Economic growth, international trade, and environmental degradation in Sub-Saharan Africa

Purpose – This study investigates the relationship between economic growth, international trade, and environmentaldegradationinSub-SaharanAfrica(SSA),focusingonthevalidityoftheenvironmentalKuznets hypothesis(EKC),thepollutionhavenshypothesis(PHH),andthefactorendowmenthypothesis(FEH). Design/methodology/approach – The study uses annual data for 41 SSA countries between 1990 and 2017 and employs the bias-corrected least square dummy variable (LSDVC) estimation techniques. Environmental degradationisindicatedbycarbondioxide(CO 2 ),delicateparticulatematter(PM2.5)emissions,anddeforestation. Findings – The results confirm the validity of the EKC hypothesis for PM2.5 emissions and deforestation but notforCO 2 emissions.Theresultsalsoindicatethatinternationaltradereducesdeforestationandthatboththe PHH and FEH are valid for CO 2 emission but not for PM2.5 emissions and deforestation. Practical implications – In this paper, the authors are able to illustrate that both economic growth and internationaltradecanharmtheenvironmentifunchecked.Therefore,theconclusionofthisstudyofferspolicy optionsthroughwhichSSAcountriescanachievedesiredeconomicgrowthgoalswithoutaffectingenvironmentalquality.Thestudycanbeabenchmarkforenvironmentalpolicyintheregion. Originality/value – The authors provide an in-depth discussion of the growth-trade-environmental degradation nexus in SSA. The EKC, PHH, and FEH ’ s validity confirm that economic growth remains a threat to the local natural environment in SSA. Hence, the need for a trade-off between economic growth needs and environmental degradation and understanding where to compromise to achieve SSA ’ s economic development priorities.


Introduction
Environmental degradation defined as air pollution is the contamination of air with the persistent emission of greenhouse gases (GHGs), fumes or odours and dust into the atmosphere (Stern et al., 1996). Data from the World Resources Institute (2019) suggest that carbon dioxide (CO 2 ) is the most abundant GHG emitted. Between 1990 and 2018, CO 2 emissions in Sub-Saharan Africa (SSA) have remained less than 5% of global CO 2 emissions (IPCC, 2018). Albeit being low, SSA's per capita CO 2 emission estimated at 3.9 tonnes per capita is slightly above the recommended 2.6 tonnes per capita level (IPCC, 2018). Therefore, the region's emissions growth rate is among the highest globally and if unchecked, SSA countries may turn into major emitters (Wang and Dong, 2019). However, most SSA countries are low-income countries that lack the required financial resources to mitigate the consequences of environmental degradation. Furthermore, according to Amegah and SSA has amplified its efforts to spur economic growth over the last 3 decades (Zeufack et al., 2021) with average annual gross domestic product (GDP) growth of 3.5% between 1990 and 2019, which is higher than the global annual average of 2.8% in the same period (World Bank, 2020). But, according to Tenaw and Beyene (2021), the downside of rapid economic growth especially in developing countries like those in SSA is environmental degradation (Kong and Khan, 2019). However, Bah et al. (2020) suggest that SSA countries can continue their efforts to spur economic growth while minimising environmental damage at the same time. Besides, numerous evidence from studies on the relationship between economic growth and environmental degradation in SSA is still inconclusive (Tenaw and Beyene, 2021). Therefore, this study helps us to understand the relationship between economic growth and environmental degradation in SSA, and it is essential in guiding policymakers to design policy options that achieve sustainable economic growth.

Literature review 2.1 The EKC hypothesis
Among the earliest studies to examine the economic growth-environmental degradation nexus is Grossman and Krueger (1991) who studied the environmental impacts of the North American Free Trade Agreement (NAFTA). They suggested that the relationship between economic growth and environmental degradation is translated through three effects, namely, scale, technique and composition effects. According to Nkengfack et al. (2019), the scale effect (scale of economic activity) suggests that other factors held constant, an increase in the size of the economy (the scale effect) raises the level of environmental degradation. On the other hand, the technique effect explains the positive impacts of income growth on environmental degradation. This implies that growth in income may influence peoples' demand for better environmental standards and protection (Cole and Elliott, 2003). In presence of the EKC, the technique effect is expected to be negative and significant (Dinda, 2005). Lastly, the composition effect explains how a country's capital and labour structure can affect the country's environment. Given the dynamic structure of the economy and the production structure, changes in the composition of capital and labour used in the production of output will determine whether the composition effect has a positive or negative impact on environment quality (Cherniwchan et al., 2017).
Looking at empirical literature on the EKC hypothesis, Grossman and Krueger (1991) use data for several cities in 42 countries between 1977 and 1988 obtained from the Global Environmental Monitoring System. They found evidence for the EKC relationship to exist for sulphur oxide (SO 2 ) and dark matter with a turning point estimated to be between USD4,000-USD6,000 per capita GDP. In addition, Cropper and Griffiths (1994), Cole et al. (1997) and Barbier (1997) are among the early studies that provide evidence for the EKC hypothesis.  Ogundari et al. (2017) confirm the validity of the EKC hypothesis using different environmental degradation indicators. On the contrary, Beyene and Kotosz (2020) find a bell-shaped relationship and conclude no evidence of the EKC in East Africa. Similarly, Ogundari et al. (2017), Ojewumi and Akinlo (2017), Amuakwa-Mensah and Adom (2017) and Zerbo (2017) fail to confirm the validity of the EKC hypothesis for CO 2 in SSA.

International trade, the PHH and FEH
From the literature, international trade can induce economic growth, and trade-induced economic growth directly affects environmental degradation (Cherniwchan et al., 2017).

Economic growth in Sub-Saharan Africa
This means that international trade may fundamentally affect how scale, technique and composition effects determine environmental degradation (Bataka, 2021). Holding other factors constant, income gains arising from an opportunity to trade can positively induce the scale of economic activities through increased production level to satisfy the newly opened markets, and this will most definitely increase environmental degradation. As such, tradeinduced economic growth can also be regarded as "trade-induced scale effect" (Cole, 2004). However, economic growth may encourage cleaner production techniques and provide funds for public expenditure in environmental abatement technologies leading to a cleaner environment (Dinda, 2005). This is regarded as the "trade-induced technique effect" which exhibits a negative relationship between economic growth and environmental degradation (Cole and Elliott, 2003). As such, Dinda (2006) suggests that the PHH is only possible due to international trade. Hence, trade has indirect effects on environmental degradation through the inflow of pollution-intensive industries into countries that have relatively lenient environmental regulations. Furthermore, when an economy participates in the global trade system, free movement of factor inputs may also change the relative capital and labour structure that represents the composition effect. Antweiler et al. (2001) suggest that free movement of factors of production means that production goes to places where resources are abundant. Because resources in developing countries are generally untouched, openness to trade attracts more exploitation and hence more production activities which leads to increased environmental pollution and damage which explains the FEH (Managi et al., 2008).

The model
The empirical model adopted in this research starts with environmental degradation as a function of real per capita income as suggested by Antweiler et al. (2001). It is written as; where; i and t indicate country and year, respectively, for i ¼ 1; 2; . . . . . . N; t ¼ 1; 2; . . . . . . T.
ED it indicates environmental degradation, α denotes country-specific intercepts, θ represents time-specific intercepts, and M represents real per capita income. ε it is the error term that is assumed to be independently and identically distributed (iid). We include the square of real per capita income to allow for diminishing marginal effects (Antweiler et al., 2001). Besides, Nkengfack et al. (2019) suggest that both M it and M 2 it capture the scale effect and the technique effects. For the validity of the EKC hypothesis, it is expected that β 1 > 0 and β 2 < 0 (Dinda, 2005). We follow Cole and Elliot (2003) to include the composition effect and international trade as well as the interaction terms for the trade-induced income effects and trade-induced composition effects to examine the PHH and FEH respectively. Other control variables include foreign direct investment (FDI) as suggested by Demena and Afesorgbor (2020), energy consumption as suggested by Kwakwa (2020) and Gulistan et al. (2020). Lastly, following Udeagha and Ngepah (2019), we include the ratification of the Kyoto Protocol (KPR) to indicate political commitments towards combatting environmental degradation. Therefore, the empirical model becomes where L at the beginning of some variables indicates a natural log. KL it is the capital-labour ratio which is used to indicate the composition effect. TI it represents international trade intensity. Further, following, Cole and Elliot (2003), we also use interaction terms of trade intensity and income (TI * M Þ it to capture the trade-induced income gains as a proxy for the PHH. We also use the interaction term of the trade intensity and the capital-labour ratio (TI * KLÞ it to capture the trade-induced composition effect which represents the FEH. The control variables include FDI (FDI), the ratification of the Kyoto Protocol (KPR) and energy consumption (EU).

Description of variables
We use three indicators for environmental degradation. First, we follow Tenaw and Beyene (2021) and Bataka (2021) to use per capita CO 2 emissions (LCO) to represent global air pollution. Data are obtained from Global Carbon Atlas by the CO 2 Information Analysis Centre (CDIAC). Second, we follow Wu (2017) to use the natural log of fine particulate matter (PM2.5) emissions (LPM) as a local air pollutant (Schwela, 2012). Finally, we follow Cropper and Griffiths (1994) and Hammig (2001, 2004) to use the natural log of deforestation (LFO) as an indicator of environmental degradation. Deforestation is calculated using where FO is deforestation, F it and F i;t−1 are the total forest area (sq. km) in the current period and the previous period respectively. Data for both LPM and LFO are obtained from World Bank's World Development Indicators (WDI). For M it and M 2 it we use real per capita GDP and the square of real per capita GDP as suggested by Nkengfack et al. (2019). For KL it the capital-labour ratio for the respective country is used as a proxy of the composition effect (Cole and Elliot, 2003). Data for both M and KL are obtained from Penn World Tables (PWT). We also use the composite trade intensity (TI) measure suggested by Squalli and Wilson (2011). This TI-based measure offers a broader consideration in regards to a country's participation in global trade and the country's significance in the global economy because TI is adjusted by the share of a country's trade level in relation to the average international trade. TI it is calculated as: where X it , I it and Y it are exports, imports and real GDP, respectively. Also, i denotes a country in SSA and j represents a country's major trading partner. The first part of the equation captures world trade while the second part represents each county's trade openness. Data used to compute TI are obtained from the World Trade Organisation (WTO). For FDI, we use FDI inflow, and data are obtained from the World Bank's WDI. For EU, we use "fossil fuels energy use (FFEU)" as a proxy for energy consumption when CO and PM are used as the dependent variables. When deforestation is used as the dependent variable, we use "firewood energy use (FWEU)" as the indicator of energy consumption. This is because fossil fuels are direct emitters of CO 2 and PM2.5 while fuelwood is a direct product of Economic growth in Sub-Saharan Africa deforestation. Data for FFEU are obtained from World Bank's WDI while that of FWEU are obtained from UNDATA. Finally, we create a dummy variable KPR that takes the value of one from the year in which the SSA country ratified the Kyoto Protocol onwards, otherwise zero.

Data and methods
For this study, we use annual panel data of 41 SSA countries for the period between 1990 and 2017. As such, the data used suggest that (T 5 28) is less than (N 5 41). To estimate equation (2), we start with a panel vector autoregressive (VAR) model in the first differences expressed as: where Δ is the first difference operator, Y i;t is a vector of the logs of dependent variables, X 0 i;t is a vector of the independent variables, ε it is a vector of idiosyncratic errors and d t is the time dummy. The panel VAR model in the first difference allows us to estimate the underlying dynamic relationships without applying any a priori restrictions. In equation (5), the lagged dependent variable on the right-hand side correlates with the new differenced error term causing a possible simultaneity problem. Besides, the dynamic panel specification in equation (2) implies an endogeneity problem where the independent variable becomes endogenous if it is correlated with the error terms (Baltagi, 2013). This is the called Nickell's bias (Nickell, 1981) which means that the non-zero correlation between the lagged dependent variable and the individual-specific effects render the estimation of equation (2) using traditional panel data estimation approaches (Fixed effects and random effects) biased for data samples with small T regardless of the size of N. However, when N is finite, the variance of the estimates may asymptotically increase thereby generating considerable bias. As such, several N-consistent estimators such as the instrumental variable (IV) approach by Anderson and Hsiao (1982) and the generalised method of moments (GMM) estimators including the difference GMM by Arellano and Bond (1991) and the system GMM by Arellano and Bover (1995), and Blundell and Bond (1998) are possible alternatives. Albeit being versatile, these estimators may suffer distinctive weaknesses. The IV and difference GMM suffer finite-sample bias in panels with small N as well as a weak instruments bias in the presence of highly persistent data (Bruno, 2005b). The system GMM can also be affected by the proliferation bias in case the number of instruments is larger than the number of cross-sectional units (Bruno, 2005b). Therefore, as time becomes infinite ðT → ∞Þ, the least-squares dummy variable (LSDV) estimator is more consistent (Nickell, 1981). But, when T < 30, Judson and Owen (1999) show that the LSDV estimator has a bias of up to 20% of the time value coefficient of interest. Kiviet (1995) suggests an asymptotic expansion technique for correcting the bias. The biascorrected LSDV (LSDVC) is further extended by Judson and Owen (1999), and Bun and Kiviet (2003). Depending on the bias approximation chosen, Bun and Kiviet (2003) demonstrate that the LSDVC estimator can be initialised through three N-consistent estimators (the IV, the difference GMM and the system GMM estimator) in terms of bias and root mean square errors for a balanced panel (Meschi and Vivarelli, 2009). Bruno (2005a, b) further illustrate that the LSDVC estimator approach can be performed with similar efficiency using unbalanced panel data. According to both root mean squared error and bias criteria, regardless of the initialising estimator and the accuracy of the bias approximation are similar (Bruno, 2005b). Monte Carlo analysis by Bruno (2005b) demonstrates that the LSDVC outperforms other estimators when T < 30. Hence, the LSDVC estimation procedure is the most appropriate for this study.

Empirical results, findings and discussions 4.1 Descriptive statistics and cross-sectional dependence
Before estimating the models, we look at the descriptive statistics as well as examine the presence of cross-sectional dependence (CSD) following Pesaran (2004). The results are presented along with the descriptive statistics in Table 1. The CSD test statistics for all the variables are significant at 1% level. Therefore, we reject the null hypothesis ðH o Þ of crosssection independence. From the descriptive statistics, it can be observed that the number of observations varies with the highest at 1,148 and the lowest at 1,060. We also conduct the correlation test amongst the independent variables, and the correlation matrix is presented in Table 2. Based on the correlation matrix, we can conclude that there is multicollinearity amongst the independent variables.

Presentation of findings
We estimate equation (2) using the LSDVC estimation which is initialised by dynamic panel data estimates of the difference GMM by Arellano and Bond (1991) in this case AB. To ensure the robustness of our estimates, we also include the estimates when the LSDVC estimator is initialised by the system GMM by Blundell and Bond (1998) in this case BB and the IV techniques by Anderson and Hsiao (1982) Bruno (2005a). Further, the statistical significance of the LSDVC coefficients was tested using bootstrapped standard errors with 200 iterations following Bruno (2005b). The results from the LSDVC estimation are presented in Table 3 and arranged in columns following how the LSDVC estimator is initialised. In columns (1)-(3), (4)-(6) and (7)- (9), the LSDVC estimator is initialised by AB, BB and AH, respectively.

Discussion of findings
In Table 3, the results for L.LED are all significant at 1%. Further, for all the columns, the coefficients for LM are positive and significant in all columns. The positive and significant coefficients for LM mean that other factors held constant, per capita income growth leads to an increase in environmental degradation in SSA. This echoes the findings by Gulistan et al. (2020), Bah et al. (2020) and Ssali et al. (2019). Further, the coefficients for LM are larger than all other significant coefficients regardless of the indicator used. This indicates that economic growth is the largest contributor to environmental degradation in SSA which justifies the findings by Aliyu and Ismail (2015). It also explains the primacy of scale effects in determining environmental degradation in SSA as suggested by Nkengfack et al. (2019). Besides, the coefficient for LM 2 is negative and significant in columns when LPM and LFO are used as dependent variables. This means that LPM and LFO exhibit an inverted U-shaped curve which confirms the validity of the EKC hypothesis. Therefore, we can confirm that in SSA, the EKC hypothesis is valid for PM2.5 emissions and deforestation but not for CO 2 emissions. While we fail to find relevant studies that examine the EKC hypothesis for SSA using PM2.5 emissions, the validity of the EKC hypothesis for deforestation is comparable to the findings by Bissoon (2018), Cropper and Griffiths (1994) and Hammig (2001, 2004) but in contrast with Ogundari et al. (2017). The implication for this result is similar to that discussed by Chiu (2012) who suggests that when countries have low per capita income, deforestation rates increase as income increases due to the transformation of forest and woodland areas into arable land as well as fuelwood. However, as per capita income increases to a certain level, coupled with the adoption of modernised agricultural technology and advanced sources of energy and fuels, deforestation reduces (Barbier, 2004). This can also stem from the encouragement to engage in sustainable forest management, afforestation and natural expansion of forests, as well as the consciousness of the effects of climate change (Chiu, 2012).
Further, the turning points for PM2.5 emissions are USD 1356 (US Dollars), USD 1397 and USD 959.8 in columns (2), (4) and (8), respectively. However, between 1990 and 2017, the average annual per capita GDP (constant 2010 US Dollar) for SSA was USD 1304.5 in 1990and USD 1667 in 2017 with the lowest at which at USD 1173 in 1994. Therefore, the turning points for PM2.5 below USD 1000 are unrealistic but those at USD 1356 and USD 1397 are more realistic and were probably achieved between 2004 and 2005. On the other hand, the turning points for deforestation are USD 1206, USD 1195, and USD 2045.1 in columns (3), (6) and (9), respectively. The turning points for deforestation in columns (3), (6) are more realistic and probably achieved between the years 1995 and 1996. Besides, the turning points for deforestation were achieved before those for PM2.5 which may suggest that SSA countries experienced a reduction in deforestation before they experienced a reduction in PM2.5 emission. The findings confirm that the evidence of the EKC hypothesis in the empirical literature is mixed (Kong and Khan, 2019), and the results vary from one study to another based on the selection of variables, the data, the model specification and methodology used (Harbaugh et al., 2002). Specifically, the mixed results of the EKC hypothesis in this study are due to using different dependent variables.
Furthermore, Cole and Elliott (2003) suggest that the coefficient for the composition effect can be either positive or negative. From the findings, the coefficients for KL are significant for both PM2.5 emissions and deforestation. The relationship between KL and PM2.5 emissions Note(s): For ***p < 0.01, **p < 0.05, *p < 0.1, standard errors in parentheses. The LSDVC initialised by Arellano and Bond (1991) (AB), Blundell and Bond (1998) (BB), and Anderson and Hsiao (1982) AH consistent estimators and the standard errors are bootstrapped with 200 iterations. L.LED is the lagged dependent variable, YDs represents the year dummies, OBS is the number of observations and C_ID is the number of countries used Table 3. The LSDVC estimates for economic growthenvironmental degradation nexus in SSA Economic growth in Sub-Saharan Africa is negative which suggests that as SSA becomes more capital intensive, there is a reduction in PM2.5 emissions. Because the majority of the population in SSA is employed in labourintensive agriculture where forest and bush burning are common practices (Rudel (2013), PM2.5 emissions are high. With more capital, agriculture modernisation is inevitable, leading to a fall in PM2.5 emissions. On the other hand, KL has a positive relationship with deforestation which means that as SSA countries become more capital intensive, the deforestation in the region increases. We can argue agriculture plays a huge role in most SSA economies, and the adoption of more capital may lead to a transformation from subsistence to commercial agriculture which requires more land for agricultural expansion thereby increasing deforestation (Soaga and Kolade, 2013). The findings also show that international trade has a negative relationship with deforestation which means that as SSA countries become more open to international trade, deforestation decreases. This is because openness to international trade provides SSA countries with access to modernised technologies and factor inputs that substitute and reduce the dependence on forests for raw materials thereby reducing deforestation in SSA (Saggi, 2002). Such technologies may include agriculture technologies and fertilisers that can improve crop yield on existing farmland without the need to cut down more forests to provide new lands for agriculture.
Furthermore, LTI*M has a positive and significant relationship with LCO in columns (1) and (4) which confirms the validity of the PHH for CO 2 emissions in SSA. This means that openness to international trade facilitates pollution-intensive industries to relocate to SSA which leads to both trade-induced per capita income gains and increase in CO 2 emissions in the region. According to Twerefou et al. (2019) and Sun et al. (2020), the relocation of pollution intensive industries can be attributed to poor environmental regulation because SSA countries are among the lowest-ranked countries on the environmental performance index (EPI) Wendling et al. (2018) with only five countries were ranked in the top 100. This is in addition to lax environmental regulations that lower production costs and increase profit margins. Besides, SSA countries are also willing to welcome pollution-intensive production because it creates more jobs and increases economic growth at the cost of the environment. In the end, such trade-induced income growth can lead to high CO 2 emissions, thereby validating the PHH.
For the FEH, the findings show a positive relationship between TI*KL and CO 2 emissions in columns (1) and (7) which means that openness to international trade induces capital growth in SSA which in turn increase CO 2 emissions in the region. Because, capital moves to where it has a comparative advantage and the comparative advantage is driven by factor abundance, factor endowments including cheap and abundant labour and natural resources in SSA attract more capital especially extractive industries that emit more CO 2 .
Among the control variables, we find that FDI has a positive relationship with PM2.5 emissions and a negative relationship with deforestation. This suggests that in SSA, PM2.5 emissions increase as FDI inflow increases which is similar to the findings by Demena and Afesorgbor (2020), and Bissoon (2018). On the other hand, deforestation falls as the inflow of FDI increases. We can argue that FDI inflow reduces overall dependence on natural forests for survival and provides alternative raw materials and technologies that substitute forestbased raw materials and goods. We also examine the relationship between energy consumption with environmental degradation. The findings suggest that both FFEU and FWEU have a positive impact on air pollution and deforestation respectively which is in line with the findings of Gulistan et al. (2020), Wang and Dong (2019) and Ssali et al. (2019). Finally, our findings suggest that ratification of the Kyoto Protocol has no impact on CO 2 and PM2.5 emissions in SSA, and we can say that the finding is valid is because, by the time SSA ratified the Kyoto protocol, GHGs emissions in the region were still very low and thus, had not so much effect (Le Qu er e et al., 2015). However, based on the findings, the ratification of the Kyoto Protocol is effective in reducing deforestation in SSA. This is because, under the Kyoto protocol, the clean development mechanism (CDM) proposes carbon credits and establishes international carbon markets for within which these carbon credits can be traded (Olsen, 2007) and because SSA has usually emitted less GHGs compared to other regions, companies have bought carbon credits from countries in the region leading to a fall in the deforestation rate thereby indirectly improving the air quality (Cushing et al., 2018).

Conclusions and recommendations
In this study, we examine the empirical relationship between economic growth and environmental degradation in SSA focussing on the validity of the EKC hypothesis, PHH and FEH in SSA. Our findings confirm the existence of the EKC for PM2.5 emissions and deforestation but not for CO 2 emissions despite CO 2 emissions having a positive relationship with economic growth. We also find a negative relationship between trade openness and deforestation which means that SSA countries increase their participation in international trade as deforestation falls. Also, the PHH is valid for CO 2 emission suggesting that tradeinduced economic growth leads to an increase in CO 2 emissions. This is owed to the fact that trade increases technology transfer which in turn increases production leading to more economic growth. However, due to lax environmental regulations in SSA, more pollutionintensive industries are attracted which also increases emissions. Besides, our findings also show that the validity of the FEH in SSA for CO 2 emissions but not for PM2.5 emissions and deforestation.
Whereas the validity of the EKC illustrates that PM2.5 emissions and deforestation will fall as per capita income increases, and it is not the same for CO 2 emissions. The positive relationship between per capita income and environmental degradation indicators illustrates that in the processes of achieving high economic growth, SSA bears the consequence of environmental degradation in the form of CO 2 and PM2.5 emissions as well as deforestation. In addition, the validity of the PHH and FEH means that SSA countries should tighten their environmental standards and enforcement of the standards. This can be through instituting regional environmental standards to act as benchmarks for individual country's environmental standards and establishing a regional body to monitor and enforce the agreed standards. Besides, policymakers in SSA countries should encourage green investments and the adoption of environmentally friendly technologies, which increase economic growth but cause less environmental damage. This can be done by increasing government spending on research and development on green technologies to boost the motivation to create and innovate environment-friendly technologies. In addition, offering tax waivers, exemptions and subsidies to both local and foreign investors bringing green investments to SSA. Further, SSA governments should facilitate the transition from fossil fuels and fuelwood to renewable energy, which causes less harm to the environment. Similarly, SSA governments should improve access to information on the environment as well as encourage knowledge management strategies as a means of creating awareness on environmental issues.
Furthermore, policymakers in SSA should encourage green investments and the adoption of environmental friendly technologies which increase economic growth without degrading the environment. This can be through offering tax waivers, exemptions and subsidies on green technologies for both local and foreign investors. In addition, SSA governments should increase spending on research and development on green technologies to boost the innovation of eco-friendly tools. SSA governments should also facilitate the transition from environmentally damaging fossil fuels and fuelwood energy to environmentally friendly renewable energy use. Further, the creation of easily accessible knowledge management strategies as a means of creating civil awareness about the consequences of environmental Economic growth in Sub-Saharan Africa degradation amongst all sectors through different media is a vital way for all citizens to change attitudes towards environmental protection. This study contributes to the literature on the EKC hypothesis in SSA by using several indicators environmental degradation and confirming that the validity of the EKC, PHH and FEH hypotheses depends on the type of indicator for environmental quality used. We also acknowledge not examining the relationship between economic growth and water pollution because countries in SSA do not have reliable data for water pollution indicators such as dissolved oxygen (DO) and biochemical oxygen demand (BOD). This is a gap in the literature that may be closed in the future when data for such indicators are made publicly available.