Does foreign direct investment asymmetrically influence carbon emissions in sub-Saharan Africa? Evidence from nonlinear panel ARDL approach

In recent decades, the relationship between foreign direct investment (FDI) and carbon emissions has garnered the extensive attention by researchers and governments across the globe. Also, for most part, empirical studies on this nexus have assumed a symmetric relationship through the imposition of linear specifications. However, such relationships do not account for asymmetries in the impact of FDI on carbon emissions. In the case of sub-Saharan Africa, such relationships are crucial and need more careful analysis given the important role FDI plays in the development of the sub-region. Thus, this paper examines the asymmetric effect of FDI on carbon emissions in 41 selected sub-Saharan African countries spanning from 1996 to 2018. In order to decompose FDI into positive and negative partial sum and examine possible asymmetric effects of the variables on carbon emissions, we used the panel nonlinear autoregressive distributed lag (NARDL) approach. This method accounts for cross-sectional variances. Our results show that in the long run, a positive shock in FDI increases carbon emissions while a negative shock lowers them. Our results also show that carbon emissions respond asymmetrically to changes in FDI. It is recommended that comprehensive investment policies aimed at encouraging clean technology and environmentally friendly investments be implemented to ensure environmental sustainability.


Introduction
Foreign direct investment (FDI) plays a vital role in the economic growth and development of an economy, particularly when the economy's domestic savings are not adequate to meet its investment needs. It also serves as a conduit for the transfer of advanced and modern technology to host countries, and a potential source of employment for both skilled and unskilled labor (Demena and Bergeijk 2019). Additionally, FDI helps in enterprise development, generates positive externalities, and increases productivity gains thereby stimulating growth in host economy (Lee 2013;Shahbaz et al. 2015). Although FDI makes all these contributions to the host country, it can have detrimental effects on the environment (Zhu et al. 2016). The economic gains that may have been achieved as a result of increases in FDI inflows could be neutralized by increases in carbon emissions. Thus, it may be a cause for concern when considering the environmental consequences of an FDI-led economic growth (Shahbaz et al. 2018). There are several empirical studies on the relationship between FDI and carbon emissions (see Koçak and Şarkgüneşi 2018;Chang 2015;Ren et al. 2021;Minh 2020;Hou et al. 2021;Ochoa-Moreno et al. 2021;Mert and Caglar 2020;Yilanci et al. 2019;Demena and Afesorgbor 2020;Odugbesan and Adebayo 2020;Jakada and Mahmood 2020;Faheem et al. 2022;Huang et al. 2022;Musah et al. 2022;Gao et al. 2022). However, empirical evidence on this relationship is mixed and inconclusive. This may be due to the fact that at the theoretical level, the relationship between FDI inflows and carbon emissions has two main dimensions, Responsible Editor: Eyup Dogan namely, pollution haven hypothesis (PHH) and the pollution halo hypothesis (PH).
According to the PHH, relaxed environmental policies in a host country may attract more and more FDI inflows by multinational companies to enhance their production and emit carbon dioxide and other externalities (Huang et al. 2022). On the other hand, the PH states that in applying a universal environmental standard, multinationals engaging in FDI come with advanced and energy efficient technology, better management concepts and works under better management practices that lead to improve environmental quality in host country (Ren et al. 2021;Mert and Caglar 2020;Bakhsh et al. 2017). More importantly, FDI is likely to reduce energy intensity and carbon emissions if the multinationals adopt advanced technologies in the production process (Lee 2013). Against this background, this paper examines the asymmetric effect of FDI on carbon emissions in sub-Saharan Africa spanning from 1996 to 2018. In this case, the study aims to asymmetrically analyze the link between FDI and carbon emissions both in the short and long runs, as opposed to the existing panel studies which investigate these relationships symmetrically. Generally, FDI depends on factors such as the size and growth potential of the national economy, natural resource endowments, openness to international trade, access to international markets, the quality of physical, financial and technological infrastructure, and the quality of workforce. An increase in these factors will lead to an increase in FDI and vice versa. Thus, it is expected that FDI inflows will have both positive and negative shocks, which will impact differently on carbon emissions. For instance, FDI may promote energy-saving technologies in the host country or region and reduce the demand for non-renewable energy, which will consequently reduce carbon emissions. However, in sub-Saharan Africa, most firms are still heavily dependent on fossil fuel sources, which are largely carbon intensive (IEA 2019). Therefore, an increase in FDI inflows in the sub-region implies an increase in non-renewable energy demand, which will ultimately lead to a rise in carbon emissions.
This paper contributes to existing literature on environmental pollution in twofold. First, while we acknowledge the existing panel studies on the symmetric relationship between FDI and carbon emissions (Bokpin 2017;Demena and Afesorgbor 2020;Huang et al. 2022;Ochoa-Moreno et al. 2021;Musah et al. 2022), nonlinear panel studies on this relationship have not been exhaustive and evidence within the sub-Saharan African context remains sparse. Secondly, although a number of studies attempt to investigate the influence of FDI on carbon emissions, most of the studies use conventional panel unit root tests (i.e., the Im-Pesaran-Shin (IPS) and the Levin-Lin-Chu (LLC) unit root tests) which fail to account for cross-sectional dependence and/or heterogeneity across the panel in their estimation processes and this may result in biased estimates and forecasting errors. Thus, this paper seeks to fill this gap by conducting the second generation unit root tests (i.e., the cross-sectional augmented Dickey-Fuller (CADF) and the cross-sectional Im-Pesaran-Shin (CIPS) unit root tests) in our estimation process. By doing so, the reported empirical results become more robust and reliable.
In order to account for asymmetric effects in the relationship between FDI and carbon emissions, we formulate a panel nonlinear autoregressive distributed lag (NARDL) model. We have chosen the panel NARDL framework over other approaches used in modeling asymmetries (for example, quantile regression) because of the following reasons. First, this model allows for the consideration of asymmetry in the carbon emission equation and for the computation of the positive and negative partial sum decompositions of the relevant exogenous variable(s) (in this case, FDI) (Salisu and Isah 2017). Secondly, the approach is suitable for capturing inherent heterogeneity effects in the slope coefficients resulting from cross-sectional differences. Thirdly, it facilitates the estimation of both the long run and short run responses of carbon emissions to changes in FDI. The choice of sub-Saharan Africa for this study is motivated by (i) the importance of foreign direct investment in the economic development outcomes of the sub-region and (ii) poor environmental management in the sub-region. Jarrett (2017) posits that even though some of the worst energy grid systems in the world can be found in sub-Saharan Africa, there is no political will to address the energy and related environmental issues. Thus, an understanding of the nexus between FDI and carbon emissions will provide governments and policymakers in sub-Saharan Africa a deeper insight of how positive and negative shocks of FDI impact the environmental component of the sustainable development goals (SDG # 13) in the sub-region. Figures 1 and 2 report the evolution of carbon emissions and FDI, respectively, across time for sub-Saharan Africa. Specifically, Figs. 1 and 2 show that to some extent, increasing FDI inflows and carbon emissions have co-movement in the same direction. The next section deals with a review of relevant literature and chronologically; the successive sections are the theoretical linkages and methodology, the discussion of research findings, and the conclusion and policy implications of the study.

Literature review
In recent decades, the effect of FDI on carbon emissions has been a debatable issue. Theoretically, FDI can either have a positive or negative impact on the environment, depending on which dimension is dominant (Shahbaz et al. 2018). As earlier discussed in the introduction, there are two main dimensions to the FDI-carbon emissions, namely, the PHH and the PH. The PHH argues that production activities that are pollution intensive are directed from developed countries to countries with more relaxed environmental regulations through FDI (Mert and Caglar 2020). On the other hand, the PH advances the view that foreign firms prefer to function in economies that are environmentally conscious (Zarsky 1999). Thus, there can be a varying effect of FDI on carbon emissions, and so for this reason, the empirical evidence suggests mixed results (Shahbaz et al. 2018). For instance, at a global level, Muhammad and Khan (2021) applied the generalized methods of moment (GMM) and the fixed effect model to examine a panel of 170 countries around the world and concluded that FDI has boosted carbon emissions. Also, in a study on high, middle, and low income countries by Shahbaz et al. (2015), the fully modified ordinary least squares (FMOLS) approach was applied and the results confirmed the PHH. The authors found evidence that FDI inflows significantly increase environmental degradation. At the regional level, a study by Bokpin (2017) across Africa for the period 1990-2013 revealed that an increase in FDI inflows significantly increases environmental degradation. Similarly, Ojewumi and Akinlo (2017) used panel vector autoregressive (PVAR) and panel vector error correction (PVEC) methodologies on a sample of 33 sub-Saharan African countries and concluded that the inflow of FDI had greater long run positive impact on carbon emissions. Again, Gao et al. (2022) investigated the linear and nonlinear impacts of FDI and terrorism on carbon emissions for ten fragile economies for the period 1973-2019 using the ARDL and NARDL approaches. The authors found that positive changes in FDI have significant positive impact on carbon emissions. Also, Teng et al. (2021) utilized the PMG estimator to investigate the effect of FDI on carbon emissions of 10 economies and found that FDI positively affects environmental degradation. Musah et al. (2022) used the dynamic common correlated effects mean group (DCCEMG) approach and the cross-sectional autoregressive distributed lag (CS-ARDL) model to investigate the effect of FDI on environmental quality in G-20 countries and found that higher FDI inflows surge carbon emissions. Furthermore, Ashraf and Umar (2022) applied the panel NARDL methodology to investigate the asymmetric effect of FDI and oil prices on carbon emissions in the Gulf Cooperative Council economies. The authors found evidence that FDI has a long run positive impact on carbon emissions. On the other hand, examining the impact of FDI and the potential of renewable energy consumption on carbon dioxide emissions in 21 Kyoto countries using an unbalanced panel data, Mert and Bölük (2016) concluded that foreign direct investment brings in clean technology and improves environmental quality. Also, Bhujabal et al. (2021) examined the effect of information communication technology (ICT) and FDI on environmental pollution in major Asia Pacific countries using the pooled mean group (PMG) and Dumitrescu-Hurlin panel causality. The results revealed that ICT and FDI negatively affect environmental pollution. Again, Duodu et al. (2021) applied system GMM to investigate the relationship between FDI and environmental quality, taking into account policies and institutions for environmental sustainability for 23 Sub-Saharan Africa countries. The results revealed that FDI improves environmental quality in the long run. Furthermore, Guoyan et al. (2022) applied the panel smooth transition regression model (PSTR) to explore the nonlinear association between FDI and carbon emissions in the Middle East and North Africa (MENA) countries. The authors concluded that at a low regime, increase in FDI increases carbon emissions, but as the economy progresses to the high regime, the relationship between the two variables becomes negative and significant. Ansari et al. (2019) applied the FMOLS approach to investigate the validity of the PHH for a panel of 29 countries in Asia over the period 1994-2014. The long-run results confirmed the presence of the PHH only in East Asia. The study results, however, found evidence that FDI reduces environmental degradation in Southeast Asia, thereby rejecting the validity of the PHH.
At the country level, Salahuddin et al. (2018) applied the ARDL approach in a study on Kuwait for the period 1980-2013 and concluded that increases in economic growth and FDI stimulate carbon emissions in both the short run and the long run. Also, investigating the PHH in Ghana for the period 1980-2012 and applying the ARDL approach, Solarin et al. (2017) found evidence that FDI has a positive and significant impact on carbon emissions. Furthermore, Zhou et al. (2018) explored the impact of economic growth, population, FDI, and other economic variables on carbon emissions in China using the stochastic impacts by regression on population, affluence, and technology (STIRPAT) model. The study found evidence that FDI increases carbon emissions in Chinese cities. For this same country, Zhang and Zhou (2016) examined the impact of FDI on carbon emissions at the national and regional levels using provincial panel data from 1995 to 2010. The authors adopted the STIRPAT model and their results indicated that the impact of FDI on carbon emissions decreases from the western region to the eastern and central regions. Furthermore, in a panel study of 112 Chinese cities, Liu et al. (2018) found evidence that FDI inflows reduce carbon emissions. Based on this result, the authors recommended the utilization of advanced clean technology acquired by means of FDI. Additionally, Mahadevan and Sun (2020) contends that total inward FDI into China shows a pollution-reducing effect in the western and eastern regions while that in the central region remains unchanged. Furthermore, a study by Tang and Tan (2015) used the cointegration and causality techniques in a study on Vietnam and concluded that FDI reduces carbon emissions. A recent study on the USA between 1970 and 2015 by Zafar et al. (2019) found evidence that FDI significantly reduces the carbon emissions in the USA. Specifically, an increase of 1% in FDI causes a reduction in 0.025% in the carbon emissions. Similarly, Haug and Ucal (2019) and Koçak and Şarkgüneşi (2018) found that for Turkey, there is evidence of the existence of an equilibrium relationship between FDI and carbon emissions, with a positive impact in the short run and a negative impact in the long run. On the other hand, in Malaysia, a study by Lau et al. (2014) found that FDI inflows promote greater economic growth and lead to further environmental degradation. The empirical findings by these authors are similar to the results of Minh (2020) for Vietnam for the period 1990-2015, which, through ARDL models, found evidence that FDI inflows contribute marginally to environmental degradation, both in the short run and long run. In the context of India, a recent study by Zameer et al. (2020)  The study results revealed that while other countries' FDIs have increased carbon emissions in these countries, China's FDIs have rather mitigated carbon emissions. In almost all the aforementioned country-specific studies, it is important to note that the linear specification of the FDI-carbon nexus was applied with no attention given to possible asymmetry between the variables.
There are other existing studies that have also claimed that there is a nonlinear relationship between FDI and carbon emissions. For instance, using the panel smooth transition regression (PSTR) model with nonlinear and dynamic features to simultaneously investigate the direct and spillover influences at work in FDI inflows and carbon emissions in selected emerging countries, Xie et al. (2020) found evidence that FDI can directly result in an increase in carbon emissions. On the other hand, the results of spillover effect through economic growth suggest that FDI can reduce carbon emissions. In the context of MINT (Mexico, Indonesia, Nigeria, and Turkey) countries, Balsalobre-Lorente et al. (2019) used the FMOLS and the dynamic ordinary least square (DOLS) techniques and found evidence that there is an inverted U-shape relationship between FDI and carbon emissions. This means that initially, FDI inflows increased carbon emissions in these countries. However, as FDI inflows rise to a certain level, it mitigates carbon emissions. Based on the review of related literature, it is clear that there is no consensus on the FDI-carbon emission nexus. While some studies validate PHH, other studies support PH and there are also others that claim a nonlinear relationship between FDI and carbon emissions. Also, with the exception of the studies by Gao et al. (2022) and Ashraf and Umar (2022) for 10 fragile economies and the Gulf Cooperative Council economies, respectively, the rest of the panel studies (including those on sub-Saharan Africa) on the FDIcarbon emissions nexus have used the linear specifications of the FDI variable, which might be inadequate. In order to, therefore, address this lacuna, the current paper specifies FDI both in its linear and nonlinear forms. The use of the NARDL framework will facilitate the estimation of both the long run and short run effects of FDI on carbon emissions in sub-Saharan Africa and provide a deeper insight into the FDI-carbon emission nexus.

Theoretical linkages and methodology
Generally, the weak environmental regulations in the developing and emerging economies attract some investors to invest in their pollution-intensive industries. As a result, carbon emissions tend to increase in these countries. Also, countries with vast fossil fuel reserves tend to engage in carbon-intensive activities and produce pollution-intensive products. Thus, these countries are likely to attract dirty FDI inflows that would positively influence carbon emissions and degrade the environment (Banerjee and Murshed 2020). On the other hand, clean FDI inflows may promote technological development in the host countries which, in turn, can be associated with the effective implementation of carbon emission-mitigating policies (Musah et al. 2022). Subsequently, the environmental pollution accompanying dirty FDI inflows can be neutralized to improve the environment. The theoretical linkages regarding FDI-carbon emissions from the perspective of the host nation are presented in Fig. 3.

Data source
In this study, a balanced annual panel data for 41 sub-Saharan African countries spanning from 1996 to 2018 was used. The study period and the countries were selected (see Appendix 1) based on the availability of data, which was obtained from the World Development Indicators (WDI) online database of the World Bank (2021). It is worthwhile to note that all variables (except foreign direct investment as a percentage of gross domestic product (GDP)) are transformed into natural logarithms to avoid heteroskedacticity and spurious regression results. Table 1 presents the sources and description of the variables used in this paper.

Empirical model
In estimating our model, the study employed robust panel methods referred to as the mean group (MG) estimator and the PMG estimator. The MG estimator relies on estimating N time series regressions and averaging coefficients while the PMG estimator has to do with the combination and pooling of coefficients (Blackburne and Frank 2007). The Hausman test was then used to determine whether there is any systematic difference between the MG and the PMG.

The symmetric panel ARDL
We began the estimation process by assuming a symmetric response of carbon emissions to changes in FDI. Thus, the symmetric version of the panel ARDL is implicitly stated as follows: where CO 2 represents carbon emissions, FDI is foreign direct investment, REC is renewable energy consumption, and RGDPPc is real gross domestic product per capita.
Equation (1) is explicitly stated as follows: As already indicated in Table 1, lnCO 2it is the natural log of carbon emissions (in kilotons) for each unit i,over a period of time t. FDI represents foreign direct investment as a percentage of GDP and lnREC and lnRGDPPc are control variables representing the natural log of renewable energy consumption as a percentage of total final energy consumption and the natural log of real GDP per capita (a proxy for economic growth), respectively. Here, the log transformation is used in order to compress the scale in which the variables are measured. This is to avoid the possibility of heteroscedasticity in the model (Gujarati 1995). i denotes the group specific effect. i represents the sampled units while t is the number of periods. The long-run slope (elasticity) coefficient for each cross section is calculated as − 2i 1i ,− 3i 1i and − 4i 1i since in the long run, the assumption is thatΔlnCO 2i,t−1 = 0 ,ΔFDI i,t−1 = 0 , ΔlnREC i,t−1 = 0 , andΔlnRGDPPc i,t−1 = 0 . On the other hand, the short-run estimate for FDI, renewable energy consumption, and real GDP per capita are obtained as ij , ij and ij , respectively. We can include an error correction term in Eq. (4) and re-write it as follows: (1) . It can be seen from Eqs. (4) and (5) that there is no decomposition of FDI into positive and negative shocks. This is due to the assumption of symmetric impact of FDI shocks on carbon emissions in this scenario.

The asymmetric panel ARDL
Under the asymmetric panel ARDL, the positive and negative shocks of FDI are not expected to have identical impacts on carbon emissions. Thus, the asymmetric panel ARDL is presented as follows: where FDI + and FDI − represent positive and negative FDI shocks, respectively. The long-run coefficients for FDI + and FDI − are computed as − + 2i 1i and − − 3i 1i , respectively. These shocks are, respectively, calculated as positive and negative partial sum decompositions of changes in FDI and stated as follows: The error-correction version of Eq. (6) is stated as follows: (3) min ΔFDI ik , 0  (2021) i,t−1 stated in Eq. (9) expresses the long-run equilibrium in the asymmetry panel ARDL, while i computes how long it would take the system to converge to its long-run equilibrium when there is a disequilibrium.

Carbon emissions (kt)
In the study, the natural log of carbon emissions ( lnCO 2 ) is used as the outcome variable. The choice of carbon emissions measured in kilotons is consistent with contemporary literature (Odhiambo 2020;Kwakwa et al. 2019;Jebli et al. 2019;Asongu et al. 2019). Several studies including Mert and Caglar (2020) and Shahbaz et al. (2015) have confirmed the link between FDI and carbon emissions. Adejumo and Asongu (2020) and Koçak and Şarkgüneşi (2018), FDI as a percentage of GDP is used as the variable of interest.

Renewable energy consumption as a percentage of total final energy consumption
Renewable energy consumption as a percentage of total final energy consumption is used as a control variable in the model and this is consistent with Yuping et al. (2021) and Hanif (2018). Renewable energy is considered pollutionfree, cleaner than fossil energy, and useful to reduce carbon emissions (Cheng et al. 2019;Asongu et al. 2019).

Real GDP per capita
Another control variable used in the study is the natural log of real GDP per capita. The use of real GDP per capita as a determinant of carbon emissions is consistent with Odhiambo (2012) and Schröder and Storm (2020).

Panel unit root tests
We used panel unit root tests to determine the data series' stationarity. As a result, the Im-Pesaran-Shin (IPS) and (7) Levin-Lin-Chu (LLC) unit root tests were used in this study. These tests are based on the assumption of cross-sectional independence.

Cross-sectional dependence test
It is significant to note that dependence in cross sections can occur due to externalities, disregarded common factors, and economic and regional associations (Kasman and Duman 2015). Thus, disregarding the possible presence of crosssectional dependence and heterogeneity across the panel can cause biased estimates and forecasting errors. In view of this, we conducted the Pesaran (2004) CD test to verify if there is the presence of cross-sectional dependence. The formula for cross-sectional dependence test is stated as follows: where T is time, N represents the sample size, and ̂ is the coefficient of residual correlation in individual ordinary least square (OLS) regression. Where cross-sectional dependence exists in the series, it is recommended that second-generation unit root tests such as CADF and CIPS unit root tests are conducted to be sure that the series are indeed, stationary either at I(0) or I(1). Pesaran (2007) proposed the CADF unit root test and it is stated as follows: where Δy t−1 represents the first difference and y t−1 is the average of all observations in the model at time t − 1 . The CIPS, on the other hand, is computed by the average of t statistics of the parameter * i in the CADF model. It is stated as follows:

Empirical findings and discussion
In Table 2, we present the results of the correlation matrix. The primary purpose of the correlation matrix is to ensure that the explanatory variables are free from the problem of multicollinearity. Secondly, it is to measure the strength of the relationship between two variables and finally, it indicates the direction of the relationship between two variables (either positive or negative). The results in Table 2 clearly suggest that the explanatory variables generally exhibit strong correlation and there is no presence of multicollinearity in the data set.
We performed the IPS and LLC unit root tests in order to ascertain the order of integration of the variables. The integration order of the variables is shown in Table 3. The results show that lnCO 2 has unit root at level but becomes stationary after first difference. However, FDI is stationary even at level. Furthermore, both lnREC and lnRGDPPc are non-stationary at level but become stationary at first difference. The panel cross-sectional unit root test results indicate that the null hypothesis of cross-sectional dependence for lnCO 2 , FDI, lnREC , and lnRGDPPc is rejected at 1% level of significance. This implies that the firstgeneration panel unit root framework is not appropriate for this study. The results are shown in Table 4. Breitung and Pesaran (2008) posit that spatial spillover effects and (un)observable factors between countries and regions are likely to cause a strong panel cross-sectional dependence. Thus, based on the fact that the variables in our model have cross-sectional dependence, the CADF and CIPS unit root tests were performed and the results are presented in Table 4. The CADF and CIPS unit root test results in Table 5 show that both lnCO 2 and FDI are stationary at least, at their first differences. We can therefore conclude that the variables are stationary at their first difference and not at their second difference.
We proceeded to estimate all the equations using both the MG and PMG estimators and the findings were then subjected to the Hausman test. If the null hypothesis is not rejected, the PMG estimator is more efficient than the MG estimator in the given situation. On the other hand, a rejection of the null hypothesis implies the adoption of the MG estimator. The Hausman test results in Table 6 are in favor of the PMG estimator as the efficient estimator for modeling the symmetric and asymmetric effects of FDI on carbon emissions. Thus, only the empirical estimates obtained from the PMG estimator are reported and discussed in this paper. The results are discussed under two sub-headings: model without asymmetry and with asymmetry. The computed coefficients suggest that FDI has a positive influence on carbon emissions in the regression findings of the model without asymmetry (see Table 6A). This finding validates the PHH and also consistent with previous research by Zhou et al. (2018), Shahbaz et al. (2015), Salahuddin et al. (2018), and Muhammad and Khan (2021). The long-run estimates reveal that a percentage increase in FDI induces a 0.006% increase in carbon emissions at 1% level of significance ceteris paribus. In addition, an increase in renewable energy consumption exerts a negative impact on carbon emissions. Holding other factors constant, a percentage increase in renewable energy consumption leads to a reduction in carbon emissions by 1.039% and this is statistically significant at 1%. The result supports the findings by Apergis et al. (2018) which contends that renewable energy consumption contributes to a reduction in carbon emissions in sub-Saharan Africa. However, carbon emissions respond positively to an increase in economic growth all other things being equal. Specifically,  a percentage increase in economic growth leads to a statistically significant increase in carbon emissions by 0.284%. This means that the environmental sustainability criterion for Sub-Saharan African countries is yet to be reached. The short-run estimates indicate that all other things being equal, an increase in FDI by1% leads to a decrease in carbon emissions by 0.004%. However, an increase in renewable energy consumption by 1% will induce a decrease in carbon emissions by 2.593%. This suggests that even in the short run, renewable energy consumption reduces carbon emissions in sub-Saharan Africa. Furthermore, carbon emissions respond positively to increases in economic growth. An increase in economic growth by 1% leads to a positive and statistically significant increase in carbon emissions by 0.224%, ceteris paribus. As expected, the error correction term is negative (− 0.314) and statistically significant, as seen in Table 6A.
Let us now turn our attention to the regression results of the model with asymmetry in Table 6B. According to the long-run estimates, a positive shock in FDI induces a positive and statistically significant effect of 0.006 on carbon emissions and this validates the PHH. On the other hand, a negative shock of FDI induces a negative effect of 0.003 on carbon emissions suggesting that any negative shock of FDI reduces carbon emissions in the selected countries. The findings demonstrate that a positive FDI shock has a marginally higher positive impact on carbon emissions as compared to the decreasing impact of a negative FDI shock. These results are consistent with the findings by Ashraf and Umar (2022) and Gao et al. (2022) which evidence that increases in FDI cause more carbon emissions. Additionally, the long-run results indicate that a percentage increase in renewable energy consumption will lead to a negative and significant impact on carbon emissions by 1.239%. Furthermore, a 0.308% increase in carbon emissions is induced by a 1% increase in economic growth and this is statistically significant at 1%. The findings further reveal that in the short run, a positive FDI shock has an insignificant interconnection on carbon emissions. However, a negative shock in FDI will lead to a statistically significant decrease in carbon emissions by 0.005. Also, an increase in renewable energy consumption by 1% induces a statistically significant decrease in carbon emissions by 2.607%, suggesting that any increase in renewable energy consumption in the short run will reduce carbon emissions in sub-Saharan Africa. Additionally, the short run results show that an increase in economic growth leads to an increase in carbon emissions. More specifically, a 1% increase in economic growth increases carbon emissions by 0.208%. Again just like the case of the symmetric model, the error-correction term for the model with asymmetry is also negative (− 0.390) and statistically significant, which shows a return to equilibrium when there is a disequilibrium. It is worthy to note that the speed of adjustment to steady state for the asymmetric model is slightly higher (by 0.015) than that of the symmetric model.

Conclusion and policy implications
This paper provides a perspective on the asymmetric relationship between FDI and carbon emissions in sub-Saharan Africa spanning from 1996 to 2018. In order to account for asymmetries, we formulated a nonlinear panel ARDL model proposed by Shin et al. (2014). In panel data analysis, there is the possibility of some variances between cross sections within the same group. Therefore, in order to ensure that such variations are accounted for in the estimation process, we allowed for heterogeneity effect. We also estimated the symmetric version of the linear panel ARDL model in order to perform relevant comparative analysis. The results from the symmetric version indicate that in the long run, there is a significant positive relationship between FDI and carbon emissions. This finding validates the PHH and coincides with studies by Zhou et al. (2018), Shahbaz et al. (2015), Salahuddin et al. (2018), and Muhammad and Khan (2021). The long-run estimates further indicate that renewable energy consumption has a decreasing effect on carbon emissions while economic growth increases them. Furthermore, the results indicate that carbon emissions respond asymmetrically to changes in FDI. In the long run, a positive FDI shock has a positive impact on carbon emissions and this validates the PHH. On the other hand, a negative FDI shock has a negative impact on carbon emissions. The positive shock, however, has a marginally higher positive impact on carbon emissions as compared to the decreasing impact of a negative FDI shock. From these findings, it is clear that the NARDL approach provides more explanatory and robust estimates for the selected countries than the ARDL approach. This study has drawn the following important policy implications. The confirmation of the PHH in the present study supports prior claims that ineffective and lax regulations, coupled with the inflow of carbon-intensive production through FDI, have turned sub-Saharan African economies into pollution havens. Unlike the developed regions, sub-Saharan African countries have a limited option in terms of FDI selection. However, in the presence of emission convergence like the EKC hypothesis, policymakers could introduce carbon taxing and carbon credits for reducing carbon-intensive technologies and industries across sub-Saharan African countries. Furthermore, sub-Saharan African governments should implement comprehensive investment policies aimed at encouraging clean technology and environmentally friendly investments. Investments in these areas will lead to a reduction in carbon emissions in the long run and also ensure environmental sustainability. FDI inflows create knowledge spillovers of environmentally friendly technologies in the invested sector and region (Koçak and Şarkgüneşi 2018). Thus, there is the need for countries in sub-Saharan Africa to make arrangements to strengthen the absorptive capacity mechanisms so as to make this process more effective and efficient. Future research should consider the asymmetric Granger causality between FDI and environmental degradation in sub-Saharan Africa for fresh insight. Data availability Data sources are outlined above in Table 1 and will be made available on demand.

Declarations
Ethics approval and consent to participate We, the authors, are giving our ethical approval and consent for this paper to be published in ESPR if found publishable.

Consent to participate
We, the authors, are giving our consent for participation in this paper to be published in ESPR if found publishable.

Consent for publication
We, the authors, are giving our consent for this paper to be published in ESPR if found publishable.

Competing interests
The authors declare no competing interests.