality of environmental problems faced by Bangladesh. Since RET is directly concerned with the mitigation of greenhouse gas emissions, it may not be sufficient to reduce the other forms of environmental hardships. Hence, policies should be undertaken to address this issue and try to utilize the foreign funds to also tackle the other dimensions of environmental adversity in Bangladesh. Moreover, since the hypothesis regarding a joint environmental impact of FDI inflows and REO was verified, Bangladesh must attract renewable energy technology development-related FDI. Consequently, both the energy security and environmental sustainability issues can be accounted for in tandem.

The findings from the diagnostic tests, as shown in Tables 2 and 3 for model (1) and model (2) respectively, suggest that both the regression models considered in this study are not subject to autocorrelation and heteroscedasticity issues. Moreover, the stability of ARDL-ECM elasticity estimates is confirmed by the CUSUM and CUSUMSQ charts.² For robustness check, the long-run elasticities are re-estimated using the Fully Modified Ordinary Least Squares (Phillips & Hansen, 1990) and the Dynamic Ordinary Least Squares (Stock & Watson, 1993) regression techniques. The corresponding results, reported in Table 9 in the appendix, conform to the ARDL long-run elasticity estimates in terms of predicted signs; thus the robustness of the long-run elasticity estimates across different regression methods is affirmed. The causality investigations follow the regression analysis.

Table 5 reports the results from the causality analysis for both model (1) and (2). In general, the results denote robustness across different causality estimation techniques which can be perceived to the identical causality estimates found from both the HH (2012) and TY (1995) tests. In the context of model (1), a unidirectional causality is found to be running from FDI inflows to REO shares in the long-run. Hence, in line with the corresponding elasticity estimates in the context of model (1), it can be said that attracting clean FDI into the Bangladesh economy can be effective in facilitating the RET process; thus, the nation's predominant reliance on fossil fuels for power generation purposes can gradually be phased out to integrate renewable electricity into the energy sector of Bangladesh. Therefore, channeling foreign investment funds towards the energy sector, particularly for the development of renewable power plants, should be a prioritized policy agenda of the government. The results are parallel to the findings in the study by Ahmad et al. (2019) for China. Besides, energy intensity levels are also found to causally influence the REO shares without feedback. Hence, it is once again assured that technological innovation, through a reduction in the intensity of energy use, is a pre-requisite to undergoing RET in Bangladesh. Song et al. (2019) also emphasized on the importance of technological progress for green innovation and sustainable resource management within the economy. This finding also implicates that FDI inflow-led technological innovation in Bangladesh can further enhance the nation’s REO shares. Among the other causal impacts concerning the REO shares in Bangladesh, the statistical significance of the test statistics certify between REO shares and per capita GDO, between REO shares and real crude oil prices, and between REO shares and CO₂ emissions. These, in line with the corresponding elasticity findings, collectively imply that not only do rising national income level, crude oil prices and CO₂ emissions facilitate the RET phenomenon, the higher shares of renewable electricity in the aggregate electricity outputs also determine economic growth, neutralize the adversities of oil-price volatilities and environmental welfare in Bangladesh.

On the other hand, the causality estimates in the context of model (2) also depict unidirectional causation, without the feedback, from FDI inflows to ecological footprints in Bangladesh. This finding, along with the positive sign of the corresponding elasticity estimate concerning FDI inflows and ecological footprints, implicates that the quality of the FDI flowing into the Bangladesh economy does not safeguard the environmental goals of the nation. Thus, it is pertinent to restructure the foreign finance policy of the government whereby emphasis must be given to attract the relatively cleaner and renewable energy-intensive forms of FDI. The unidirectional causal finding contradicts the bidirectional association between FDI inflows and ecological footprints reported by Khan et al. (2020) for China, India, and Pakistan. Moreover, unidirectional causality from REO shares to ecological footprints is also ascertained from the causality analysis. This, in line with the corresponding elasticity estimate, implies that enhancing the REO shares to phase out the fossil fuel dependency is a plausible solution to the environmental hardships in Bangladesh. Similarly, Sharif et al. (2020) found unidirectional causality stemming from renewable energy use to ecological footprints in the case of Turkey. Besides, unidirectional causation from energy use intensity levels to ecological footprints is also revealed. Therefore, this finding further asserts the need for technological innovation, synonymous with a decline in the energy use intensity, for mitigating the ecological footprints figures in Bangladesh. Moreover, shocks to real crude oil prices are also estimated to causally influence the ecological footprints. In line with the corresponding elasticity estimate, this unidirectional causation between oil price and ecological footprints suggests that higher oil prices induce the substitution of fossil fuels by renewable alternatives for electricity generation processes which, in turn, can effectively reduce the ecological footprints levels in Bangladesh. Furthermore, the causality estimates also predict a unidirectional causality running from urbanization to ecological footprints which, along with the corresponding elasticity estimate, highlights the adverse environmental impacts associated with the unplanned urbanization problems in Bangladesh. Similar causality in the context of G7 countries was also stated in the study by Ahmed et al. (2020). On the other hand, a feedback effect between per capita GDP and ecological footprints is also witnessed. This implies that affluence plays a key role in influencing the environmental indicators in Bangladesh. Moreover, higher levels of national income, in turn, can ideally empower the Bangladesh economy to control environmental pollution to a large extent. Identical bidirectional causation between GDP per capita and ecological footprints in the context of Turkey was opined by Ozcan et al. (2018).