key: cord-0993162-ng002ydl
authors: Marques, António Cardoso; Caetano, Rafaela
title: The impact of foreign direct investment on emission reduction targets: evidence from high- and middle-income countries
date: 2020-08-29
journal: nan
DOI: 10.1016/j.strueco.2020.08.005
sha: 9414661c4d609ac2952e11a7b6fb7c0d228ad301
doc_id: 993162
cord_uid: ng002ydl

Besides bringing countries closer, the effects of globalization can help increase the production of goods and services, and foster economic growth. Foreign direct investment (FDI) is one of the processes of globalization. One aspect of globalization that has piqued the interest of economists, is the transfer of polluting industries between countries. A principal factor in this are discrepancies of environmental regulations, and these have also been instrumental in a failure to control pollution worldwide. With this impasse in mind, a Panel Autoregressive Distributed Lag was applied to evaluate the impacts of FDI on the carbon dioxide emissions of 21 countries divided by income level, for a period from 2001 to 2017. This methodology allowed the analysis of the resulting dynamics of pollution into the short-run and long-run. The characteristics of efficiency, innovation, and regulation are crucial to better understand the consequences of flows in FDI. Regulation seems to increase pollution in high-income countries, which merits further discussion. FDI decreases emissions in high-income countries, while increasing them in the short-run in middle-income countries, which supports the Pollution Haven Hypothesis. Nonetheless, the capacity of middle-income countries to absorb technology is crucial for them to benefit in the long-run. Trade openness is also highly influenced by environmental regulation in middle-income countries. Since our aim is to understand the transfer of polluting industries, an analysis of emissions from the industrial sector provided a robustness check. It also revealed that policymakers do not seem to be paying sufficient attention to innovation and controlling the environmental degradation that this sector causes.

A country that wants to transfer green technologies and knowledge through FDI will evaluate the capabilities of prospective recipient countries, such as their capacity for innovation and efficiency. Conversely, a country that wants to relocate its polluting industries to another country to circumvent environmental taxes, will search for a country with lower environmental regulations. This paper contributes by expanding knowledge in this area, by providing empirical evidence that reveals that regulatory measures seem to be less effective in reducing emissions in high-income countries. This surprising finding strongly suggests that further discussion of the regulatory structure of these countries is needed, and should be further investigated. The study also found that trade openness is highly influenced by regulation in middle-income countries. Apparently, these countries are almost entirely focused on developing new patents with the aim of increasing their income, and not to improve environmental performance. Furthermore, the dichotomy found between short-and long-run impacts, is further indication of the need to use the ARDL model, as a failure to use this type of dynamic analysis could well result in misleading results. Indeed, FDI was shown to decrease overall pollution and from the pollution from the industry sector in high-income countries, while increasing it in middle-income countries in the short-run. This suggests that the transfer of polluting industries from high to middle-income countries confirms the PHH. However, middle-income countries could also be receiving clean FDI but, as will be discussed in this paper, their capacity to absorb technology also influences the impact of FDI on the environment. The ARDL model also provides better evidence on this technology-absorptive-capacity of countries, as it analyses the impacts over time.

The rest of the paper is organized as follows: Section 2 presents a theoretical background; Section 3 discloses the data and methodology; Section 4 reveals the results and discussion about overall CO 2 emissions and those from industry, with a brief comparison between models; and Section 5 concludes.

The industrial structure of countries has been continually evolving since the industrial revolution. As industrial output has increased, so have CO 2 emissions, and with that, environmental degradation has emerged. FDI has an important role in economic growth (Omri et al., 2014) . However, besides the effect on economic growth, FDI also impacts the environment of host countries. The relationship between FDI and the environment was initially analysed by academics through the impact of international trade on the environment, and subsequently, by considering the effect of FDI on the environmental quality of host countries (Shahbaz et al., 2015) . International trade (reflected in trade openness) is still of great value and its effects are commonly evaluated in this field (e.g., Essandoh et al., 2020; Sbia et al., 2014) . In addition to its high correlation and Granger causality with FDI, trade openness also has an effect on pollution. Essandoh et al., (2020) find that trade openness is environmentally favourable for developed countries, with no evidence of impact in developing countries. However, as noted by Ren et al., (2014) , trade openness could harm the environment if countries have a comparative advantage in dirty production due to weak environmental regulations. The irresponsible behaviour of certain countries in order to increase their income, increases pollution, impacting the global environment, and ultimately leads to climate change. However, unlike the economic recessions the world encounter, climate changes may be irreversible (Doytch & Uctum, 2016) .

The characteristics of a country matter when evaluating the impact of FDI on the environment, because FDI, per se, does not have an impact. There is contrasting empirical evidence in the literature about the effect of FDI on the environment. For some countries, FDI has a positive and significant role in reducing emissions due to the transfer and adoption of greener technology. This technology boosts efficiency (Pao & Tsai, 2011) and improves environmental quality. However, in other countries, FDI increases CO 2 emissions (Ren et al., 2014) contributing to environmental degradation. In addition to the different characteristics of the countries, the adoption of different empirical methodologies or different periods may also impact the resulting conclusions (Zhang & Zhou, 2016) .

In several studies, the effect of FDI on the environment is divided into three categories: technique, structural, and scale effects (see Liang, 2009; He, 2008; Cole & Elliott, 2003) . The technique effect is based on the diffusion of new and more efficient machines, for example, though FDI (Pazienza, 2019) , which decreases the emissions per unit of a good produced. This effect also suggests that the introduction of environmental regulations can improve the environment by decreasing emissions (Shahbaz et al., 2018) . The structural effect is related to the characteristics of an economy. For instance, an economy whose production of goods is energy-intensive, consumes more energy than an economy specializing in the services sector (Pazienza, 2019) . Succinctly, the effect of FDI depends on the comparative advantage and specialization of the sectors of an economy (Shahbaz et al., 2018) . Lastly, the scale effect states that as FDI increases the industrial output of the host country, it also increases energy consumption and CO 2 emissions (Pao & Tsai, 2011) .

A country's characteristics may amplify the impacts of FDI. It can harm the environment in countries with lower environmental awareness (Xing & Kolstad, 2002) , or improve environmental quality in countries more conscious of its importance as referred by Demena & Afesorgbor (2019) . These characteristics are more exploited in the three main hypotheses associated with the FDI-environment nexus: the Porter, Pollution Halo, and Pollution Haven hypotheses. The Porter hypothesis states that FDI can improve the environmental quality of host countries by introducing new technologies that consume less energy; the so-called eco-friendly technologies (C. Zhang & Zhou, 2016; Mielnik & Goldemberg, 2002) . The main impetus for this are environmental regulations that encourage firms to invest in green innovation (Shen et al., 2020) to improve their efficiency. However, both regulation and the development of new technologies lead to an increase in costs. Thus, it is important to bear in mind the cost-benefit analysis, because regulations can only promote innovation when their benefits exceed their costs. In other words, to benefit the environment, regulation should offset the cost of environmental compliance by improving the competitiveness of firms (Shen et al., 2020) .

The Pollution Halo Hypothesis is based on the positive effect that FDI has on the environment. The transfer of new technologies that can decrease energy consumption (Mielnik & Goldemberg, 2002) , and the transfer of business knowledge-so-called "know-how"- (Shahbaz et al., 2015) are some examples of benefits that FDI can bring to countries if multinationals are less pollution-intensive (Cole et al., 2011) . The countries' characteristics are relevant to determining this impact, because countries with higher levels of environmental awareness are unlikely to accept polluting FDI. Finally, the PHH states that FDI harms the environment in host countries (e.g., Baek, 2016; Al-mulali, 2012) . Countries with strict environmental regulations transfer their polluting industries to countries with more relaxed environmental laws, to avoid additional costs and taxes (Shahbaz et al., 2015) . However, the PHH only occurs when it is relatively easy and inexpensive to transfer the industries. This divides industries into two groups as Dou & Han, (2019) state: strongly-, and weakly-mobile pollution industries. Strongly-mobile industries will be relocated when environmental regulations become more stringent, whereas weakly-mobile industries will invest in R&D to improve efficiency; and effect known as "Innovation Compensation" (Dou & Han, 2019).

Thus, environmental regulation is the main focus of these two hypotheses, which state that the effect of FDI on pollution depends on the level of environmental regulation of host countries. Some literature links the environmental regulation and environmental performance of the countries, mainly reflected in their levels of pollution. On the one hand, environmental regulation may improve environmental quality, by increasing the productivity and efficiency of the country's firms, mainly through saving energy (N. Zhang & Choi, 2013) , which means that regulation can decrease pollution (Hashmi & Alam, 2019) . On the other hand, environmental regulation could produce an inhibitory effect on green innovation for some firms and industries, by imposing additional costs (Gray & Shadbegian, 2003) . This negative effect could result in the transfer of industries to countries with lower environmental standards (Z. Dong et al., 2020) , as suggested by the PHH. Energy-efficiency and innovation emerge as important factors in the debate about the effects of environmental regulation on pollution. The extensive literature about energy consumption states that it is detrimental to the environment, which means that advances in energy-efficiency could help reduce pollution (Balsalobre-Lorente et al., 2019) . Inflows of clean FDI, and investment in R&D targeting innovation, could both improve energy-efficiency. Therefore, policies on energy innovation are required, since economic growth cannot improve pollution by itself, and the continued promotion of energy innovation can reduce emissions (Balsalobre-Lorente et al., 2017) . Furthermore, innovation can help countries improve their environmental performance by reducing the cost of environmental compliance, and could also help countries boost their sustainable economic growth (Balsalobre-Lorente et al., 2019) .

To analyse the full impact of FDI on the environment, it is crucial to consider variables that represent the specific characteristics of the host countries, complemented with a transversal consideration of their impacts over time. In this paper, the ARDL is used to distinguish between short-and long-run impacts. To better understand the effect of inward FDI on host countries, this study empirically considers their levels of innovation, efficiency, and regulation, thus filling an important gap in the literature. Source countries evaluate specific features of a country before providing investment there, and these features are considered in this paper. Countries that want to transfer polluting industries will check if environmental restrictions are more relaxed in the potential host country. However, countries that want to increase global access to technology, by transferring their green technologies to another country, will check the efficiency, and innovation capabilities of host countries. As polluting industries are generally transferred between countries with different levels of development and income, the countries in the study were divided by income level to better investigate this transfer. Analysis of the industry sector was found to produce stronger evidence of the PHH and could serve as a robustness check. The results suggest that policymakers should realize that policies directly related to FDI should be carefully considered, because they not only affect FDI, but also the environment.

In this paper, a panel of 21 countries was studied, namely: Argentina, Austria, Bulgaria, Croatia, the Czech Republic, France, Germany, Greece, Hungary, Ireland, Japan, the Netherlands, Norway, Peru, Philippines, Portugal, Romania, South Africa, Spain, Turkey, and the United Kingdom. The period studied was from 2001 to 2017. The sample was selected according to the availability of data to create as larger a panel as possible. Data on the environmentally-related tax revenue variable is less available for low-middle and upper-middle countries, as is that for industrial production. The lack of data on these variables reduced the number of countries under scrutiny, particularly middle-income countries. The countries were divided into high-or middleincome countries, according to the World Bank's classification (see more in Table Appendix 1 (A1)). Given the limited availability of data on the variables for lower-middle-income and upper-middle-income countries, both the upper-and lower-middle countries were combined in the same group. As mentioned above, the efficiency level of the countries is important for evaluating the impacts of FDI. In this study, the energy efficiency index represents the industrial efficiency of the countries and was calculated using equation (1). This concept was developed by Patterson (1996) and reveals how many units of input are necessary to produce an output unit (Marques et al., 2019) .

(1) Industrial production was used as a proxy of the Industrial Production Index (IPI) to represent the output. Energy consumption from industry in kWh was also used. Recent studies indicate that the incorporation of energy consumption in CO 2 emissions regressions could produce biased results (Jaforullah & King, 2017) , and for this reason, energy efficiency was considered more suitable.

In the literature, CO 2 emissions are commonly used as a proxy of environmental pollution. CO 2 emissions from the industrial sector represent the environmental pollution derived from the industry. Gross fixed capital formation (GFCF) was used as a proxy of a country's economic performance. GFCF could be related to capital intensive industries, as increasing the capital invested in a production process generally results in higher energy consumption that may, in turn, increase pollution (Sapkota & Bastola, 2017) . Trade as a share of GDP (TO) represents trade openness, as is usual in the literature (e.g., Essandoh et al., 2020; Sbia et al., 2014) , and has a higher correlation with FDI, and also Granger Causality, meaning that trade is, most probably, related to FDI. Environmentally-related tax revenue (REG) is used as a proxy for environmental regulation (Hashmi & Alam, 2019) . Environmental regulations may stimulate innovation as suggested by the literature (Kneller & Manderson, 2012; Lee et al., 2011; Johnstone et al., 2010) , which could mean that countries with a high level of innovation do not admit polluting or inefficient industries. To evaluate this, patents (PAT) were used to measure the innovation level of the countries (Burhan et al., 2017).

All numerical variables are converted into per capita, and then into natural logarithms. As they represented a share of GDP, the variables TO and REG are directly converted into natural logarithms. The descriptive statistics of the variables are shown in Table A2 .

Preliminary tests were carried out to assess the presence of multicollinearity, collinearity, and the cross-sectional dependence of variables. To do this, correlation matrices, VIF, and cross-sectional dependence tests were used. Cross-sectional dependence (CD) must be checked in panel data studies since, if it is present, this means that the observations on individual countries are no longer independent, but affect each other's outcomes. This must be corrected as it can produce misleading results (De Hoyos & Sarafidis, 2006) . The null hypothesis of the crosssectional dependence test proposed by Pesaran (2004) is cross-sectional independence. Tables A3 and A4 reveal that none of these phenomena are a concern. First-generation unit root tests may not be effective for assessing the order of integration of the variables in the presence of individual CD, as stated by Pesaran (2007) . Therefore, both first-generation unit roots tests (Levin et al., 2002; Maddala & Wu, 1999 ) and a second-generation unit roots test, crosssectional augmented IPS (CIPS) (Pesaran, 2007) , were carried out, following Shahbaz et al., (2015) . These tests suggested that all variables are stationary in their levelsi.e. I(0) -and on their first differences -I(1) -, which further vindicates the use of the ARDL model.

The ARDL model was proposed by Pesaran et al., (2001) . The main motivation for using this methodology is that it allows an analysis of the dynamic effects of the variables, by analysing effects in the short-and long-run. The specification of the ARDL model is the following:

(2)

To capture the dynamic relationships between variables, the parameters of Equation (2) were reparametrized to the following equation:

(3)

The prefix "D" represents the first differences of variables, and "L" the natural logarithm. is an intercept. are the short-run coefficients of the explanatory variables, are the long-run outputs, t refers to the period analysed in years, i represents the cross (countries), and is the error term. LCO2 it-1 represents the Error Correction Mechanism (ECM), that is, the long-run coefficient of the lagged dependent variable.

To avoid biased results, the Robust Hausman test was carried out (see, e.g., Neves et al., 2017) with 20 bootstrap repetitions to check for the presence of the individual effects by countries. This test was carried out instead of a traditional Hausman test, because it more appropriated in the presence of heteroskedasticity and/or serial correlation (Neves et al., 2017) . The results showed that the fixed effects estimator was suitable to use, and it also highlighted the existence of individual effects. Moreover, in this section, three other diagnostic tests were carried out on the residuals, to further analyse the data's characteristics, namely: (i) the Modified Wald test to verify the existence heteroskedasticity with the null hypothesis of homoskedasticity; (ii) the Breusch Pagan LM test to analyse the cross-sectional correlation with the null hypothesis of cross-sectional independence; and (iii) the Wooldridge to test existence of first-order serial autocorrelation with the null hypothesis of no first-order serial autocorrelation. The existence of cross-section dependence, first-order serial correlation, and heteroskedasticity in the high-income countries model, allowed the use of the Driscoll & Kraay (1998) estimator (DK), as this estimator produces robust standard errors with these characteristics and allows the utilization of fixed effects (Neves et al., 2017) . The DK was also used in the middle-income countries model.

This section consists of three subsections. The first two reveal the results and discussion for both overall CO 2 emissions and CO 2 emissions from the industry. The majority of the results from the analyses of CO 2 emissions from the industry sector corroborated the results from overall CO 2 emissions, thereby providing a robustness check. Furthermore, the analysis of the industry sector provides useful additional information about the impact on pollution of the sector's levels of innovation and regulation, which merits considerable attention; This issue is further explored in the final subsection in a comparison between the models.

The long-run elasticities were derived from the ratio between the coefficient of the respective variable and the ECM; both lagged once, and this ratio was subsequently multiplied by -1. The socio-economic context of the countries was considered and controlled through the inclusion of impulse dummies. Failing to consider socio-economic events could produce misleading results. Consequently, the Zivot & Andrews (1992) (ZA) unit root test was performed to verify the existence of any structural breaks, and the results are presented in Table  A6 , in the appendix. Jointly considering the results of the ZA test, an analysis of each country's socio-economic context, and an analysis of the residuals, the milestones were identified and evaluated. In addition, a test of overall significance was carried out, with the null hypothesis that the coefficients of dummies are equal to 0 (see table A7 ).

Following the United Nations Framework Convention on Climate Change (UNFCCC), Norway registered an increase in its CO 2 emissions of 33% in 2010, compared to 1990. In 2010, Portugal installed 10% more renewable energy capacity, and its emissions reduced by 5.5%. According to Statistics Portugal (INE) 1 , Portugal registered 44 green patents that year. These improvements are allied to the first commitment period of the Kyoto protocol. Bulgaria increased its energy consumption in 2011 and was considered the economy with the highest level of energy intensity in 2010, according to the European Commission 2 . As stated in the UNFCCC inventory 3 , more than half of the emissions from Bulgaria are related to energy supply. From 2000, emissions in Bulgaria started to rise, and in 2011 reached 1990 levels of pollution. Spain dealt with an economic crisis in 2013, which could explain the slowdown in its emissions. 1 https://www.ine.pt/xportal/xmain?xpid=INE&xpgid=ine_publicacoes&PUBLICACOEStipo=ea&PUBLICACOEScoleccao=107664 &selTab=tab0&xlang=pt 2 https://ec.europa.eu/clima/sites/clima/files/strategies/progress/reporting/docs/bg_2014_en.pdf 3 https://unfccc.int/sites/default/files/resource/Bulgaria%20bg_br2.pdf Gross fixed capital formation was found to increase pollution in both the short-and long-run, which was not unexpected. As stated by Sapkota & Bastola (2017) , an increase in the level of capital of a production process will consume more energy, which could also be related to the scale effect: an increase in investment will increase production and energy consumption. If an increase in energy consumption leads to an upsurge in pollution, this rekindles the argument about energy sources, suggesting that these countries are not using enough renewable energy sources (RES), which can help in reduce emissions (e.g., Ben Jebli & Ben Youssef, 2015; Apergis & Payne, 2012). Trade openness has been found to contribute to environmental degradation, which could be linked to increased energy consumption (Sbia et al., 2014) , for example, due to the exports of energy-intensive goods that consequently increases pollution (Sun et al., 2017) . Once again these are related to the question of energy sources.

Energy-efficiency contributes to reducing emissions in high-income countries, but it is only effective in reducing pollution in the long-run in middle-income countries. As a proxy of environmental regulation, environmentally-related tax revenue produced an unanticipated result, as it was expected to reduce CO 2 emissions. On the contrary, it appeared to increase pollution in the short-run in high-income countries. This outcome could mean that these environmental taxes are not a good instrument for reducing emissions. For middle-income countries, environmental regulations seemed to decrease emissions as they were supposed to do. This effect does not necessarily mean that environmental regulation is more effective in middle-income countries.

Even though regulation decreases pollution in the long-run, the implementation of new environmental laws does not have much effect in the short-run.

With respect to patents, they were not shown to be statistically significant for highincome countries model, which was also unforeseen, given the higher levels expenditure on Research and Development (R&D) and human capital, which are a feature in these countries. For instance, countries develop new technologies (and register their intellectual property) with the specific aim of decreasing pollution. This unexpected effect could be connected to an eventual decrease in the number of patents, as explained by Su & Moaniba (2017) . This does not mean that these countries have become less environmentally aware, but rather that they develop new technologies to decrease emissions to a certain level, and then suspend further research in new technologies. In contrast, middle-income countries seem to develop new patents in direct proportion to their economic growth, with the aim of growing as quickly and inexpensively as possible. The significant current obstacles to the transfer of the carbon mitigation technologies usually developed by high-income countries (Cheng et al., 2019), could also explain this negative effect.

High-income countries were found to benefit from FDI, as it reduced CO 2 emissions, both in the short-and long-run, thereby supporting the Pollution Halo Hypothesis. This effect has a strong linkage with the level of a country's environmental protection and efficiency, policies that tend to preclude the admission of dirty FDI. One can observe that, in middleincome countries, FDI caused an increase in pollution in the short-run, while decreasing it in the long-run. The short-run effect supports the PHH. However, in the long-run, FDI decreased emissions. Initially, the effect of FDI on the environment depends on a country's environmental awareness, economic development, and above all, its capacity to absorb technology. This capacity reflects the country's ability to learn quickly, as noted by Adom et al., (2019). Even if high-income countries want to transfer their eco-friendly technologies through FDI, if the capacity of the host countries do absorb them is low, their effect will only be gradual, and the costs of adjustment will be greater (Adom et al., 2019). In brief, middle-income countries could be receiving both dirty and clean FDI. In the short-run, clean FDI does not improve the environment due to the lower capacity of middle-income countries to absorb technology. However, in the long-run, these countries seem to incorporate the new technologies and techniques, applying imported knowledge in their domestic firms, and, consequently, decreasing pollution.

The impact that FDI has on CO 2 emissions could be due to the impact on energy consumption of the resulting increased industrial activity, such as stated by Salim et al., (2017) . Furthermore, as previously stated, the main objective of this paper is the analysis of the transfer of polluting industries between countries through FDI, as this transfer does not embody environmental improvements, only a reallocation of emissions sources. To address this, an analysis was made of the impact on emissions of the industry sector, involving all the previously-described tests (see Tables A8 to A10 ). The socio-economic context was also considered, paying particular attention to the industrial sector.

A more detailed analysis of the idiosyncrasies of certain countries showed that in 2003, for example, the price of oil caused a favourable shock in demand in Norway, which increased oil investment and fiscal receipts. This shock could also have been responsible for a 26% growth in emissions in 2003, compared to 1990. In 2009, Bulgaria faced a difficult year. In addition to the global economic crisis, gas supplies were cut during the Russia-Ukraine gas dispute. Moreover, the production of industrial minerals, such as cement, for example, registered a significant decrease, accompanied by a significant reduction in refined lead exports according to International Business Publications 4 . These changes were reflected in a decrease in industrial production and emissions from the industrial sector. The Hungarian Central Statistical Office 5 stated that, in 2001, a substantial increase in the external trade of transport equipment arose. In this year, the gross fixed capital formation rose by 1.7%, and industrial performance achieved an increase in volume of 5.7%, which explains the substantial emissions. Both gross fixed capital formation, and trade openness increase pollution. Energy efficiency enhances environmental quality and defines the environmental performance of the industrial. Neither patents nor environmental regulations were shown to be statistically significant in explaining emissions from industry in either high-and middle-income countries. This absence of statistical significance could mean that governments are not paying enough attention to increasing innovation in the industry sector or considering the pollution emitted by this sector.

One observes that FDI presents different degrees of statistical significances in the shortand long-run. As detailed by Baek (2015), FDI could be considered as a long-run phenomenon, but in the short-run, the environmental benefits of introduction new technologies may be insufficient to mitigate the negative impact that FDI has on pollution (Shahbaz et al., 2019) .

Thus, FDI causes a decrease in CO 2 emissions from the industry in high-income countries, and it increases them in middle-income countries. These results also support the PHH for middleincome countries.

The findings of this paper shed light on the impact of FDI on the environment and addresses two aspects of this relationship: overall CO 2 emissions and CO 2 emissions from the industry sector. Overall, the results were high consistent, not only in relation to the literature but also between each other. They suggest that gross fixed capital formation and trade openness should be treated as the main drivers of pollution, probably due to them causing an increase in energy consumption. These countries affected should improve their RES capacity. Given that energy-efficiency helps in reducing emissions, these countries should rethink their environmental regulations related to energy production sources to move away from fossil fuels. At the same time, policymakers should encourage investment in R&D to increase industrial efficiency in countries where the sector is a major source of pollution. Where it is more costeffective, some firms prefer to pay extra taxes to keep pollute, rather than invest in innovation, as it is more countervailing, which explains why pollution may increase despite regulation in high-income countries.

One observes that FDI can be seen to improve environmental quality in high-income countries, but harms in middle-income countries in the short-run. However, in the long-run, FDI could also help middle-income countries reduce their emissions. These contrasting effects reveal these countries' lower capacity to absorb technology. Regarding the contrasting effects on highincome and middle-income countries, there is no doubt about the effect of transferring of polluting industries in the short-term; however, the positive effect in the long-run could provoke some doubts. Broadly speaking, if middle-income countries receive new technology, it must be absorbed and applied in their industries to increase the influence of the technique effect. The results suggest that this absorption is happening, but only very slowly. However, regarding the industry, the FDI only impacts their emissions in the short-run, which means that they are not applying or absorbing enough green technology to benefit from the phenomenon as much as high-income countries. In summary, the Pollution Halo Hypothesis is supported for highincome countries, and the PHH is sustained for middle-income countries.

Although high-income countries increase their external dependency by importing finished goods, they are still more profitable, because they circumvent stricter environmental regulation. Nonetheless, this only happens because middle-income countries have lower levels of environmental regulation. The shift of FDI away from the industrial sector should not be an option, since it could lead to deindustrialization (Doytch & Uctum, 2011) . This means that environmental policies encouraging clean FDI are required to achieve sustainable development (Essandoh et al., 2020) . If this is the case, these countries must rethink their regulatory structure, and encourage investment in R&D and the development of human capital, as this can improve their capacity to absorb technology, which would help them benefit more from FDI. Without sufficient capability to absorb technology, firms cannot apply imported knowledge quickly, and its benefits will be delayed. Policymakers must pay more attention to the industrial sector to encourage the development of green patents linked to industries, to increase their efficiency. Furthermore, policymakers should also tighten their environmental laws, especially concerning the admission of new industries, to avoid dirty FDI. Moreover, increasing the stringency of these countries' environmental regulation will help them improve their environmental performance. Different countries need different policy frameworks to reduce environmental emissions (Soytas & Sari, 2006a , 2006b . It is extremely important to establish a logical linkage between outcomes. Trade openness was once considered a main driver of pollution in middle-income countries, but does not appear to be statistical significance in the long-run. Middle-income countries may not have enough trade openness to have an impact on pollution in the long-run, but a more probable explanation has emerged. In the short-run, in an attempt to increase income, these countries increase their trade openness without major environmental concerns. This is reflected in the statistical non-significance of regulation in the short-run. Then, both FDI and trade openness increase pollution, reflecting the arrival of new polluting industries and the production of energy-intensive goods. However, perhaps due to the pressure of international agreements, in the long-run, these countries improve their environmental awareness and regulations to control polluting emissions. To begin with, these countries have a comparative advantage in polluting industries, but as they raise their environmental standards, foreign investments decrease. This brings a corresponding decrease in the significance of trade openness in determining the emissions of these countries. The foreign investments that middle-income countries receive in the short-run are mainly due to their more relaxed environmental laws. But, in the long-run, they increase their environmental standards, and efficiency becomes more important for attract more FDI inflows and reducing emissions.

Finally, what if suddenly and unexpectedly, countries, irrespective of their income level, were all confronted by a symmetrical crisis? For example, a crisis resulting from a pandemic, like that of COVID-19. Generally, during an economic crisis, automatic stabilizers are triggered without additional efforts by governments to diminish its impacts, in an attempt to harmony the government budget balance. But this does not act in a symmetric crisis. International trade suffers, and globalization reduces substantially, as reflected in areas such as tourism and the flows of FDI.

A major question concern is whether the current pandemic will have a lasting impact on globalization; something that cannot be answered yet. An optimistic post-crisis scenario would anticipate the speedy return of international trade to pre-crisis levels, presenting a V-shaped recovery, but this may not happen. The same is true for FDI flows. This crisis, although symmetrical among countries, has greater consequences for less-developed countries. For some countries, FDI is one of the major sources of income, productivity, and development. The consumption and the exportations highly reduced, the production stalled, and consequently, a sharp downward trajectory for these countries. This symmetrical crisis exposes the fragilities of the dominant strategy of exploiting the comparative advantages of countries, particularly in production. Besides the peak effect of pollution in countries with a comparative advantage in polluting industries, this exploitation also exposes the debilities, external exposure, and dependency of the countries.

At a time of economic uncertainty, multinational enterprises rethink their priorities, and limit capital expenditure related to foreign investments, delaying the flow of FDI flows, or even cancelling it. More developed countries are major sources of outbound FDI, which means that profits in their foreign subsidiaries will be substantially reduced. According to the UNCTAD (2020b), FDI flows will be hitting their lowest level for the past two decades (UNCTAD, 2020a) . This means that the effect of the PHH will be "weakened", not because of stricter or more relaxed environmental regulations, but as a consequence of the reduction in FDI. Although less frequent, industry transferrals will probably still continue, source countries will probably switch their investment to closer countries. This will lead to changes in those countries most commonly targeted as recipients of FDI, which are generally less developed, and extremely dependent on FDI inflows. Consequently, countries must invest more than ever in R&D to increase innovation and efficiency. This can reduce the costs (of both the firms, and the public health response). FDI is crucial to help middle-income countries soften the impact of the pandemic crisis. Middle-income countries should not reduce their environmental standards to attract FDI, but should increase the attractiveness of their workforce, through human development (investing more in R&D and reducing the uncertainty of corruption, ill health, and terrorism, for example). Furthermore, source countries of FDI should make a greater effort to transfer improved technologies and techniques to these countries. This bidirectional benefit could lead to: cheaper labor, reduced costs, and improved environmental quality worldwide. This is obviously a major issue needing urgent and extensive investigation in further research.

This paper focuses on the analysis of the impact that FDI has on pollution. In this study, countries were divided by income levels, because polluting industries are mainly transferred between countries with different levels of development and income. The ARDL model provided a useful disaggregation of the impacts, making it possible to better understand the impacts extended over time. This paper contributes by expanding the literature on the FDI-environment nexus with empirical evidence of a linkage between variables whose may vary over time. Another, innovative aspect of this paper is its consideration of factors such as the levels of regulation, innovation, and efficiency in the countries under scrutiny. Furthermore, the discussion about overall emissions and those from the industrial sector provides robust support for the PHH in middle-income countries, whereas high-income countries benefit from FDI phenomena. The capacity of middle-income countries to absorb technology plays a critical role in analysing the impacts of FDI, although the positive effect of FDI on overall emissions takes time to develop. The countries in this study face a trade-off between FDI and meeting pollution reduction targets. The findings of this paper can provide policymakers with useful guidance to help understand how they can increase the income of a country through an inflow of FDI while, at the same time, preserving the environment. With this in mind, is it crucial to establish a stable legal structure, as regulation plays an important role in this theme.

Regulation can shift attitudes, encourage investment in R&D, and increase the use of RES. Given the unexpected findings on the impact of regulation in high-income countries, these countries must combine different policy tools to obtain the goal of decreased emissions. For instance, it is available to not only regulate in the form of fees and taxes, but also to provide subsidies. For example, the awarding of subsidies to efficient firms with a high level of innovation and highly qualified workers could encourage investment in R&D. These subsidies must reward the investment that companies are obliged to carry out. The creation of direct subsidies for researchers and the foundation of research centres is strongly recommended. Increasing the level of human capital in countries will increase their environmental awareness. An increase in the use of RES is also required. However, given that renewable energy involves higher costs than fossil fuels, policymakers should introduce policies to increase the competitiveness of RES, by decreasing investment costs.

The environmental laws of the middle-income countries under analysis must be tightened. If the quality of their human capital increases, this will be reflected in greater innovation and efficiency. Furthermore, these countries must improve their evaluation criteria for FDI quality, and make them more attractive for the entry of new multinational enterprises, enterprises that could bring with them advanced and eco-friendly technologies, and efficient management skills. Recipient countries must absorb these technologies to change their industrial structure. Co-operation between countries is also essential to guarantee the transfer of knowledge and efficiency, and it is important to remember that corruption is a serious concern and is difficult to control. Policymakers from high-income countries must impose stricter controls on outflow FDI, applicable to the foreign subsidiaries of their parent companies, and ensure that these companies are investing in innovation and transferring their knowledge, and not just avoiding environmental compliance costs or relocating their emissions. With this cooperation and the mutual combination of different policy tools, high-income countries would be discouraged from transferring their polluting industries, and middle-income countries would not have to accept them. 

Globalization and the environmental impact of sectoral FDI

Consistent Covariance Matrix Estimation with Spatially Dependent Panel Data

Linking international trade and foreign direct investment to CO2 emissions: Any differences between developed and developing countries?

Plant vintage, technology, and environmental regulation

Dynamic relationship among environmental regulation, innovation, CO2 emissions, population, and economic growth in OECD countries: A panel investigation

Foreign Direct Investment and Air Pollution in China: Evidence From Chinese Cities. Region et Developpement

The econometric consequences of an energy consumption variable in a model of CO2 emissions

Renewable energy policies and technological innovation: Evidence based on patent counts

Environmental regulations and innovation activity in UK manufacturing industries

Linking induced technological change, and environmental regulation: Evidence from patenting in the U.S. auto industry

Unit root tests in panel data: Asymptotic and finite-sample properties

Does Foreign Direct Investment Harm the Host Country's Environment? Evidence from China. Ssrn

A Comparative Study of Unit Root Tests with Panel Data and a New Simple Test

Energy efficiency and sustainable growth in industrial sectors in European Union countries: A nonlinear ARDL approach

Foreign direct investment and decoupling between energy and gross domestic product in developing countries

Is energy consumption in the transport sector hampering both economic growth and the reduction of CO2 emissions? A disaggregated energy consumption analysis

Causal interactions between CO2 emissions, FDI, and economic growth: Evidence from dynamic simultaneous-equation models

Multivariate Granger causality between CO2 emissions, energy consumption, FDI (foreign direct investment) and GDP (gross domestic product): Evidence from a panel of BRIC (Brazil, Russian Federation, India, and China) countries. Energy

What is energy efficiency? Concepts, indicators and methodological issues

The impact of FDI in the OECD manufacturing sector on CO2 emission: Evidence and policy issues

General diagnostic tests for cross section dependence in panels

A simple panel unit root test in the presence of cross-section dependence

Bounds testing approaches to the analysis of level relationships

International trade, FDI (foreign direct investment) and embodied CO2 emissions: A case study of chinas industrial sectors

Can foreign direct investment harness energy consumption in China? A time series investigation

Foreign direct investment, income, and environmental pollution in developing countries: Panel data analysis of Latin America

A contribution of foreign direct investment, clean energy, trade openness, carbon emissions and economic growth to energy demand in UAE

Foreign direct Investment-CO 2 emissions nexus in Middle East and North African countries: Importance of biomass energy consumption

Environmental degradation in France: The effects of FDI, financial development, and energy innovations

Does foreign direct investment impede environmental quality in high-, middle-, and low-income countries?

The effect of environmental policy tools on regional green innovation: Evidence from China

Can China contribute more to the fight against global warming

Energy consumption and income in G-7 countries

Does innovation respond to climate change? Empirical evidence from patents and greenhouse gas emissions

Investigation of pollution haven hypothesis for China: An ARDL approach with breakpoint unit root tests

Investment promotion agencies: striving to overcome the COVID-19 challenge

When does FDI matter? The roles of local institutions and ethnic origins of FDI

Do lax environmental regulations attract foreign investment? Environmental and Resource Economics

Does foreign direct investment lead to lower CO2 emissions? Evidence from a regional analysis in China

Total-factor carbon emission performance of fossil fuel power plants in China: A metafrontier non-radial Malmquist index analysis

*** 2005 DLCO 2 I -5.835*** 2006 -7.286*** 2010 -7.678*** 2010 LFDI -1.705 2013 -2.150 2014 -1.943 2007 DLFDI -6.213*** 2005 -6.183*** 2011 -7.048***

DLCO 2 I -5.835*** 2006 -7.286*** 2010 -7.678*** 2010 LFDI -1.705 2013 -2.150 2014 -1.943 2007 DLFDI -6.213*** 2005 -6.183*** 2011 -7.048*** 2015 LGFCF -7.368*** 2008 -3.888 2006 -7.254*** 2007 DLGFCF -4.126* 2011 -5.561*** 2008 -5.202**

DLFDI -5.326*** 2012 -7.450*** 2010 -8.297*** 2010 LGFCF -1

LFDI -7.496*** 2008 -2.532 2006 -6.419***

Note: the lag selection criteria of Zivot and Andrews test is based in a TTest

The financial support of the NECE-UBI, Research Unit in Business Science and Economics, sponsored by the Portuguese Foundation for the Development of Science and Technology, project UIDB/04630/2020, is gratefully acknowledged. We also acknowledge the anonymous reviewers' comments that greatly contributed to improving the quality of the paper.