5 Pollution Haven Hypothesis Revisited A Comparison of The BRICS and MINT Countries Based On VECM Approach 2019

Journal of Cleaner Production 227 (2019) 724e738
Contents lists available at ScienceDirect
Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro
Pollution haven hypothesis revisited: A comparison of the BRICS and

MINT countries based on VECM approach
Qinglong Shao a, Xiaoling Wang b, *, Qian Zhou c, La
szlo
Balogh b
a
China Center for Special Economic Zone Research, Shenzhen University, Shenzhen, China
b
School of Economics and Management, University of Science and Technology Beijing, Beijing, China
c
School of Shanghai Development & Institute of Free Trade Zone, Shanghai University of Finance and Economics, China
a r t i c l e i n f o a b s t r a c t
Article history: The aim of this study is to investigate the existence of pollution haven hypothesis (PHH) and test the
Received 21 December 2018 casual relationship between inward foreign direct investment (FDI) and environmental pollutions, as
Received in revised form well as other potential influencing factors of environmental performance. To this end, a panel vector
23 March 2019
error correction model (VECM) and panel co-integration test is conducted for two country groups - the
Accepted 18 April 2019
Available online 23 April 2019
BRICS (i.e., Brazil, Russia, India, China, and South Africa) and the MINT (i.e., Mexico, Indonesia, Nigeria,
and Turkey) - during 1982e2014. Empirical results illustrate a bidirectional and positive causality be-
tween FDI inflows and GDP per capita for both groups, implying a virtuous circle of FDI-growth nexus.
Keywords:
Pollution haven hypothesis (PHH)
More importantly, both the BRICS and MINT countries illustrate bidirectional and negative causality
Vector error correction model (VECM) between FDI inflows and environmental pressures (i.e., per capita energy consumption for BRICS and per
Causality test capita carbon emissions for MINT). Such findings indicate the pollution haven hypothesis does not stand
Foreign direct investment (FDI) in this case. In addition, it is also found that the opening degree of trade can be promoted by FDI inflows,
which has rarely been deeply analyzed in prior studies. Concluding remarks and policy suggestions are
proposed in the last section.
© 2019 Elsevier Ltd. All rights reserved.
1. Introduction discounts this effort (Zhang et al., 2017).

One of the main drivers of climate change is the rapid low-
Climate change is one of the greatest challenges in the 21st quality growth of developing countries who emit amounts of
century and it has garnered wide public attention (Almer and GHG through inefficient fossil fuel use, so it is important to identify
Winkler, 2017; Espeland and Kettenring, 2018). Governments and the underlying reasons of this heavy pollution. Among these
organizations are collaborating to combat this issue. At the 2012 developing countries, the BRICS (i.e., Brazil, Russia, India, China and
United Nations Climate Change Conference in Doha the KP was South Africa) country group is expected to be a driving and vital
amended to decrease total emissions by at least 18% below 1990 force of global economic growth in coming decades (O'Neill, 2001).
levels during the second commitment period from 2013 to 2020 Nevertheless, there is another group of countries, the MINT (i.e.,
(UN, 2012). As of 29 August 2018, 115 countries have ratified the Mexico, Indonesia, Nigeria and Turkey) countries, which are rapidly
Doha Amendment, but a total of 144 countries are necessary for the gaining significance (Boesler, 2013). The BRICS and MINT econo-
amendment to get forced through (UNFCCC, 2018). The next sig- mies' combined share of the global GDP is more than 25%, while
nificant step in climate negotiations is the Paris Agreement, which their population accounts for more than 50% of the world popula-
sets a common cause for each nation to undertake against climate tion since 1982 (World Bank, 2018). These countries are increas-
change for the first time and attempts to limit the global temper- ingly embedded in the global economy and receive a growing share
ature increase to 1.5 C, although the withdrawal of the U.S. of global foreign direct investments (FDI), which is an essential
determining factor of pollution in less developed countries
regardless of the lack of environmental-friendly technologies and
management practice (Yang and Li, 2017; Li and Wang, 2017).
* Corresponding author.
E-mail addresses: [email protected] (Q. Shao), wangxiaoling@ustb. FDI inflows have been on the rise since the 1970s, and as gov-
edu.cn (X. Wang), [email protected] (Q. Zhou), laszlo.balogh@outlook. ernments liberalize policy barriers to trade and investment, some
com (L. Balogh).
https://doi.org/10.1016/j.jclepro.2019.04.206
0959-6526/© 2019 Elsevier Ltd. All rights reserved.
Q. Shao et al. / Journal of Cleaner Production 227 (2019) 724e738 725
even start competing for FDI, which is increasingly regarded as a 2. Literature review: empirical evidences of the pollution
key element of their development (Velde, 2006). The size of FDI net haven hypothesis
inflows (in current USD) grew from 10.17 billion in 1970 to 196.32
billion in 1990, reaching its peak in 2007 with 3.10 trillion and then The hypothesis of pollution havens has attracted much attention
declining, with fluctuation, to 1.86 trillion in 2017 (World Bank, in past decades and scholars have used various methods to examine
2018). Inward FDI is essential for domestic construction particu- the validity based on different research perspectives. This section
larly in developing countries, as it can stimulate economic growth presents a brief review of related literatures by dividing them into
in a short period of time. On one hand, capital accumulation can three types of research samples: country-, industrial- and firm-
create positive externalities through technology transfer (bring levels (see Table 1). This is meaningful because different sample
advanced clean technologies), spillover effects (cross-border intra- levels can reveal different findings.
industry technology transfer), productivity gains (more efficient
production process) and the introduction of new managerial skills 2.1. National-level summary statistics
(Lee, 2013). On the other hand, FDI inflows lead to environmental
degradation such as investment in pollution-intensive enterprises, Most of the related empirical studies use cross-country samples
or directly extracting natural resources without much compensa- to test the hypothesis. For example, Kellenberg (2009) examines
tion. Therefore, a deep understanding of the nexus between envi- the underlying reasons for multinational affiliate production of the
ronment, economy and FDI inflows should be the basis of policy- U.S. to the top and bottom 20th percentiles for all countries in terms
making (Omri et al., 2014). of GDP growth, and finds that declining environmental stringency
Numerous empirical studies have been conducted aiming to of the developing and transition economies is the main reason,
explain the causal association between FDI and carbon emissions which corresponds to the PHH. Further, the author revealed that
and to test whether inward FDI may harm the host countries’ do- except for the pollution-intensive industries, electronics and
mestic environments, i.e., the pollution haven hypothesis (PHH). appliance manufacturers are more sensitive to environmental
PHH assume that multinational enterprises (MNEs) tend to locate regulations. Later Kellenberg (2010) focused on waste trade in in-
the production of their pollution-intensive goods in developing ternational markets from the consumption side. He developed a
countries with lower environmental standards, thus less- USA-Asia trade model to justify the existence of PHH, and found
developed countries become the “haven” of “dirty” goods, or even that exporting the waste to the South was economically beneficial
the waste dumps of advanced economies (Kellenberg, 2010; for the North. This result can be regarded as an extension of the
Doytch, 2012; Zugravu-Soilita, 2017; Yang et al., 2018). However, hypothesis of pollution haven that advanced countries not only
research gaps remain despite the numerous studies. First, prior export “dirty” goods to poor countries, but also have economic
studies usually employ input-output analysis (Dietzenbacher and incentives to export the waste directly. A recent case confirmed this
Mukhopadhyay, 2007) or spatial econometric model (Jiang et al., study. As the largest importer of waste, in early 2018 China banned
2018) to analyze the FDI-environment nexus; little research uti- 24 types of scraps from entering the country, making the waste
lizes the VECM approach to check the bidirectional or unidirec- exporters (mostly developed countries such as the U.S., EU and
tional relationships between the variables. Second, it is difficult to Japan) have to look for alternatives with less strict requirements,
capture the whole picture by using a single country (such as India because recyclers cannot dispose the materials well enough within
or China, refer to Dietzenbacher and Mukhopadhyay, 2007; Jiang their own countries to meet domestic contamination standards
et al., 2018) or a group of countries (such as Belt and Road coun- (Nogrady, 2018). This implies that pollution-intensive goods and
tries, refer to Cai et al., 2018) as the research sample, because they manufacturing, as well as waste, tend to flow into countries with
do not represent emerging economies. Third, except for FDI and less stringent environmental regulations, and the case of China il-
environmental indicators, few studies discuss other potential lustrates that enhancing environmental standards is an effective
impact factors. way to stop some of this international flow.
Thus, findings are still inconclusive and further studies are A number of studies confirmed the PHH based on different
needed. This study tries to test this hypothesis by conducting a perspectives or by including new constraint factors. For instance,
comparative panel data analysis on BRICS and MINT country groups taking the institutional factor into account, Cole and Fredriksson
using the latest available data from 1982 to 2014. Three contribu- (2009) revealed that countries with few legislative units (such as
tions can be illustrated here: First, vector error correction model presidents, prime ministers and parties) are more likely to become
(VECM), which is mostly applied in financial and economic studies, pollution havens, because it is easier for the industrial lobby to
has been introduced to environmental studies to check the bidi- successfully pressure for lower environmental standards when the
rectional or unidirectional relationships of the dependent and in- host countries have fewer legislative units. In view of corporate
dependent variables; Second, there is no existing comparative social responsibility (CSR), Bu et al. (2014) confirmed the PHH and
analysis between the BRICS and MINT countries, dealing with the proposed that forcing social responsibility on multinational cor-
causal relationship between the variables; Lastly, empirical results porations in China will deter the foreign investments. Sapkota and
show that FDI inflows promote an increase in the opening to Bastola (2017) used 14 Latin American developing countries during
foreign trade, which was rarely discussed in prior studies. The re- 1980e2010 as research sample and their findings support the PHH.
sults of this study refute the pollution haven hypothesis that inward The same applies to Bakirtas and Cetin (2017) who used MIKTA
FDI may damage host countries’ environments, because this study (Mexico, Indonesia, South Korea, Turkey and Australia) from 1982
finds a bidirectional and negative causality between FDI inflows to 2011as a study sample to prove the existence of PHH. Solarin
and per capita energy consumption for BRICS countries and a et al. (2017) generated similar results using carbon emissions as
bidirectional and negative causality between FDI inflows and per an indicator of air pollution for the period of 1980e2012 in Ghana.
capita carbon emissions for MINT countries. By dividing FDI into two types of local - market-oriented and
The remainder of the paper is organized as follows. Section 2 export-oriented - Tang (2015) found that local environmental
presents a literature review on pollution haven hypothesis. Sec- standards have a great impact on FDI inflow, and the latter type of
tion 3 describes the data and methodology. Empirical results are FDI is more sensitive to local environmental regulations than the
presented in Section 4 while the concluding remarks and policy former. Koçak and Şarkgüneşi (2018) even found a bi-directional
implications are reported in Section 5. relationship between FDI inflow and local carbon emissions: FDI
726 Q. Shao et al. / Journal of Cleaner Production 227 (2019) 724e738
Table 1
Summary of literature review on empirical evidences of the Pollution Haven Hypothesis in chronological order.
Author(s) Method(s) Data structure Type of Main results

(country sample; data
year) source
Letchumanan and Panel regressions 4 developing and 3 Industrial No strong evidence of PHH are found, and compared to environmental
Kodama (2000) developed level regulations, technology may have a larger impact on overseas re-location.
countries;
1987e1994.
Smarzynska and Probit model 24 transition Firm level Found evidence of pollution haven hypothesis, but not robust.
Wei (2004) economies;
1989e1994.
Cole and Elliott GMM, IV U.S., Mexico and Industrial In addition to lax environmental regulations, capital intensity is also an
(2005) Brazil; level important precondition for becoming a pollution haven.
1989e1994.
Dietzenbacher Input-output analysis India; Country As a developing country, India is not a pollution haven. This result
and 1991/1992 and level contradicts the hypothesis.
Mukhopadhyay 1996/1997.
(2007)
Cave and Fixed effect panel regression. EU; Country Energy intensive trade conforms to PHH, but not for toxic intensive trade.
Blomquist 1970e1999. level
(2008)
Cole and Fixed effect panel regression, three-stage 13 OECD and 20 Country Except for the stringency of environmental standards, countries tend to
Fredriksson lobby game developing level become pollution havens when they have fewer legislative units.
(2009) countries;
1982e1992.
Kellenberg (2009) OLS and GMM-IV estimations U.S.; Country Foreign investment of U.S. tends to flow to countries with lower
1999e2003. level environmental standards, which is in line with the pollution haven
hypothesis.
Wagner and GMM estimation Six industries of Industry Environmental regulatory stringency has significant impact on FDI.
Timmins Germany; level
(2009) 1996e2003.
Kellenberg (2010) A two country trade model U.S.; Country The cost of export waste to poor countries is lower than domestic disposal.
1996e2007. level
Kearsley and Panel regression with country and time 27 OECD member Country In general, no significant correlation of trade openness and pollution
Riddel (2010) fixed effects countries; 1980 level emissions, thus the PHH is not valid.
e2004.
Ben Kheder and Economic geography model, conditional French firms; Firm level PHH is confirmed for developed and most emerging economies and Central
Zugravu (2012) logit model. 1996e2002. and Eastern European countries, but not for Commonwealth of
Independent States and the developing counterparts.
Manderson and Conditional logit model with fixed effect. UK; Firm level Outward FDI is not a determinant factor for UK firms' international
Kneller (2012) 2005 relocation decision.
Dam and Logit regression. 540 companies from Firm level Inbound FDI only has a small influence on local environmental regulations.
Scholtens 233 countries PHH is valid in resource abundant countries while this is not the case for
(2012) worldwide; countries with fewer or no resources.
2004.
Al-mulali and Non-stationary panel techniques. GCC countries; Country In the short-term FDI inflow has no impact on carbon emissions, but in the
Foon Tang 1980e2009. level long-term there is a significant negative correlation.
(2013)
Asghari (2013) Fixed and random effect models. MENA countries; Country FDI is found negatively correlated with carbon emissions, which is opposite
1980e2011. level to PHH.
Rezza (2013) OLS Norway; Firm level Investment of Norwegian multinationals tends to flow to regions with lax
1999e2005. environmental regulations.
Bu et al. (2014) Conditional logit model. China; Country Lax environmental regulations can attract foreign investment.
1998e2007. level
Tang (2015) Cournot and Bertrand models U.S.; Country Confirmed the existence of PHH, but local environmental regulations show
1999e2003. level different impacts on different types of FDI.
Zugravu-Soilita Panel regression with first and second France, Germany, Industrial Pollution haven hypothesis is valid for a certain amount of capital intensity
(2017) order interaction terms. Sweden and UK. level and degrees of environmental stringency, while pollution halo hypothesis
1995e2008. stands in other cases.
Sapkota and Panel fixed and random effect model. 14 Latin American Country Found strong evidence to support PHH.
Bastola (2017) countries; level
1980e2010.
Zhou et al. (2017) Dynamic panel regression. China; Industrial PHH and PH coexist in Chinese polluting industries.
1998e2008. level
Sun et al. (2017) The ARDL method, breakpoint unit root China; Country One percent increase of FDI inflow will enhance carbon emissions at about
tests. 1980e2012. level 0.06 percent.
Solarin et al. The ARDL method Ghana; Country PHH exists in Ghana.
(2017) 1980e2012. level
Bakirtas and Cetin PVAR model MIKTA countries; Country FDI has a detrimental effect on the environment.
(2017) 1982e2011. level
Koçak and Maki structural break cointegration test, Turkey; Country Found a bi-directional relationship between FDI and carbon emissions in
Şarkgüneşi DOLS, Hacker and Hatemi-J bootstrap 1974e2013. level the long run. This supports the HPP.
(2018) causality test.
Jiang et al. (2018) Spatial econometric models. China; Country FDI is found to be negatively correlated with air pollution, justifying the
2014. level pollution halo hypothesis.
Liu et al. (2018)
Table 1 (continued )
Author(s) Method(s) Data structure Type of Main results

(country sample; data
year) source
Global and local spatial autocorrelation China; Country FDI has distinct impacts on different pollutions, it reduces waste soot and
and spatial panel data model. 2004e2015. level dust, but increases wastewater and sulfur dioxide.
Yang et al. (2018) Conditional logit model. Jiangsu, China; Firm level Firms with pollution-intensive products tend to relocate to the north of
2006e2010. Jiangsu Province which has comparably lax environmental regulations.
Cai et al. (2018) Multi-regional input-output model. China and 66 Belt Country Among countries in the Belt and Road, 22 developed countries make China
and Road countries; level their pollution haven and 19 developing countries have been China's
2013 pollution haven.
pez et al. (2018) Multi-regional input-output model.
Lo Seven regions Country International trade reduces overall carbon emissions, but this is not the
around the world; level case for certain regions.
1995e2009.
Note: This summary is based on existing literatures. For the abbreviations: OLS e ordinary least square, IV e instrumental variable, GMM e generalized method of moments,
PHH e pollution haven hypothesis, MIKTA e Mexico, Indonesia, South Korea, Turkey and Australia, MENA e Middle East and North Africa, GCC e Gulf Cooperation Council, PH
e Porter Hypothesis, ARDL e autoregressive distributed lag, PVAR e panel vector auto regressive, DOLS e dynamic ordinary least square estimator.
increases the emissions and emissions conversely affect FDI entries. extended their research into the manufacturing industry of devel-
Focusing on the countries along the Belt and Road, Cai et al. (2018) oped countries including France, Germany, Sweden and the UK,
calculated that China has been the pollution haven for 22 devel- with results that showed countries with lower capital intensity and
oped countries, and at the same time has made 19 developing strict environmental regulations tended to become pollution halos,
countries its pollution havens. whereas countries with average capital endowments and low
However, other scholars raise different viewpoints and propose environmental standards tended to become pollution havens. In
that PHH is not true under any circumstances. Cave and Blomquist light of this, not all economies with low environmental standards
(2008) found international trade of energy-intensive goods in EU fit have the potential of becoming pollution havens; capital intensity
for PHH, while this is not the case for toxic-intensive trade. of the FDI host country should be fully considered in the analysis,
Empirical tests for most of the developed countries around the and this enriched the PHH.
world show no significant sign of correlation between trade In addition, companies and investments that relocate to coun-
openness (calculated as exports plus imports divided by GDP) and tries with less stringent environmental policies may cause
per capita emissions (Kearsley and Riddel, 2010). Liu et al. (2018) agglomeration externalities, which can promote knowledge/tech-
argued that FDI has distinct impacts on different kinds of pollu- nology transfer and enhance labor skills on one side, while on the
tions in China. other side, price competition of similar products may lower prod-
Furthermore, according to PHH, developing countries lose by ucts’ quality and the path-dependence effect will threaten the
receiving pollution-intensive industries while developed countries innovative capacity of the whole industry. These effects have been
gain through exporting “dirty” goods, but evidence shows that ignored by previous studies but are obviously important. To address
certain developing countries such as India have gained from in- this issue, Wagner and Timmins (2009) controlled for agglomera-
ternational trade and FDI inward (Dietzenbacher and tion externalities represented by total stock of inward FDI and
Mukhopadhyay, 2007). This is “Pollution Halo Hypothesis”, which confirmed the existence of PHH, that higher environmental stan-
argues that host developing countries benefit from the knowledge dards may impede FDI inflow especially for the chemical industries.
spillover (such as management practice) and technology transfer of Except for environmental regulations, technologies are proposed to
multinational corporations. This hypothesis is confirmed by be a determining factor in facilitating transboundary location
Asghari (2013), who found in MENA (Middle East and North Africa) (Letchumanan and Kodama, 2000).
countries FDI inflow promotes carbon mitigation; inward FDI of the
gulf cooperation council (GCC) countries reduced carbon emissions
2.3. Firm-level microdata analysis
in the long run (Al-mulali and Foon Tang, 2013); and findings have
even shown that international trade has been reducing global
Microdata on the firm level can reveal some new findings, as
emissions on the whole (Lo pez et al., 2018). Moreover, using China
macroscopic analysis at country- and industrial-level samples
as case study, Jiang et al. (2018) found a negative correlation for FDI
cannot capture the whole picture. This analysis will lead to some
and air pollution in 150 Chinese cities. Conversely, Sun et al. (2017)
omitted defects that can cause biased results. For example, some
argued that FDI inflow will significantly increase carbon emissions
featured characteristics of host countries may affect the FDI-
in the long run, which supports the hypothesis of pollution haven.
pollution nexus, such as bureaucratic corruption which can deter
Different research samples and methods may be the underlying
FDI but lower environmental standards. Further, it is no easy task to
reason for the opposing conclusion.
measure the pollution intensity of multinational enterprises as well
as the environmental policy stringency of the FDI host countries. To
2.2. Industrial-level statistical analysis tackle these problems, Smarzynska and Wei (2004) tested the
pollution haven hypothesis using firm-level data of 24 transition
Until now, only limited studies have used FDI indicator to economies. Their results showed weak evidence of the hypothesis,
examine pollution havens at the industrial level. By focusing on the and were not valid for robustness checks. In fact, the hypothesis
capital intensity of pollution-intensive industries, Cole and Elliott would be more robust domestically, because re-location barriers
(2005) revealed the reason why many related studies find little or are much lower within a country than for international movement -
no evidence of pollution haven. In most cases, developing countries a recent empirical study in Jiangsu Province of China justified this
with lax environmental standards do not have enough capital viewpoint (Yang et al., 2018).
accumulation to attract pollution-intensive investments, so capital Similarly, Ben Kheder and Zugravu (2012) employed firm-level
intensity is a prerequisite for countries to become a pollution ha- data to explore 1,374 French investments in 74 countries during
ven. Following this line of thought, Zugravu-Soilita (2017) further the period 1996e2002. Their results showed that wealthy and fast-
growing economies conformed to PHH while poor countries did used for the empirical analysis. The purpose of this study, as
not, and concluded that stricter regulations could attract more mentioned earlier, is to explore the causal relationships between
French investments. This implies that other invisible potential net FDI inflows and carbon emissions per capita, as well as other
factors may affect this nexus, such as tax incentives and land use potential influencing explanatory variables, for the BRICS and MINT
rights for foreign investments. As for UK firms, domestic stringent countries. All relevant data are obtained from the World Develop-
policies on environmental protection do not significantly affect a ment Indicator (WDI) (World Bank, 2018). Certain countries' FDI
firm's decision to establish overseas, this is another case that con- data are not available before 1982, and data on per capita CO2
tradicted the PHH (Manderson and Kneller, 2012). Dam and emissions from the WDI are available to 2014, so the research
Scholtens (2012) determined stocks of natural resources matter, period was extended from 1982 to 2014 to achieve a balanced
as they found that polluting firms tend to locate in countries with panel. Russia's data are only available since 1992, after the collapse
weak environmental regulations when the countries are abundant of the Soviet Union. Before conducting the empirical analysis, all
in natural resources, but not in countries with fewer or no natural the data have been converted into natural logarithms. Based on the
resources. This result can be linked to the “resource curse”, wherein theories and previous studies discussed in earlier sections, the
resource abundant countries are not only prone to growth decline, selected variables are explained as follows.
but also gradually become a haven for pollution intensive goods.
Rezza (2013) looked for the determinants of Norwegian outward 3.1.1. Foreign direct investment
FDI in various motives, and found that parent companies in Norway Foreign direct investment (FDI) is a type of international in-
reduced investment to their foreign affiliates with higher levels of vestment undertaken by a resident in one economy (i.e., direct
environmental stringency; this case proved the hypothesis from investor) whose objective is the formation of a long-term interest in
another perspective. Zhou et al. (2017) innovatively confirmed the an enterprise (i.e., direct investment enterprise) that resides in an
co-existence of pollution haven hypothesis (PHH) using Chinese economy different from that of the direct investor. The direct
firm-level data of polluting industries. investor is motivated in establishing a strategic and long-term as-
In general, summarizing the aforementioned literatures, there is sociation with the direct investment enterprise, which is accom-
no consensus among scholars of the effects of FDI on the environ- plished by guaranteeing a substantial level of influence in the
ment of host countries. Pollution haven hypothesis is more likely to management of the direct investment enterprise. As the goal is to
be valid in certain countries, industries and research periods have a “lasting interest”, the direct investor has to own at least 10%
(Zaman and Moemen, 2017), and dependent on the type of envi- of the voting power of the direct investment enterprise. Thus, direct
ronmental policies and industrial characteristics in those countries investment and portfolio investment are different in that portfolio
(Millimet and List, 2004; Zheng and Shi, 2017). Some local features investors generally do not expect to have an effect on the man-
(such as corruption, see Candau and Dienesch (2017)), laws (such as agement of the endeavor in which they invest.
commercial laws on social responsibility, see Bu et al. (2014)) and Generally speaking, FDIs have double effects on a host country's
even domestic reduced competitiveness (Bommer, 1999) have po- economy. FDI, in most cases, is undertaken by MNEs. In this way the
tential impacts on the FDI-Pollution nexus. Thus, there is no one- MNEs contribute to the dissemination of technology from devel-
size-fits-all answer, instead a case-by-case study is necessary. oped to developing countries that is otherwise not available
This study uses the two largest developing country groups during through the market (because of legal or financial constraints for
1982e2014, BRICS and MINT, to examine the existence of PHH, example) (Maher and Christiansen, 2001). Therefore, FDI is the
explore what factors drive FDI inflows and determine whether the most important means by which international technology transfer
domestic environment was worsened by the inward FDI. is achieved. In addition, managerial skills and practices are trans-
ferred through labor training. A spin-off effect of this is trained
3. Data description and model specification personnel leaving the MNEs can work at other local firms or
establish new startups. As well, MNEs may improve the manage-
This section presents the variables employed and data sources, ment skills of suppliers, distributors or competitors through busi-
as well as the model specification. Considering the numerous ness activities. However, there are also negative impacts. For
technical terms used in this study, the terms are presented them in example, the controlling stake in the company may be in foreign
Table 2 for easier reference. hands which may damage national sovereignty; outflow of earn-
ings from the subsidiary can cause a balance of payment deficits;
the strong economic power of subsidiaries can create monopoli-
3.1. Data description
zation in the market; and jobs created in FDI host countries enable
job losses in home countries. Further, inward FDI may pose harmful
The aim of this section is to give a detailed description the data
environmental impacts on host countries, i.e. the pollution haven
hypothesis, which is the theme of this study.
Table 2 Taking all this into consideration, this study used foreign direct
Description of the terms used in this study. investment net inflows (measured in percentage of GDP, henceforth
Terms Descriptions FDI inflows) from 1982 to 2014 as the substitute for inward FDI,
extracted from the World Bank (2018). In specific, the largest per-
BRICS Brazil, Russia, India, China and South Africa
MINT Mexico, Indonesia, Nigeria and Turkey
centage changes of FDI inflow among the BRICS and MINT countries
FDI Foreign direct investment are in China and India, which have a 295-fold and a 227-fold in-
PHH Pollution haven hypothesis crease during the research period, respectively. The smallest
VECM Vector error correction model amount of FDI net inflows in 1982 is to Turkey, because liberaliza-
FMOLS Fully modified ordinary least squares
tion of the Turkish economy only started in the 1980s (Nas and
VAR Vector autoregressive model
DF-test Dickey-Fuller test Odekon, 1992). The low investment inflow to South Africa is due
ADF-test Augmented Dickey-Fuller test to the Apartheid era between 1948 and 1994 (South African History
PP-test Phillips-Perron test Online, 2018), when financial inflows to the country were limited.
IPS-test Im, Pesaran and Shin test Between 1982 and 2014, FDI inflows to the BRICS and MINT coun-
LLC-test Levin-Lin-Chu test
tries increased from 14.15 billion USD (without Russia, in constant
2010 USD) to 463.71 billion USD. Moreover, the BRICS and MINT Table 3 presents the descriptive statistics for the whole panel
countries attracted 27.7% of the total global FDI inflows in 2014. and for the panels of the BRICS and MINT countries for the period
1982e2014 (Russia: 1992e2014). As can be seen, the highest rela-
3.1.2. Carbon emissions per capita tive volatility is in FDI and per capita carbon emissions while urban
Carbon dioxide emissions are those stemming from the burning population has the lowest relative volatility. The mean values of per
of fossil fuels and the manufacture of cement. They include carbon capita GDP, trade openness and urban population is lower than the
dioxide produced during consumption of solid, liquid, and gas fuels, median values, implying that the data is skewed to the left. In the
and gas flaring. In this study, per capita carbon dioxide emissions panel of BRICS countries urban population and trade openness
(CO2.p.c., measured in metric tons) are used as a substitute for illustrate the highest volatility, while energy consumption per
environmental indicator. Per capita carbon emissions have been capita is the lowest. The panel of MINT is similar to the panel of
relatively stable in the BRICS and MINT countries, with a clearly BRICS in urban population and trade openness which show the
visible upward trend in China and Turkey. By comparison, Russia, highest volatility, while energy consumption per capita is the least
South Africa, Mexico and Nigeria have achieved a decrease in per volatile.
capita carbon emissions between 1982 and 2014 (Russia between
1992 and 2014), decreased percentages are 15.18%, 1.38%, 7.92% and 3.2. Model specification
35.57%, respectively.
Based on empirical findings of previous studies, it is possible to
3.1.3. Gross domestic product per capita establish the long-run relationship between FDI, carbon emissions,
Gross domestic product is the sum of gross value added by all economic growth, energy consumption, trade, and urban popula-
resident producers in the economy plus any product taxes and tion using panel data framework:
minus any subsidies not included in the value of the products. GDP
could have been used as a substitute for real income, but since a lnXit ¼ b0i þ b1i lnY1it þ b2i lnY2it þ b3i lnY3it þ b4i lnY4it
relative measure of emissions (i.e., carbon emissions per capita) þ b5i lnY5it þ uit (1)
instead of an absolute measure of emissions is used as one of the
variables, it is more suitable to use GDP per capita (GDP.p.c., in where i ¼ 1, …, N denotes the country and t ¼ 1, …, T represent the
constant 2010 USD) in this analysis. Growth rates of per capita GDP time period. In this study T ¼ 33 (i.e., 1982e2014), and in different
from 1982 to 2014 are 55.04% (Brazil), 48.94% (Russia, 1992e2014), models N denotes different country numbers. lnX and lnY1, lnY2,
403.30% (India), 1575.32% (China), 15.93% (South Africa), 23.10% lnY3, lnY4 and lnY5 denote the variables lnFDI, lnCO2.p.c., lnGDP.p.c.,
(Mexico), 284.82% (Indonesia), 60.68% (Nigeria), and 257.29% lnEC.p.c., lnTRA and lnURB in various combinations, respectively. The
(Turkey), respectively. Turkey had the highest per capita GDP in notation ln represents the natural logarithms of per capita FDI in-
2014 (13312.03 USD), while the lowest was India (1646.78 USD). flows (as percentage of GDP), per capita carbon emissions (in ki-
lotons), per capita GDP (in constant 2010 USD), per capita energy
3.1.4. Energy consumption per capita consumption (in kilogram of oil equivalent), trade openness (as the
Energy consumption refers to the use of primary energy before sum of imports and exports divided by the GDP), and urban pop-
transformation to other end-use fuels. This study used energy ulation (as the percentage of the total population), respectively. The
consumption per capita (EC.p.c., measured in kilogram of oil error term (uit) is presumed to be independent and identically
equivalent, kgoe) to represent the use of energy. The highest per distributed with zero mean and constant variance. The estimation
capita energy consumption among the BRICS and MINT countries in of Eq. (1) can be used to establish the presence of a long-run
2014 was in Russia with 4942.876 kgoe, which is a 7.63% decrease connection between the variables, along with the error correction
compared to 1992; while the lowest per capita consumption of model (ECM), which is utilized to determine the short- and long-
energy was in India with 637.43 kgoe, which is an increase of run dynamic forces.
213.15% compared to 1982. The largest increase between 1982 and The empirical analysis divides into four steps. The first step is to
2014 was found in China, from 606.83 kgoe to 2236.73 kgoe (an establish the order of integration for the variables. This has to be
increase of 368.59%) and in Indonesia, from 388.83 kgoe to 883.91 checked because the following co-integration tests are valid only if
kgoe (an increase of 227.32%). the variables are integrated at the same order. The second step,
after the validation of the order of integration, is to test the panel
3.1.5. Trade openness co-integration relationships. The third step can only be undertaken
Trade openness (TRA) is defined as the sum of imports and ex-
ports divided by the GDP. Indonesia witnessed a slight decrease of Table 3
Descriptive statistics for the whole panel, BRICS and MINT during 1982e2014.
trade openness from 48.68% in 1982 to 48.08% in 2014, Nigeria also
decreased from 37.75% to 30.89%, while other countries all expe- Statistics FDI CO2.p.c. GDP.p.c. EC.p.c. TRA URB
rienced a trade openness increase during the period. For example, Panel Mean 1.749 3.705 5133.719 1431.542 43.122 52.751
India increased from 14.28% to 49.01%, China increased from 19.74% Median 1.317 2.334 5795.198 991.441 46.667 53.003
to 45.61%. In 1992, the trade openness value in Russia was 110.58%, Max 10.833 13.980 13312.030 5351.077 110.577 85.433
Min 2.757 0.308 387.746 299.053 12.352 20.902
which was due to the 3.0%, 5.05% and 14.53% GDP decreases in
Std. Dev. 1.639 3.452 3511.727 1140.807 16.409 19.588
1990, 1991 and 1992, respectively.
BRICS Mean 1.721 4.966 5181.975 1799.079 38.451 52.937
Median 1.068 2.566 5876.145 1136.968 39.123 54.004
3.1.6. Urbanization
Max 6.187 13.980 11912.150 5351.077 110.577 85.433
Urban population ratio (URB) refers to people living in urban Min 0.654 0.479 387.746 299.053 12.352 20.902
areas as defined by national statistical offices. The rate of urbani- Std. Dev 1.580 4.119 3615.093 1415.564 17.433 21.033
zation is conveyed as the ratio of the total population. The share of MINT Mean 1.783 2.224 5077.055 999.965 48.607 52.532
urban population between 1982 and 2014 increased in all of the Median 1.547 1.957 4433.418 844.196 49.345 51.100
sample countries. China's increase was the fastest at 33.51% and Max 10.832 4.491 13312.030 1659.516 96.186 78.971
Russia's increase was the slowest at 0.54% during the research Min 2.757 0.308 1151.126 387.097 23.609 23.389
Std. Dev 1.712 1.407 3399.171 371.703 13.207 17.819
period.
if all variables are integrated at the same order and co-integrated. In

this case, short- and long-run relationship among the variables X
k
trace statistic: LRðltrace Þ ¼ T log 1 b
li (2)
(Granger-causality) can be calculated using the VECM. The final
i¼rþ1
step is to check the elasticity in the cases where Granger-causality
has been confirmed.
maximum eigenvalue statistic: LRðlmax Þ ¼ Tlog 1 b
l rþ1
3.2.1. Stationarity and unit root testing (3)
Stationary models assume that the stochastic process remains in
statistical stability with probabilistic properties that do not change where T is the number of observations, b l i are the estimated ei-
over time. Subsequently, parameters such as mean and variance genvalues of P. The null and alternative hypotheses in the trace test
also remain unchanged over time (Box et al., 2016). Stationarity are:
applies to a time series if the auto-covariance of the series is in- H0: rank (P) is at most r co-integration vectors,
dependent from the time. However, economic and financial time H1: P has full rank, rank (P) ¼ k co-integration vectors, while for
series tend to display trending behavior, or non-stationarity, in the the maximum eigenvalue test:
mean. Before conducting a co-integration test, a necessary first step H0: rank (P) is r co-integration vectors,
is to check whether the variables contain a unit root. The presence H1: rank (P) ¼ (rþ1) co-integration vectors.
of a unit root means non-stationarity and the data has to be The Johansen co-integration test can be conducted as an indi-
changed into stationary form before any further analysis can be vidual co-integration test, but it cannot be used for co-integration
conducted. Two commonly used techniques of trend removal are in panel settings. For panel setting the Johansen Fisher panel co-
first differencing and time-regression. Unit root tests can be used to integration test can be used. It was developed by Maddala and
determine which technique should be applied (Zivot and Wang, Wu (1999), and is where the co-integration test outcomes for the
2006). panel data are achieved by combining the Johansen co-integration
The Dickey-Fuller (DF) test was developed by Dickey and Fuller test and Fisher's method of merging individual cross-section co-
(1979) and is one of the best known and most widely used unit root integration test results of maximum eigenvalues and trace statis-
tests. The Augmented Dickey-Fuller test (ADF test) is created by the tics. In this case, under the null hypothesis, the test statistic for the
autocorrelation of non-systematic component in DF models panel is:
(Dickey and Fuller, 1981). Compared to DF, the ADF test can used for
time series models that are more extensive and complicated. The X
k
JF ¼ 2 logðpi Þ c22k (4)
statistic used in the ADF test is a negative number, and the rejection
i¼1
of the null hypothesis (the presence of a unit root) depends on its
negativity, i.e. the more negative it is, the stronger the rejection. where JF denotes Johansen Fisher test, p-values (pi) are the co-
However, DF and ADF tests have only been used in a single- integration test results for each of the k individual cross-sections.
country setting and more recent studies have put forward that
panel-based unit root tests can be more reliable. For panel data
3.2.3. Granger causality
analysis the most widely used unit root tests are: LLC test (Levin
Proving a co-integrating relationship between the variables is
et al., 2002), Breitung test (Breitung, 2015), IPS test (Im et al.,
not enough to define the direction of causality. Co-integration in-
2003), Fisher-type test using augmented DickeyeFuller equation
dicates that there are long-run equilibrium relationships among
(Fisher-ADF) (Maddala and Wu, 1999) and Phillips-Perron Fisher-
the variables, thus there must be Granger causality among them in
type test (Fisher-PP) (Phillips and Perron, 1988). The LLC and Brei-
at least one direction (Engle and Granger, 1987). Assuming a co-
tung tests assume that there is a common unit root through the
integrating relationship exists between the variables, it is possible
cross-sections, while the remaining three tests assume individual
to conduct the augmented Granger causality test by approximating
unit root processes across the cross sections. These tests take the
the following VECM, based on Pao and Tsai (2011):
presence of a unit root (non-stationarity) as the null hypothesis and
2 3 2 3 2 3
test against alternatives involving stationarity. DlnCO2it a1 b11p b12p b13p b14p b15p b16p
6 DlnGDPit 7 6 a2 7 6 b21p b22p b23p b24p b25p b26p 7
6 7 6 7 X 6 7
6 DlnFDI 7 6 a3 7 r 6
b36p 7
3.2.2. Co-integration 6 it 7 6 7 6 b31p b32p b33p b34p b35p 7
6 DlnEC 7 ¼ 6 a4 7 þ 6 b46p 7
Co-integration was first proposed by Granger (1980) and points 6 it 7 6 7 p¼16 b41p b42p b43p b44p b45p 7
4 DTRADE 5 4 a5 5 4 b51p b52p b53p b54p b55p b56p 5
to the presence of a long-run connection amongst economic vari- it
ables. The basis of co-integration testing is to see whether two or
DURBit a6 b61p b62p b63p b64p b65p b66p
2 3
more integrated variables diverge considerably from a certain DlnCO2itp
relationship (Abadir and Taylor, 1999). When series are integrated 6 DlnGDPitp 7
6 7
at order one (become stationary after being differenced once), it is 6 DlnFDIitp 7
6
6 DlnECitp 7þ
7
possible that they have linear combinations that are stationary 6 7
without differencing - these series are called co-integrated series 4 DTRADEitp 5
(Granger, 1981). DURBitp
Engle and Granger (1987) introduced a two-step test to identify
2 3 2 3
the existence of co-integration, but this test cannot handle situa- q1 ε1it
tions where more than one co-integrating relationship is possible. 6 q2 7 6 ε2it 7
6 7 6 7
6 q3 7 6 7
6 7ECTit1 þ 6 ε3it 7
To overcome this problem, Johansen (1988) proposed two different
likelihood ratio tests, which allow the determination of up to r 6 q4 7 6 ε4it 7 (5)
6 7 6 7
linearly independent co-integrating vectors (or co-integration 4 q5 5 4 ε5it 5
ranks, r g-1, where g stands for the number of variables tested q6 ε6it
for co-integration). They are the trace test and maximum eigen-
value test as elaborated in Ong and Sek (2013): where D is the first difference operator; i ¼ 1, …, N indicates the
country; t ¼ 1, …, T indicates the time period; the various a, b, and q co-integration and causality is to check the order of integration by
are parameters have to be estimated; εit is assumed to be serially unit root testing. As macroeconomic variables have an inclination
uncorrelated error term; ECT is the one period lagged error- for exhibiting a trend over time, it is customary to employ a unit
correction term derived from the co-integration vector. As the root test with only an intercept and with an intercept and trend at
VECM structure is used, all variables are considered being endog- level form. First differencing is expected to eliminate any deter-
enous variables. ministic trend in the variables, thus unit root test with an intercept
should be applied at first difference. The unit root test results for
3.2.4. Fully modified OLS the individual countries can be found in Table 4. The results of the
Phillips and Hansen (1990) designed the fully modified ordinary ADF test statistics for the first two models (with constant, with
least squares (FMOLS) regression with the aim of offering optimal constant and trend) indicate that the null hypothesis of a unit root
estimates of co-integrating regressions. The method modifies least at level for the majority of the variables in the individual countries
squares to account for serial correlation effects and for the endo- cannot be rejected at the 5% significance level. After transforming
geneity in the regressors that results from the existence of a co- the variables into first difference, the ADF statistics reject the null
integrating relationship. Consider the (nþ1) dimensional time se- hypothesis of a unit root at the 5 or 1% significance level in most of
0
ries vector process ðyt ; X t Þ with the cointegrating equation: the cases, but not all of the variables are stationary at first differ-
ence, i.e. the variables are not integrated at the same order. Vari-
0 0
yt ¼ bX t þ g1 Dt þ u1t (6) ables that are not stationary at first difference include: lnCO2.p.c. in
China, lnFDI in Indonesia, lnTRA in Nigeria and lnURB in all of the
0 0 0
where Dt ¼ ðD1t ; D2t Þ0 are deterministic trend regressors and the countries except for South Africa and Turkey. The test results for the
0
stochastic regressors X t are governed by the system of equations, individual countries do not make it possible to check the co-
b 1t is the estimated residual.
u integration between the variables in all of the individual countries.
Results of the panel unit root test can be found in Table 5a and
4. Empirical results and discussions 5b. As mentioned earlier, five types of panel unit root tests (i.e., LLC,
Breitung, IPS, ADF and PP) were performed to check the time series
The aim of this section is empirical analysis of the casual rela- properties of the variables. Results indicate that except for lnURB,
tionship between FDI inflow and CO2 emissions. The first part of the null hypothesis of a unit root can be rejected at the 1% signifi-
this section will discuss results of unit root tests. The following cance level in all cases. Considering the panel unit root test results,
parts will examine and discuss the co-integration, Granger-cau- it can be established that lnFDI, lnCO2.p.c., lnGDP.p.c., lnEC.p.c. and
sality and elasticity test results. Two different data sets will be used lnTRA for the BRICS and MINT panels are integrated at order one,
for testing, i.e., the panels of BRICS and MINT countries. Of note is i.e., I(1). As the test results for the variable lnURB are inconclusive, it
that the unit root test and panel co-integration test are pre- has to be removed from further analysis.
regression tests: the former is to make sure the panel data are
stale and the latter is to confirm the long-run relationship between 4.2. Panel co-integration tests
dependent and independent variables.
Since five out of six variables are integrated at order one, com-
4.1. Unit root tests bined panel co-integration tests can be performed to examine the
presence of a long-run relationship between the variables. The
A preliminary step of the panel data analysis before analyzing panel co-integration were checked using Johansen Fisher, Pedroni
Table 4
ADF unit root test results for BRICS and MINT countries.
Variables Brazil Russia India China South Africa Mexico Indonesia Nigeria Turkey
Model with constant (level form)
lnCO2.p.c. 0.826 (0) 2.799* (0) 0.092 (0) 0.440 (1) 2.089 (0) 4.065*** (0) 1.106 (0) 2.190 (0) 1.361 (0)
lnGDP.p.c. 0.591 (0) 0.225 (0) 2.890 (0) 0.445 (2) 0.236 (1) 0.242 (0) 0.053 (0) 0.577 (0) 0.100 (0)
lnFDI 1.398 (0) 1.988 (0) 1.352 (0) 2.744* (0) 2.679* (0) 1.330 (2) 2.212 (6) 2.922* (0) 2.068 (0)
lnEC.p.c 3.687 (8) 1.887 (0) 2.305 (0) 0.223 (1) 1.840 (0) 1.523 (0) 1.555 (0) 0.972 (0) 1.025 (0)
lnTRA 1.565 (0) 2.795* (0) 0.277 (0) 2.009 (1) 1.217 (0) 1.433 (0) 2.881* (0) 1.928 (0) 2.051 (0)
lnURB 2.146 (1) 0.401 (1) 1.698 (2) 1.642 (1) 2.466 (2) 1.538 (2) 2.121 (1) 0.473 (1) 0.881 (2)
Model with constant and trend (level form)
lnCO2.p.c. 2.870 (4) 4.658*** (0) 0.765 (0) 2.304 (1) 2.048 (0) 4.862*** (0) 3.377* (1) 2.107 (0) 2.957 (0)
lnGDP.p.c. 1.511 (0) 3.714** (0) 1.488 (0) 3.759** (1) 1.458 (1) 4.001** (0) 2.190 (1) 2.041 (0) 2.272 (0)
lnFDI 2.753 (0) 1.044 (0) 3.013 (0) 1.874 (1) 3.377* (0) 0.747 (2) 0.388 (6) 2.679 (0) 3.119 (0)
lnEC.p.c. 1.021 (0) 5.418*** (1) 0.117 (0) 1.749 (1) 2.056 (0) 3.028 (0) 1.143 (0) 2.858 (0) 3.127 (0)
lnTRA 2.001 (0) 4.127*** (0) 2.520 (0) 1.964 (1) 2.567 (0) 3.762** (1) 2.801 (0) 0.597 (7) 3.187 (0)
lnURB 1.379 (1) 0.479 (2) 2.091 (1) 2.190 (1) 2.418 (2) 3.284* (2) 0.949 (1) 3.519* (1) 3.700** (2)
Model with constant (first difference)
lnCO2.p.c. 5.460*** (0) 3.254** (0) 5.211*** (0) 2.586 (0) 5.680*** (0) 8.717*** (0) 5.255*** (1) 5.472*** (0) 6.255*** (0)
lnGDP.p.c. 5.405*** (0) 2.681* (0) 4.067*** (0) 4.284*** (1) 3.749*** (0) 6.058*** (0) 4.179*** (0) 4.419*** (0) 5.881*** (0)
lnFDI 6.560*** (0) 4.874*** (0) 8.422*** (0) 3.712*** (0) 7.491*** (0) 8.713*** (1) 2.217 (5) 9.540*** (0) 6.551*** (0)
lnEC.p.c. 5.844*** (0) 3.210** (0) 4.381*** (0) 2.786* (0) 5.731*** (0) 6.582*** (0) 5.852*** (0) 5.061*** (0) 6.211*** (0)
lnTRA 5.072*** (0) 8.194*** (0) 5.333*** (0) 4.513*** (0) 5.875*** (0) 4.911*** (1) 8.082*** (0) 1.180 (7) 5.196*** (0)
lnURB 0.966 (0) 0.023 (0) 1.343 (1) 1.049 (1) 6.675*** (1) 1.833 (0) 0.378 (0) 1.875 (0) 2.825* (3)
Notes: (1) The unit root tests are performed using automatic lag length selection based on the Schwarz information criterion.(2) Numbers in parentheses indicate the lag
length.(3) *, ** and *** indicate significance at 10%, 5% and 1%, respectively. Same apply to the following.
Table 5a 4.3. Granger causality regressions

Panel unit root test results for BRICS countries.
Variable Common unit root Individual unit root The presence of a panel long-run co-integration relationship
LLC Breitung IPS Fisher-ADF Fisher-PP
among lnFDI, lnCO2.p.c., lnGDP.p.c., lnEC.p.c. and lnTRA in the BRICS
and MINT countries confirms the existence of at least one causal
Model with constant (level form)
relationship, but it does not indicate its direction. Using VECM, it is
lnCO2.p.c. 0.460 1.508 a 8.328 8.158 possible to check the presence of Granger causality in three
b 1.403 1.722
different ways. First short-run Granger causality can be examined
LnGDP.p.c. 2.380 5.664 a 0.287 0.495
b 5.295 5.544 by testing the significance of the sum of lagged differences of
lnFDI 2.455*** 1.268 a 14.519 15.087 explanatory variables by using the partial F-statistic. Second, long-
b 1.291a 1.159 run Granger causality can be tested by analyzing the significance of
lnEC.p.c. 3.378 4.877 a 4.339 4.839 the coefficient ECTi,t-1 based on t-statistic. Third, the strong Granger
b 4.374 3.400
lnTRA 0.376 0.045 a 9.552 17.093*
causality, which means a joint significance of the two sources
b 0.040 0.551 (short- and long-run), can be checked by testing the joint hypoth-
lnURB 1.743** 1.603 a 8.618 34.958*** esis through the Wald test, acting as a long-run causality test for the
b 1.462 2.393 individual variables. In this line, The VECM equations, without ur-
Model with constant and trend (level form) banization, can be written as follows:
lnCO2.p.c. 0.700 0.358 1.145 a 15.456 19.461**
b 0.936 0.809
lnGDP.p.c. 1.241 1.256 0.339 a 14.898 7.732
b 0.101 0.845 X
r X
r
lnFDI 1.572* 0.253 0.641 a 13.037 12.897 DlnCO2it ¼ a1 þ b11p DlnCO2itp þ b12p DlnGDPitp
b 0.668 0.483 p¼1 p¼1
lnEC.p.c. 0.391 1.593 0.264 a 15.140 21.167**
X
r X
r
lnTRA 2.091 0.738 1.537*

b
a
0.696
16.029*
0.308
15.481
þ b13p DlnFDIitp þ b14p DlnECitp
p¼1 p¼1
b 1.292* 0.945
lnURB 0.639 1.820 1.076 a 5.148 6.578 X
r
b 1.193 2.935 þ b15p DTRADEitp þ q1 ECTit1 þ ε1it (7)
p¼1
Model with constant (first difference)
lnCO2.p.c. 6.171*** 3.856*** 7.292*** a 67.178*** 67.645***

b 6.406*** 6.456***
lnGDP.p.c. 6.921*** 3.436*** 6.257*** a 56.302*** 48.977***
b 5.805*** 5.132*** X
r X
r
lnFDI 9.890*** 4.243*** 11.676*** a 106.521*** 106.866*** DlnGDPit ¼ a2 þ b21p DlnCO2itp þ b22p DlnGDPitp
b 8.824*** 8.822*** p¼1 p¼1
lnEC.p.c. 5.836*** 4.184*** 7.172*** a 65.694*** 66.234***
b 6.332*** 6.351*** X
r X
r
lnTRA 6.961*** 4.606*** 8.361*** a 79.207*** 92.050*** þ b23p DlnFDIitp þ b24p DlnECitp
b 7.469*** 8.260*** p¼1 p¼1
lnURB 1.088 0.292 0.780 a 26.587*** 10.932 X
r
b 0.194 0.746 þ b25p DTRADEitp þ q2 ECTit1 þ ε2it (8)
Note: a: ADF - Fisher Chi-square; b: ADF - Choi Z-stat. Same apply to the following. p¼1
and Kao tests, and the results are presented in Table 6a and 6b.
The Johansen Fisher panel co-integration test was conducted on X
r X
r
two different models; the first one allows for an intercept but no DlnFDIit ¼ a3 þ b31p DlnCO2itp þ b32p DlnGDPitp
trend in the co-integration equation and vector autoregression, p¼1 p¼1
while the second one allows for both an intercept and trend in the X
r X
r
co-integration equation but for no trend in VAR. The null hypoth- þ b33p DlnFDIitp þ b34p DlnECitp
esis of no co-integration is rejected at the 1% in both models, p¼1 p¼1
indicating the presence of at least one co-integrating vector in the X
r
panel of BRICS and MINT countries. þ b35p DTRADEitp þ q3 ECTit1 þ ε3it (9)
The testing results of Pedroni for both the BRICS and MINT p¼1
countries indicate that panel co-integration in two out of the four
panels is at a 5% or lower level of significance. Furthermore, two out
of the three group tests indicate the presence of panel co-
integration, also at a 5% or lower level of significance. In total, X
r X
r
four of the seven tests indicate that there is panel co-integration DlnECit ¼ a4 þ b41p DlnCO2itp þ b42p DlnGDPitp
among the variables. p¼1 p¼1
The Kao test also suggests panel co-integration at a 1% level of X

r X
r
significance for the BRICS countries and at a 5% level of significance þ b43p DlnFDIitp þ b44p DlnECitp
p¼1 p¼1
for the MINT countries. Overall, there is robust statistical evidence
supporting panel co-integration in both cases, which excludes the X
r
possibility of the estimated relationship being spurious. þ b45p DTRADEitp þ q4 ECTit1 þ ε4it (10)
p¼1
Table 5b
Panel unit root test results for MINT countries. X
r X
r
Variable Common unit root Individual unit root

DlnTRADEit ¼ a5 þ b51p DlnCO2itp þ b52p DlnGDPitp
p¼1 p¼1
LLC Breitung IPS Fisher-ADF Fisher-PP
X
r X
r
Model with constant (level form) þ b53p DlnFDIitp þ b54p DlnECitp
* p¼1 p¼1
lnCO2.p.c. 1.832** 1.478 a 16.139** 17.374**
b 1.367* 1.565* X
r
lnGDP.p.c. 1.435 3.636 a 0.385 0.351 þ b55p DTRADEitp þ q5 ECTit1 þ ε5it
b 3.488 3.494 p¼1
lnFDI 2.765*** 1.252 a 11.831 13.171
b 1.269 1.572*
(11)
lnEC.p.c. 2.103** 0.033 a 6.676 5.018
b 0.090 0.446 where D is the first-difference operator, r is the lag-length based on
lnTRA 2.081** 1.237 a 11.816 10.830 the Schwarz information criterion, ε is the serially uncorrelated
b 1.270 1.002 error term. The first step is to analyze the short-run causal rela-
lnURB 1.878** 0.734 a 4.780 70.619***
tionship from the independent to the dependent variable on an
b 0.803 6.572***
individual basis. The next step is to check the presence of a long-run
Model with constant and trend (level form)
causal relationship from the combined independent to the
lnCO2.p.c. 2.983*** 1.990** 2.601*** a 21.116*** 20.098*** dependent variable. The panel Granger causality results are shown
b 2.441*** 2.246** in following tables. Note that carbon emissions, GDP and energy
lnGDP.p.c. 1.342* 0.102 1.094 a 12.284 12.666
b 1.078 0.912
consumption are all in per capita values, even where it is not stated.
lnFDI 2.684*** 2.012** 0.337 a 9.863 15.336* In the BRICS countries, results of the short-run test presented in
b 0.314 1.966** Table 7a indicate unidirectional causality running from GDP to CO2,
lnEC.p.c. 1.274 1.041 0.887 a 11.742 11.510 from CO2 to FDI and from FDI to energy use. There is also a bidi-
b 0.918 0.737
rectional causality between FDI and GDP. The coefficient on the
lnTRA 1.077 0.239 0.827 a 17.449** 9.423
b 0.774 0.718 lagged ECT points to a long-run relationship among the variables in
lnURB 1.255 1.543 2.924 a 26.772*** 51.890*** Eqs. (8) and (9), as the ECT is statistically significant at the 5% and
b 2.516*** 2.962*** 1% level, respectively. The significance of the ECT coefficients in
Model with constant (first difference) these equations implies that there are two long-run panel causality
relationships running from CO2, FDI, energy consumption and trade
lnCO2.p.c. 12.196*** 5.051*** 10.972*** a 92.444*** 105.150***
b 8.451*** 9.162*** openness to GDP, and from CO2, GDP, energy consumption and
lnGDP.p.c. 8.836*** 6.476*** 8.135*** a 67.589*** 76.765*** trade openness to inward FDI.
b 6.952*** 7.428*** As for the MINT countries, the results indicate unidirectional
lnFDI 13.886*** 6.992*** 12.923*** a 104.689*** 145.359*** causality running from CO2 and energy use to FDI inflows, and from
b 8.813*** 11.135***
lnEC.p.c. 10.653*** 4.408*** 9.921*** a 84.388*** 91.950***
CO2 and FDI to trade. The coefficient on the lagged ECT indicates a
b 8.060*** 8.498*** long-run relationship among the variables in Eqs. (7), (8) and (11),
lnTRA 12.621*** 1.283* 11.161*** a 94.400*** 120.695*** as the ECT is statistically significant at the 5% and 1% levels,
b 8.517*** 10.000*** respectively. The significance of the ECT coefficient in these equa-
lnURB 2.481*** 0.394 0.832 a 12.812 23.167***
tions implies that there are three long-run panel causality linkages
b 0.829 1.575*
running from GDP, FDI, energy use and trade openness to CO2; from
CO2, FDI, energy use and trade openness to GDP; and from CO2,
Table 6a
Panel co-integration test results for BRICS countries.
Johansen Fisher test
Hypothesized number of co-integrating equations Model 1 Model 2
Trace Maximum eigenvalue Trace Maximum eigenvalue
None 78.88*** 66.55*** 94.18*** 61.01***

At most 1 30.37*** 21.08** 46.29*** 28.81***
At most 2 14.65 13.30 26.08*** 16.42*
At most 3 7.067 7.141 14.38 13.99
At most 4 8.604 8.604 6.309 6.309
Pedroni (Engle-Granger based) test
Idividual intercept
Panel v-Statistic (weighted) 0.579
Panel rho-Statistic (weighted) 0.647
Panel PP-Statistic (weighted) 2.602***
Panel ADF-Statistic (weighted) 2.601***
Group rho-Statistic 0.258
Group PP-Statistic 2.248**
Group ADF-Statistic 2.505***
Kao test
ADF-statistic 4.912***
Notes: (1) The Johansen Fisher panel cointegration test results heavily depend on the number of lags of the vector autoregression (VAR) system. The results obtained here use
one lag according to the lag length selection based on Schwarz information criterion.(2) Model 1: Intercept (no trend) in cointegration equation and VAR.Model 2: Intercept
and trend in cointegration equation - no trend in VAR.(3) Pedroni and Kao tests: Automatic lag selection Schwartz information criterion. Newey-West automatic bandwidth
selection and Bartlett kernel.
Table 6b
Panel co-integration test results for MINT countries.
Johansen Fisher test
Hypothesized number of co-integrating equations Model 1 Model 2
Trace Maximum eigenvalue Trace Maximum eigenvalue
None 26.96*** 21.51*** 29.40*** 35.61***

At most 1 11.22 8.627 6.958 4.908
At most 2 6.981 6.083 4.372 5.361
At most 3 4.230 4.399 1.868 2.977
At most 4 6.517 6.517 2.221 2.221
Pedroni (Engle-Granger based) test
Idividual intercept
Panel v-Statistic (weighted) 0.516
Panel rho-Statistic (weighted) 0.271
Panel PP-Statistic (weighted) 1.027**
Panel ADF-Statistic (weighted) 1.703**
Group rho-Statistic 1.021
Group PP-Statistic 0.512**
Group ADF-Statistic 2.356***
Kao test
ADF-statistic 2.076**
GDP, FDI and energy use to opening degree of trade. inflows, CO2, GDP and opening degree of trade. Fig. 1 and Fig. 2
Following the individual short- and combined long-run causal present the short-run and long-run Granger causalities among
tests, the strong Granger causality (joint significance of short- and the variables.
long-run causality) was tested by analyzing the joint hypothesis
through the Wald test. Results are presented in Table 7b. The joint
Wald F-test results for the BRICS countries imply that there is a 4.4. Elasticity tests
strong unidirectional panel causality running from CO2, energy use
and trade openness to GDP; and from CO2 and trade openness to The long-run coefficients of FDI inflows have been estimated
FDI inflows. There is also indication of a strong bidirectional panel where the Granger causality test results indicate the presence of a
causality between FDI inflows and GDP, energy use and FDI inward. strong causal relationship between FDI inflows and the dependent.
Similarly, the joint Wald F-test results for the MINT countries These variables are lnGDP.p.c and lnEC.p.c in the BRICS countries,
indicate the presence of a strong unidirectional panel causality lnCO2.p.c, lnGDP.p.c and lnTRA in the MINT countries.
running from energy consumption to CO2, FDI and energy con- Results for the BRICS countries in Table 8a reveal that when
sumption to GDP and trade. The results also indicate the existence holding the other variables fixed, a 1% increase of CO2 will decrease
of a strong bidirectional panel causality between GDP and CO2, FDI GDP by approximately 0.010%. An increase in FDI inflows by one
inflows and CO2, trade openness and CO2, as well as trade openness percent will increase GDP by 0.001%. An increase in energy con-
and GDP. sumption by one percent will increase GDP by 0.026%. Therefore,
The focus of this study is an analysis of the effects of FDI inflows energy use has the greatest impact on the economy in the long
on the other selected variables. According to the Granger causality term. If energy consumption is considered as the dependent, 1%
test results, there is a strong causal relationship between FDI in- increases of CO2 and GDP pose positive significant impacts at
flows, GDP and energy consumption in the BRICS countries. In 0.005% and 0.003% levels, respectively. By comparison, a 1% in-
MINT countries, there is a strong causal relationship between FDI crease in FDI inflows will have a negative 0.0001% effect on energy
consumption, which corresponds to Pao and Tsai (2011)'s findings
Table 7a
Individual short- and long-run Granger causality for BRICS and MINT countries.
Dependent variable Source of causation (independent variables)
Short-run Long-run
DlnCO2 DlnGDP DlnFDI DlnEC DlnTRA ECT
F-statistics t-statistics
BRICS countries
DlnCO2 e 3.808* 0.348 0.756 1.471 0.662
DlnGDP 0.397 e 3.081* 0.375 1.896 ¡2.489**
DlnFDI 3.403* 3.067* e 1.832 0.776 ¡6.131***
DlnEC 1.144 1.269 8.917*** e 2.692 0.587
DlnTRA 0.361 0.0617 0.106 1.753 e 0.324
MINT countries
DlnCO2 e 0.998 0.910 1.365 0.0232 2.460**

DlnGDP 2.537 e 1.509 0.169 0.011 ¡3.166***
DlnFDI 7.697*** 0.206 e 3.123*** 0.108 0.417
DlnEC 0.022 0.377 0.076 e 0.099 1.324
DlnTRA 9.658*** 0.300 4.581** 0.873 e ¡3.877***
Note: ***, ** and * indicate significance at 1%, 5% and 10%, respectively. Same apply to the following.
Table 7b
Individual strong Granger causality for BRICS and MINT countries.
Dependent variable Source of causation (independent variables)
Joint (strong)
DlnCO2, ECT DlnGDP, ECT DlnFDI, ECT DlnEC, ECT DlnTRA, ECT
F-statistics
BRICS countries
DlnCO2 e 1.962 0.566 0.567 0.931

DlnGDP 3.619** e 3.640** 3.381*** 4.153**
DlnFDI 22.808*** 19.768*** e 20.357*** 19.029***
DlnEC 0.664 1.492 4.518** e 1.549
DlnTRA 0.208 0.162 0.152 0.908 e
MINT countries
DlnCO2 e 3.148** 3.079*** 4.128** 3.641***

DlnGDP 5.338*** e 5.100*** 5.341*** 5.896***
DlnFDI 4.539** 0.241 e 1.766 0.105
DlnEC 1.004 1.292 0.879 e 1.233
DlnTRA 9.902*** 8.488*** 8.200*** 7.622*** e
Fig. 1. Short-run and long-run Granger causalities for BRICS country group.
Fig. 2. Short-run and long-run Granger causalities for MINT country group.
that there exist negative and significant correlation between FDI 0.013%. Considering GDP as dependent, a one percent increase of
inflows and energy consumption. In addition, the increase of car- FDI and energy use have 0.0003% and 0.0222% positive effects on
bon emissions per capita will significantly reduce the GDP per
capita, illustrating the harmful impact of environmental degrada-
tion on the economy in the long-term (Hao et al., 2018), and thus Table 8a
Panel group mean FMOLS long-run elasticity results for BRICS countries.
carbon mitigation policies implemented in developing economies
are encouraged (Azam, 2016). The bidirectional positive relation- Dependent variable: Dependent variable:
ship of energy consumption and economic growth is confirmed by lnGDP.p.c lnEC.p.c
numerous prior studies (Ouyang et al., 2013; Wang et al., 2016; Coefficient t-statistics Coefficient t-statistics
Destek, 2016; Zaman and Moemen, 2017). lnCO2.p.c 0.963 ¡12.965*** lnCO2.p.c 0.487 20.630***
The results for the MINT countries in Table 8b reveal that a one lnFDI 0.093 21.144*** lnGDP.p.c 0.323 13.918***
percent increase in FDI will decrease CO2 by 0.0006%, while the LnEC.p.c 2.586 24.798*** lnFDI 0.013 ¡7.328***
same amount of increase in trade openness increases CO2 by lnTRA 0.006 0.307 lnTRA 0.006 0.451
Table 8b
Panel group mean FMOLS long-run elasticity results for MINT countries.
Dep. var: lnCO2.p.c Dep. var: lnGDP.p.c Dep. var: lnTRA
Coef. t-stat. Coef. t-stat. Coef. t-stat.
lnGDP.p.c 0.107 0.733 lnCO2.p.c 0.127 1.208 lnCO2.p.c 0.304 0.810

lnFDI 0.064 ¡1.960*** lnFDI 0.027 1.583** lnGDP.p.c 0.325 1.306
lnEC.p.c 0.051 1.875 lnEC.p.c 2.215 12.487*** lnFDI 0.150 4.643***
lnTRA 1.270 ¡0.610*** lnTRA 0.143 ¡3.451*** lnEC.p.c 2.474 3.805***
GDP, respectively, while a one percent increase in trade openness 6. Policy implication
has a negative 0.0014% effect. Looking at the opening degree of
trade, a one percent increase of FDI increases trade openness The results for both BRICS and MINT countries refute the
0.0015% and a one percent increase of energy use increases trade pollution haven hypothesis, that foreign direct investments
openness 0.0247%. In this light, the MINT case again confirms the attracted by BRICS and MINT countries not only promote domestic
pollution halo hypothesis - that inward FDI leads to carbon miti- economic growth, but also alleviate environmental pressure (rep-
gation, but not to the creation of a pollution haven. Further, FDI resented by energy consumption per capita and carbon emissions
inflows help to increase the degree of trade openness and promote per capita respectively) in the long-run. Thus, FDI inflows are
economic growth, which is corresponds to the findings of Pegkas beneficial for sample countries’ economic development and also
(2015) and Hong (2014). environment protection, environmental-friendly technologies and
As can be seen, a higher opening degree of trade can increase managerial practices may play important roles in this process
carbon emissions by generating more economic activities, but in (Asghari, 2013; Lo pez et al., 2013). Based on the aforementioned
certain cases other air pollutants such as particulate matter and analysis, the following policy suggestions are proposed.
sulfur dioxide can be mitigated through trade liberalization policies
(Cherniwchan, 2017). Moreover, trade openness can lower per (ⅰ) Remove barriers to trade to attract foreign investments. Both
capita income, which is in line with Yanikkaya (2003) but contra- the BRICS and MINT country groups show bidirectional and
dictory to Frankel and Romer (1999) and Fetahi-Vehapi et al. (2015) positive causality between FDI inflows and economic growth.
who use 10 South-East European countries as samples. This dif- This indicates a virtuous circle, where an increase of FDI in-
ference may be because countries with higher per capita income flows can contribute to economic growth, which in turn can
levels in this study, such as Turkey, have lower opening degrees of create positive settings for attracting more FDI. Thus, gov-
trade, while poor countries always show higher trade openness ernments of developing economies should endeavor to
(World Bank, 2018). reduce and eventually eliminate harmful legal and non-legal
barriers so that MNEs can conveniently invest in host
countries. Moreover, this would also be beneficial for local
5. Conclusions firms who wish to access advanced technologies and inter-
national financing from foreign markets. Governments could
This study examines the casual relationships of foreign direct attract FDI through tax reduction or investment subsidies
investment, carbon emissions per capita, GDP per capita, energy according to host countries' economic conditions (Tian,
consumption per capita, trade openness, as well as urbanization for 2018), and they also should pay attention to the quality of
BRICS and MINT countries during 1982e2014, using the vector er- FDI, i.e., whether these investments can bring employment,
ror correction model. Specifically, the panel data analysis is con- new technologies and know-how, as well as the managerial
ducted in four steps, i.e., order of integration, co-integration, experiences. These efforts have to be taken with the pre-
Granger-causality, elasticity. The first step of the data analysis requisite of low or no trade barriers in the form of laws and
revealed that the variables (except for urbanization) for the BRICS regulations as well as other non-legal barriers.
and MINT panels are integrated at order one. The second step, (ⅱ) Absorption of foreign capital is an effective way to alleviate
combined panel co-integration tests, confirmed a strong statistical environmental burdens for developing economies. Both
evidence supporting panel co-integration in both panels. BRICS and MINT countries illustrate bidirectional and nega-
The Granger causality analysis (the third step) is the main focus tive causality between FDI inflows and environmental in-
of this study. FDI inflows were found to be affected by CO2 in the dicators (per capita energy consumption for BRICS and per
short term and influenced by CO2, GDP, energy use and trade capita carbon emissions for MINT). These results indicate the
openness in the long term, as well as having a long run impact on existence of pollution halo hypothesis with a self-reinforcing
the economy of BRICS countries. For MINT countries, FDI inflows feedback loop for these two country groups. In other words,
were affected by CO2 and energy consumption, in the long run it pollution would be reduced rather than increased through
had remarkable influence on CO2, GDP and trade openness, but no attracting foreign investments for BRICS and MINT countries,
factors showed a decisive impact on FDI in the long run. As for and the improvement of environmental conditions would
further elasticity analysis (the last step), BRICS group revealed that attract more investments. It may be that the application of
FDI inflows have a positive long-run effect on the economy and a new clean technologies attracts and helps create clean in-
negative long-run effect on energy consumption; while for MINT dustries, which in turn helps to improve the composition of
countries, FDI inward has a negative long-run effect on CO2, and a national output. Economic growth induced by FDI inflows
positive long-run effect on per capita GDP and trade openness. In may result in an increase of environmental pollution, but this
sum, in the short term carbon emissions affect BRICS and MINT can be offset by improved energy efficiency. In light of this,
countries’ FDI inflows; while FDI inward may improve rather than governments should subsidize local firms who use
contaminate the environment in the long run. Therefore, the environmentally-friendly technologies and try to reduce
pollution haven hypothesis does not stand in BRICS and MINT energy intensity by applying renewable energies.
countries.
Furthermore, governments should evaluate the environ- halo or pollution haven hypothesis. Int. J. Sci. Res. Environ. Sci. 1, 92e100.
Azam, M., 2016. Does environmental degradation shackle economic growth? A
mental impacts of foreign investment before implementa-
panel data investigation on 11 Asian countries. Renew. Sustain. Energy Rev. 65,
tion, and make the technology standards easier for domestic 175e182. https://doi.org/10.1016/j.rser.2016.06.087.
companies to adopt. It is worth noting that although FDI Bakirtas, I., Cetin, M.A., 2017. Revisiting the environmental Kuznets curve and
flows to BRICS are not directly Granger-causing carbon pollution haven hypotheses: MIKTA sample. Environ. Sci. Pollut. Res. 24,
18273e18283. https://doi.org/10.1007/s11356-017-9462-y.
emissions, a decline of per capita energy use with the com- Belloumi, M., 2014. The relationship between trade, FDI and economic growth in
bined effects of energy-intensive industries upgrade and Tunisia: an application of the autoregressive distributed lag model. Econ. Syst.
transformation through technological progress will more or 38, 269e287. https://doi.org/10.1016/j.ecosys.2013.09.002.
Ben Kheder, S., Zugravu, N., 2012. Environmental regulation and French firms
less contribute to the reduction of carbon emissions. location abroad: an economic geography model in an international comparative
(ⅲ) Foreign trade can be promoted by FDI inward. Very few study. Ecol. Econ. 77, 48e61. https://doi.org/10.1016/j.ecolecon.2011.10.005.
studies focus on the effect of inward FDI on the opening Boesler, M., 2013. The economist who invented the BRICs just invented a whole new
group of countries: the MINTs. Bus. Insid. 2.
degree of trade, so this result is a new finding. Possible ex- Bommer, R., 1999. Environmental policy and industrial competitiveness: the
planations for this result may be that new technologies and pollution-haven hypothesis reconsidered. Rev. Int. Econ. 7, 342e355. https://
managerial practices introduced by FDI inflows can be doi.org/10.1111/1467-9396.00168.
Box, G.E.P., Jenkins, G.M., Reinsel, G.C., 2016. Time series analysis: forecasting &
transferred into local firms through knowledge spillover ef-
control. In: Book. Prentice Hall, New Jersey, Fifth. https://doi.org/10.1016/
fects that improve their productivities, and this coupled with j.ijforecast.2004.02.001.
the export activities of MNEs, increases foreign trade vol- Breitung, J., 2015. In: Ba, Badi H. (Ed.), The Local Power of Some Unit Root Tests for
Panel Data. merald Group Publishing Limited.
umes. Moreover, trade patterns can be upgraded through
Bu, M., Liu, Z., Wagner, M., Yu, X., 2014. Corporate social responsibility and the
this effect, as the FDI host countries' export become more pollution haven hypothesis: evidence from multinationals' investment decision
diversified and sophisticated, in particularly for the devel- in China. Front. Econ. Glob. 14, 109e127. https://doi.org/10.1108/S1574-
oping countries (Iwamoto and Nabeshima, 2013). In fact, 871520140000014017.
Cai, X., Che, X., Zhu, B., Zhao, J., Xie, R., 2018. Will developing countries become
China is a good case for supporting this argument. After the pollution havens for developed countries? An empirical investigation in the
“Open-Door” policy enacted in 1978 investment attraction Belt and Road. J. Clean. Prod. 198, 624e632. https://doi.org/10.1016/
gradually become the main task of local governments and j.jclepro.2018.06.291.
Candau, F., Dienesch, E., 2017. Pollution haven and corruption paradise. J. Environ.
net FDI inflows grew from 430 million USD to 168 billion USD Econ. Manag. 85, 171e192. https://doi.org/10.1016/j.jeem.2017.05.005.
from 1982 to 2017, a 391 times growth. During the same Cave, L.A., Blomquist, G.C., 2008. Environmental policy in the European Union:
period, net trade in goods and services increased from 4.737 fostering the development of pollution havens? Ecol. Econ. 65, 253e261.
https://doi.org/10.1016/j.ecolecon.2007.12.018.
billion USD to 210.728 billion USD, almost 45 times growth Cherniwchan, J., 2017. Trade liberalization and the environment: evidence from
(World Bank, 2018). It should be clarified that this FDI-trade NAFTA and U.S. manufacturing. J. Int. Econ. 105, 130e149. https://doi.org/
unidirectional relationship does not suit all developing 10.1016/j.jinteco.2017.01.005.
Cole, M.A., Elliott, R.J.R., 2005. FDI and the capital intensity of “dirty” sectors: a
economies alike, as BRICS countries showed no sign of cor-
missing piece of the pollution haven puzzle. Rev. Dev. Econ. 9, 530e548. https://
relation in this study and Tunisia revealed the opposite doi.org/10.1111/j.1467-9361.2005.00292.x.
relation (Belloumi, 2014). In this light, more empirical ana- Cole, M.A., Fredriksson, P.G., 2009. Institutionalized pollution havens. Ecol. Econ. 68,
1239e1256. https://doi.org/10.1016/j.ecolecon.2008.08.011.
lyses and discussions are needed in future studies.
Dam, L., Scholtens, B., 2012. The curse of the haven: the impact of multinational
enterprise on environmental regulation. Ecol. Econ. 78, 148e156. https://
In addition, while this study and its conclusions offer insights doi.org/10.1016/j.ecolecon.2012.04.011.
into the PHH debate, there are several limitations that could be Destek, M.A., 2016. Natural gas consumption and economic growth: panel evidence
from OECD countries. Energy 114, 1007e1015. https://doi.org/10.1016/
addressed in future studies. First, more countries in addition to the j.energy.2016.08.076.
BRICS and MINT economics could be selected as research samples Dickey, B.Y.D.A., Fuller, W.A., 1981. Likelihood ratio statistics for autoregressive time
to conduct more comprehensive empirical tests on PHH. Second, series with a unit root. Econometrica 49, 1057e1072.
Dickey, D.A., Fuller, W.A., 1979. Distribution of the estimators for autoregressive
more environmental performance indicators could be incorporated time series with a unit root. J. Am. Stat. Assoc. 74, 427e431. https://doi.org/
to reflect the impacts of inwards FDI to enrich the existing discus- 10.1080/01621459.1979.10482531.
sion. Last but not least, research on individual countries would Dietzenbacher, E., Mukhopadhyay, K., 2007. An empirical examination of the
pollution haven hypothesis for India: towards a green Leontief paradox? En-
make a good supplement for studies based on group analysis. viron. Resour. Econ. 36, 427e449. https://doi.org/10.1007/s10640-006-9036-9.
Doytch, N., 2012. FDI halo vs. Pollution haven hypothesis. In: Vogel, R. (Ed.), Pro-
Acknowledgement ceedings of the New York State Economics Association. New York State Eco-
nomics Association, New York, pp. 47e49.
Engle, R.F., Granger, C.W.J., 1987. Co-integration and error correction: representa-
The authors are grateful for the support provided by the Na- tion, estimation, and testing. Econometrica 55, 251e276. https://doi.org/
tional Natural Science Foundation of China (71704010, 71771024, 10.2307/1913236.
Espeland, E.K., Kettenring, K.M., 2018. Strategic plant choices can alleviate climate
71673022, and 71602008), the Humanities and Social Science change impacts: a review. J. Environ. Manag. 222, 316e324. https://doi.org/
project of Ministry of Education of China (17YJC630163), the Social 10.1016/j.jenvman.2018.05.042.
Science Research Foundation of Beijing (18JDYJB021), Fundamental Fetahi-Vehapi, M., Sadiku, L., Petkovski, M., 2015. Empirical analysis of the effects of
trade openness on economic growth: an evidence for South East european
Research Funds for the Central Universities (FRF-TP-18-007A2), the countries. Procedia Econ. Financ. 19, 17e26. https://doi.org/10.1016/S2212-
National Social Science Research Project (18CGL008), and China 5671(15)00004-0.
Postdoctoral Science Foundation (2019M653047). Frankel, J.A., Romer, D., 1999. Does trade cause Growth ? Am. Econ. Rev. 89,
379e399.
Granger, C.W.J., 1981. Some properties of time series data and their use in econo-
References metric model specification. J. Econom. 16, 121e130. https://doi.org/10.1016/
0304-4076(81)90079-8.
Abadir, K.M., Taylor, A.M.R., 1999. On the Definitions of (Co-)integration. J. Time Ser. Granger, C.W.J., 1980. Long memory relationships and the aggregation of dynamic
Anal. 20, 129e138. models. J. Econom. 14, 227e238.
Al-mulali, U., Foon Tang, C., 2013. Investigating the validity of pollution haven hy- Hao, Y., Peng, H., Temulun, T., Liu, L., Mao, J., Lu, Z., 2018. How harmful is air
pothesis in the gulf cooperation council (GCC) countries. Energy Policy 60, pollution to economic development? New evidence from PM2.5 concentrations
813e819. https://doi.org/10.1016/j.enpol.2013.05.055. of Chinese cities. J. Clean. Prod. 172, 743e757. https://doi.org/10.1016/
Almer, C., Winkler, R., 2017. Analyzing the effectiveness of international environ- j.jclepro.2017.10.195.
mental policies: the case of the Kyoto Protocol. J. Environ. Econ. Manag. 82, Hong, L., 2014. Does and how does FDI promote the economic growth? Evidence
125e151. https://doi.org/10.1016/j.jeem.2016.11.003. from dynamic panel data of prefecture city in China. IERI Procedia 6, 57e62.
Asghari, M., 2013. Does FDI promote MENA region's environment quality? Pollution https://doi.org/10.1016/j.ieri.2014.03.010.
Im, K.S., Pesaran, M.H., Shin, Y., 2003. Testing for unit roots in heterogeneous panels. China) countries. Energy 36, 685e693. https://doi.org/10.1016/
J. Econom. 115, 53e74. https://doi.org/10.1016/S0304-4076(03)00092-7. j.energy.2010.09.041.
Iwamoto, M., Nabeshima, K., 2013. Can FDI Promote Export Diversification and Pegkas, P., 2015. The impact of FDI on economic growth in Eurozone countries.
Sophistication of Host Countries? Dynamic Panel System GMM Analysis, p. 347. J. Econ. Asymmetries 12, 124e132. https://doi.org/10.1016/j.jeca.2015.05.001.
Chiba. Phillips, P., Perron, P., 1988. Testing for a unit root in time series regressions. Bio-
Jiang, L., Zhou, H. Feng, Bai, L., Zhou, P., 2018. Does foreign direct investment drive metrika 75, 335e346. https://doi.org/10.2307/2336182.
environmental degradation in China? An empirical study based on air quality Phillips, P.C.B., Hansen, B.E., 1990. Statistical inference in instrumental variables
index from a spatial perspective. J. Clean. Prod. 176, 864e872. https://doi.org/ regression with I(1) processes. Rev. Econ. Stud. 57, 99. https://doi.org/10.2307/
10.1016/j.jclepro.2017.12.048. 2297545.
Johansen, S., 1988. Statistical analysis of cointegration vectors. J. Econ. Dyn. Control Rezza, A.A., 2013. FDI and pollution havens: evidence from the Norwegian
12, 231e254. https://doi.org/10.1016/0165-1889(88)90041-3. manufacturing sector. Ecol. Econ. 90, 140e149. https://doi.org/10.1016/
Kearsley, A., Riddel, M., 2010. A further inquiry into the pollution haven hypothesis j.ecolecon.2013.03.014.
and the environmental kuznets curve. Ecol. Econ. 69, 905e919. https://doi.org/ Sapkota, P., Bastola, U., 2017. Foreign direct investment, income, and environmental
10.1016/j.ecolecon.2009.11.014. pollution in developing countries: panel data analysis of Latin America. Energy
Kellenberg, D., 2010. Consumer waste, backhauling, and pollution havens. J. Appl. Econ. 64, 206e212. https://doi.org/10.1016/j.eneco.2017.04.001.
Econ. 13, 283e304. https://doi.org/10.1016/S1514-0326(10)60013-X. Smarzynska, B.K., Wei, S.-J., 2004. Pollution havens and foreign direct investment:
Kellenberg, D.K., 2009. An empirical investigation of the pollution haven effect with dirty secret or popular myth? Contrib. Econ. Anal. Policy 3, 1e32. https://
strategic environment and trade policy. J. Int. Econ. 78, 242e255. https:// doi.org/10.2202/1538-0645.1244.
doi.org/10.1016/j.jinteco.2009.04.004. Solarin, S.A., Al-Mulali, U., Musah, I., Ozturk, I., 2017. Investigating the pollution
Koçak, E., Şarkgüneşi, A., 2018. The impact of foreign direct investment on CO2 haven hypothesis in Ghana: an empirical investigation. Energy 124, 706e719.
emissions in Turkey: new evidence from cointegration and bootstrap causality https://doi.org/10.1016/j.energy.2017.02.089.
analysis. Environ. Sci. Pollut. Res. 25, 790e804. https://doi.org/10.1007/s11356- Sun, C., Zhang, F., Xu, M., 2017. Investigation of pollution haven hypothesis for
017-0468-2. China: an ARDL approach with breakpoint unit root tests. J. Clean. Prod. 161,
Lee, J.W., 2013. The contribution of foreign direct investment to clean energy use, 153e164. https://doi.org/10.1016/j.jclepro.2017.05.119.
carbon emissions and economic growth. Energy Policy 55, 483e489. https:// Tang, J., 2015. Testing the pollution haven effect: does the type of FDI matter?
doi.org/10.1016/j.enpol.2012.12.039. Environ. Resour. Econ. 60, 549e578. https://doi.org/10.1007/s10640-014-9779-
Letchumanan, R., Kodama, F., 2000. Reconciling the conflict between the “pollution- 7.
haven” hypothesis and an emerging trajectory of international technology Tian, Y., 2018. Optimal policy for attracting FDI: investment cost subsidy versus tax
transfer. Res. Pol. 29, 59e79. https://doi.org/10.1016/S0048-7333(99)00033-5. rate reduction. Int. Rev. Econ. Financ. 53, 151e159. https://doi.org/10.1016/
Levin, A., Lin, C.-F., Chu, C.-S.J., 2002. Unit root tests in panel data: asymptotic and j.iref.2017.10.018.
finite sample properties. J. Econom. 108, 1e24. https://doi.org/10.1016/S0304- UN, 2012. Doha Amendment to the Kyoto Protocol. UNFCCC. https://doi.org/
4076(01)00098-7. 10.1289/image.ehp.v119.i03.
Li, M., Wang, Q., 2017. Will technology advances alleviate climate change? Dual UNFCCC, 2018. Status of the Doha amendment [WWW Document]. United Nations
effects of technology change on aggregate carbon dioxide emissions. Energy Framew. Conv. Clim. Chang. http://unfccc.int/kyoto_protocol/doha_
Sustain. Dev. 41, 61e68. https://doi.org/10.1016/j.esd.2017.08.004. amendment/items/7362.php. accessed 9.7.18.
Liu, Q., Wang, S., Zhang, W., Zhan, D., Li, J., 2018. Does foreign direct investment Velde, D.W. Te, 2006. Foreign Direct Investment and Development: an Historical
affect environmental pollution in China's cities? A spatial econometric Perspective. World Economic and Social Survey for, p. 2006.
perspective. Sci. Total Environ. 613e614, 521e529. https://doi.org/10.1016/ Wagner, U.J., Timmins, C.D., 2009. Agglomeration effects in foreign direct invest-
j.scitotenv.2017.09.110. ment and the pollution haven hypothesis. Environ. Resour. Econ. 43, 231e256.
Lopez, L.A., Arce, G., Kronenberg, T., Rodrigues, J.F.D., 2018. Trade from resource-rich https://doi.org/10.1007/s10640-008-9236-6.
countries avoids the existence of a global pollution haven hypothesis. J. Clean. Wang, S., Li, Q., Fang, C., Zhou, C., 2016. The relationship between economic growth,
Prod. 175, 599e611. https://doi.org/10.1016/j.jclepro.2017.12.056. energy consumption, and CO2 emissions: empirical evidence from China. Sci.
Lopez, L.A., Arce, G., Zafrilla, J.E., 2013. Parcelling virtual carbon in the pollution Total Environ. 542, 360e371. https://doi.org/10.1016/j.scitotenv.2015.10.027.
haven hypothesis. Energy Econ. 39, 177e186. https://doi.org/10.1016/ World Bank, 2018. World development indicators Database [WWW Document].
j.eneco.2013.05.006. http://data.worldbank.org/news/new-country-classifications accessed 8.1.18.
Maddala, G.S., Wu, S., 1999. A comparative study of unit root tests with panel data Yang, J., Guo, H., Liu, B., Shi, R., Zhang, B., Ye, W., 2018. Environmental regulation and
and a new simple test. Oxf. Bull. Econ. Stat. 61, 631e652. https://doi.org/10.1111/ the pollution haven hypothesis: do environmental regulation measures matter?
1468-0084.0610s1631. J. Clean. Prod. https://doi.org/10.1016/J.JCLEPRO.2018.08.144.
Maher, M., Christiansen, H., 2001. OECD Global Forum on International Investment: Yang, L., Li, Z., 2017. Technology advance and the carbon dioxide emission in China
New Horizons and Policy Challenges for Foreign Direct Investment in the 21st e empirical research based on the rebound effect. Energy Policy 101, 150e161.
Century. Organisation for Economic Co-operation and Development (OECD), https://doi.org/10.1016/j.enpol.2016.11.020.
Mexico City. Yanikkaya, H., 2003. Trade openness and economic growth: a cross-country
Manderson, E., Kneller, R., 2012. Environmental regulations, outward FDI and het- empirical investigation. J. Dev. Econ. 72, 57e89. https://doi.org/10.1016/
erogeneous firms: are countries used as pollution havens? Environ. Resour. S0304-3878(03)00068-3.
Econ. 51, 317e352. https://doi.org/10.1007/s10640-011-9500-z. Zaman, K., Moemen, M.A. el, 2017. Energy consumption, carbon dioxide emissions
Millimet, D.L., List, J.A., 2004. The case of the missing pollution haven hypothesis. and economic development: evaluating alternative and plausible environ-
J. Regul. Econ. 26, 239e262. https://doi.org/10.1007/s11149-004-7550-7. mental hypothesis for sustainable growth. Renew. Sustain. Energy Rev. 74,
Nas, T.F., Odekon, M., 1992. Economics and Politics of Turkish Liberalization. Lehigh 1119e1130. https://doi.org/10.1016/j.rser.2017.02.072.
University Press , Associated University Presses. Zhang, Y.X., Chao, Q.C., Zheng, Q.H., Huang, L., 2017. The withdrawal of the U.S. from
Nogrady, B., 2018. Waste crisis: where's your recycling going now? Guardian 2. the Paris Agreement and its impact on global climate change governance. Adv.
O'Neill, J., 2001. Building Better Global Economic BRICs, Global Economics Paper. Clim. Change Res. 8, 213e219. https://doi.org/10.1016/j.accre.2017.08.005.
Omri, A., Nguyen, D.K., Rault, C., 2014. Causal interactions between CO2emissions, Zheng, D., Shi, M., 2017. Multiple environmental policies and pollution haven hy-
FDI, and economic growth: evidence from dynamic simultaneous-equation pothesis: evidence from China's polluting industries. J. Clean. Prod. 141,
models. Econ. Modell. 42, 382e389. https://doi.org/10.1016/ 295e304. https://doi.org/10.1016/j.jclepro.2016.09.091.
j.econmod.2014.07.026. Zhou, Y., Zhu, S., He, C., 2017. How do environmental regulations affect industrial
Ong, S.M., Sek, S.K., 2013. Interactions between economic growth and environ- dynamics? Evidence from China's pollution-intensive industries. Habitat Int. 60,
mental quality: panel and non-panel analyses. Appl. Math. Sci. 7, 687e700. 10e18. https://doi.org/10.1016/j.habitatint.2016.12.002.
Online, S.A.H., 2018. A History of Apartheid in South Africa [WWW Document]. Zivot, E., Wang, J., 2006. Modelling Financial Time Series with S-PLUS, Second.
South African Hist. Online. https://www.sahistory.org.za/article/history- Springer Science & Business Media. https://doi.org/10.1007/978-0-387-32348-
apartheid-south-africa. accessed 9.8.18. 0.
Ouyang, W., Xie, D., Zhang, T., 2013. The Energy Consumption and Economic Growth Zugravu-Soilita, N., 2017. How Does Foreign Direct Investment Affect Pollution?
in China - Decoupling Analysis Based on IPAT Decoupling Index, 1, pp. 21e27. toward a Better Understanding of the Direct and Conditional Effects, Environ-
Pao, H.T., Tsai, C.M., 2011. Multivariate Granger causality between CO2 emissions, mental and Resource Economics. Springer, Netherlands. https://doi.org/
energy consumption, FDI (foreign direct investment) and GDP (gross domestic 10.1007/s10640-015-9950-9.
product): evidence from a panel of BRIC (Brazil, Russian Federation, India, and

5 Pollution Haven Hypothesis Revisited A Comparison of The BRICS and MINT Countries Based On VECM Approach 2019

Uploaded by

Copyright:

Available Formats

5 Pollution Haven Hypothesis Revisited A Comparison of The BRICS and MINT Countries Based On VECM Approach 2019

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

5 Pollution Haven Hypothesis Revisited A Comparison of The BRICS and MINT Countries Based On VECM Approach 2019

Uploaded by

Copyright:

Available Formats

Journal of Cleaner Production 227 (2019) 724e738

Contents lists available at ScienceDirect

Journal of Cleaner Production

Pollution haven hypothesis revisited: A comparison of the BRICS and

1. Introduction discounts this effort (Zhang et al., 2017).

Author(s) Method(s) Data structure Type of Main results

Author(s) Method(s) Data structure Type of Main results

if all variables are integrated at the same order and co-integrated. In

Model with constant (level form)

Model with constant and trend (level form)

Model with constant (ﬁrst difference)

Table 5a 4.3. Granger causality regressions

lnTRA 2.091 0.738 1.537*

lnCO2.p.c. 6.171*** 3.856*** 7.292*** a 67.178*** 67.645***

The Kao test also suggests panel co-integration at a 1% level of X

Variable Common unit root Individual unit root

Johansen Fisher test

Hypothesized number of co-integrating equations Model 1 Model 2

Trace Maximum eigenvalue Trace Maximum eigenvalue

None 78.88*** 66.55*** 94.18*** 61.01***

Johansen Fisher test

Hypothesized number of co-integrating equations Model 1 Model 2

Trace Maximum eigenvalue Trace Maximum eigenvalue

None 26.96*** 21.51*** 29.40*** 35.61***

Dependent variable Source of causation (independent variables)

DlnCO2 DlnGDP DlnFDI DlnEC DlnTRA ECT

DlnCO2 e 0.998 0.910 1.365 0.0232 2.460**

Dependent variable Source of causation (independent variables)

DlnCO2 e 1.962 0.566 0.567 0.931

DlnCO2 e 3.148** 3.079*** 4.128** 3.641***

Dep. var: lnCO2.p.c Dep. var: lnGDP.p.c Dep. var: lnTRA

Coef. t-stat. Coef. t-stat. Coef. t-stat.

lnGDP.p.c 0.107 0.733 lnCO2.p.c 0.127 1.208 lnCO2.p.c 0.304 0.810

You might also like

lnCO2.p.c. 6.171* 3.856* 7.292* a 67.178* 67.645***

None 78.88* 66.55* 94.18* 61.01*

None 26.96* 21.51* 29.40* 35.61*

DlnCO2 e 3.148 3.079* 4.128 3.641*