Do human capital and governance thresholds matter for the environmental impact of FDI? The evidence from MENA countries

This paper studies whether foreign direct investment (FDI)-CO2 emissions relationship may change depending on the data-driven estimated threshold levels for the country characteristics (CC) including human capital and governance in a sample of 13 Middle East and North Africa (MENA) economies during the 1996–2019 period. Our results strongly suggest that endogenously estimated CC thresholds matter for the impact of FDI on CO2 emissions. The pollution haven hypothesis, which maintains that FDI is associated with higher levels of pollution, appears to be valid for economies with weak CC. In addition to this, the pollution halo argument suggesting FDI lowers the emissions appears to be hold in countries with strong CC. The results in this study may indicate that policies aiming to improve human capital and governance may be expected not only to increase the economic benefits of FDI in terms of growth but also mitigate the negative environmental impacts of FDI in the MENA region.


Introduction
Environmental deterioration has been one of the serious concerns for developing countries containing the Middle East and North Africa (MENA) economies. Abumoghli and Goncalves (2020) draws attention to the substantial utilization of not only fossil fuels but also non-renewable energy sources in production along with prevailing environmental issues such as air pollution, loss of biodiversity, and inadequate waste management. According to World Development Indicators (WDI) database, CO 2 emissions per capita in 1975 were 0.48 kg in the MENA and it is around 0.56 in Organization for Economic Co-operation and Development (OECD) countries. However, the carbon emissions of MENA have increased to 0.73 in 2017 while the carbon emissions of the OECD decreased to 0.22.
High CO 2 emissions may prevent developing countries to achieve their development goals, particularly the Sustainable Development Goals which focus on promoting a green economy. "Greening" is associated with the low-carbon energy transitions aiming not only to access renewable energy sources but also to reduce poverty along with job creation (Siciliano et al. 2021). Therefore, mitigation of CO 2 emissions is crucially important to provide the sustainability of environment. Implementing proactive strategies to mitigate ecosystem vulnerability and ensure the environmental sustainability matter for policymakers in developing countries, including MENA.
The environmental economics literature investigates the possible determinants of environmental degradation. The earliest of these studies consider income per capita Responsible Editor: Eyup Dogan (Grossman and Krueger 1995) and energy consumption (Pao and Tsai 2010) as the important determinants of CO 2 emissions. Afterwards, the effects of many other variables such as technological innovation (Khan et al. 2020), trade openness (Copeland and Taylor 2005), financial development (Ozturk and Acaravci 2013), urbanization (Zhang et al. 2017), and foreign direct investment, FDI, (Levinson and Taylor 2008;Lee 2009) on pollution are investigated. Among these variables, one of the most important, especially for developing countries, is the FDI. The international economics literature maintains that FDI often leads to better growth episodes by providing efficient allocation of capital, access to financial markets and new technology and increasing total factor productivity. While FDI inflows have often been associated with high growth rates, the environmental economics literature often remarks that FDI may provide deleterious results for the environment. The purpose of this article is to study FDIpollution relationship for MENA economies.
The literature investigating the environmental effects of FDI has been centered around two views: The first one is the "pollution haven" hypothesis maintaining that FDI leads to environmental degradation. This may be consistent with the fact that advanced economies locate pollution-intensive activities in developing countries with lax environmental restrictions and regulations by FDI linkages (Levinson and Taylor 2008). The second one is the "pollution halo" hypothesis suggesting that the impact of FDI is environmental improvement (Cole et al. 2011). This may be in line with the argument that international firms with high environmental quality may bring sophisticated, energy-efficient, environmentally cleaner technologies to host economies along with better environmental management systems (Wang and Chen 2014). Therefore, the investigation of FDI inflows and pollution relationship is still one of the important research topics in the literature.
The literature often suggests that the FDI-pollution relationship is invariant to the country characteristics (CC) including human capital and institutional quality and governance levels. However, the investment decisions of multinational firms may be affected by the prevailing CC (Mengistu and Adhikary 2011; Noorbakhsh et al. 2001;Cantwell et al. 2010). The theoretical basis of this argument is based on the endogenous growth model (Romer 1990) and new institutional economics theory (Coase 1937;North 1990;Kostova and Hult 2016).
The endogenous growth model maintains that human capital plays a key leading role in the growth process through technological diffusion and total factor productivity. Technological diffusion and total factor productivity may also be achieved by FDI inflows. Accordingly, Nelson and Phelps (1966) establish technological diffusion model which indicates that the host country needs a high level of human capital for jobs that require learning, adapting and following new technologies. Therefore, it can be assumed that human capital plays a role in FDI inflows as suggested by the empirical studies. For instance, Sekkat and Veganzones-Varoudakis (2007) argue that a certain level of human capital in the host country is necessary to attract FDI inflows. According to Xu (2000), a country's level of human capital determines how intensively technology is transferred to that country through FDI. Borensztein et al. (1998) find that a threshold level of human capital is also required to obtain the growth effect of FDI inflows.
On the other hand, the motivation to focus on the role of institutional quality in the host country in determining FDI inflows is based on the new institutional economics theory. The theory maintains that institutions are one of the indispensable parts of market-based economies to function efficiently (Rutherford 2001). North (1990, p. 3) notes that "institutions are the rules of the game in a society or, more formally, are the humanly devised constraints that shape human interaction." Daniele and Marani (2006) report that the institutional quality is an important driver of FDI inflows as better institutions provide higher productivity, reduce investment-related transaction costs resulting from corruption practices, and prioritize property rights. Also, Kaufmann et al. (2011) andBuckley et al. (2016) suggest that FDI tends to flow to countries where there is a fair and transparent institutional framework and low political risk.
All these theoretical arguments suggest that human capital and governance i.e., CC are crucially important determinants of FDI inflows. Based on these arguments, it can be argued that CC may also affect the impact of FDI on pollution with different transmission mechanisms. First, we consider how the impact of FDI on pollution changes with the level of human capital. In this context, conventional wisdom maintains that more educated labor demands clean environment, encourages the use of renewable energy products, promotes energy efficiency and tends to better adopt environmental regulation as well as greener technology. Then, we regard how the sensitivity of pollution to FDI varies with the level of governance. Considering the theoretical arguments by new institutional economics, we may suggest that economies with well-established rules, norms and regulations allow them to implement environmental protection policies. In this vein, the empirical findings by Wang and Chen (2014), Bokpin (2017), Omri and Hadj (2020), and Bouchoucha (2021) show that improvements in institutions mitigate the negative effect of FDI on pollution. The empirical findings by Tang et al. (2021) suggest the crucial importance of human capital and institutions for sustainable development that prioritize lower pollution. Considering all these issues, we may plausibly assume that the level of CC matters for FDI-pollution relationship. Furthermore, the CC may provide a threshold for the effect of FDI on pollution.
In this context, we analyze FDI inflows and environmental degradation relationship by focusing on the role of CC. First, we investigate the direct effect of FDI on environmental degradation by utilizing panel fixed effects model. Based on the arguments suggested by the literature and theory, it may be misleading to assume that the impact of FDI on pollution is the same in economies with weak country characteristics (CC) and the others. To analyze whether the effect of FDI on pollution can vary with the level of CC, we examine the thresholding effect of CC on FDI-pollution relation by employing panel fixed effects threshold estimation method of Hansen (1999) and dynamic panel threshold estimation procedure of Kremer et al. (2013) for 13 MENA countries during the 1996-2019 period. To our knowledge, this is the first study that maintains data-driven estimated threshold level of CC may affect the FDI-pollution relationship.
Our panel fixed effects estimation results suggest that there is a positive and significant association between FDI inflows and CO 2 emissions. The empirical findings indicate the validity of pollution haven hypothesis. On the other hand, our panel fixed effects threshold estimation results suggest that CC including human capital and institutional quality and governance provide data-driven estimated thresholds in explaining the impact of FDI on pollution. Accordingly, FDI leads to pollution in weak CC economies with less educated labor and unfavorable institutional quality and governance. FDI, on the other hand, enhances environment in economies with better CC including more educated labor and favorable institutional quality and governance. We also employ dynamic panel threshold estimation procedure as a robustness check. By considering the first principal component of human capital and governance as threshold, we find that our panel fixed effects threshold estimation results are robust to dynamic panel threshold estimation procedure. The empirical findings in this study indicate that pollution haven hypothesis stands for economies with weak CC while pollution halo is the case for the economies with strong CC. The results suggest that to reap the environmental enhancing effect of FDI, MENA countries whose growth requires the promotion of FDI inflows, may implement policies aiming to improve human capital and institutional environment.
This article is planned as follows. A brief literature review on FDI-pollution relation is provided in "A Brief Literature Review". "FDI-CO2 Emissions: The Data and Some Descriptive Statistics" explains the data and reports some descriptive statistics. "Empirical Methodology and Estimation Results" introduces the empirical methodology and estimation results. Panel fixed effects estimation results are reported in "FDI-CO2 Relationship: Panel Fixed Effects Estimation Results". In "Panel Fixed Effects Threshold Estimation Method and Empirical Results", we employ panel fixed effects threshold estimation procedure and present the results. "Thresholding Effect of Human Capital" reports the thresholding effect of human capital and "Thresholding Effect of Governance" presents the thresholding effect of governance. As a robustness check, we utilize dynamic panel threshold estimation method which considers endogeneity and report estimation results in "Robustness Check: Dynamic Panel Threshold Estimation Results". Finally, we conclude and provide some policy implications in "Conclusion and Policy Implications".

A brief literature review
Most of the studies in the literature note that FDI is crucially important for growth, albeit the environmental effect of FDI is mixed. The literature on FDI-pollution relation points to the "pollution haven" and "pollution halo" hypotheses. Bashir (2022) briefly reviews the literature on pollution haven hypothesis.
The "pollution haven" hypothesis maintains that FDI inflows may increase growth and thus CO 2 emissions in the host country. This is mainly related with not only higher production caused by FDI but also prevailing lax environmental restrictions and exemptions in developing economies. The former increases energy consumption and CO 2 emissions. The latter suggests that developed countries shift their emission intensive economic activities to developing countries causing them to have a pollution haven (Aliyu 2005). It is argued that FDI inflows from developed to developing countries are used to finance pollution-intensive, environmentally inefficient production and infrastructure projects. (Jorgenson 2009).
FDI, on the other hand, may lower emissions by the transmission of environment-friendly production and management techniques from advanced to developing economies, as suggested by the "pollution halo" hypothesis. FDI inflows may lead to the expansion of less polluting labor-intensive industries by taking advantage of the cheap labor in emerging market economies (He 2006).
The empirical literature often studies whether FDI is related with pollution haven or halo hypotheses. However, this relationship is mixed. Some studies find that FDI increases pollution and thus "pollution haven" hypothesis holds. Mukhopadhyay (2006) finds that Thailand was a pollution haven for OECD countries in 2000. The cointegration-based empirical results by Acharyya (2009) suggest that FDI augments pollution in India during the 1980-2003 period. Zhang (2011) notes that FDI and carbon emissions are positively associated in China. This appears also to be the case for Africa (Kivyiro and Arminen 2014;Bokpin 2017). The empirical findings by Seker et al. (2015), Salahuddin et al. (2018), and Abdouli and Hammami (2017) suggest that environmental degradation effect of FDI is also the case for the members of MENA. Shahbaz et al. (2015) report that FDI lowers pollution in high-income countries while increases it in low-income countries. For middleincome countries, they indicate that thresholding effect of FDI matters for FDI-pollution relation. The empirical results by Wang and Chen (2014) show that FDI inflows to China from OECD countries support the pollution haven hypothesis while those from Hong Kong, Macao, and Taiwan do not lead to environmental degradation. Shahbaz et al. (2019) report that N-shaped relation holds among the FDI and CO 2 emissions for MENA countries. Xie et al. (2020) investigates both direct and indirect effects of FDI on CO 2 emissions. Their results suggest that the direct effect of FDI is to increase CO 2 emissions while FDI lowers emissions indirectly by economic growth.
Some other studies find that "pollution halo" hypothesis holds (List and Co 2000;Mielnik and Goldemberg 2002;Wheeler 2001;Zhu et al. 2016). These studies often suggest that "pollution halo" hypothesis prevails by arguing that FDI inflows from developed economies contribute to energy efficiency in developing countries. Similarly, Wheeler (2001) demonstrates that FDI decelerates air pollution in Brazil, China, and Mexico. Using panel quantile regressions, Zhu et al. (2016) also support the "pollution halo" hypothesis for the Association of Southeast Asian Nations (ASEAN) economies. Muhammad and Long (2021) show that pollution halo is the case for high income economies. The panel fixed effects threshold estimation results by Aluko et al. (2021) show that the level of income and globalization provide thresholds for the effect of FDI on pollution. They find that FDI lowers pollution in economies with lower levels of income and globalization while it leads to environmental degradation in higher levels of income and more globalized countries.
The empirical literature often ignores the effects of human capital and institutional quality and governance for FDI-pollution relationship. Conventional wisdom maintains that more educated labor demands clean environment, promotes the use of renewable energy products, energy efficiency and tends to better adopt environmental regulation as well as greener technology. As consistent with this argument, the empirical literature investigating the human capital-environment relationship suggests that pollution is lower in economies with better educated labor. For example, human capital leads to less pollution in Pakistan according to Bano et al. (2018). Also, the results by Ahmed and Wang (2019) show that human capital decreases ecological footprint in India. Shobande and Asongu (2021) find that human capital development mitigates CO 2 emissions. Lan et al. (2012) report that FDI diminishes pollution for the provinces of China with better educated labor.
According to institution-based approach, the institutional structure of the host country is effective for the environmental impacts of FDI inflows (Wang and Chen 2014;Abid 2016;Ali et al. 2019). In economies with better institutions, environmental rules are clear, transparent, consistent and strict. (Wang and Chen 2014). Shojaeenia et al. (2022) find that better institutions promote the environment-friendly energy consumption as compared to conventional dirty energy sources. The presence of a good institutional environment which is an indicator for the implementation of better environmental protection policies, has been led multinational corporations to invest in environmentally friendly technologies and implement more responsible waste creation and management (King and Shaver 2001;Christmann 2004). In this vein, it may be plausible to assume the impact of FDI is environmental enhancing in economies with better institutions. In a more institutionally sound environment, local businesses tend to increase their efficiency and innovation to compete with multinational firms (Wang and Chen 2014). Bokpin (2017), Omri and Hadj (2020), Bakhsh et al. (2021) and Bouchoucha (2021) find that good governance is essential for FDI inflows to decrease carbon emissions.

FDI-CO 2 emissions: the data and some descriptive statistics
This article examines the relationship between foreign direct investment (FDI) inflows and CO 2 emissions. We analyze this crucially important question for a balanced panel of 13 Middle East and North Africa (MENA) economies (Algeria, Bahrain, Egypt, Iran, Israel, Jordan, Kuwait, Morocco, Qatar, Saudi Arabia, Tunisia, Turkey and the United Arab Emirates) over the 1996-2019 period. Our sample is determined by the availability of the data. Considering that MENA cannot be treated as a single unit because of the heterogeneity in natural resource endowments, we study the FDI-CO 2 emissions relationship also for the samples of oil-exporting (Algeria, Bahrain, Egypt, Iran, Kuwait, Qatar, Saudi Arabia, and the United Arab Emirates) and oil-importing (Israel, Jordan, Morocco, Tunisia, and Turkey) MENA.
In this study, CO 2 stands for the log. of CO 2 emissions in terms of per tones per capita, GDPpc shows the log. of real GDP per capita, REC represents the renewable energy consumption as a percent of total energy consumption, GOV corresponds to the institutional quality and governance, HC is the human capital index, and FDI denotes the foreign direct investment inflows as a percent of GDP. The data for CO 2 emissions are from Joint Research Centre Emissions Database for Global Atmospheric Research.
GDPpc and FDI data are taken from United Nations Conference on Trade and Development database. The data for GOV are taken from World Bank Governance Indicators. The governance and institutional quality data consider the six characteristics including voice and accountability, political stability and violence, government effectiveness, regulatory quality, rule of law and control of corruption (Kaufmann et al. 2005). These variables are between − 2.5 and 2.5 with higher values denoting better institutional quality and governance. We consider a simple average of six variables to provide an indicator for institutional quality and governance. HC is the human capital index measured as not only the schooling years but also educational returns. The data for HC are from Penn World Table (Feenstra et al. 2015). Table 1 presents some descriptive statistics for our variables of interest. Accordingly, the mean of CO 2 emissions per tones per capita 1 is around 13.14 for the whole sample, 18.83 for the oil-exporting MENA, and 4.05 for the oil-importing MENA. As compared to the oil-importing MENA, CO 2 emissions are much higher and more volatile in oil-exporting MENA. The mean FDI inflows is around 2.7 for the whole sample, albeit it is slightly much higher in oil-importing MENA economies. The volatility of FDI is almost the same both in oil-exporting and oil-importing MENA. The mean income per capita is around 19,000$ for the whole sample. In comparison to the oil-importing MENA, the mean and volatility of income are much higher for the oil-exporting MENA. On the other hand, the mean renewable energy consumption is higher in oil-importing MENA economies. The means of human capital and governance are almost the same in both sub-samples.

FDI-CO 2 relationship: panel fixed effects estimation results
We first analyze the drivers of CO 2 emissions. As consistent with the environmental economics literature, our benchmark equation is: In Eq. (1), the subscript i represents the countries, the subscript t shows the years, CO2 shows the log. of CO 2 emissions, GDPpc stands for the logarithm of real income per capita, REC indicates the renewable energy consumption as a percent of total energy consumption, GOV presents the average of six aspects of institutional quality and governance, HC is the human capital index, and FDI shows the foreign direct investment inflows as a percent of GDP.
In accordance with the literature, Eq.
(1) maintains that CO 2 emissions can be explained by income, renewable energy consumption, governance and institutional quality, human capital, and FDI inflows. To analyze the relationship between FDI and pollution, we control the impacts of income, renewable energy consumption, governance, and human capital as suggested by the environmental economics literature. Considering the conventional environmental Kuznets curve maintaining that CO 2 emissions first increase and then decrease with income, real GDP per capita may be potentially endogenous for CO 2 emissions. To take this issue into account, we consider lagged income per capita. Table 2 reports the panel fixed effects estimation results of Eq. (1). Considering the heterogeneity in natural resource endowments, we estimate Eq. (1) also for the samples of (1) oil-exporting and oil-importing MENA. According to all the estimation results in Table 2, there is a positive and significant association between FDI inflows and CO 2 emissions. As compared to oil-importing MENA, this relationship is much higher in oil-exporting MENA. This suggests that the increase in FDI leads to environmental degradation by increasing CO 2 emissions. This result proves that the pollution haven hypothesis is valid for whole sample, oil-exporter, and oil-importer MENA economies. Table 2 also shows that the income elasticity of pollution is positive and significant. This indicates that an increase in income leads to higher CO 2 emissions. This result is consistent with the scale effect argument by Grossman and Krueger (1995). On the other hand, the estimated coefficient of REC is negative and significant for the oil-importing MENA economies. This suggests that renewable energy consumption appears to lower CO 2 emissions. The direct effect of institutional quality and governance is associated with a decrease in CO 2 emissions, except the sample of oil-exporting MENA. An increase in human capital tends to diminish CO 2 emissions in oil-importing MENA and whole sample.

Panel fixed effects threshold estimation method and empirical results
In Eq.
(1), we analyze the direct effect of FDI inflows on CO 2 emissions. The estimation results in Table 2 show that there is a positively significant relationship between FDI inflows and pollution. This suggests the validity of the pollution haven hypothesis. Equation (1) assumes that FDI-CO 2 emissions relationship is invariant to the country characteristics (CC) including human capital and institutional quality levels. However, the conventional wisdom maintains that more educated labor demands clean environment, promotes the use of renewable energy products, energy efficiency and tends to better adopt environmental regulation. Kwon (2009) notes that better educated labor provides efficiency related solutions like employment of emission reduction technologies. On the other hand, wellestablished rules, norms, and regulations for the environment may affect the sensitivity of CO 2 emissions to FDI inflows. In this vein, economies having better institutions implement policies to protect the environment. All of these may indicate that FDI-pollution relationship may not be the same in economies with better CC and weak CC. Furthermore, the CC consisting of human capital (HC) and institutional quality and governance (GOV) may provide data-driven estimated thresholds in explaining the impact of FDI on pollution. Thus, we regard the following equation: Alternatively, we can also specify Eq.
(2) as the following: Under the null hypothesis that α 5 = α 6 in Eqs. (2) and (3), there is no significant thresholding effect of CC for the FDI-pollution relation, and we obtain Eq. (1). First, we trim 5% of the smallest and largest observations and investigate the threshold by considering the rest of the observations as potential candidate. For each one of the potential candidates, we utilize panel least squares and determine the threshold that gives the minimum sum of squared residuals. The endogenously (data-driven) estimated CC threshold, λ, allows the whole sample to partition into the two regimes. For instance, if CC ≤ λ, the estimated parameter α 5 represents FDI-pollution relation in the low regime containing weak CC. Otherwise, if CC > λ, the estimated parameter α 6 indicates the FDI-CO 2 emissions relationship in the high regime consisting of better CC. We estimate Eq. (2) for 13 MENA economies during the 1996-2019 period by utilizing panel threshold estimation procedure by Hansen (1999). (2)

Thresholding effect of human capital
First, we investigate the thresholding effect of HC. Our estimated equation is as follows: Equations (4.1), (4.2), and (4.3) in Table 3 provide the panel fixed effects threshold estimation results of Eq. (4), respectively, for the whole sample, oil-exporting, and oilimporting MENA. Accordingly, human capital provides data-driven estimated threshold for the effect of FDI on CO 2 emissions. The threshold level of human capital is around 2.7 for the whole sample, 2.6 for the oil-exporting MENA and 3.4 for the oil-importing MENA. Table 1 reports that the mean of human capital is around 2.4. The data-driven estimated threshold is approximately the same with the mean of human capital for the whole sample and oil-exporting MENA, albeit it is slightly higher for oil-importing MENA. Almost 20% of the observations belong to the high regime including more educated labor episodes.
The estimation results in Table 3 suggest that FDI and CO 2 emissions are positively associated in the low regime including observations with less educated labor. On the other hand, for all equations, the sensitivity of pollution to FDI is negative and significant in the high regime including more educated labor episodes. Consistent with the results by Lan (4) al. (2012), this empirical finding may show that pollution haven tends to be the case for economies with less educated labor while pollution halo holds in economies with better educated labor. 2 An increase in income which is the aggregated measure of economic activities leads to higher CO 2 emissions. This is consistent with the scale effect explanations by Grossman and Krueger (1995) maintaining higher income is associated with deterioration in environment. A rise in REC lowers CO 2 emissions for the oil-importing MENA. An improvement in institutional quality and governance decreases CO 2 emissions in oil-importing MENA economies and whole sample. Human capital and pollution are negatively associated for the oil-importing MENA and whole sample. This empirical finding is in concordance with the results by Lan et al. (2012), Bano et al. (2018), and Ahmed and Wang (2019) suggesting that better educated labor tends to diminish the pollution potentially by advocating the use of renewable energy products, energy efficiency and adopting environmental regulation.

Thresholding effect of governance
We now proceed with the investigation of thresholding effect of governance (GOV) for the impact of FDI on CO 2 emissions. In this vein, we estimate: Equations (5.1), (5.2), and (5.3) in Table 4 provide the results of Eq. (5), respectively, for the whole sample, oilexporting, and oil-importing MENA economies. The results by Table 4 suggest that FDI-pollution relationship may change with respect to the threshold values for GOV. 3 The endogenously estimated threshold level is around 0.4 for the whole sample and oil-exporting MENA and 0.6 for the oilimporting MENA. These threshold values are almost the same with the mean of GOV as reported by Table 1. Almost 20% of the observations belong to the high regime including observations with better institutional environment. FDI inflows lead to higher pollution in the low regimes including weak governance episodes. However, FDI diminishes CO 2 emissions in the high regime containing better governance (5) observations. Our estimation results may indicate that pollution haven hypothesis holds for the countries with weak institutional levels while pollution halo hypothesis appears to be hold in economies with better institutional environment. 4 This finding is mainly in accord with the results by Bokpin (2017) suggesting that institutions have a responsibility to guarantee that FDI's environmental effects remain within a regulated framework. We find that income elasticity of pollution is positively significant indicating the impact of income is associated with higher pollution. For the oil-importing MENA economies, the higher the renewable energy consumption, it is lower the CO 2 emissions. The direct impact of GOV is associated with lower pollution in oil-importing MENA economies and whole sample. This finding may suggest that better institutional quality causes the countries to implement environmental protection policies and diminishes the emissions. This result is mainly in accord with the findings by Wang and Chen (2014), Bokpin (2017), Omri and Hadj (2020), and Bouchoucha (2021). Better educated labor tends to decrease the pollution in oil-importing MENA and whole sample.

Robustness check: dynamic panel threshold estimation results
This section aims to provide a robustness check for our earlier results. To investigate the robustness of panel fixed effects threshold results, we employ dynamic panel threshold estimation method 5 that explicitly considers the endogeneity (Kremer et al. 2013). First, we specify the dynamic version of Eq. (2). Then, we consider the first principal component of human capital and governance, PC, 6 as the thresholding variable. 7 We estimate the following dynamic equation: In Eq. (6), we examine whether the sensitivity of CO 2 emissions to FDI may change with respect to the thresholding effect of PC. To examine this important issue, we control the impacts of income per capita, renewable energy consumption, institutional quality and governance and human capital. Table 5 reports dynamic panel threshold estimation results. Accordingly, the endogenously estimated threshold level of PC is around 1.04 for the oil-exporting MENA and whole sample, albeit it is slightly higher for oil-importing MENA and it is around 1.08. The endogenously estimated threshold levels are contained within the 95% confidence interval. This can show the statistically significant thresholding effect of PC.
The estimated coefficient for lagged CO 2 emissions is positive and significant for all equations in Table 5. This may suggest that there is a convergence in CO 2 emissions for the (6)  Wald test [p-value] = 0.00 5 The first step of dynamic panel threshold procedure is to remove fixed effects by forward orthogonal deviation transformation. Then, a reduced form regression is obtained for the endogenous variables. By substituting the predicted values into Eq. (6), Hansen (1999) method is applied to find the threshold. Following the identification of threshold, we estimate slope parameters by employing the generalized method of moments procedure. 6 The principal component analysis retains the variations and linear combinations of the variables and reduces the dimension of the data. In the Appendix, Table 7 shows our panel fixed effects threshold estimation results by using the PC as the thresholding variable. The results in Table 7 are almost the same with our earlier findings. 7 The selection of thresholding variable that affects the sensitivity of pollution to FDI inflows have been determined based on the theory and literature. Accordingly, human capital and institutional quality and governance are crucially important variables that influence the foreign investment decisions of multinational companies. Trade openness, natural resource endowments, wages, and tax rates etc. can be considered the other variables that effect FDI inflows. The empirical investigation of whether these variables provide data-driven estimated thresholds for the sensitivity of pollution to FDI inflows is an encouraging research agenda for the forthcoming studies.
Footnote 7 (continued) MENA economies. In a similar vein to our earlier findings, the sensitivity of CO 2 emissions to FDI inflows changes with respect to the thresholding effect of PC. Accordingly, FDI increases pollution in the low regime including observations with less educated labor and poor institutions. However, FDI lowers pollution in the high regime including observations with better educated labor and better institutions. As consistent with panel fixed effects threshold estimation results, dynamic panel threshold estimation results indicate that pollution haven hypothesis is valid in economies with weak country characteristics while pollution halo hypothesis is the case in countries with better country characteristics. The estimated parameters for the rest of the variables are essentially the same with our previous findings.

Conclusion and policy implications
The international economics literature maintains that foreign direct investments (FDI) bring many benefits including better growth episodes, access to financial markets, and new technology along with higher total factor productivity. However, the environmental economics literature often reports the validity of either pollution haven or halo hypothesis for the impact of FDI inflows on pollution. Hence, this article examines the FDI-pollution relationship for MENA, which suffers from low growth rates on the one hand, and various environmental problems, such as fossil fuel use, water scarcity, air pollution, and degradation of coastal ecosystems on the other hand (Abumoghli and Goncalves 2020). Our study explores whether country characteristics (CC) such as human capital and institutional quality and governance levels play a role in the effects of FDI on pollution, a topic often ignored by the literature. First, we employ panel fixed effects estimation procedure to examine the direct impact of FDI inflows on pollution. Our estimation results show that FDI inflows lead to an increase in CO 2 emissions in MENA economies, demonstrating the validity of the pollution haven hypothesis. The estimation results also show that the direct effects of CC on CO 2 emissions are significantly negative. This finding indicates that an increase in years of schooling and improvement in institutions i.e., better CC decrease pollution.
Then, we consider the thresholding effects of CC for the impact of FDI on CO 2 emissions. Our panel fixed effects threshold estimation results support an argument that FDIpollution relationship may change depending on the level of country characteristics (CC), including human capital and governance. According to our estimation results, CC provides data-driven estimated thresholds for FDI-pollution relation in MENA. This appears also be the case for the samples of oil-exporting and oil-importing MENA. Accordingly, FDI leads to more pollution in economies with weak CC containing less educated labor and worse institutional environments. On the other hand, FDI enhances the environmental quality in countries with strong CC including better educated labor and institutional environment. These empirical findings may suggest that pollution haven holds in weak CC economies while pollution halo is the case for the economies with strong CC.
The findings of this paper imply that policies aiming to improve human capital and institutional environment may be expected to enrich not only the economic benefits of FDI in terms of growth but also mitigate negative environmental effects of FDI in MENA. Investing in human capital also eases the employment of environment-friendly technologies and increases the environmental awareness. Moreover, human capital is one of the essential ingredients for green growth (Acemoglu et al. 2012). Consistent with an argument maintaining good institutional environment is closely associated with the implementation of better environmental protection policies, improvement in institutional quality and governance is expected to reduce both the emissions and the degradational effect of FDI inflows.
Based on the empirical findings in this paper, the policies aiming to promote energy efficiency, energy conservation and emissions diminishing technologies may alleviate the procyclicality of pollution to income. These policies may lead the countries to succeed in the sustainable development goals which promote the green economy. "Greening" may also be considered the low-carbon energy transitions aiming access to renewable energy sources and reduction in poverty along with the job creation. All these may contribute to environmental sustainability and sustainable development goals.
The environmental management systems aiming to reduce emissions may require the institutional and regulatory reforms, green investment, better governance, regional cooperation, and participation of all stakeholders (Abumoghli and Goncalves 2020). Xing and Kolstad (2002) suggests the necessity of cooperative solutions to overcome the pollution since the environmental policy gap leads the movement of pollutive production activities to economies with lax environmental regulations. Implementing proactive strategies to mitigate pollution is critical for policymakers in developing countries, including MENA countries. Future studies investigating whether our empirical findings for MENA are robust to different samples like developed, emerging market and developing economies may be considered a promising research topic. This may be extended by employing sector-specific FDI inflows and other pollution measures. Furthermore, investigating the thresholding effect of country characteristics for the sensitivity of pollution to FDI by using some other alternative endogenously estimated threshold procedures appears to be an important research topic most potentially covering our empirical findings presented by this study. Tables 6 and 7   Table 6 Thresholding effect of the main components of governance * < 10%, ** < 5%, *** < 1% F B is the bootstrapped F-test results with p-value in the brackets. The robust standard errors are in parentheses. N and NT are, respectively, the numbers of countries and observations