Journal of Cleaner Production xxx (xxxx) xxx

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

The effects of environmental regulation on outward foreign direct

investment’s reverse green technology spillover: Crowding out or
facilitation?
Liyun Liu a, Zhenzhi Zhao a, Mingming Zhang a, *, Changxiao Zhou a, Dequn Zhou b
a
College of Economics and Management, China University of Petroleum (East China), Qingdao, 266580, China
b
College of Economics and Management, Nanjing University of Aeronautics and Astronautics, Nanjing, 211106, China

a r t i c l e i n f o a b s t r a c t

Article history: This study provided statistical evidence for the moderating effect of environmental regulations on the
Received 30 April 2020 reverse green technology spillover of outward foreign direct investment (OFDI). Static and dynamic panel
Received in revised form data model and three estimation methods were used to control for potential endogeneity and ensure the
11 October 2020
robustness of the results. Chinese provincial-level data were used in the empirical study to reveal the
Accepted 13 October 2020
Available online xxx
existence, nature, and regional heterogeneity of the moderating effect of environmental regulations. The
results showed that environmental regulations served as a moderator to facilitate the reverse green
^ as de
Handling editor Cecilia Maria Villas Bo technology spillover of OFDI from the national perspective. After switching to the regional perspective,
Almeida environmental regulations changed into a moderator that produced a significant crowding-out effect on
the reverse green technology spillover of OFDI in eastern regions. In contrast, in the mid-west regions,
Keywords: environmental regulations had no significant moderating effect on this knowledge transfer process.
Environmental regulation Taken together, our findings indicate that the reverse green technology spillover brought by OFDI can be
Reverse green technology spillover amplified through well-designed environmental regulations and improved knowledge transfer ability of
Moderating effect
OFDI.
Regional heterogeneity
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction Porter (1991), Porter and Van de Linde (1995) proposed that well-
designed environmental regulations can protect the environment
The continuous development of the world economy has led to and accelerate technology innovation, offsetting compliance costs.
environmental problems that threaten the lives of human beings. This hypothesis has triggered a wide debate on the impact of
These problems have drawn international attention. To address environmental regulation, with some researchers further focusing
environmental problems in China, the Chinese government has on its impacts on green technology innovation (Chan et al., 2016;
enacted several environmental regulations since the 1970s. Ac- Horbach et al., 2012; Hu et al., 2018; Kesidou and Demirel, 2012; Xie
cording to the China Environment Protection Database (Laws and et al., 2017). Kneller and Manderson (2012) proposed that envi-
regulations, 2020), 8210 laws and regulations related to environ- ronmental regulation may be a key force driving green technology
mental protection have been published from 1970 to the present. Of innovation. Given that green innovation can serve as an effective
these, nearly 40% were published after 2007. Environmental pro- instrument for improving the ecological environment, some re-
tection is becoming increasingly intensive and environmental searchers have further proposed that the influence of environ-
regulations are becoming stricter. mental regulation on green technology innovation could be a
The strengthening of environmental regulations reduces envi- dominant factor for achieving sustainable ecological and economic
ronmental contamination, impacts economic development pat- development (Brunnermeier and Cohen, 2003; Rennings, 2000;
terns, and accelerates green technology progress. The crucial role of Rennings et al., 2006).
environmental regulation has been discussed since the 1990s. Environmental regulations can influence green technology
innovation in many ways. Outward foreign direct investment
(OFDI) is one source of pivotal technology (Buckley et al., 2007;
* Corresponding author. Pathak et al., 2013). Advanced foreign technology can be obtained
E-mail address: [email protected] (M. Zhang). and transferred to the receiving or “home” country through OFDI.

https://doi.org/10.1016/j.jclepro.2020.124689
0959-6526/© 2020 Elsevier Ltd. All rights reserved.

Please cite this article as: L. Liu, Z. Zhao, M. Zhang et al., The effects of environmental regulation on outward foreign direct investment’s reverse
green technology spillover: Crowding out or facilitation?, Journal of Cleaner Production, https://doi.org/10.1016/j.jclepro.2020.124689
L. Liu, Z. Zhao, M. Zhang et al. Journal of Cleaner Production xxx (xxxx) xxx

This technology transfer process is referred to as reverse technol- reveal the internal mechanisms involved. Second, the study applied
ogy spillover (Child and Rodrigues, 2005; Li et al., 2012; Pathak the number of granted green patents to represent green technology
et al., 2013). Eco-technology can also be transferred through innovation performance. This maximized the accuracy in evalu-
OFDI, accelerating green innovation in the receiving country (Zhou ating both innovation and research and development (R&D) per-
et al., 2019). This process is known as the reverse green technology formance. Third, the study conducted sub-regional analyses to
spillover effect of OFDI. However, eco-technology seeking is not the describe the features of the moderating effects of environmental
main motivation for participating in OFDI (Chen et al., 2012; Li et al., regulation. The empirical results provided supportive evidence for
2016; Luo and Tung, 2007, 2018), unless the investor is under the distortion produced by regional heterogeneity. This empha-
pressure to meet specific environmental protection standards. sized the importance of integrating regional heterogeneity into the
Environmental regulations produce this type of pressure. When analysis paradigm when evaluating the impacts of environmental
pressured by environmental regulations, the investor tends to ac- regulation.
quire advanced eco-technology in the process of OFDI, and then use The rest of this paper is organized as follows. Section 2 reviews
the technology to compensate for emission disadvantages in the earlier studies. Section 3 describes data sources, regression models,
home country. Thus environmental regulation may play a moder- estimation methods, and the selection of variables. Section 4 pre-
ating effect in the reverse technology spillover effect of OFDI. sents the empirical results. Section 5 concludes the study with
Most studies that have explored the role of environmental important findings and policy implications.
regulation have been set against a broad background (Porter and
Van de Linde, 1995; Ramanathan et al., 2017; Wang et al., 2019) 2. Literature review
or have explored the impacts on green technology innovation
(Horbach et al., 2012; Hu et al., 2018; Kesidou and Demirel, 2012). There have been extensive investigations about the effects of
However, very little attention has been given to the specific impact environmental regulations on green technology innovation, with
of environmental regulations on green innovation in the reverse mixed results. Neoclassical approaches assumed that increasing
green technology spillover of OFDI (Zhou et al., 2019). Actually, the environmental regulations may lead to additional compliance
reverse green technology spillover effect of OFDI can provide the costs, crowd out R&D investments, and limit the progress of green
home country with the opportunity to learn-by-doing, saving technology innovation (Harrison and Eskeland, 1997; Palmer et al.,
research and development (R&D) investment outlay. However, the 1995; Wang et al., 2019; Xie et al., 2017). A few studies have argued
reverse green technology spillover effect brought about by OFDI is that environmental regulatory pressures do not directly affect
an implicit process that is difficult to control and adjust. If envi- green technology innovation and other environmentally-friendly
ronmental regulations do have a moderating effect, they can be an behaviors (Bigliardi et al., 2012; Frondel et al., 2008; Ren et al.,
important means to adjust the knowledge transfer process of OFDI. 2018). However, extensive theoretical and empirical studies have
This leads to the first question: Is there a positive or negative proposed other viewpoints. The most representative of these is the
moderating effect of environmental regulation in the reverse green “Porter hypothesis.” Porter (1991) and Porter and Van de Linde
technology spillover process of OFDI? (1995) proposed that a Pareto Improvement could be accom-
Furthermore, different regions may respond differently to the plished through a well-designed environmental regulation,
same level of environmental regulation stringency (ERS). This may because this approach can protect the environment, while also
be due to significant regional differences in economic development, enhancing competitiveness through accelerated innovation and
investment preferences, industrial structure, and the effectiveness compliance cost offsets.
of environmental monitoring and enforcement. Further, the Many studies have tested the Porter hypothesis (Ambec and
regional distribution of OFDI has been found to be uneven and Barla, 2006; Filbeck and Gorman, 2004; Ford et al., 2014;
heterogeneous, which could result in diverse technology spillover Ramanathan et al., 2017; Wang et al., 2019) and explored the
effects (Cheung and Ping, 2004; Li et al., 2016, 2018). Thus, the eco- relationship between environmental regulation and green tech-
innovations transferred through OFDI could be spatially heteroge- nology innovation (Chan et al., 2016; Horbach et al., 2012; Hu et al.,
neous under the influence of environmental regulations. So far, we 2018; Kesidou and Demirel, 2012; Xie et al., 2017). These studies
believe few studies have assessed the regional heterogeneity of the have demonstrated that green technology innovation can be driven
possible moderating effects. Accordingly, this leads to this paper’s by domestic regulations (Frondel et al., 2007; Ja €nicke and Jacob,
second question: how does the effect of environmental regulation 2004; Popp, 2006) and the need for a globalized national econ-
on the reverse green technology spillover effect of OFDI vary from omy (Ja €nicke and Jacob, 2004). Some studies have further illus-
the regional perspective? trated that green technology innovations may have significantly
To address these research questions, this paper integrated the different designs and outcomes because of the differences in
static and dynamic panel data model and analyzed provincial panel environmental technology fields (Frondel et al., 2007), existing and
data for 29 Chinese provinces from 2007 to 2016. Multiple esti- anticipated environmental regulations (Khanna et al., 2009), and
mation methods were applied, including Ordinary Least Squares target areas affected by environmental regulations (Horbach et al.,
(OLS), Differenced Generalized Method of Moments (GMM), and 2012; Kammerer, 2009).
System GMM. These approaches addressed the endogeneity issue The United Nations Conference on Trade and Development
and ensured the robustness of the empirical estimates. The (UNCTAD) defined foreign direct investment (FDI) as an investment
empirical results demonstrated the existence and inherent nature activity representing long-term interest and control by a resident
of the moderating effect of environmental regulations on the entity of an economy in another entity outside the investor’s
reverse green technology spillover effect of OFDI in China. The country (UNCTAD, 2017). The circulation of green technology can be
study also captured the regional heterogeneity of the moderating achieved with economic activities and capital exchanges of FDI
effect of environmental regulations, allowing an in-depth analysis €nicke and Jacob, 2004). Thus FDI can lead to reverse technology
(Ja
of differentiated green technology innovation process and the spillover and has been identified as a source of technology (Cheung
development of relevant policy implications. and Ping, 2004; Sinani and Meyer, 2004; Zhou et al., 2019). FDI can
This paper makes three key contributions to the field. First, the be divided into two categories: inward FDI (IFDI) and OFDI. IFDI
study integrated environmental regulation, OFDI, and green tech- refers to an investment by a foreign entity with effective control,
nology innovation into a unified analysis framework, to better investing in a domestic entity. OFDI refers to an investment by a
2
L. Liu, Z. Zhao, M. Zhang et al. Journal of Cleaner Production xxx (xxxx) xxx

domestic entity with effective control investing in a foreign entity when exploring the influence of environmental regulation on the
(Buckley et al., 2007; Li et al., 2012). As such, while it is useful to reverse green technology spillover of OFDI.
explore the effects of environmental regulation on the knowledge Briefly, the literature has already illustrated the impact of
transfer process of FDI, only a few studies have focused on this topic environmental regulation on green innovation (Harrison and
(Hu et al., 2018; Zeng and Eastin, 2012; Zhou et al., 2019). Eskeland, 1997; Palmer et al., 1995; Wang et al., 2019; Xie et al.,
Focusing on less-developed countries with stringent environ- 2017). Few studies have explored this topic with respect to the
mental regulations, Zeng and Eastin (2012) identified the positive knowledge transfer process created by OFDI (Zhou et al., 2019).
effect of IFDI on the level of environmental stewardship for firms in Further, the functions of regional heterogeneity remain a question
their home country. Certification to the ISO 14001 standards has to explore when analyzing the effects of environmental regulation
generally been used to signal a firm’s commitment to environ- on the reverse green technology spillover effect of OFDI. It is
mental policy. Hu et al. (2018) argued that industries with different important to identify systematic and useful policy recommenda-
levels of environmental regulation may perform differently when tions and to apply environmental regulations in a way that ad-
impacted by IFDI green technology spillover. Using green total vances green technological innovation. This study conducts an in-
factor productivity as a measure of green technology spillover, they depth and systematic study by addressing the problem above.
demonstrated that in the high-intensity environmental regulation
industry, capital-based IFDI may produce a positive green tech- 3. Methodology and data
nology spillover effect. Jiao et al. (2020) selected Hong Kong-owned
factories as the provider of FDI and Guangdong as the host region of The strength of environmental protection has gradually
FDI. They analyzed the environmental effects of the Cleaner Pro- increased over time. At the same time, the Chinese government has
duction Partnership Programme on the host region. The results further proposed the “The Green Belt and Road initiative,”
indicated positive improvements in the host region’s environ- emphasizing the strengthening of environment protection. Local
mental performance through FDI. governments have concurrently established specific environmental
The studies above focused on IFDI. The question remains: what protection objectives and requirements appropriate for their pre-
effects can environmental regulations achieve through the reverse fecture. Governments require the timely disclosure of data about
green technology spillover effect of OFDI? Zhou et al. (2019) pollutant emissions and ecological investments. These disclosed
addressed this question using a threshold approach. Consistent publications offer an unprecedented opportunity to identify ways
with Hu et al. (2018), Zhou et al. (2019) applied green total factor of absorbing the green technology provided by OFDI. Investors
productivity to measure the green technology spillover effect of working on OFDI may adjust their business activities to align with
OFDI, identifying a threshold effect of environmental regulations. the environmental protection process. This provides powerful data
The results suggest that OFDI produces a green spillover effect for to conduct in-depth analyses, strongly supporting an empirical
the home country only when the ERS reaches a specific level. study focusing on China.
However, according to the existing literature, the effect of ERS on
OFDI’s reverse green technology spillover could be more complex
3.1. Model specification
than the threshold effect. The underlying interaction mechanisms
between environmental regulations and the reverse green tech-
3.1.1. The static model
nology spillover effect of OFDI is complex, highlighting the need for
To test the effects of environmental regulation on OFDI’s reverse
further exploration. There is similar literature that discussed this
green technology spillover, we first built a static panel data model.
topic with regard to the home country’s carbon productivity (Pan
The natural logarithm of granted green patents count represented
et al., 2020) and environmental quality (Hao et al., 2020), yet
green technology innovation performance. The model also used the
none of them have related this topic with green technology
level of economic development and its square term. The square
innovation.
term of economic development is a covariate that accounts for a
Different regions have differences in economic development,
possible curvilinear relationship between economic development
performance levels, investment preferences, industrial structures,
and eco-innovation. This usage was based on the Environmental
and the effectiveness of environmental monitoring and enforce-
Kuznets Curve (Grossman and Krueger, 1991). Gross Domestic
ment. As such, different regions will respond differently to the
Product (GDP) of each province was used to represent the level of
same levels of ERS. The actual level of ERS also differs among
economic development. The static model was established as fol-
provinces. Thus, environmental regulations have heterogeneous
lows:
effects at a regional level. Previous studies have considered this

Greenpatentit ¼ C þ a1 OFDIit þ a3 GDPit þ a4 GDPit 2 þ a5 ESit þ a6 HUMANit þ a7 RDit

þa8 TRADEit þ a9 LABOURit þ εit (1)

heterogeneity when discussing similar topics, however, they where Greenpatentit is the natural logarithm of granted green
focused on qualitatively analyzing the type or intensity of the patents count for province i in year t; C is a constant term; OFDIit is
environmental regulations (Dean et al., 2009; Xie et al., 2017), the standardized OFDI flow from province i in year t; GDPit reflects
environmental innovation types (Rexha €user and Rammer, 2014), the economic development level of province i in year t; ESit is the
and different firm characteristics (Manderson and Kneller, 2012; energy structure of province i in year t; HUMANit is the human
Wang et al., 2019). Previous studies have also identified the spatial capital of province i in year t; RDit is the standardized R&D expense
spillover impact of FDI (Bai et al., 2012; Jiang et al., 2014). This of province i in year t; TRADEit is the level of trade openness of
highlights the importance of considering regional heterogeneity province i in year t; LABOURit is the labor productivity of province i

3
L. Liu, Z. Zhao, M. Zhang et al. Journal of Cleaner Production xxx (xxxx) xxx

in year t; a19 represents the parameters to be estimated; and εit is previous studies were more likely to be loosely defined and may
an independent and identically distributed error term. have been overly optimistic to justify more subsidies and a more
Porter’s Hypothesis states that environmental regulations can favorable company image (Li, 2009).
advance the R&D level and innovation performance (Porter and The model had two independent variables. The first was pro-
Van der Linde, 1995). To further test the moderating effect of ERS, vincial OFDI, measured by the annual incremental value of the
the interactive term between ERS and OFDI was defined as follows: provincial OFDI flow, deflated with provincial GDP. We applied

Greenpatentit ¼ C þ 4ðOFDIit  ERSit Þ þ a1 OFDIit þ a2 ERSit þ a3 GDPit þ a4 GDPit2 þ a5 ESit

þa6 HUMANit þ a7 RDit þ a8 TRADEit þ a9 LABOURit þ εit (2)

OFDI flow instead of stock for two main reasons. First, it is the
where ERSit is the level of ERS in province i in year t; 4 is the flowing investment that can achieve knowledge transfer. Second,
parameter of the interactive term. OFDI flows can better reflect the dynamic development of inter-
national trade in a given year. The second independent variable was
3.1.2. The dynamic model the ERS. To measure this variable, the ratio of environmental
In light of economic inertia, the evolution of innovations in pollution control cost to GDP for each province was calculated.
green technology may have certain path dependence over several Pollutant disposal in China is now subject to detailed regulations
years. To control for this possible dynamic effect, the lagged form of and is costly for regions. To be specific, the Chinese National Bureau
the logarithmic green patent count was included based on equa- of Statistics publishes environmental pollution control costs that
tions (1) and (2). Therefore, dynamic panel models were built as referencing the European Union’s Classification of Environmental
follows: Protection Activities (Eurostat, 2001) and statistics for international
environmental protection expenditures. As a result, the data con-

Greenpatentit ¼ C þ bGreenpatenti;t1 þ a1 OFDIit þ a3 GDPit þ a4 GDPit2 þ a5 ESit þ a6 HUMANit

þa7 RDit þ a8 TRADEit þ a9 LABOURit þ εit (3)

struction of environmental pollution control costs is consistent

with international standards and can facilitate international com-
where b is the parameter of the lagged form of the green patent parisons in terms of statistical caliber.
count. The interactive term between the ERS and OFDI was the product

Greenpatentit ¼ C þ bGreenpatenti;t1 þ 4ðOFDIit  ERSit Þ þ a1 OFDIit þ a2 ERSit þ a3 GDPit þ a4 GDPit2

þa5 ESit þ a6 HUMANit þ a7 RDit þ a8 TRADEit þ a9 LABOURit þ εit (4)

3.2. Variables definition of the multiplication of the two variables. To avoid significant
multicollinearity, both the OFDI and ERS were centralized before
The dependent variable was the regional green innovation multiplying. Based on previous studies, the control variables
performance, measured using the natural logarithm of the sum of included provincial GDP, R&D expense, trade openness, labor pro-
green invention patents and green utility model patents. A green ductivity, energy structure, and human capital. The log value was
patent portfolio represents an assemblage of distinct units of computed for the provincial GDP. R&D expenses were standardized
ecological knowledge (Ahuja and Katila, 2001). Identifying the against GDP. Trade openness was calculated using the total export-
green patent portfolio can help identify the features of an ecological import volume divided by GDP. Labor productivity was calculated
knowledge reserve. Compared with other variables, the number of as the ratio of total employment to GDP. Based on Wang and Feng
green patents reflects innovation and R&D performance more (2003), Energy structure was calculated as follows:
accurately, because the patent registration is closely overseen by
the state intellectual property office and other relevant de- Pm
partments. Patents represent ecological technologies (Jaffe et al., j¼1 HCIitj
ESit ¼ (5)
1993; Jiao et al., 2020) that reflect innovations in both production GDPit
and process (Hong and Su, 2013; Wang and Lin, 2013; Usai, 2011). In
contrast, other variables used to reflect innovation performance in where HCIitj is the output value of the high coal consumption in-
dustry ranked as j for province i in year t; and m is the number of
4
L. Liu, Z. Zhao, M. Zhang et al. Journal of Cleaner Production xxx (xxxx) xxx

Table 1
Variable descriptions and sources.

Variables Acronym Operationalization Sources

Green innovation Greenpatentit Natural log of green patent count China National Intellectual Property
performance Administration (CNIPA)
OFDI OFDIit OFDI flow/provincial GDP Statistical Bulletin of China’s Outward
Foreign Direct Investment
ERS ERSit 1 if the ratio of investment in environmental pollution control to provincial GDP is higher China Environment Yearbook
than average, 0 if it is lower than average
Gross Domestic GDPit Natural log of provincial GDP China Statistical Yearbook
Product
R&D Expense RDit R&D expense/provincial GDP China Statistical Yearbook
Trade openness TRADEit Total export and import volume/provincial GDP China Statistical Abstract
Labor productivity LABOURit Total employment/provincial GDP China Statistical Yearbook
Energy Structure ESit Sum of the output value of high coal consumption industry/provincial GDP China Industry Statistical Yearbook
Human Capital HUMANit See formula (6) China Statistical Yearbook

high coal consumption industry. According to the Chinese National the original horizontal equation. The lagged variables are used as
Bureau of Statistics, there are six high coal consumption industries, instruments of the differenced equation, and the lagged differenced
including the petroleum, coal and other fuel processing industry; variables are used as the instruments of the horizontal equation.
chemical raw materials and chemical products manufacturing in- The system GMM is an extension of the differenced GMM. More
dustry; non-metallic mineral products industry; ferrous metal importantly, the system GMM can address the problem of a weak
smelting and rolling processing industry; non-ferrous metal instrumental variable. There are certain problems with the differ-
smelting and rolling processing industry; and the power, heat, gas enced GMM. First, the estimation process eliminates the individual
and water production and supply industry (“Industrial Statistics effect of the non-observed section and other variables that do not
(20) - frequently asked questions,” 2020). change with time. Second, sometimes, the lagged variable is not an
Following Barro and Lee (1993), Bermejo Carbonell and Werner ideal instrument. As such, this study mainly depended on the
(2018), human capital was computed as follows: regression results of the system GMM. The results of the differ-
enced GMM are provided for comparison where relevant. To ach-
HUMAN ¼ ð6  Npsg þ 9  Njhsg þ 12  Nshsg þ 16  NgwdaÞ ieve consensus estimates of the GMM estimates, the Arellano-Bond
 0:01 test was further applied to detect whether there was a first or
second-order serial correlation. A Sargan test was used to assess the
(6)
validity of instrumental variables.
where Npsg is the number of graduates from primary school, Njhsg
is the number of graduates from junior high school, Nshsg is the 3.4. Data selection
number of graduates from senior high school, and Ngwda is the
number of graduates with a college degree or above. Table 1 pro- Many studies have applied provincial-level data to analyze
vides the measurements and sources of these variables. ecological performance in China, including Bai et al. (2012), Dean
et al. (2009), Xie et al. (2017), and Zhou et al. (2019). This study
3.3. Estimation methods used panel data from 29 Chinese provinces (except for Tibet, Xin-
jiang, Hong Kong, Macao, and Taiwan) for the period from 2007 to
This study focuses on testing the moderating role of environ- 2016. This ten-year sample span covers the Eleventh and Twelfth
mental regulations on the reverse green technology spillover ef- Five-Year Plan periods. This sample span is of particular interest
fects of OFDI. The result of the Hausman Test indicated that the because it is a period when environmental regulations were being
fixed effect (FE) model was suitable. As such, we first used the static rapidly established. The beginning of the sample span is the year
panel data model and the FE OLS estimation method to depict the after the Six National Conference on Environmental Protection of
relationships among variables and provide comparisons with other China, which marked the start of the increased emphasis on envi-
methods. However, a possible endogeneity issue was identified: the ronmental protection. Since then, a relatively complete environ-
independent variable may influence green innovation perfor- mental regulation system was formed. In addition, pollution
mance, and green innovation performance may influence some or controls were widely deployed during this period. The investment
all of the independent variables. In our case, this could cause a two- in environmental protection during the Eleventh Five-Year Plan
direction effect between OFDI and green innovation through self- period was equal to the total environmental investment over the
selection or learning effects (Li et al., 2016). The OLS was ineffi- previous 20 years (Wang, 2018). Therefore, this period represents a
cient to address this endogeneity issue. As such, next, we applied fast-growing period of environmental regulation, facilitating the
the dynamic panel data model and two GMM estimation methods exploration of the influence of environmental regulation, and
in the model solution. providing practical and significant results.
The system GMM was developed by Blundell and Bond (1998) The data related to OFDI were collected from the Statistical
and the differenced GMM was developed by Arellano and Bond Bulletin of China’s OFDI, compiled by the Ministry of Commerce,
(1991). Both are suitable for generating short dynamic panel esti- National Bureau of Statistics and State Administration of Foreign
mates and are simultaneously applied to generate more robust Exchange. The number of granted green patents was collected from
results. Instrument selection is the main difference between the the CNIPA database. We searched for patents related to environ-
two GMM estimation methods (Chen et al., 2018). The differenced mentally sound technologies, including biomass gasification,
GMM is the differential form of the original equation, using lagged renewable energy, and other energy not generated by combustion.
variables as instruments. However, the system GMM is equivalent The search criteria were defined by the seven green inventory clas-
to the simultaneous establishment of the differenced equation and sification numbers in the Green Patent List published by the World
5
L. Liu, Z. Zhao, M. Zhang et al. Journal of Cleaner Production xxx (xxxx) xxx

Fig. 1. ERS, OFDI flow, and green patents in China.

Intellectual Property Organization (IPC Green Inventory, 2012). Other differed due to regional heterogeneity. In contrast, ERS, another key
data were assembled from the China Statistical Yearbook, China variable, experienced a relatively low variance. This meant that there
Statistical Abstract, China Industry Statistical Yearbook and China was a similar level of stringency in environmental regulation and
Environment Yearbook. were few differences in the efforts to achieve comprehensive
ecological improvements among regions. To control the problem of
4. Empirical results multicollinearity, GDPit and GDPit2 were standardized, after which the
mean variance inflation factor (VIF) for all variables became 3.32.
4.1. Descriptive statistics This indicated there was no serious multicollinearity.

Fig. 1 shows the ERS, OFDI flow, and green patents in China from 4.2. Main results
2007 to 2016. In general, the number of green patents and OFDI in
China increased significantly during the time span, while the ERS Based on the panel data of 29 Chinese provinces from 2007 to
remained steady with only minor fluctuations. The number of 2016, the static panel model was used first to generate an estimate.
green patents increased from 13,381 in 2007 to 115,189 in 2016. The regression results are shown in Column (1) and Column (2) of
Fig. 2 shows the geographic distribution of the ERS, OFDI flow, Table 3. The two variables OFDI and ERS did not significantly impact
and green patents in four specific years. The national increase in green innovation performance before the interaction term was
green patents was not evenly distributed among provinces. Only a added. After adding the interaction term, OFDI became significant
few regions, including Beijing, Jiangsu, Guangdong, and Zhejiang, at the 0.05 level, and the interaction term was also significant. This
are shaded darker green, indicating that they had a higher green means that environmental regulation played a positive moderating
patent count. Fig. 3 also shows this trend. Fig. 2 shows that the most role in the reverse green technology spillover effect of OFDI. Be-
active regions on OFDI were the coastal ones with higher yellow sides, when comparing the results before and after the interaction
bars. Beijing, Tianjin, and Shanghai belong to the most significant term was added, the significance of the other control variables in
regions. The grey circles in Fig. 2 represent the ERS, with larger the model remained the same. Further, the corresponding co-
circles indicating stricter regulations. Some mid-west regions were efficients remained at the same level, except that the R2 values
stricter than the coastal regions. Therefore, the data clearly indicate were 0.725 and 0.724, respectively. These two aspects demonstrate
that there were regional differences in the ERS, the flow of OFDI, certain rationality in variable selection, which was re-confirmed in
and the number of green patents. These may have significantly the subsequent dynamic panel regression model (see Table 3).
impacted the process of reverse green technology spillover. As noted in section 3.3, GMM estimation methods can better
Table 2 provides an analysis of the study’s descriptive statistics. address the problem of endogeneity. As such, we applied both the
Some key variables experienced significant variability, including the system GMM and differenced GMM to ensure the robustness of
OFDI, number of green patents, and GDP. The flow volume of OFDI, results. The results of the system GMM are described in Column (3)
ecological innovation performance, and economic development and (4) of Table 3, and the results of the differenced GMM are
6
L. Liu, Z. Zhao, M. Zhang et al. Journal of Cleaner Production xxx (xxxx) xxx

Fig. 2. Geographic distribution of green patents, ERS, and OFDI flow in China

shown in Column (5) and (6). Blundell and Bond (2000) noted that knowledge is then absorbed and is represented as the number of
the consistency of GMM estimates relies on a valid instrument, as green patents. The overall results of Column (1) and (2), estimated
well as there being no autocorrelation of second-order or higher for using the static FE method, were consistent with the results of Col-
the residual sequence in the difference equation. In this study, we umn (3) and (4), with one interesting difference. When comparing
used the 1e3 years lagged dependent variable as instruments and the significance of the OFDI before and after introducing the inter-
applied the Sargan test to assess the overall validity of instruments action term, it can be seen that the influence of OFDI changed to
in the estimation process. The Sargan values of Column (3) to being significant, after having been insignificant, after introducing
Column (6) all exceeded 0.1. This indicates that the results of the the interaction term. This change reveals that OFDI enhanced green
Sargan test were all significant at a 10% significance level. This innovation performance under the pressure of environmental
supported the research design, verifying the validity of all instru- regulation. The same kind of change was seen in the OFDI co-
ment variables. In addition, the results of the Arellano-Bond test, efficients in Column (5) and (6). Before the interaction term was
labeled as AR (2), showed no second-order (or higher) correlation considered, the OFDI coefficient was significant but negative. After
in the residual sequence of the differenced equation. The results of adding the interaction term, the coefficient became positive and was
the above tests further supported the use of GMM for the estima- significant. The moderating effect of environmental regulation was
tion. The regression results using the GMM estimation method larger when using differenced GMM as the estimation method,
were reliable. compared to when using the static FE method.
The three estimation methods all captured the moderating effect The analyses above show overall consistency in the moderating
of environmental regulation on the reverse green technology spill- effect of environmental regulation on the reverse green technology
over effect of OFDI. Column (3) shows that, separately, the OFDI and spillover effect of OFDI. There were differences in the details, likely
environmental regulation had no significant influence on green caused by different estimation methods, missing variables, weak
technology innovation. This was measured before the interaction variables, and endogeneity. As illustrated in section 3.3, we based
term was included. Column (4) shows the results after the interac- our analysis of the interaction mechanism on the results of the
tion term was included. In that case, the interaction term between system GMM.
the OFDI and ERS was positive and significant. This means that OFDI The results of the system GMM demonstrated that environ-
can transfer green technology spillover back to the home country mental regulation had a positive and significant influence on the
using the moderating effect of environmental regulation. The reverse technology spillover effect of OFDI. This aligns with Porter’s
7
L. Liu, Z. Zhao, M. Zhang et al. Journal of Cleaner Production xxx (xxxx) xxx

Fig. 3. Accumulated green patent distribution from 2007 to 2016

Hypothesis (Porter and van de Linde, 1995). The mechanism et al., 2018; Huang and Zhang, 2017; Xie et al., 2017). However,
influencing environmental regulations can be discussed from two there are diverse states in each province, both in terms of eco-
perspectives. First, the ultimate goal of investment is to generate nomics and environmental protection. Even if environmental reg-
profits, while complying with environmental regulations. Green ulations are equally strict in each province, the efficiency and
technology innovation can improve cost efficiency and reduce oversight of ERS may differ. This is referred to as regional hetero-
costs. Meanwhile, the technological advantages produced by green geneity. This section analyzes the difference in the effect of envi-
technology innovation can offset the constraints and additional ronmental regulations on OFDI’s reverse green technology spillover
costs of environmental regulation on investments. Second, envi- in eastern and mid-west regions.
ronmental regulation represents the government’s orientation, and Table 4 shows the regional results of the system GMM. The re-
investors need to follow government requirements to ensure a sults passed the Arellano-Bond test and the Sargan test, indicating
smooth and effective investment process. In this way, the pressures that the regional regression results using the system GMM esti-
created by the government drive green innovation. Investors will in mation method were reliable. For eastern regions, the OFDI and
turn develop more advanced green technology to meet regulatory environmental regulations had no significant influence on green
requirements. At the same time, if the investors meet the re- innovation before considering the interaction term. After intro-
quirements, their OFDI activities can earn more government sup- ducing the interaction term, the influence of ERS on green tech-
port, including financial and institutional resources. nology innovation became significant, with the coefficient
changing from 0.275 to 4.095. The influence of OFDI on green
4.3. Sub-regional results analyses technology innovation became significant but negative. The inter-
action term between OFDI and ERS also showed a significant and
The recent continuous development of China’s economy has negative impact on green innovation. This indicates that environ-
also increased awareness about environmental protection (Chen mental regulations had a significant crowding out effect on OFDI
and green innovation.
Table 2 For the mid-west regions, the result with the interaction term
Descriptive statistics. between OFDI and ERS indicated the presence of a reverse green
Variable Mean Std.Dev Min Max Observations technology spillover effect of OFDI. However, without the interac-
tion term, neither OFDI nor ERS were significant. Environmental
Greenpatent 6.692 1.481 2.485 9.675 290
OFDI 0.005 11.740 5.25E-06 0.070 290
regulation appears to have encouraged the use of OFDI to produce a
ERS 0.390 0.489 0 1.000 290 reverse green technology spillover effect, based on the change in
GDP 9.451 0.920 6.664 11.300 290 the OFDI coefficients. However, the coefficient of the interaction
ES 0.575 0.197 0.217 1.231 290 term was not significant, highlighting the need for further evidence.
HUMAN 9.656 1.389 6.784 14.410 290
Therefore, we plotted the marginal effect of OFDI on green inno-
RD 0.016 0.011 0.002 0.061 290
TRADE 0.296 0.338 0.013 1.708 290 vation, with environmental regulation serving as a moderator.
LABOUR 0.183 0.106 0.048 0.683 290 Fig. 4 shows that the confidence intervals contain 0 throughout the

8
L. Liu, Z. Zhao, M. Zhang et al. Journal of Cleaner Production xxx (xxxx) xxx

Table 3
Regression results.

Variables Green patent

Static panel model Dynamic panel model

FE SYS-GMM DIFF-GMM

(1) (2) (3) (4) (5) (6)

OFDI 0.003 0.004** 0.001 0.001 0.002** 0.006*

(0.002) (0.002) (0.001) (0.002) (0.001) (0.003)
ERS 0.004 0.010 0.011 0.012 0.015 0.020
(0.041) (0.041) (0.017) (0.041) (0.012) (0.048)
OFDI  ERS 0.004** 0.003* 0.007***
(0.002) (0.001) (0.002)
GDP 1.339*** 1.324*** 0.165 0.106 0.698* 0.300
(0.120) (0.119) (0.146) (0.175) (0.360) (0.303)
GDP 2 0.640*** 0.632*** 0.065 0.017 0.436 0.187
(0.082) (0.081) (0.083) (0.110) (0.308) (0.215)
ES 0.323 0.319 0.109 0.298 0.990** 0.491
(0.263) (0.261) (0.131) (0.258) (0.454) (0.384)
HUMAN 0.014 0.014 0.025*** 0.020** 0.016*** 0.012*
(0.015) (0.015) (0.007) (0.008) (0.005) (0.007)
RD 10.175*** 10.360*** 0.177 1.705 1.958 2.292
(3.859) (3.831) (1.523) (2.286) (1.707) (2.122)
TRADE 0.856*** 0.893*** 0.297*** 0.219** 0.116 0.343
(0.244) (0.242) (0.074) (0.097) (0.193) (0.239)
LABOUR 4.722*** 4.760*** 0.307 0.268 3.251*** 2.949***
(0.352) (0.350) (0.302) (0.489) (0.820) (1.029)
Greenpatentt1 0.636*** 0.578*** 0.489*** 0.569***
(0.063) (0.130) (0.081) (0.090)
Constant 7.969*** 7.990*** 1.611*** 2.081*** 4.862*** 4.033***
(0.222) (0.221) (0.323) (0.687) (0.845) (0.899)

R2 0.725 0.724
AR (2) 0.1661 0.1566 0.1199 0.1633
Sargan Test 0.9971 1.0000 0.9924 1.0000
Observations 290 290 232 232 232 232
Number of provinces 29 29 29 29 29 29
Number of years 10 10 8 8 8 8

Standard errors in parentheses.

***p < 0.01, **p < 0.05, *p < 0.1.

changes in ERS. This means that the difference created by ERS did a good choice. Most provinces and municipalities in the eastern
not significantly differ from zero. The regression results and Fig. 4 region have introduced relatively lenient levels of environmental
show that environmental regulations had no significant moder- regulation. Investors in the eastern region are capable of managing
ating effect on the reverse green technology spillover effect of OFDI environmental regulation at the current level. As a result, there are
in the mid-west regions. insufficient incentives for investors in the eastern region to switch
There are specific reasons for the differences between the main to green innovation in the process of allocating OFDI, even under
results and sub-regional results. The main focus of the research the pressure of environmental regulation. The question remains:
analysis was the 29 Chinese provinces. The models generated for why does ERS play a negative role in the interactions between OFDI
the main analysis, however, may not have fully reflected all the and green technology innovation? We believe a higher level of ERS
differences in regional characteristics. As such, the sub-regional may encourage regions to shift the budgets originally used for
analyses divided the research populations into several groups, green R&D to one-time actions that work for a long period. For
revealing different features. More specifically, the models in the example, this could include constructing an end-of-pipe control
main analysis assumed a constant slope coefficient across regions. facility.
This may not have been an accurate assumption. Regions in the In the mid-west regions, environmental regulations do not have a
sample were widely distributed, with regional differences in their statistically significant moderating effect on the reverse green
response to environmental regulations. This explains the differ- technology spillover effect of OFDI, even though the environmental
ences between the main results and sub-regional results. regulations in the mid-west regions are stricter (see Fig. 2). This
When comparing the eastern regions and mid-west regions, result could be due to underdeveloped economies and over-stressed
environmental regulations had a different effect on the reverse environmental regulations. The main motivation for mid-west re-
green technology spillover effect of OFDI. First, the lock-in effect gions to engage in OFDI is for economic development. Green inno-
and ERS can help explain why environmental regulations had a vation is less emphasized. Mid-west regions may not have sufficient
significant crowding out effect in the eastern region. Because of budgets and resources to develop green R&D and introduce enough
their economic head-start, eastern regions have developed a rela- technical personnel. Thus these regions have limited absorptive ca-
tively fixed mode of OFDI. Based on this, changing the investment pacity to employ the foreign green technologies acquired from OFDI
mode could have created high switching costs, market risks, and to further enhance domestic innovation performance. From the
uncertainties. This invoked the lock-in effect (Farrell and perspective of environmental regulation, the regulatory levels are
Klemperer, 2007). quite strict in the mid-west regions, possibly leading to excessive
A powerful external force is needed to break the lock-in effect compliance burdens. When environmental regulations are difficult
(Chignell and Gross, 2013). Strict environmental regulations may be to implement, there are few opportunities to focus on green
9
L. Liu, Z. Zhao, M. Zhang et al. Journal of Cleaner Production xxx (xxxx) xxx

Table 4 address lax regulations when there is no innovation. When ERS rises,
Regression results in the eastern and mid-west regions. however, the cost of compliance also increases, offsetting the po-
Variables Green patent tential for green innovation.
Eastern regions Mid-west regions

(1) (2) (3) (4) 5. Conclusions and policy recommendations

OFDI 0.004 0.039** 0.003 0.029*
(0.005) (0.018) (0.005) (0.016) This study identified the moderating effect of environmental
ERS 0.275 4.095* 0.121 0.028 regulation in the reverse green technology spillover effect of OFDI,
(0.335) (2.468) (0.095) (0.168)
and explored the regional heterogeneity of these effects. Two panel
OFDI  ERS 0.374* 0.019
(0.216) (0.014) data models and three estimation methods were used to ensure the
GDP 0.240 65.925 0.239 0.031 validity and robustness of the results. Panel data from 29 provinces
(0.426) (40.989) (0.359) (1.045) from 2007 to 2016 in China were applied. The number of granted
GDP 2 0.104 56.938 0.216 0.785 green patents, an underused method for representing green tech-
(0.373) (34.987) (0.457) (1.179)
nology innovation performance, facilitated an analysis of the real-
ES 0.113 150.034 0.697* 2.185*
(1.298) (92.572) (0.406) (1.239) ized green technology reserve.
RD 3.150 51.979* 22.286 39.704 This study yields interesting findings. First, it demonstrated that
(8.480) (31.123) (17.563) (26.734) environmental regulations played a positive and significant
TRADE 0.300 56.791* 1.391*** 3.089** moderating role in the reverse technology spillover of OFDI from
(0.562) (34.490) (0.522) (1.366)
LABOUR 5.483 244.122* 2.060** 4.923*
the national perspective. Second, in the eastern regions, environ-
(10.192) (147.651) (0.872) (2.719) mental regulations had a significant crowding-out effect on the
Greenpatentt1 0.649 10.768 0.853*** 0.507*** reverse green technology spillover effect of OFDI. This crowding-
(0.573) (6.755) (0.104) (0.166) out effect may have been the result of the combined action of a
Constant 3.533 134.423* 1.499** 4.624***
lock-in effect of regional developing patterns and a relatively low
(6.036) (78.035) (0.651) (1.767)
AR (2) 0.5953 0.5392 0.1582 0.1836 level of ERS. Third, in the mid-west regions, environmental regu-
Sargan test 1.0000 1.0000 1.0000 1.0000 lations had no significant moderating effect on the reverse green
Observations 99 99 162 162 technology spillover effect of OFDI. This may be due to the com-
Number of Provinces 11 11 18 18 bination of an underdeveloped economy, less attention to green
Number of years 9 9 9 9
technology innovation, a low volume of OFDI flow, and insufficient
Standard errors in parentheses. budgets and resources.
***p < 0.01, **p < 0.05, *p < 0.1.
The results above highlight specific practical implications for
innovation. Porter and van de Linde (1995) interpreted this outcome. China and other emerging market countries with similar features.
They posited that secondary treatment solutions can be used to On the one hand, environmental regulations should be well-
designed. As shown in the empirical results, OFDI can only

Fig. 4. Marginal effect of OFDI on green innovation.

10
L. Liu, Z. Zhao, M. Zhang et al. Journal of Cleaner Production xxx (xxxx) xxx

enhance eco-innovation when there is a well-designed environ- panel data models. J. Econom. 87 (1), 115e143.
Blundell, R., Bond, S., 2000. GMM Estimation with persistent panel data: an
mental regulation system. Regional heterogeneity is an important
application to production functions. Econom. Rev. 19, 321e340.
factor that should not be ignored to design environmental regula- Brunnermeier, S.B., Cohen, M.A., 2003. Determinants of environmental innovation
tion. For eastern regions of China that have usually experienced a in US manufacturing industries. J. Environ. Econ. Manag. 45 (2), 278e293.
lock-in effect with respect to their developing patterns, it is sug- Buckley, P.J., Clegg, L.J., Cross, A.R., Liu, X., Voss, H., Zheng, P., 2007. The determinants
of Chinese outward foreign direct investment. J. Int. Bus. Stud. 38 (4), 499e518.
gested to design stricter environmental regulations, and ensure the Chan, H.K., Yee, R.W., Dai, J., Lim, M.K., 2016. The moderating effect of environ-
implementation process correct, orderly and efficient. mental dynamism on green product innovation and performance. Int. J. Prod.
On the other hand, the government should take measures to Econ. 181, 384e391.
Chen, V., Li, J., Shapiro, D., 2012. International reverse spillover effects on parent
improve the knowledge transfer ability of OFDI according to firms: evidences from emerging-market MNEs in developed markets. Eur.
regional features. Due to the low OFDI volume and the lack of high- Manag. J. 30 (3), 204e218.
tech professionals, developing regions usually have shortages in the Chen, H., Hao, Y., Li, J., Song, X., 2018. The impact of environmental regulation,
shadow economy, and corruption on environmental quality: theory and
technology transfer channels and a low absorptive capacity for empirical evidence from China. J. Clean. Prod. 195, 200e214.
green technology. Thus it is needed to promote international trade Cheung, K.Y., Ping, L., 2004. Spillover effects of FDI on innovation in China: evidence
and introduce high-tech professionals. For developed regions, it is from the provincial data. China Econ. Rev. 15 (1), 25e44.
Chignell, S., Gross, R.J., 2013. Not locked-in? The overlooked impact of new gas-fired
suggested to improve the structure of OFDI. For example, the gov- generation investment on long-term decarbonisation in the UK. Energy Pol. 52,
ernment can increase technology-motivated OFDI and promote 699e705.
pollution-intensive enterprise’s participation. Child, J., Rodrigues, S., 2005. The internationalization of Chinese firms: a case for
theoretical extension? Manag. Organ. Rev. 1, 381e410, 03.
Like all studies, this study had some limitations given the
Dean, J.M., Lovely, M.E., Wang, H., 2009. Are foreign investors attracted to weak
complexity of the knowledge transfer process resulting from OFDI. environmental regulations? Evaluating the evidence from China. J. Dev. Econ.
First, this study was not able to distinguish different types of OFDI 90 (1), 1e13.
because of data limitations. Different investment destinations, Eurostat, 2001. Classification of Environmental Protection Activities and Expendi-
ture (2000). Conference of European Statisticians, Luxembourg.
motivations, and entry pathways for OFDI may cause distinct Farrell, J., Klemperer, P., 2007. Coordination and lock-in: competition with switching
reverse technology spillover effects. Second, some innovation costs and network effects. Handb. Ind. Organ. 3, 1967e2072.
achievements cannot be codified and are not reflected in the Filbeck, G., Gorman, R.F., 2004. The relationship between the environmental and
financial performance of public utilities. Environ. Resour. Econ. 29 (2), 137e157.
number of patents. Future research could identify alternative ap- Ford, J.A., Steen, J., Verreynne, M.L., 2014. How environmental regulations affect
proaches that extend beyond normal eco-innovation outputs to innovation in the Australian oil and gas industry: going beyond the Porter
more clearly depict the transfer channels of OFDI, for example, big Hypothesis. J. Clean. Prod. 84, 204e213.
Frondel, M., Horbach, J., Rennings, K., 2007. End-of-pipe or cleaner production? An
data analytics (Maheshwari et al., 2020). empirical comparison of environmental innovation decisions across OECD
countries. Bus. Strat. Environ. 16 (8), 571e584.
CRediT authorship contribution statement Frondel, M., Horbach, J., Rennings, K., 2008. What triggers environmental man-
agement and innovation? Empirical evidence for Germany. Ecol. Econ. 66 (1),
153e160.
Liyun Liu: Methodology, Software, Formal analysis, Writing - Grossman, G.M., Krueger, A.B., 1991. Environmental Impacts of a North American
original draft. Zhenzhi Zhao: Project administration, Funding Free Trade Agreement (No. W3914). National Bureau of Economic Research.
Hao, Y., Guo, Y., Guo, Y., Wu, H., Ren, S., 2020. Does outward foreign direct invest-
acquisition. Mingming Zhang: Writing - review & editing, Visual-
ment (OFDI) affect the home country’s environmental quality? The case of
ization. Changxiao Zhou: Data curation, Validation. Dequn Zhou: China. Struct. Change Econ. Dynam. 52, 109e119.
Supervision. Harrison, A.E., Eskeland, G., 1997. Moving to Greener Pastures? Multinationals and
the Pollution-Haven Hypothesis. The World Bank.
Horbach, J., Rammer, C., Rennings, K., 2012. Determinants of eco-innovations by
Declaration of competing interest type of environmental impactd the role of regulatory push/pull, technology
push and market pull. Ecol. Econ. 78, 112e122.
The authors declare that they have no known competing Hong, W., Su, Y., 2013. The effect of institutional proximity in non-local university-
industry collaborations: an analysis based on Chinese patent data. Res. Pol. 42
financial interests or personal relationships that could have (2), 454e464.
appeared to influence the work reported in this paper. Huang, Y., Zhang, Y., 2017. How does outward foreign direct investment enhance
firm productivity? A heterogeneous empirical analysis from Chinese
manufacturing. China Econ. Rev. 44, 1e15.
Acknowledgments Hu, J., Wang, Z., Lian, Y., Huang, Q., 2018. Environmental regulation, foreign direct
investment and green technological progressdevidence from Chinese
The authors gratefully acknowledge the financial support pro- manufacturing industries. Int. J. Environ. Res. Publ. Health 15 (2), 221.
Industrial Statistics (20) - Frequently Asked Questions, 2020. Stats.gov.cn. http://
vided by the National Natural Science Foundation of China (No. www.stats.gov.cn/tjzs/cjwtjd/201311/t20131105_455942.html. (Accessed 6 July
71804190) and the Fundamental Research Funds for the Central 2020)
Universities (No. 19CX04034B). IPC Green Inventory, 2012. https://www.wipo.int/classifications/ipc/en/green_
inventory/. (Accessed 6 February 2020).
Jaffe, A.B., Trajtenberg, M., Henderson, R., 1993. Geographic localization of knowl-
References edge spillovers as evidenced by patent citations. Q. J. Econ. 108 (3), 577e598.
€nicke, M., Jacob, K., 2004. Lead markets for environmental innovations: a new role
Ja
Ahuja, G., Katila, R., 2001. Technological acquisitions and the innovation perfor- for the nation state. Global Environ. Polit. 4 (1), 29e46.
mance of acquiring firms: a longitudinal study. Strat. Manag. J. 22 (3), 197e220. Jiang, L., Folmer, H., Ji, M., 2014. The drivers of energy intensity in China: a spatial
Ambec, S., Barla, P., 2006. Can environmental regulations be good for business? An panel data approach. China Econ. Rev. 31, 351e360.
assessment of the Porter hypothesis. Energy Stud. Rev. 14 (2), 42. Jiao, J., Chen, C., Bai, Y., 2020. Is green technology vertical spillovers more significant
Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo in mitigating carbon intensity? Evidence from Chinese industries. J. Clean. Prod.
evidence and an application to employment equations. Rev. Econ. Stud. 58 (2), 120354.
277e297. Kammerer, D., 2009. The effects of customer benefit and regulation on environ-
Bai, C.E., Ma, H., Pan, W., 2012. Spatial spillover and regional economic growth in mental product innovation: empirical evidence from appliance manufacturers
China. China Econ. Rev. 23 (4), 982e990. in Germany. Ecol. Econ. 68 (8e9), 2285e2295.
Barro, R.J., Lee, J.W., 1993. International comparisons of educational attainment. Kesidou, E., Demirel, P., 2012. On the drivers of eco-innovations: empirical evidence
J. Monetary Econ. 32 (3), 363e394. from the UK. Res. Pol. 41 (5), 862e870.
Bermejo Carbonell, J., Werner, R.A., 2018. Does foreign direct investment generate Khanna, M., Deltas, G., Harrington, D.R., 2009. Adoption of pollution prevention
economic growth? A new empirical approach applied to Spain. Econ. Geogr. 94 techniques: the role of management systems and regulatory pressures. Environ.
(4), 425e456. Resour. Econ. 44 (1), 85e106.
Bigliardi, B., Bertolini, M., Doran, J., Ryan, G., 2012. Regulation and firm perception, Kneller, R., Manderson, E., 2012. Environmental regulations and innovation activity
eco-innovation and firm performance. Eur. J. Innovat. Manag. 15 (4), 1e24. in UK manufacturing industries. Resour. Energy Econ. 34 (2), 211e235.
Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic Laws and regulations, 2020. China Environment Protection Database. http://hbk.cei.

11
L. Liu, Z. Zhao, M. Zhang et al. Journal of Cleaner Production xxx (xxxx) xxx

cn/aspx/default.aspx. (Accessed 12 August 2020). hypothesis. J. Clean. Prod. 155, 79e92.

Li, J., Li, Y., Shapiro, D., 2012. Knowledge seeking and outward FDI of emerging Ren, S., Li, X., Yuan, B., Li, D., Chen, X., 2018. The effects of three types of environ-
market firms: the moderating effect of inward FDI. Global Strategy Journal 2 (4), mental regulation on eco-efficiency: a cross-region analysis in China. J. Clean.
277e295. Prod. 173, 245e255.
Li, J., Strange, R., Ning, L., Sutherland, D., 2016. Outward foreign direct investment Rennings, K., 2000. Redefining innovationdeco-innovation research and the
and domestic innovation performance: evidence from China. Int. Bus. Rev. 25 contribution from ecological economics. Ecol. Econ. 32 (2), 319e332.
(5), 1010e1019. Rennings, K., Ziegler, A., Ankele, K., Hoffmann, E., 2006. The influence of different
Li, J., Xia, J., Shapiro, D., Lin, Z., 2018. Institutional compatibility and the interna- characteristics of the EU environmental management and auditing scheme on
tionalization of Chinese SOEs: the moderating role of home subnational in- technical environmental innovations and economic performance. Ecol. Econ. 57
stitutions. J. World Bus. 53, 641e652. (1), 45e59.
Li, X., 2009. China’s regional innovation capacity in transition: an empirical Rexha €user, S., Rammer, C., 2014. Environmental innovations and firm profitability:
approach. Res. Pol. 38 (2), 338e357. unmasking the Porter hypothesis. Environ. Resour. Econ. 57 (1), 145e167.
Luo, Y., Tung, R., 2007. International expansion of emerging market enterprises: a Sinani, E., Meyer, K.E., 2004. Spillovers of technology transfer from FDI: the case of
springboard perspective. J. Int. Bus. Stud. 38 (4), 481e498. Estonia. J. Comp. Econ. 32 (3), 445e466.
Luo, Y., Tung, R., 2018. A general theory of springboard MNEs. J. Int. Bus. Stud. 49 (2), UNCTAD, 2017. Methodological Note. World Investment Report 2017: Investment
129e152. and the Digital Economy. United Nations, New York and Geneva.
Maheshwari, S., Gautam, P., Jaggi, C.K., 2020. Role of Big Data Analytics in supply Usai, S., 2011. The geography of inventive activity in OECD regions. Reg. Stud. 45 (6),
chain management: current trends and future perspectives. Int. J. Prod. Res. 711e731.
1e26. Wang, X., Feng, Z., 2003. Energy consumption with sustainable development in
Manderson, E., Kneller, R., 2012. Environmental regulations, outward FDI and het- developing country: a case in Jiangsu, China. Energy Pol. 31, 1679e1684.
erogeneous firms: are countries used as pollution havens? Environ. Resour. Wang, C., Lin, G., 2013. Dynamics of innovation in a globalizing China: regional
Econ. 51 (3), 317e352. environment, inter-firm relations and firm attributes. J. Econ. Geogr. 13 (3),
Palmer, K., Oates, W.E., Portney, P.R., 1995. Tightening environmental standards: the 397e418.
benefit-cost or the no-cost paradigm? J. Econ. Perspect. 9 (4), 119e132. Wang, Y., 2018. The historical change of environmental protection policy in China,
Pan, X., Li, M., Wang, M., Chu, J., Bo, H., 2020. The effects of outward foreign direct 04 Environ. Sustain. Dev. 43, 5e9 [in Chinese].
investment and reverse technology spillover on China’s carbon productivity. Wang, X., Zhang, C., Zhang, Z., 2019. Pollution haven or porter? The impact of
Energy Pol. 145, 111730. environmental regulation on location choices of pollution-intensive firms in
Pathak, S., Xavier-Oliveira, E., Laplume, A.O., 2013. Influence of intellectual property, China. J. Environ. Manag. 248, 109248.
foreign investment, and technological adoption on technology entrepreneur- Xie, R.H., Yuan, Y.J., Huang, J.J., 2017. Different types of environmental regulations
ship. J. Bus. Res. 66 (10), 2090e2101. and heterogeneous influence on “green” productivity: evidence from China.
Popp, D., 2006. International innovation and diffusion of air pollution control Ecol. Econ. 132, 104e112.
technologies: the effects of NOX and SO2 regulation in the US, Japan, and Zeng, K., Eastin, J., 2012. Do developing countries invest up? The environmental
Germany. J. Environ. Econ. Manag. 51 (1), 46e71. effects of foreign direct investment from less-developed countries. World Dev.
Porter, M.E., 1991. America s green strategy. Sci. Am. 264 (4), 168. 40 (11), 2221e2233.
Porter, M.E., Van der Linde, C., 1995. Toward a new conception of the environment- Zhou, Y., Jiang, J., Ye, B., Hou, B., 2019. Green spillovers of outward foreign direct
competitiveness relationship. J. Econ. Perspect. 9 (4), 97e118. investment on home countries: evidence from China’s province-level data.
Ramanathan, R., He, Q., Black, A., Ghobadian, A., Gallear, D., 2017. Environmental J. Clean. Prod. 215, 829e844.
regulations, innovation and firm performance: a revisit of the Porter

12