Content uploaded by Tingting Sun
Author content
All content in this area was uploaded by Tingting Sun on Dec 28, 2023
Content may be subject to copyright.
Vol:.(1234567890)
Environmental Science and Pollution Research (2023) 30:123948–123965
https://doi.org/10.1007/s11356-023-31076-5
1 3
RESEARCH ARTICLE
Does artificial intelligence (AI) reduce ecological footprint? The role
ofglobalization
QiangWang1,2· TingtingSun1· RongrongLi1,2
Received: 21 September 2023 / Accepted: 13 November 2023 / Published online: 23 November 2023
© The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2023
Abstract
This article explores the impact of artificial intelligence (AI) on global ecological footprints, which has important implica-
tions for global sustainability in the digital age. Using the comprehensive evaluation index of AI constructed by the entropy
method and the dataset at the global national level, we find that from 2010 to 2019, the overall level of global AI shows an
upward trend, in which the growth rate of AI in developed countries is more pronounced and exhibits a stable growth trend,
while the growth rate of AI in developing countries displays a trend of instability. The research results show that AI has a
significant inhibitory effect on ecological footprints. This conclusion holds even after endogeneity and robustness tests. In
addition, under the effect of globalization, the impact of AI on ecological footprints shows nonlinear characteristics. As
globalization deepens, the marginal effect of AI in reducing the ecological footprint shows an increasing trend. These find-
ings emphasize the important role of AI in environmental governance and provide a new and comprehensive perspective for
policymakers. Therefore, the government should continue to support the research and application of AI, promote the cross-
industry integration of AI, and play a positive role in the process of globalization to promote global sustainable development.
Keywords Artificial intelligence· Ecological footprint· Globalization· Panel threshold model
Introduction
As the global economy continues to expand and the process
of industrialization accelerates, humanity is facing increas-
ingly severe environmental problems, such as excessive con-
sumption of natural resources, continuous degradation of the
ecological environment, frequent extreme weather events,
and a sharp decrease in biodiversity (Sun and Dong 2023).
These issues not only pose a direct threat to human survival
and well-being (Danish etal. 2019), but also pose a serious
challenge to global ecological security (Castells-Quintana
etal. 2021). Therefore, finding a balance between economic
development and environmental preservation to achieve
sustainable development has become a key issue that the
international community urgently needs to resolve (Coteur
etal. 2019; Saqib etal. 2023). In this context, scholars have
continuously explored various methods to accurately assess
the impact of human activities on the environment (Sun and
Dong 2022). In 1998, Wackernagel and Rees (1998) intro-
duced the concept of “ecological footprint,” which compre-
hensively evaluates environmental quality by quantifying
human resource consumption and pollution emissions in
various key ecological subsystems (Appiah etal. 2023). Due
to its comprehensiveness and scientific nature, this method
has rapidly become a hot topic in the field of sustainable
development and has received widespread attention (Ahmad
etal. 2020).
According to data from the Global Ecological Footprint
Network, the Earth has been in a state of ecological over-
shoot since 1970, indicating that human demand for natu-
ral resources far exceeds the earth’s capacity to regenerate
these resources. What is more concerning is that this eco-
logical deficit has not been mitigated but rather hasshown
a trend of continuous increase, as depicted in Fig.1. This
not only illustrates the crisis we currently face but also
highlights the urgent need to address this issue (Dong etal.
Responsible Editor: Philippe Garrigues
* Qiang Wang
wangqiang7@upc.edu.cn
1 School ofEconomics andManagement, China
University ofPetroleum (East China), Qingdao266580,
People’sRepublicofChina
2 School ofEconomics andManagement, Xinjiang University,
Wulumuqi830046, People’sRepublicofChina
123949Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
2023). Furthermore, the latest statistics indicate that by
2022, nearly 75% of the 200 countries surveyed are expe-
riencing an ecological deficit (Footprint Data Foundation
etal. 2023). This reflects that the problem of ecological
deficit has become global, extending far beyond any single
region or country. Such a situation not only exacerbates
the destruction of ecosystems, but also disrupts the bal-
ance of ecosystems and perpetuating destructive cycles
(Chen etal. 2022). Therefore, how to reduce the ecological
footprint and achieve sustainable development has become
a critical issue of global concern.
Meanwhile, artificial intelligence (AI), as the core
technology of the Fourth Industrial Revolution, is increas-
ingly becoming an important driving force behind global
economic development and social progress (Borges etal.
2021). AI not only breaks through temporal and spatial
boundaries to foster technological progress (Liu etal.
2020), but also plays a pivotal role in global economic
growth (Vinuesa etal. 2020; Borges etal. 2021). In terms
of environmental protection, AI has shown tremendous
potential (Kar etal. 2022; Marvin etal. 2022). Zhang etal.
(2022a) indicate that AI constructs scientific and efficient
prediction models by processing massive amounts of his-
torical data, thereby enhancing the utilization efficiency of
renewable energy. In addition, AI can also collect and ana-
lyze user behavior data to predict product selection prefer-
ences, encouraging users to make energy-saving choices
and thereby improving environmental quality (Hua etal.
2021). Nevertheless, we must also recognize that AI in its
initial stages of development may entail significant energy
consumption and cost expenses (Hao 2019; Strubell etal.
2019). This situation prompts us to contemplate an intrigu-
ing question: whether AI can bolster sustainable develop-
ment, ameliorate the adverse environmental consequences
of human activities, and reduce our ecological footprint.
Against the backdrop of globalization, the cross-border
flow of production factors has enhanced the efficiency of
resource allocation, spurred economic growth and tech-
nological advancement, deepened the interconnections
between nations, and driven social transformation (Mur-
shed etal. 2022; Ojekemi etal. 2022; Wang etal. 2023e). In
environmental terms, globalization has been instrumental in
improving the environment through heightened environmen-
tal awareness, the introduction of advanced technologies,
and the upgrading of industrial structures (Shi etal. 2023;
Cutcu etal. 2023; Sun etal. 2023). However, it has also
intensified resource consumption, leading to environmental
degradation (Kirikkaleli etal. 2021). Therefore, its impact
on the environment is complex and dual-faceted (Ahmed
etal. 2019). In such a context, the role of AI becomes par-
ticularly significant. As a disruptive innovative technology,
AI contributes to environmental improvement and the reduc-
tion of ecological footprints by optimizing resource alloca-
tion, increasing production efficiency, and promoting green
technology innovation (Li etal. 2021). With the advance
of globalization, the dissemination and application of AI
technologies are becoming more widespread, prompting
governments worldwide to increase their support for this
domain in the hope of securing a dominant position in the
technological revolution, which may further amplify AI’s
impact on ecological footprints (Wang etal. 2023d). It is
imperative to stress that a comprehensive consideration of
the interplay between economic, social, and technological
factors will enable policymakers to understand and address
this complex issue with a more holistic vision, achieving
sustainable development.
In this context, this study attempts to explore the fol-
lowing questions: Does AI have an impact on the ecologi-
cal footprint? Under the influence of globalization, does
AI exhibit a nonlinear impact on the ecological footprint?
To address these two questions, we used a sample of 60
Fig. 1 Trend chart of global
ecological footprint and bioca-
pacity from 1961 to 2021. Data
comes from: (https:// data. footp
rintn etwork. org/#/ compa reCou
ntries? cn= all& type= EFCpc &
yr= 2019)
123950 Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
countries globally from 2010 to 2019. Firstly, we applied the
entropy method to measure the comprehensive evaluation
index of AI. Secondly, we employed fixed-effects models
to determine the direction of AI’s impact on the ecologi-
cal footprint. Finally, we further examined the nonlinear
impact of AI on the ecological footprint under the influ-
ence of globalization by using a panel threshold model. The
research findings contribute to a deeper understanding of the
impact of AI on the ecological footprint, which holds sig-
nificant implications for accelerating digital transformation
and achieving global sustainability. Additionally, the major
global economies offer new insights and recommendations
for addressing environmental issues in other economies.
These studies can assist other nations in harnessing AI for
automation, digitization, and intelligence, thereby easing
resource and environmental constraints and achieving sus-
tainable development.
This article is structured into 5 sections: the “Literature
review” section presents a literature review; the “Methods
and data” section introduces the research methods and data;
the “Empirical results and discussion” section comprises the
empirical analysis; and the final section offers conclusions
and recommendations.
Literature review
AI andecological footprint
The rapid advancement of AI has attracted widespread
attention from the academic community, and its socioeco-
nomic impacts have become a focal point of research (Zador
etal. 2023; Fosso Wamba 2022), especially in areas such
as the labor market (Damioli etal. 2021; Acemoglu and
Restrepo 2020), corporate innovation (Wang etal. 2023b),
and economic growth (Graetz and Michaels 2018; Aghion
etal. 2017). As related research deepens, some scholars
have noticed the potential environmental implications of AI
(Kar etal. 2022; Al-Sharafi etal. 2023; John etal. 2022).
Vinuesa etal. (2020) delved into the impacts of AI on the
17 goals and 169 specific targets outlined in the UN’s “2030
Agenda for Sustainable Development,” revealing that AI can
contribute to the achievement of most of these goals. Dhar
(2020) discussed the dual role of AI in the carbon reduction
process, highlighting that it is both a tool for addressing
climate change and a significant emitter of carbon. Addition-
ally, regional studies have shed light on the diverse envi-
ronmental impacts of AI. For instance, Ding etal. (2023)
examined the influence of AI on carbon emissions across
30 provinces in China from 2006 to 2019, finding that AI
significantly reduced local carbon emissions and also had
spillover effects on adjacent areas. Employing an interactive
three-stage network DEA model, Liang etal. (2022) utilized
data from China spanning 2016 to 2019 to study the perfor-
mance of AI in carbon reduction within the manufacturing
sector. The results indicated that China’s performance in this
area is still lacking, with substantial room for improvement.
Liu etal. (2022) explored the impact of AI on carbon emis-
sion intensity, discovering heterogeneity in effects across
different stages of development and industrial phases. Fur-
thermore, AI has demonstrated considerable potential in the
fields of renewable energy and environmental monitoring
(Ahmad etal. 2021; Chen etal. 2021; Ye etal. 2020).
Nevertheless, existing literature has devoted only limited
attention to comprehending the impact of AI on the ecologi-
cal footprint, with only a handful of scholars delving into the
effects of digital technologies on this crucial environmen-
tal metric. For instance, Huang etal. (2022) scrutinized the
influence of Information and Communication Technology
(ICT) on the ecological footprint across two distinct cat-
egories of nations, developed and developing, from 1995 to
2018. Their findings revealed that over the long term, ICT
technologies significantly reduced the ecological footprint in
both groups of countries. Taking Saudi Arabia as an exam-
ple, Kahouli etal. (2022) investigated the impact of ICT on
the ecological footprint and found that in the short term,
ICT significantly mitigated the ecological footprint. Albaity
and Awad (2023) studied the influence of ICT technology
on the ecological footprint of high, upper-middle, and low-
income economies from 2000 to 2020. They observed that in
high-income and upper-middle-income countries, ICT tech-
nology had a significant restraining effect on the ecological
footprint of countries with poor environmental quality. Abid
etal. (2023), adopting a consumption-based perspective on
ecological footprints, further corroborated the negative rela-
tionship between ICT technology and ecological footprints.
Karlilar etal. (2023) posited that digital transformation con-
tributes significantly to environmental sustainability, par-
ticularly when coupled with effective environmental poli-
cies, thereby exerting a stronger impact on environmental
quality. Zhang etal. (2022b) discovered that digital trade
facilitates technology spillovers and knowledge diffusion,
driving green innovation, which contributes to a reduction
in the ecological footprint. Given that AI serves as a critical
vehicle for intelligent transformation, its judicious applica-
tion becomes pivotal in curtailing ecological footprints (Ni
etal. 2022). In a related vein, Chen etal. (2022) regarded
industrial robots as a primary manifestation of AI and dis-
cerned that industrial robots impact the ecological footprint
through a complex interplay of four effects: time-saving,
green employment, industrial growth, and energy upgrading.
AI andglobalization
AI’s rapid development has become a significant force in
globalization (Borges etal. 2021). Scholars have delved
123951Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
deeply into the complex relationship between AI and glo-
balization. Firstly, in terms of global economic develop-
ment, Korinek and Stiglitz (2021) explored the relationship
between AI, globalization, and economic development strat-
egies, emphasizing the potential of AI to improve productiv-
ity, create job opportunities, and promote innovation. Garg
etal. (2022) noted that AI has led to the flourishing of digital
trade, which is quickly reshaping the pattern of global trade.
Benabed etal. (2022) discussed the role of AI in helping
small and medium-sized enterprises (SMEs) to cope with
internationalization and globalization processes.
Secondly, regarding social transformation, Hilbert (2020)
indicated in his research that, in the era of digital globali-
zation, AI is becoming a significant force driving social
change. Puaschunder (2019) similarly emphasized the dis-
ruptive role of AI in the market, profoundly affecting tra-
ditional industries and business models. Haluza and Jun-
gwirth (2023) conducted exploratory research using GPT-3
to investigate the relationship between AI and major social
trends.
Finally, in the political realm, Dauvergne (2021) included
AI in the analytical framework of global environmentalist
politics, highlighting the impact of AI on international polit-
ical relations. The rapid development of AI requires govern-
ments to update and formulate relevant policies and regula-
tions to address the global ethical and legal challenges it
brings, such as data privacy, data security, and ethical issues
related to AI (Safdar etal. 2020; Zhang etal. 2021). Addi-
tionally, the global digital divide remains an issue (Lythrea-
tis etal. 2022), and with the ongoing proliferation of AI, the
digital divide between countries is widening, exacerbating
issues of inequality (Lutz 2019).
Globalization andecological footprint
Regarding the relationship between globalization and the
ecological footprint, the academic community has conducted
extensive discussions, summarized into three main hypoth-
eses: the Pollution Haven Hypothesis, the Pollution Halo
Hypothesis, and the Neutral Hypothesis. First, the Pollution
Halo Hypothesis posits that globalization has a significantly
negative impact on the ecological footprint (Jahanger etal.
2022; Balsalobre-Lorente etal. 2019; Li etal. 2022). Glo-
balization accelerates the flow of information, knowledge,
and technology through spillover effects and enhances envi-
ronmental awareness and governance through demonstration
effects (Cutcu etal. 2023; Zhou etal. 2023). Hassan etal.
(2023), using data from OECD countries from 1990 to 2019,
investigated the impact of globalization on the ecological
footprint. Their findings showed that globalization signifi-
cantly reduced the ecological footprint and improved envi-
ronmental quality. Similarly, Alvarado etal. (2022), employ-
ing the AMG panel model and CCE-MG model, found a
negative correlation between globalization and the ecologi-
cal footprint. Saud etal. (2020) discovered that globalization
significantly alleviates the ecological footprint of countries
participating in the “Belt and Road” initiative. Based on data
from BRICS countries, Sun etal. (2023) found that globali-
zation has nonlinear effects on the ecological footprint of
different nations. Based on the STIRPAT model, Yang etal.
(2021) found a unidirectional causal relationship between
globalization and the ecological footprint, with globaliza-
tion contributing to long-term reductions in the ecologi-
cal footprint. Jahanger etal. (2022) used sample data from
Asia, Africa, Latin America, and the Caribbean, indicating
that globalization has a significant inhibitory effect on the
ecological footprint of developing countries in Africa and
Asia. Employing Japan as a case study, Ahmed etal. (2021)
employed an asymmetric ARDL approach to investigate
the long-term relationship between economic globalization
and the ecological footprint. The results demonstrated that
globalization has a negative impact on Japan’s ecological
footprint through the transfer of clean technology.
Secondly, the Pollution Haven Hypothesis suggests that
globalization can enlarge the ecological footprint of certain
countries (Solarin etal. 2017). Developed nations, for the
purpose of industrial transformation, shift highly polluting
and energy-intensive industries to regions with lower envi-
ronmental regulatory requirements, leading to an increase in
the ecological footprint in these areas (Sultana etal. 2023).
Based on data from BRICS countries spanning from 1971
to 2016, Pata (2021) found that globalization significantly
stimulates the ecological footprint, which is detrimental to
improving ecological well-being. Khan etal. (2019), using
data from Pakistan, similarly discovered a positive correla-
tion between globalization and carbon emissions. Rudolph
and Figge (2017) pointed out that for every 1% increase in
economic globalization, the consumption ecological foot-
print increases by 0.07%, further confirming the detrimental
impact of economic globalization on environmental quality.
Langnel and Amegavi (2020), employing the ARDL method
to study the impact of globalization on the ecological foot-
print in Ghana, concluded that comprehensive globalization
significantly elevates Ghana’s ecological footprint. Eco-
nomic globalization and social globalization have positive
effects on the ecological footprint, while political globaliza-
tion has a negative impact. Finally, a minority of scholars
support the Neutral Hypothesis, suggesting that the impact
of globalization on the environment can be disregarded.
Using data from 130 countries over the period from 1980 to
2016, Apaydin etal. (2021) corroborated this perspective.
In summary, existing research has discussed the potential
impact of AI and globalization on the ecological footprint.
While some scholars have noticed the impact of digital tech-
nology on the ecological footprint and studied the potential
impact of AI on the ecological environment, discussions on
123952 Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
how AI affects the ecological footprint are still relatively
limited. Currently, most scholars have relied on ordinary
panel models to measure the direct impact of AI on envi-
ronmental quality, often overlooking the nonlinear relation-
ship between the two variables, which can potentially dis-
tort estimation results. Furthermore, given that authoritative
measures in the field of AI are still lacking, existing litera-
ture commonly uses proxy variables such as the number of
industrial robots, the number of published papers related to
AI, and the number of patents related to AI for analysis (Liu
etal. 2022, 2020; Shen and Zhang 2023). These singular
indicators can hardly reflect the comprehensive development
level of a country in the field of AI comprehensively.
In comparison to existing literature, the marginal con-
tributions of this study are as follows: Firstly, by using the
entropy method, this paper has constructed a comprehensive
AI indicator system that encompasses four crucial dimen-
sions: AI-related technology, network infrastructure, gov-
ernment support, and the international competitiveness of
digital products. This not only overcomes the limitations of
relying on a single indicator found in previous studies but
also offers a more realistic and objective method of evalua-
tion. Secondly, we integrate AI into the analytical framework
of the global ecological footprint, delving into AI’s impact
on the environment. Through empirical testing, we reveal
not only the direct impact of AI on the ecological footprint
but also provide new empirical evidence for the role of AI in
promoting global sustainable development. Lastly, by apply-
ing a panel threshold model, we comprehensively examine
how AI and globalization together affect the environment
and further confirm that the impact of AI on the ecological
footprint changes as the globalization process deepens. This
provides valuable practical experience and theoretical sup-
port for countries to advance the development of AI technol-
ogy against the backdrop of globalization.
Methods anddata
Econometric methods
First, before performing regressions, we conducted a bat-
tery of tests on all variables, encompassing unit root tests
and panel cointegration tests. Subsequently, we employed a
fixed-effects model to examine the direct impact of AI on the
ecological footprint. Finally, employing globalization as a
threshold, we delved into the nonlinear impact of AI on the
ecological footprint.
Variable test
We employed unit root tests to verify the stability of the
variables chosen in this study, including the LLC test (Levin
etal. 2002), the IPS test (Im etal. 2003), the Fisher-ADF test
(Maddala and Wu 1999), and the Fisher-PP test (Phillips and
Perron 1988). If the null hypothesis is rejected, it signifies
that the variables have no unit root. Conversely, it means that
the selected variable has a unit root.
After confirming that the selected variables were station-
ary, we employed the Pedroni test (Pedroni 2001) and the
Kao cointegration test (Kao 1999) to test whether there is a
long-term relationship between variables. If the null hypoth-
esis is rejected, it indicates the presence of a long-term coin-
tegrating relationship among the variables. Otherwise, there
is no cointegration relationship. The relevant formulas are
in Appendix B.
Baseline model
The purpose of this study is to analyze the impact of AI
on ecological footprint. We take AI as the input variable,
ecological footprint as the output variable, and urbaniza-
tion, foreign direct investment, renewable energy transition,
and economic development level as control variables. The
functional expression is as follows:
In this paper, the ordinary least square (OLS) model and
fixed-effect (FE) model are respectively constructed, which
are shown in formulas (2)-(3):
Among them, i represents the country;
t
represents the
year; EF is the explained variable, representing the eco-
logical footprint of each country; AI is the core explana-
tory variable, representing the level of AI development;
∑n
j=1
X
jit
is the control variable, including urbanization,
foreign direct investment, renewable energy transforma-
tion, and economic development level;
𝛼0
,…,
𝜑j
are coef-
ficients to be estimated;
𝜇i
represents individual fixed
effects,
𝛿t
represents time-fixed effects, and
𝜀it
represents
random error items.
Panel threshold model
This paper employs a panel threshold model to further inves-
tigate the nonlinear relationship between AI and ecological
footprint. The initial equation of the panel threshold model
proposed by Hansen (1999) is as follows:
(1)
EF =f(AI
,
UR
,
FDI
,
RE
,
ED)
(2)
EF
it =𝛼0+𝛼1AIit +
∑n
j=
1𝜑jXjit +𝜀
it
(3)
EF
it =𝛼0+𝛼1AIit +
∑n
j=
1𝜑jXjit +𝛿t+𝜇i+𝜀
it
(4)
yit =ui+𝜀it,qit +𝜃1xit ≤𝛾
123953Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
where
yit
is the explained variable,
xit
is the explanatory
variable,
qit
is the threshold variable, and
𝜀it
is the random
error term.
On this basis, we constructed a panel threshold model to
study the nonlinear relationship between AI and ecological
footprint, as shown in the formula (6).
Among them, I(·)is an indicator function, which takes a
value of 1 or 0;
qit
represents the threshold variable corre-
sponding to globalization; and
𝛾
denotes the threshold value.
The estimated value
𝛽i(𝛾)
and the sum of squared residu-
als
S1(𝛾)
in formula (6) can be obtained by using the least
square method. Finally, sort the value of
S1(𝛾)
in order and
select the smallest
𝛾
value as the estimated threshold.
Before conducting the threshold model regression, it
needs to be tested, which is divided into the following two
steps:
(1) Threshold effect test. After determining the threshold
value, the sample is divided into two parts, with estimated
coefficients
𝛽1
and
𝛽2
for each part. The null hypothesis for
the threshold effect test is
𝛽1=𝛽2
, indicating no threshold
effect. It is tested using the LR statistic, which is con-
structed as formula (7):
where
S0
is the residual sum of squares when there is no
threshold effect;
S1(𝛾)
is the residual sum of squares when
there is a threshold effect, and
𝜎2
is the variance of the
residual when there is a threshold effect.
If the model does not have the threshold effect, then
the threshold value cannot be identified, and the test
no longer adheres to a standard distribution. Therefore,
following the Bootstrap method mentioned by Hansen
(1999), we can obtain the value of LR statistics. For a
given significance level
𝛼
, we can determine the critical
value of the LR test by searching for the 1-
𝛼
quantile. As
long as the LR value exceeds this critical value, it signi-
fies the presence of a significant threshold effect.
(2) Threshold value test. The null hypothesis for this test
is
𝛾=𝛾
, which indicates that the threshold value aligns
with the actual value; and the alternative hypothesis is
𝛾≠𝛾
. The corresponding test expression is as follows:
(5)
yit =ui+𝜀it,qit +𝜃2xit >𝛾
(6)
EF
it =𝛼0+𝛽1AIit ×I
(
qit ≤𝛾
)
+𝛽2AIit ×I
(
qit >𝛾
)
+∑n
j
=1𝜑jXjit +𝛿t+𝜇i+𝜀it
(7)
LR
=
S
0
−S
1
(𝛾)
𝜎
2
(8)
LR
1(𝛾)=
S1(𝛾)−S1
(
𝛾
)
𝜎
2
The critical value level h(α) can be determined using the
formula
h(
𝛼
)=−
2
log(
1
−√
1
−
𝛼
)
. If the value of LR
exceeds
h(𝛼)
, the null hypothesis is rejected, indicating
that the selection of the threshold value is incorrect. If the
value of LR does not exceed
h(𝛼)
, the null hypothesis is
not rejected, indicating that the threshold choice is correct.
Artificial intelligence index
Our research focuses on a critical question: how to accu-
rately measure a country’s level of AI development. Existing
studies often use relatively simple single indicators, such as
industrial robot intensity, AI-related patents, and regional
policy variables (Liu etal. 2022, 2020; Shen and Zhang
2023). However, this approach is relatively narrow and can-
not comprehensively assess the multi-dimensional develop-
ment of AI. In contrast, we believe that the development of
AI is a multifaceted and comprehensive concept, and relying
solely on a single indicator can be challenging for accurate
measurement. Therefore, we have drawn on the research
methods of Li etal. (2023b) and Ding etal. (2023), and on
this basis, we have established a comprehensive indicator
system that reflects the global level of AI. Li etal. (2023b)
used principal component analysis to construct an ICT index
from three aspects: ICT infrastructure construction, insti-
tutional guarantees, and the competitiveness of ICT prod-
ucts and services. Ding etal. (2023) measured the level of
China’s AI development from multiple perspectives of the
innovation support environment, vitality and efficiency, and
output. However, they did not comprehensively consider the
important roles of AI technology and network infrastructure.
Taking into account data availability and research neces-
sity, we measure the Artificial Intelligence Index based on
four aspects: AI-related technology, network infrastructure,
international competitiveness of digital products, and gov-
ernment institutional support (as shown in Table1). Relevant
data is sourced from the International Federation of Robot-
ics (IFR), World Intellectual Property Organization (WIPO),
World Development Indicators (WDI), United Nations Con-
ference on Trade and Development (UNCTAD), SCIMAGO
Journal and Country Ranking Database, World Bank-Global
Doing Business Huaning Report, World Economic Forum,
and United Nations E-Government Questionnaires, with
missing data supplemented using interpolation methods.
Regarding the evaluation methods of AI, the existing
research has primarily fallen into two types: subjective
weighting method and objective weighting method. Com-
pared with other methods, the entropy method can not only
make full use of the information contained in the index, but
also avoid the issue of information redundancy. Thus, we
use the entropy method to assess the development of AI.
The calculation steps of the entropy method are as follows:
123954 Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
(1) Data standardization processing
(2) Calculate the weight of the index value of the jth
index of the ith country in the year t
(3) Calculate the information entropy of the jth index
(4) Calculate the difference coefficient of the jth index
(5) Calculate the weight of the jth index
(6) Calculation of artificial intelligence development index
Based on the above indicators and evaluation methods,
we calculated the AI development index of 60 countries
from 2010 to 2019. To intuitively depict the dynamic evo-
lution of AI, we plotted the annual average and spatial dis-
tribution of AI in all sample countries (see Figs.2 and 3).
(9)
x
�
ijt =
x
ijt
−𝑚𝑖𝑛(x
jt
)
𝑚𝑎𝑥(x
jt
)−𝑚𝑖𝑛(x
jt)
(10)
pijt =
x
�
ijt
m
i=
1x�
ijt
0≤pij ≤1
(11)
E
jt =−
1
𝑙𝑛n
×
∑
n
i=1pijt ×𝑙𝑛p
ijt
(12)
gjt =1−Ejt
(13)
w
jt =
g
jt
∑
n
j=
1gjt
(14)
S
it =
∑n
j=
1wjt ×x�
ijt
Our findings reveal that, from the perspective of time, the AI
index exhibits an overall upward trend from 2010 to 2019.
The global average value of the AI index rose from 0.088 in
2010 to 0.140 in 2019, with an average annual growth rate of
5.37%. The average AI value of both developing and devel-
oped countries has increased steadily, with the AI growth
rate of developing countries showing a fluctuating trend.
This could be attributed to developed countries rapidly seiz-
ing the AI market and implementing technology lock-ins in
recent years. From a spatial perspective, there is a significant
spatial variation in the development of AI, and it exhibits
significant disparities.
Data
This study examines the impact of AI on the ecological
footprint using national-level data from 2010 to 2019.
Given data availability for empirical analysis, the study
has selected a sample of 60 countries, which are listed
in Table13 of Appendix A. The explained variable in
this analysis is the ecological footprint (EF), measured in
global hectares (gha) per capita in the Global Footprint
Network (GNF) database. The core explanatory variable
is artificial intelligence (AI), which is calculated by the
entropy method above. Control variables employed in the
analysis include urbanization (UR), foreign direct invest-
ment (FDI), renewable energy transition (RE), and eco-
nomic development level (ED). Urbanization is measured
as the ratio of urban population to the total population. The
economic development level is represented by per capita
GDP at constant prices in 2015. Foreign direct investment
is expressed as the ratio of net inflows of foreign direct
investment to GDP. Renewable energy consumption is
Table 1 Index selection of artificial intelligence
Elemental indicators Basic indicators Source
AI-related technology Industrial robot installation IFR
Artificial intelligence patent WIPO
Frontier Technology Readiness Index UNCTAD
AI Papers SCIMAGO Journal and Country Ranking Database
Network infrastructure Fixed broadband per 100 people WDI
Fixed telephone number per 100 people WDI
Percentage of individuals using the Internet WDI
Secure web servers per 10,000 people WDI
International competitiveness of digital
products HHigh-tech export WDI
Proportion of ICT goods exports WDI
Proportion of ICT service exports WDI
Government institutional support Business Environment Index World Bank-Doing Business Huaning Report
Intellectual Property Protection World Economic Forum
E-participation United Nations e-government survey
online service United Nations e-government survey
123955Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
expressed by the proportion of renewable energy in total
final energy consumption. The threshold variable is glo-
balization (GL), measured using the globalization index
from the KOF database (Gygli etal. 2019). Table2 lists
the definitions and sources of the variables.
To mitigate the impact of heteroscedasticity on the regres-
sion results in this study, we logarithmize some variables.
Table3 provides descriptive statistics for the variables.
Table4 presents the VIF values and correlation coefficients
for the variables in this study. It can be observed that the
Fig. 2 Annual average of AI in
all samples from 2010 to 2019
Fig. 3 AI index distribution
map of the full sample in 2010
and 2019. Note: For the Peo-
ple’s Republic of China, Hong
Kong, Macau, and Taiwan are
not included in the sample.
Therefore, these regions are not
colored in the figure
123956 Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
mean values of the variables are all positive, with RE having
the highest mean and standard deviation, while AI has the
lowest mean and standard deviation. This indicates that the
data selected have sufficient variability for linear analysis.
Additionally, the VIF of each variable is below 5, and the
correlation coefficients among variables are all below 0.8,
suggesting that there is no serious multicollinearity and cor-
relation among variables.
Empirical results anddiscussion
Sample test results
To avoid spurious regression, we first conducted unit root
tests on all variables to assess whether the data were station-
ary. The test results are presented in Table5. It is evident
that EF, AI, UR, FDI, RE, ED, and GL are non-stationary.
However, the first differences of variables become station-
ary. This suggests that the variables used in this study are
stationary processes.
After determining that the first-order difference of the
selected variables is stationary, we used Pedroni and Kao
tests to verify whether there exists a long-term equilibrium
relationship among the variables. The test results in Table6
show that there is a long-term equilibrium relationship
among the variables we selected, satisfying the prerequisites
for subsequent empirical analysis.
Table 2 Definition of main variables
variable Variable symbol Variable definitions Source
Ecological footprint EF Per capita ecological footprint: an assessment of the impact of human activities on the
ecosystem. It can be used to compare the demand of humans on resources with the
capacity of the earth to provide renewable resources
GFN
Artificial intelligence AI The comprehensive evaluation index of AI, calculated by the entropy method based on
the dimensions of AI-related technology, network infrastructure, international com-
petitiveness of digital products, and government institutional support
-
Urbanization UR Urban population/total population. Urban population refers to the population residing
in urban areas as defined by the National Bureau of Statistics WDI
Foreign direct investment FDI Foreign direct investment, net inflows (% of GDP). Foreign direct investment refers to
the acquisition of a lasting management interest of 10% or more in enterprises oper-
ating within an economy outside of that of the investor
WDI
Renewable energy transition RE Renewable energy consumption/total energy consumption. Renewable energy typically
refers to various sources of energy that can be sustainably regenerated from natural
processes (mainly including solar energy, wind energy, hydro energy, biomass
energy, etc.)
WDI
Economic development level ED GDP per capita (2015 constant price). The data are calculated in constant 2015 U.S.
dollars WDI
Globalization GL The KOF Globalization Index is an indicator used to measure the degree of globaliza-
tion, including economic, social, and political globalization. It was proposed by the
KOF Research Center of the University of Zurich in Switzerland
KOF
Table 3 Descriptive statistics of the main variables
Variable Obs Mean Std. Dev Min Max
EF 600 1.675 0.401 0.534 2.774
AI 600 0.107 0.075 0.014 0.665
UR 600 4.280 0.245 3.464 4.615
FDI 600 4.060 9.686 − 40.087 102.314
RE 600 17.985 13.766 0.000 62.370
ED 600 9.692 1.040 7.042 11.375
GL 600 4.345 0.120 3.990 4.523
Table 4 VIF and correlation
coefficient table of variables EF AI UR FDI RE ED FG VIF
EF 1.000 -
AI 0.311 1.000 1.29
UR 0.675 0.230 1.000 2.25
FDI 0.076 0.095 0.129 1.000 1.04
RE − 0.243 − 0.173 − 0.281 − 0.154 1.000 1.17
ED 0.848 0.432 0.695 0.121 − 0.099 1.000 4.35
GL 0.632 0.361 0.437 0.131 − 0.002 0.769 1.000 2.59
123957Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
Benchmark regression result analysis
Table7 reports the baseline regression results for the impact
of AI on the ecological footprint. Among them, column (1)
presents the results of the OLS model, while columns (2)
to (4) present the results of the fixed-effects models. In
column (4), the regression coefficient of AI is − 0.265, sig-
nificant at the 1% level, indicating a negative correlation
between AI and ecological footprint. This finding suggests
that AI can significantly mitigate the ecological footprint.
According to the statistics, for every 1% increase in AI, the
ecological footprint decreases by 0.265%. This aligns with
the research conducted by Chen etal. (2022) and Mor etal.
(2021). Specifically, utilizing data from 72 countries over
the period of 1993–2019, Yang Chen etal. (2022) found
that the application of industrial robots across various indus-
tries can enhance production efficiency and resource utiliza-
tion, thereby reducing the ecological footprint. Focusing on
the agricultural sector in India, Mor etal. (2021) explored
the important roles of AI in agriculture. They suggest that
AI can be widely applied in scenarios such as crop variety
selection, extreme weather prediction, and smart agricultural
advisory, to reduce the consumption of human resources,
improve production efficiency, and thus lower the carbon
footprint of agriculture.
The reasons behind this finding may be due to its out-
standing environmental characteristics; AI has become an
effective solution to the difficulties of resource allocation
encountered in the processes of urbanization and industriali-
zation (Qiang Wang etal. 2023c). Advanced AI technologies
such as industrial robots, machine learning, deep learning,
and image recognition have demonstrated their enormous
potential in enhancing the efficiency of production and
transactions across various industries. These technologies
directly alleviate the pressure of human activities on natural
resources by increasing the utilization rate of raw materials
Table 5 Panel unit root test
*** Denote that the test levels of 1% reject the null hypothesis
Variable Test method At level At 1st difference
t-Statistic Prob t-Statistic Prob
EF LLC − 8.587*** 0.000 − 23.611*** 0.000
IPS − 2.821*** 0.002 − 12.531*** 0.000
Fisher-ADF 172.978*** 0.001 408.637*** 0.000
Fisher-PP 215.141*** 0.000 533.538*** 0.000
AI LLC 6.055 1.000 − 17.876*** 0.000
IPS 12.823 1.000 − 4.243*** 0.000
Fisher-ADF 51.216 1.000 195.106*** 0.000
Fisher-PP 75.301 1.000 207.096*** 0.000
UR LLC − 31.850*** 0.000 − 6.934*** 0.000
IPS − 35.805*** 0.000 − 11.086*** 0.000
Fisher-ADF 359.841*** 0.000 261.987*** 0.000
Fisher-PP 401.535*** 0.000 304.219*** 0.000
FDI LLC − 17.434*** 0.000 − 27.087*** 0.000
IPS − 9.356*** 0.000 − 15.160*** 0.000
Fisher-ADF 318.212*** 0.000 470.512*** 0.000
Fisher-PP 337.101*** 0.000 701.545*** 0.000
RE LLC − 10.914*** 0.000 − 16.305*** 0.000
IPS − 1.549* 0.061 − 7.820*** 0.000
Fisher-ADF 172.410*** 0.001 297.913*** 0.000
Fisher-PP 145.466** 0.044 342.413*** 0.000
ED LLC − 4.079*** 0.000 − 17.752*** 0.000
IPS 4.753 1.000 − 8.818*** 0.000
Fisher-ADF 115.292 0.604 318.422*** 0.000
Fisher-PP 174.575*** 0.001 345.959*** 0.000
FG LLC − 13.852*** 0.000 − 16.964*** 0.000
IPS − 3.843*** 0.000 − 11.952*** 0.000
Fisher-ADF 193.476*** 0.000 388.465*** 0.000
Fisher-PP 219.932*** 0.000 495.581*** 0.000
Table 6 Results of Pedroni and
Kao panel cointegration tests Pedroni panel cointegration test
Group Statistic Prob Weighted Statistic Prob
Within-dimension
Panel v-Statistic − 7.414 1.000 − 8.944 1.000
Panel rho-Statistic 8.307 1.000 8.731 1.000
Panel PP-Statistic − 27.745*** 0.000 − 33.922*** 0.000
Panel ADF-Statistic − 11.363*** 0.000 − 10.553*** 0.000
Between-dimension
Group rho-Statistic 11.840 1.000 - -
Group PP-Statistic − 46.578*** 0.000 - -
Group ADF-Statistic − 13.527*** 0.000 - -
Kao residual cointegration test
t-Statistic Prob
ADF − 5.537*** 0.000
123958 Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
and reducing energy consumption in the production pro-
cess, which ultimately leads to a reduction in the ecological
footprint.
Robustness test
Endogeneity test: IV estimation
When analyzing the impact of AI on the ecological foot-
print, problems like reverse causality, model selection, and
omitted variables could potentially introduce endogeneity,
leading to biased regression results in the abovementioned
baseline model. Thus, we employ the instrumental variables
(IV) method to address endogeneity concerns.
It should be noted that in many previous studies, his-
torical information has been used to address endogeneity
(Au and Henderson 2006). Wang etal. (2023a) in their
study on the development of AI selected the number of
micro-electronic computers produced per 10,000 people
in 1993 as the instrumental variable for the develop-
ment of AI in China. Drawing on the approach of exist-
ing research, we use the installation volume of indus-
trial robots in 1993 to construct an instrumental variable
for the national level of AI development. In 1993, the
installation of industrial robots represented a significant
technological advancement in the field of automation.
Regions with higher rates of industrial robot installation
often experienced more effective knowledge spillover and
technology transfer, leading to higher levels of automa-
tion. Since the development of AI relies on cutting-edge
technologies, the adoption of advanced industrial robots
often indicates technological innovation. The installation
of industrial robots becomes a potential driver for the
development of AI (Li etal. 2023a). Therefore, select-
ing industrial robots as the instrumental variable for the
AI development index meets the relevance requirement.
Moreover, compared to the speed of development of AI
technology, the historical number of industrial robot
installations has a diminishing impact on the current
ecological footprint, satisfying the requirement of exo-
geneity. Since the level of AI development of a coun-
try is a time-variant variable, and historical variables do
not change over time, we use the interaction between the
number of industrial robot installations in 1993 and the
lagged terms of AI as the instrumental variable.
Table8 reports the results of the IV method. As shown
in column (1), the coefficient of
IV
is significantly posi-
tive, indicating a significant positive correlation between
the IV and AI. Additionally, the p-value of the overiden-
tifying restrictions test is 0.000, and the Cragg-Donald
Wald F-statistic is greater than the critical value at the
10% level of the Stock-Yogo weak identification test,
which indicates that the selection of IV is effective.
Based on the second-stage regression results reported in
column (2), it is found that after addressing the endogeneity
issue through the IV method, the coefficient of AI remains
significantly positive. This result indicates that the develop-
ment of AI significantly reduces the ecological footprint of
the sample countries.
Other robustness tests
To further examine the robustness of the impact of AI
on the ecological footprint, four methods were employed
and the results are presented in Table9. (1) Change in
measurement of explanatory variable: In column (1), we
calculated the AI assessment index using principal com-
ponent analysis and regressed it as the core explanatory
variable. (2) Adjustment of sample period: In column (2),
the sample period was adjusted to 2011–2018, and the
regression was conducted again. (3) One-period lag of
explained variable: In column (3), the explanatory vari-
ables will be lagged for one period, and then, the regres-
sion will be performed again. The estimated results in
columns (1)–(3) are basically consistent with the results
of the baseline regression, indicating that the conclusion
that AI can significantly reduce the ecological footprint
is robust.
Table 7 Benchmark regression
results of AI on ecological
footprint
Variables (1) (2) (3) (4)
AI − 0.440*** (− 3.609) − 0.368*** (− 2.915) − 0.324*** (− 3.565) − 0.265*** (− 3.154)
UR 0.154*** (3.149) 0.163*** (3.325) − 0.307* (− 1.794) 0.429** (2.561)
FDI − 0.002** (− 2.429) − 0.002** (− 2.584) 0.001* (1.891) 0.000* (1.800)
RE − 0.005*** (− 7.281) − 0.004*** (− 6.946) − 0.005*** (− 4.719) − 0.001 (− 0.911)
ED 0.312*** (25.987) 0.310*** (25.659) 0.274*** (7.731) 0.414*** (11.890)
Constant − 1.869*** (− 11.944) − 1.856*** (− 11.742) 0.455 (0.683) − 4.072*** (− 5.592)
Year FE No Yes No Yes
Nation FE No No Yes Yes
N 600 600 600 600
R2 0.758 0.761 0.132 0.309
123959Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
Nonlinear analysis
In order to further investigate the nonlinear impact of AI on
the ecological footprint, we employed the panel threshold
model to analyze whether the effect of AI on the ecological
footprint varies under the influence of globalization.
Firstly, we tested whether there is a threshold effect and
determined the number of thresholds. We conducted 300
bootstrapped tests and obtained specific F-values and P-val-
ues. The results are presented in Tables10 and 11. It appears
that globalization passed the single-threshold test at a 10%
significance level, while the F-statistic for a double threshold
was not significant. From these results, we conclude that
under the influence of globalization, there is a single-thresh-
old effect of AI on the ecological footprint. In the subsequent
analysis, we employed the single-threshold model.
Next, we conducted a threshold value test. Figure4 shows
the likelihood ratio function plot for the threshold value. It
can be seen that the threshold value is below the dashed line
(with a value of 7.35). Therefore, in the nonlinear relation-
ship between AI and ecological footprint, the choice of the
threshold value of globalization is appropriate.
Finally, based on the threshold value obtained above,
we constructed a single-threshold model with ecological
footprint as the explained variable, AI as the explanatory
variable, and globalization as the threshold variable. The
regression results are displayed in Table12. With other vari-
ables held constant, there are variations in the impact of AI
on the ecological footprint when globalization falls within
different threshold ranges.
As globalization falls within the first threshold range, the
regression coefficient of AI is − 0.180. When globalization
falls within the second threshold range, the regression coeffi-
cient of AI is − 0.703. It is evident that as globalization inten-
sifies, the inhibitory effect of AI on the ecological footprint
becomes more pronounced. This is consistent with the study
by Bibi etal. (2023) and Wang and Zhang (2021). Based
on data from China from 1987 to 2020, Bibi etal. (2023)
found that the combined effects of ICT and globalization help
reduce carbon emissions and ecological footprint by improv-
ing energy utilization efficiency and resource use rates.
Table 8 IV test regression results
Variables (1) (2)
AI − 0.427** (− 2.466)
IV 0.178*** (11.053)
UR 0.929*** (10.532) 0.562*** (2.711)
FDI − 0.000 (− 0.226) 0.000 (0.861)
RE 0.001** (2.324) − 0.001 (− 0.632)
ED 0.086*** (4.763) 0.405*** (10.808)
Constant − 4.990*** (− 12.763) − 4.884*** (− 4.860)
Year FE Yes Yes
Nation FE Yes Yes
Overidentifying
restrictions test 111.976***
Cragg-Donald test 122.173
[16.38]
N 540 540
R20.934 0.989
Table 9 Robustness test
regression results Variables (1) (2) (3)
AI − 0.026*** (− 3.580) − 0.279*** (− 2.852) − 0.215** (− 2.325)
UR 0.507*** (2.938) 0.477** (2.493) 0.379* (1.853)
FDI 0.000 (1.634) 0.001** (2.276) 0.000 (1.393)
RE − 0.001 (− 1.042) − 0.001 (− 0.582) − 0.002** (− 2.014)
ED 0.422*** (12.044) 0.425*** (10.626) 0.396*** (9.578)
Constant − 4.451*** (− 5.875) − 4.393*** (− 5.255) − 3.678*** (− 4.179)
Year FE Yes Yes Yes
Nation FE Yes Yes Yes
N 600 480 540
R20.313 0.331 0.262
Table 10 Threshold effect estimation
Threshold effect test F p-value Crit10 Crit5 Crit1
Single threshold 28.47 0.060 26.335 32.150 41.494
Double threshold 7.55 0.807 24.348 27.292 37.125
Table 11 Threshold value estimation
Threshold effect test Threshold values 95% confidence
interval
Lower Upper
First threshold (
𝛾1
) 4.4590 4.4577 4.4604
123960 Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
Globalization plays a key role in promoting AI to mitigate
ecological footprint for several reasons. Firstly, globalization
facilitates international communication and cooperation,
enabling the rapid spread of AI technologies. Enterprises
and research institutions around the world can share research
and development outcomes, accelerating technological inno-
vation. Secondly, globalization brings an abundance of data
resources, which allows AI to more accurately analyze and
predict environmental changes, providing a scientific basis
for resource optimization. Finally, with the increase in inter-
national trade and investment, AI has played a significant
role in optimizing production processes, reducing energy
consumption, and improving resource use efficiency. This
not only reduces carbon emissions but also helps allevi-
ate the over-exploitation of natural resources (Hassan etal.
2023). Therefore, globalization not only provides a favorable
environment for the dissemination and application of AI
technology but also enhances its capability to alleviate eco-
logical footprints globally, making it a key force in address-
ing this issue.
Conclusions andpolicy implications
AI plays an essential role in promoting sustainable devel-
opment. Based on relevant data from 60 countries world-
wide between 2010 and 2019, we use the entropy method to
measure the level of AI development from four dimensions:
AI-related technology, network infrastructure, government
institutional support, and the international competitiveness
of digital products. It systematically studies its impact on
the ecological footprint. The research results show that (1)
from 2010 to 2019, the global AI development level showed
an upward trend, especially in developed countries where
the AI growth rate is steadily rising, while in developing
countries, the AI growth rate fluctuated. (2) AI significantly
reduced the global ecological footprint, which highlights its
key role in promoting sustainable development. After con-
structing instrumental variables to address potential endoge-
neity issues and performing a series of robustness tests, this
conclusion remains robust. (3) By building a panel thresh-
old model, we explored the nonlinear relationship between
AI and ecological footprint, and found that as globalization
deepens, the impact of AI on the ecological footprint shows
a significant threshold effect, with its marginal effects gradu-
ally increasing. Compared with previous studies, this article
provides a new perspective on how AI affects the ecologi-
cal footprint. Previous research has mostly focused on the
Fig. 4 LR diagram of threshold
value test
Table 12 Threshold regression results
Variable Coefficient
UR − 0.032 (− 0.185)
FDI 0.000 (1.641)
RE − 0.000 (− 0.131)
ED 0.533*** (14.859)
AI(GL ≤
𝛾
1)
− 0.180** (− 2.268)
AI(GL >𝛾
1)
− 0.703*** (− 5.431)
Constant − 3.263*** (− 4.561)
Year FE Yes
Nation FE Yes
N 570
R20.407
123961Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
impact of economic growth, energy consumption, and indus-
trialization on the ecological footprint, with less discussion
on the role of new technologies like AI. Our study enriches
this area, not only providing fresh insights for global econo-
mies to vigorously develop AI and sustainable development,
but also enhancing the understanding of the intricate synergy
between AI and the ecological footprint.
Based on these research findings, we propose the follow-
ing policy recommendations to reduce the global ecological
footprint and achieve sustainable development:
(1). Given that this research has confirmed the significant
inhibitory effect of AI on the ecological footprint of
major global economies, governments at all levels
should uphold and improve policies related to AI pat-
ent protection. Combined with the development direc-
tion of AI, further enhancements should be made to
regulations and systems related to privacy protection
and data security, providing a stable legal environ-
ment for AI development.
(2). Considering the findings of the panel threshold model,
the impact of AI on the ecological footprint is based
on a certain globalization. Governments should for-
mulate measures to take integration into globalization
as a crucial direction, actively participate in global
governance, join global cooperative organizations,
and strengthen economic cooperation and research
projects with other countries.
This paper has provided a fundamental examination of the
role of AI in mitigating the ecological footprint. However,
additional work remains to be done. Firstly, due to com-
pleteness and availability of data, our analysis was limited
to data of 60 countries from 2010 to 2019, and our sample
time and number of countries can be further expanded in the
future. Secondly, while we studied the nonlinear impact of
globalization on AI’s influence on the ecological footprint,
we did not consider potential factors such as the level of
financial development and environmental regulations, which
could also play a significant role in shaping this relation-
ship. Future research can build upon this study by addressing
these limitations and exploring additional variables that may
influence the complex interplay between AI and ecological
sustainability.
Appendix
See Table 13
Appendix B Related formulas forvariable
testing
The LLC test is expressed by formula (1).
Among them,
ΔYmn
represents the difference of variable
Y for the m-th individual at time n,
Ymn−1
is the observed
value of variable Y for the mth individual at time n-1,
cm
is the autoregressive coefficient,
aml
is the coefficient of
explanatory variable X for the mth individual,
Xmn−l
is the
value of the explanatory variable for the mth individual at
time point n-l,
Ipm
is the coefficient for the trend term of
the m-th individual,
Tpn
is the time trend at time point n,
and E_mn is the error term. If
cm<0
, it indicates that the
variable has no unit root; otherwise, it suggests the pres-
ence of a unit root.
The IPS test is shown in Eq.(2).
Among them,
lm
represents the lag order for the mth
individual in the panel. The rest of the variables mentioned
are consistent with those in Eq.(1). If the IPS test statistic
is significant, it suggests that the panel data is stationary,
and hence, there is no unit root for the panel as a whole.
Fisher-ADF and Fisher-PP tests are displayed in for-
mulas (3)–(4).
(1)
Δ
Ymn =cmYmn−1+
c
∑
l=1
amlXmn−l+Ipm Tpn +E
mn
(2)
Δ
Ymn =cmYmn−1+
l
m
∑
t=1
amtXmn−t+E
mn
Table 13 List of countries IOS3
ARG EGY KOR POL
AUS ESP KWT PRT
AUT EST LTU QAT
BEL FIN LVA RUS
BGR FRA MAR SAU
BIH GBR MDA SGP
BRA GRC MEX SVK
CAN HRV M LT SVN
CHE HUN MYS SWE
CHL IDN NLD THA
CHN IND NOR TUN
COL IRL NZL TUR
CZE ISR OMN UKR
DEU ITA PAK USA
DNK JPN PER ZAF
123962 Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
Among them,
w
−1
represents the inverse normal distribu-
tion.
Rm
represents the P value of the unit root.
Sm−1
refers to
the standard deviation.
The formula of the Pedroni cointegration test is shown
in formula (5).
Among them, T is the length of the time series, N repre-
sents the number of individuals in the panel,
Yit
represents
the explained variable,
M
represents the number of inde-
pendent variables,
𝜸i
represents the individual fixed effect,
and
𝜗i
is the coefficient of the time trend.
Xmit
represents
the value of the mth explanatory variable,
𝜷mi
represents
the coefficient of the explanatory variable, and
𝜏it
repre-
sents the deviations in the long-term relationship. Pedroni
test contains 7 statistical measures: panel v-statistic, panel
rho-statistic, panel PP-statistic, panel ADF-statistic, group
rho-statistics, group PP-statistics, and group ADF-statistics.
The structures of these seven statistics are illustrated in Eqs.
(6) to (12).
Panel v-statistic (within-dimension)
Among them,
eit−1
represents the residual of individual
i at time t, and
L11i
reflects the long-term variance of the
residual sequence of individual i.
Panel rho-statistic (within-dimension)
Δeit
represents the first difference of the residuals, while
𝜆i
is the adjustment parameter.
Panel PP-statistic (within-dimension)
𝜎 2
NT
represents the heteroskedasticity-robust long-term
variance of residuals.
Panel ADF-statistic (within-dimension)
(3)
ADF
−Fisher =−2
M
∑
m=1
log(Rm)→
P
(4)
DF
−Choi =1
S
m−1
S
m=
1
w−1Rm→S(0,1
)
(5)
Y
it =𝛾i+𝜗it+
M
∑
m=1
𝛽mixmi,t+𝜏it ,t=1, …,T;i=1, …,N;m=1, …,
M
(6)
Z
vNT =
1
(
∑
N
i=1∑
T
t=1
L−2
11i
e2
it−1)
(7)
Z
𝜌 NT−1=N
i=1T
t=1
L−2
11i
eit−1Δeit −
𝜆i
N
i=1
T
t=1
L−2
11ie2
it−1
(8)
Z
tNT =∑
N
i=1∑
T
t=1
L−2
11i(eit−1Δeit −
𝜆i
)
�
𝜎 2
NT (
∑
N
i=1
∑
T
t=1
L−2
11ie2
it−1)
e
∗
it−1
and
Δ
e
∗
it
represent the adjusted residuals and their
first differences, respectively, while
S
∗2
NT
denotes the heter-
oskedasticity-robust long-term variance of residuals.
Group rho-statistic (between-dimension)
Group PP-statistic (between-dimension)
𝜎 2
i
represents the residual variance of individual i.
Group ADF-statistic (between-dimension)
s
∗2
i
is the adjusted residual standard deviation.
The equation of the Kao cointegration test is shown in
formula (13).
where
Dbn =fb
,
n−
1
+Ubn,Zbn =zb
,
n−
1
+Vbn,n=1…N;b=1…B.
Dbn
is the test statistic,
ab,b,cbn
are the coefficients of the
regression equation,
Zbn
is the explanatory variable, and
Ubn,Vbn
are the error terms.
Author contribution Qiang Wang: conceptualization, methodology,
software, data curation, writing—original draft preparation, supervi-
sion, writing—reviewing and editing; Tingting Sun: methodology,
software, investigation, writing—original draft, writing—reviewing
and editing; Rongrong Li: conceptualization, methodology, software,
investigation, writing—original draft, writing—reviewing and editing.
Funding The author thanks the following funds for the support:
National Natural Science Foundation of China (Grant No. 72104246).
Data availability The datasets used and/or analysed during the cur-
rent study are available from the corresponding author on reasonable
request.
Declarations
Ethics approval and consent to participate Not applicable.
Consent for publication Not applicable.
(9)
Z
∗
tNT =∑
N
i=1∑
T
t=1
L−2
11i
e∗
it−1Δe∗
it
�
S∗2
NT
(
∑
N
i=
1
∑
T
t=
1
L−2
11
i
e2
it−
1
)
(10)
Z
𝜌 NT−1=N
i=1
T
t=1
eit−1Δeit −
𝜆i
(
T
t=1
e2
it−1
)
(11)
Z
tNT =N
i=1
T
t=1
eit−1Δeit −
𝜆i
𝜎 2
i
(
T
t=
1e2
it−
1)
(12)
Z
∗
tNT =N
i=1
T
t=1
e∗
it−1Δe∗
it
T
t=
2
s∗2
i
e2∗
it−1
(13)
Dbn =ab+bZbn +cbn
123963Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
Competing interests The authors declare no competing interests.
References
Abid N, Ceci F, Razzaq A (2023) Inclusivity of information and com-
munication technology in ecological governance for sustain-
able resources management in G10 countries. Resour Policy
81:103378. https:// doi. org/ 10. 1016/j. resou rpol. 2023. 103378
Acemoglu D, Restrepo P (2020) Robots and jobs: evidence from US
labor markets. J Polit Econ 128(6):2188–2244
Aghion P, Jones BF, Jones CI (2017) Artificial Intelligence and Eco-
nomic Growth. National Bureau of Economic Research Working
Paper Series No. 23928. https:// doi. org/ 10. 3386/ w23928
Ahmad M, Jiang P, Majeed A, Umar M, Khan Z, Muhammad S (2020)
The dynamic impact of natural resources, technological innova-
tions and economic growth on ecological footprint: An advanced
panel data estimation. Resour Policy 69:101817. https:// doi. org/
10. 1016/j. resou rpol. 2020. 101817
Ahmad T, Zhang D, Huang C, Zhang H, Dai N, Song Y etal (2021)
Artificial intelligence in sustainable energy industry: status quo,
challenges and opportunities. J Clean Prod 289:125834. https://
doi. org/ 10. 1016/j. jclep ro. 2021. 125834
Ahmed Z, Wang Z, Mahmood F, Hafeez M, Ali N (2019) Does globali-
zation increase the ecological footprint? Empirical evidence from
Malaysia. Environ Sci Pollut Res 26(18):18565–18582. https://
doi. org/ 10. 1007/ s11356- 019- 05224-9
Ahmed Z, Zhang B, Cary M (2021) Linking economic globalization,
economic growth, financial development, and ecological foot-
print: evidence from symmetric and asymmetric ARDL. Ecol
Ind 121:107060. https:// doi. org/ 10. 1016/j. ecoli nd. 2020. 107060
Albaity M, Awad A (2023) The heterogeneous effect of ICT on coun-
tries with different levels of ecological degradation and income:
a panel quantile approach. J Open Innov: Technol Mark Complex
9(2):100055. https:// doi. org/ 10. 1016/j. joitmc. 2023. 100055
Al-Sharafi MA, Al-Emran M, Arpaci I, Iahad NA, AlQudah AA, Iran-
manesh M etal (2023) Generation Z use of artificial intelligence
products and its impact on environmental sustainability: a cross-
cultural comparison. Comput Hum Behav 143:107708. https://
doi. org/ 10. 1016/j. chb. 2023. 107708
Alvarado R, Tillaguango B, Murshed M, Ochoa-Moreno S, Rehman
A, Işık C etal (2022) Impact of the informal economy on the
ecological footprint: The role of urban concentration and globali-
zation. Econ Anal Policy 75:750–767. https:// doi. org/ 10. 1016/j.
eap. 2022. 07. 001
Apaydin Ş, Ursavaş U, Koç Ü (2021) The impact of globalization on
the ecological footprint: do convergence clubs matter? Environ
Sci Pollut Res 28(38):53379–53393. https:// doi. org/ 10. 1007/
s11356- 021- 14300-y
Appiah M, Li M, Naeem MA, Karim S (2023) Greening the globe:
Uncovering the impact of environmental policy, renewable energy,
and innovation on ecological footprint. Technol Forecast Soc Chang
192:122561. https:// doi. org/ 10. 1016/j. techf ore. 2023. 122561
Au C-C, Henderson JV (2006) How migration restrictions limit
agglomeration and productivity in China. J Dev Econ
80(2):350–388
Balsalobre-Lorente D, Gokmenoglu KK, Taspinar N, Cantos-
Cantos JM (2019) An approach to the pollution haven and
pollution halo hypotheses in MINT countries. Environ Sci
Pollut Res 26(22):23010–23026. https:// doi. org/ 10. 1007/
s11356- 019- 05446-x
Benabed A, Miksik O, Baldissera A, Gruenbichler R (2022) Small
and medium-sized enterprises’ status in the perspectives of
internationalization, globalization and artificial intelligence.
IBIMA Bus Rev 2022:1–15
Bibi M, Khan MK, Tufail MMB, Godil DI, Usman R, Faizan M
(2023) How ICT and globalization interact with the environ-
ment: a case of the Chinese economy. Environ Sci Pollut Res
30(3):8207–8225. https:// doi. org/ 10. 1007/ s11356- 022- 22677-7
Borges AFS, Laurindo FJB, Spínola MM, Gonçalves RF, Mattos CA
(2021) The strategic use of artificial intelligence in the digital
era: Systematic literature review and future research directions.
Int J Inf Manag 57:102225. https:// doi. org/ 10. 1016/j. ijinf omgt.
2020. 102225
Castells-Quintana D, Dienesch E, Krause M (2021) Air pollution in an
urban world: a global view on density, cities and emissions. Ecol
Econ 189:107153. https:// doi. org/ 10. 1016/j. ecole con. 2021. 107153
Chen X, Wu X, Lee KY (2021) The mutual benefits of renewables
and carbon capture: achieved by an artificial intelligent sched-
uling strategy. Energy Convers Manag 233:113856. https:// doi.
org/ 10. 1016/j. encon man. 2021. 113856
Chen Y, Cheng L, Lee C-C (2022) How does the use of industrial
robots affect the ecological footprint? International evidence. Ecol
Econ 198:107483. https:// doi. org/ 10. 1016/j. ecole con. 2022. 107483
Coteur I, Marchand F, Debruyne L, Lauwers L (2019) Structuring
the myriad of sustainability assessments in agri-food systems:
a case in Flanders. J Clean Prod 209:472–480. https:// doi. org/
10. 1016/j. jclep ro. 2018. 10. 066
Cutcu I, Beyaz A, Gerlikhan SG, Kilic Y (2023) Is ecological foot-
print related to foreign trade? Evidence from the top ten fast-
est developing countries in the global economy. J Clean Prod
413:137517. https:// doi. org/ 10. 1016/j. jclep ro. 2023. 137517
Damioli G, Van Roy V, Vertesy D (2021) The impact of artificial
intelligence on labor productivity. Eurasian Bus Rev 11(1):1–25
Danish, Hassan ST, Baloch MA, Mahmood N, Zhang J (2019) Link-
ing economic growth and ecological footprint through human
capital and biocapacity. Sustain Cities Soc 47:101516. https://
doi. org/ 10. 1016/j. scs. 2019. 101516
Dauvergne P (2021) The globalization of artificial intelligence: con-
sequences for the politics of environmentalism. Globalizations
18(2):285–299. https:// doi. org/ 10. 1080/ 14747 731. 2020. 17856 70
Dhar P (2020) The carbon impact of artificial intelligence. Nat Mach
Intell 2(8):423–425. https:// doi. org/ 10. 1038/ s42256- 020- 0219-9
Ding T, Li J, Shi X, Li X, Chen Y (2023) Is artificial intelligence
associated with carbon emissions reduction? Case of China.
Resour Policy 85:103892. https:// doi. org/ 10. 1016/j. resou rpol.
2023. 103892
Dong F, Li J, Huang J, Lu Y, Qin C, Zhang X etal (2023) A reverse dis-
tribution between synergistic effect and economic development:
An analysis from industrial SO2 decoupling and CO2 decoupling.
Environ Impact Assess Rev 99:107037. https:// doi. org/ 10. 1016/j.
eiar. 2023. 107037
Fosso Wamba S (2022) Impact of artificial intelligence assimilation on
firm performance: the mediating effects of organizational agility
and customer agility. Int J Inf Manag 67:102544. https:// doi. org/
10. 1016/j. ijinf omgt. 2022. 102544
Foundation FD, Initiative YUEF, Network GF (2023) National Foot-
print and Biocapacity Accounts. Downloaded [2023/10/25] from
https:// data. footp rintn etwork. org. Accessed 2023
Garg S, Mahajan N, Ghosh J (2022) Artificial Intelligence as an Emerg-
ing Technology in Global Trade: The Challenges and Possibilities.
In: Garg V and Goel R (eds) Handbook of Research on Innovative
Management Using AI in Industry 5.0 (pp. 98–117). IGI Global.
https:// doi. org/ 10. 4018/ 978-1- 7998- 8497-2. ch007
Graetz G, Michaels G (2018) Robots at work. Rev Econ Stat
100(5):753–768. https:// doi. org/ 10. 1162/ rest_a_ 00754
Gygli S, Haelg F, Potrafke N, Sturm J-E (2019) The KOF Globalisation
Index – revisited. Rev Int Organ 14(3):543–574. https:// doi. org/
10. 1007/ s11558- 019- 09344-2
123964 Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
Haluza D, Jungwirth D (2023) Artificial intelligence and ten soci-
etal megatrends: an exploratory study using GPT-3. Systems
11(3):120. https:// doi. org/ 10. 3390/ syste ms110 30120
Hansen MT (1999) The search-transfer problem: the role of weak ties
in sharing knowledge across organization subunits. Adm Sci Q
44(1):82–111
Hao KJMtR (2019) Training a single AI model can emit as much car-
bon as five cars in their lifetimes. MIT Technology Review 75:103
Hassan ST, Batool B, Wang P, Zhu B, Sadiq M (2023) Impact of eco-
nomic complexity index, globalization, and nuclear energy con-
sumption on ecological footprint: first insights in OECD context.
Energy 263:125628. https:// doi. org/ 10. 1016/j. energy. 2022. 125628
Hilbert M (2020) Digital technology and social change: the digital
transformation of society from a historical perspective. Dialogues
Clin Neurosci 22(2):189–194. https:// doi. org/ 10. 31887/ DCNS.
2020. 22.2/ mhilb ert
Hua Y, Dong F, Goodman J (2021) How to leverage the role of social
capital in pro-environmental behavior: a case study of resi-
dents’ express waste recycling behavior in China. J Clean Prod
280:124376. https:// doi. org/ 10. 1016/j. jclep ro. 2020. 124376
Huang Y, Haseeb M, Usman M, Ozturk I (2022) Dynamic association
between ICT, renewable energy, economic complexity and eco-
logical footprint: is there any difference between E-7 (developing)
and G-7 (developed) countries? Technol Soc 68:101853. https://
doi. org/ 10. 1016/j. techs oc. 2021. 101853
Im KS, Pesaran MH, Shin Y (2003) Testing for unit roots in heteroge-
neous panels (Article). J Econom 115(1):53–74. https:// doi. org/
10. 1016/ s0304- 4076(03) 00092-7
Jahanger A, Usman M, Murshed M, Mahmood H, Balsalobre-Lorente
D (2022) The linkages between natural resources, human capi-
tal, globalization, economic growth, financial development, and
ecological footprint: the moderating role of technological innova-
tions. Resour Policy 76:102569. https:// doi. org/ 10. 1016/j. resou
rpol. 2022. 102569
John N, Wesseling JH, Worrell E, Hekkert M (2022) How key-enabling
technologies’ regimes influence sociotechnical transitions: the
impact of artificial intelligence on decarbonization in the steel
industry. J Clean Prod 370:133624. https:// doi. org/ 10. 1016/j. jclep
ro. 2022. 133624
Kahouli B, Hamdi B, Nafla A, Chabaane N (2022) Investigating the rela-
tionship between ICT, green energy, total factor productivity, and
ecological footprint: empirical evidence from Saudi Arabia. Energ
Strat Rev 42:100871. https:// doi. org/ 10. 1016/j. esr. 2022. 100871
Kao C (1999) Spurious regression and residual-based tests for coin-
tegration in panel data. J Econom 90(1):1–44. https:// doi. org/ 10.
1016/ S0304- 4076(98) 00023-2
Kar AK, Choudhary SK, Singh VK (2022) How can artificial intelli-
gence impact sustainability: a systematic literature review. J Clean
Prod 376:134120. https:// doi. org/ 10. 1016/j. jclep ro. 2022. 134120
Karlilar S, Balcilar M, Emir F (2023) Environmental sustainability in
the OECD: the power of digitalization, green innovation, renew-
able energy and financial development. Telecommun Policy
47(6):102568. https:// doi. org/ 10. 1016/j. telpol. 2023. 102568
Khan MK, Teng J-Z, Khan MI, Khan MO (2019) Impact of globaliza-
tion, economic factors and energy consumption on CO2 emissions
in Pakistan. Sci Total Environ 688:424–436. https:// doi. org/ 10.
1016/j. scito tenv. 2019. 06. 065
Kirikkaleli D, Adebayo TS, Khan Z, Ali S (2021) Does globalization
matter for ecological footprint in Turkey? Evidence from dual
adjustment approach. Environ Sci Pollut Res 28(11):14009–
14017. https:// doi. org/ 10. 1007/ s11356- 020- 11654-7
Korinek A, Stiglitz JE (2021) Artificial Intelligence, Globalization,
and Strategies for Economic Development. National Bureau of
Economic Research Working Paper Series No. 28453. https:// doi.
org/ 10. 3386/ w28453
Langnel Z, Amegavi GB (2020) Globalization, electricity consump-
tion and ecological footprint: an autoregressive distributive lag
(ARDL) approach. Sustain Cities Soc 63:102482. https:// doi. org/
10. 1016/j. scs. 2020. 102482
Levin A, Lin CF, Chu CSJ (2002) Unit root tests in panel data: asymp-
totic and finite-sample properties (Article). J Econom 108(1):1–
24. https:// doi. org/ 10. 1016/ s0304- 4076(01) 00098-7
Li R, Wang Q, Liu Y, Jiang R (2021) Per-capita carbon emissions in
147 countries: the effect of economic, energy, social, and trade
structural changes. Sustain Prod Consum 27:1149–1164. https://
doi. org/ 10. 1016/j. spc. 2021. 02. 031
Li R, Wang X, Wang Q (2022) Does renewable energy reduce ecologi-
cal footprint at the expense of economic growth? An empirical
analysis of 120 countries. J Clean Prod 346:131207. https:// doi.
org/ 10. 1016/j. jclep ro. 2022. 131207
Li J, Ma S, Qu Y, Wang J (2023a) The impact of artificial intelligence
on firms’ energy and resource efficiency: empirical evidence from
China. Resour Policy 82:103507. https:// doi. org/ 10. 1016/j. resou
rpol. 2023. 103507
Li X, Zhang C, Zhu H (2023b) Effect of information and commu-
nication technology on CO2 emissions: an analysis based on
country heterogeneity perspective. Technol Forecast Soc Chang
192:122599. https:// doi. org/ 10. 1016/j. techf ore. 2023. 122599
Liang S, Yang J, Ding T (2022) Performance evaluation of AI driven
low carbon manufacturing industry in China: an interactive net-
work DEA approach. Comput Ind Eng 170:108248. https:// doi.
org/ 10. 1016/j. cie. 2022. 108248
Liu J, Chang H, Forrest JY-L, Yang B (2020) Influence of artificial intel-
ligence on technological innovation: evidence from the panel data
of china’s manufacturing sectors. Technol Forecast Soc Chang
158:120142. https:// doi. org/ 10. 1016/j. techf ore. 2020. 120142
Liu J, Liu L, Qian Y, Song S (2022) The effect of artificial intelli-
gence on carbon intensity: evidence from China’s industrial sector.
Socioecon Plann Sci 83:101002. https:// doi. org/ 10. 1016/j. seps.
2020. 101002
Lutz C (2019) Digital inequalities in the age of artificial intelligence
and big data. Hum Behav Emerg Technol 1(2):141–148
Lythreatis S, Singh SK, El-Kassar A-N (2022) The digital divide: a
review and future research agenda. Technol Forecast Soc Chang
175:121359. https:// doi. org/ 10. 1016/j. techf ore. 2021. 121359
Maddala GS, Wu S (1999) A comparative study of unit root tests
with panel data and a new simple test. Oxf Bull Econ Stat
61(S1):631–652
Marvin HJP, Bouzembrak Y, van der Fels-Klerx HJ, Kempenaar C,
Veerkamp R, Chauhan A etal (2022) Digitalisation and artifi-
cial intelligence for sustainable food systems. Trends Food Sci
Technol 120:344–348. https:// doi. org/ 10. 1016/j. tifs. 2022. 01. 020
Mor S, Madan S, Prasad KD (2021) Artificial intelligence and car-
bon footprints: roadmap for Indian agriculture. Strateg Chang
30(3):269–280. https:// doi. org/ 10. 1002/ jsc. 2409
Murshed M, Apergis N, Alam MS, Khan U, Mahmud S (2022) The
impacts of renewable energy, financial inclusivity, globaliza-
tion, economic growth, and urbanization on carbon productivity:
evidence from net moderation and mediation effects of energy
efficiency gains. Renew Energy 196:824–838. https:// doi. org/ 10.
1016/j. renene. 2022. 07. 012
Ni Z, Yang J, Razzaq A (2022) How do natural resources, digi-
talization, and institutional governance contribute to ecological
sustainability through load capacity factors in highly resource-
consuming economies? Resour Policy 79:103068. https:// doi.
org/ 10. 1016/j. resou rpol. 2022. 103068
Ojekemi OS, Rjoub H, Awosusi AA, Agyekum EB (2022) Toward
a sustainable environment and economic growth in BRICS
economies: do innovation and globalization matter? Environ
Sci Pollut Res 29(38):57740–57757. https:// doi. org/ 10. 1007/
s11356- 022- 19742-6
123965Environmental Science and Pollution Research (2023) 30:123948–123965
1 3
Pata UK (2021) Linking renewable energy, globalization, agricul-
ture, CO2 emissions and ecological footprint in BRIC coun-
tries: a sustainability perspective. Renew Energy 173:197–208.
https:// doi. org/ 10. 1016/j. renene. 2021. 03. 125
Pedroni P (2001) Fully modified OLS for heterogeneous cointegrated
panels. In: Baltagi BH, Fomby TB and Carter Hill R (eds) Non-
stationary Panels, Panel Cointegration, and Dynamic Panels.
Emerald Group Publishing Limited 15:93–130. https:// doi. org/
10. 1016/ S0731- 9053(00) 15004-2
Phillips PC, Perron P (1988) Testing for a unit root in time series
regression. Biometrika 75(2):335–346
Puaschunder JM (2019) Artificial intelligence market disruption.
Proceedings of the International RAIS Conference on Social
Sciences and Humanities organized by Research Association for
Interdisciplinary Studies (RAIS) at Johns Hopkins University,
Montgomery County Campus, Rockville, MD, United States,
pp 1–8. https:// doi. org/ 10. 2139/ ssrn. 33984 70
Rudolph A, Figge L (2017) Determinants of ecological footprints:
what is the role of globalization? Ecol Ind 81:348–361. https://
doi. org/ 10. 1016/j. ecoli nd. 2017. 04. 060
Safdar NM, Banja JD, Meltzer CC (2020) Ethical considerations in
artificial intelligence. Eur J Radiol 122:108768. https:// doi. org/
10. 1016/j. ejrad. 2019. 108768
Saqib N, Ozturk I, Usman M, Sharif A, Razzaq A (2023) Pollution
haven or halo? How European countries leverage FDI, energy,
and human capital to alleviate their ecological footprint. Gond-
wana Res 116:136–148. https:// doi. org/ 10. 1016/j. gr . 2022. 12. 018
Saud S, Chen S, Haseeb A, Sumayya (2020) The role of financial
development and globalization in the environment: accounting
ecological footprint indicators for selected one-belt-one-road
initiative countries. J Clean Prod 250:119518. https:// doi. org/
10. 1016/j. jclep ro. 2019. 119518
Shen Y, Zhang X (2023) Intelligent manufacturing, green techno-
logical innovation and environmental pollution. J Innov Knowl
8(3):100384. https:// doi. org/ 10. 1016/j. jik. 2023. 100384
Shi M, Jia Z, Mehmood U (2023) Exploring the roles of green
finance and environmental regulations on CO2es: defining the
roles of social and economic globalization in the next eleven
nations. Environ Sci Pollut Res 30(22):62967–62980. https://
doi. org/ 10. 1007/ s11356- 023- 26327-4
Solarin SA, Al-Mulali U, Musah I, Ozturk I (2017) Investigating the
pollution haven hypothesis in Ghana: an empirical investigation.
Energy 124:706–719. https:// doi. org/ 10. 1016/j. energy. 2017. 02. 089
Strubell E, Ganesh A, McCallum A (2019) Energy and policy
considerations for deep learning in NLP. arXiv preprint
arXiv:1906.02243. https:// doi. org/ 10. 48550/ arXiv. 1906. 02243
Sultana T, Hossain MS, Voumik LC, Raihan A (2023) Does globali-
zation escalate the carbon emissions? Empirical evidence from
selected next-11 countries. Energy Rep 10:86–98. https:// doi.
org/ 10. 1016/j. egyr. 2023. 06. 020
Sun J, Dong F (2022) Decomposition of carbon emission reduc-
tion efficiency and potential for clean energy power: evidence
from 58 countries. J Clean Prod 363:132312. https:// doi. org/ 10.
1016/j. jclep ro. 2022. 132312
Sun J, Dong F (2023) Optimal reduction and equilibrium carbon
allowance price for the thermal power industry under China’s
peak carbon emissions target. Financ Innov 9(1):12. https:// doi.
org/ 10. 1186/ s40854- 022- 00410-0
Sun Q, Ma R, Xi Z, Wang H, Jiang C, Chen H (2023) Nonlinear
impacts of energy consumption and globalization on ecological
footprint: empirical research from BRICS countries. J Clean
Prod 396:136488. https:// doi. org/ 10. 1016/j. jclep ro. 2023. 136488
Vinuesa R, Azizpour H, Leite I, Balaam M, Dignum V, Domisch
S etal (2020) The role of artificial intelligence in achieving
the Sustainable Development Goals. Nat Commun 11(1):233.
https:// doi. org/ 10. 1038/ s41467- 019- 14108-y
Wackernagel M, Rees W (1998) Our ecological footprint: reducing
human impact on the earth (Vol. 9). New Society Publishers
Wang Q, Zhang F (2021) The effects of trade openness on decoupling
carbon emissions from economic growth – evidence from 182
countries. J Clean Prod 279:123838. https:// doi. org/ 10. 1016/j.
jclep ro. 2020. 123838
Wang K-L, Sun T-T, Xu R-Y (2023a) The impact of artificial intel-
ligence on total factor productivity: empirical evidence from
China’s manufacturing enterprises. Econ Chang Restruct
56(2):1113–1146. https:// doi. org/ 10. 1007/ s10644- 022- 09467-4
Wang L, Zhou Y, Chiao B (2023b) Robots and firm innovation:
Evidence from Chinese manufacturing. J Bus Res 162:113878.
https:// doi. org/ 10. 1016/j. jbusr es. 2023. 113878
Wang Q, Zhang F, Li R (2023e) Revisiting the environmental
kuznets curve hypothesis in 208 counties: the roles of trade
openness, human capital, renewable energy and natural resource
rent. Environ Res 216:114637. https:// doi. org/ 10. 1016/j. envres.
2022. 114637
Wang Q, Sun J, Pata UK, Li R, Kartal MT (2023c) Digital economy and
carbon dioxide emissions: examining the role of threshold variables.
Geosci Front 101644. https:// doi. org/ 10. 1016/j. gsf. 2023. 101644
Wang Q, Zhang F, Li R (2023d) Free trade and carbon emissions
revisited: the asymmetric impacts of trade diversification and
trade openness. Sustainable Development, n/a(n/a). https:// doi.
org/ 10. 1002/ sd. 2703
Yang B, Usman M, Jahanger A (2021) Do industrialization, eco-
nomic growth and globalization processes influence the ecologi-
cal footprint and healthcare expenditures? Fresh insights based
on the STIRPAT model for countries with the highest healthcare
expenditures. Sustain Prod Consum 28:893–910. https:// doi. org/
10. 1016/j. spc. 2021. 07. 020
Ye Z, Yang J, Zhong N, Tu X, Jia J, Wang J (2020) Tackling envi-
ronmental challenges in pollution controls using artificial intel-
ligence: a review. Sci Total Environ 699:134279. https:// doi. org/
10. 1016/j. scito tenv. 2019. 134279
Zador A, Escola S, Richards B, Ölveczky B, Bengio Y, Boahen K
etal (2023) Catalyzing next-generation artificial intelligence
through NeuroAI. Nat Commun 14(1):1597. https:// doi. org/ 10.
1038/ s41467- 023- 37180-x
Zhang L, Ling J, Lin M (2022a) Artificial intelligence in renewable
energy: a comprehensive bibliometric analysis. Energy Rep
8:14072–14088. https:// doi. org/ 10. 1016/j. egyr. 2022. 10. 347
Zhang X, Song X, Lu J, Liu F (2022b) How financial development
and digital trade affect ecological sustainability: the role of
renewable energy using an advanced panel in G-7 Countries.
Renew Energy 199:1005–1015. https:// doi. org/ 10. 1016/j.
renene. 2022. 09. 028
Zhang Y, Wu M, Tian GY, Zhang G, Lu J (2021) Ethics and privacy
of artificial intelligence: understandings from bibliometrics.
Knowledge-Based Systems 222:106994. https:// doi. org/ 10.
1016/j. knosys. 2021. 106994
Zhou H, Awosusi AA, Dagar V, Zhu G, Abbas S (2023) Unleashing the
asymmetric effect of natural resources abundance on carbon emis-
sions in regional comprehensive economic partnership: what role
do economic globalization and disaggregating energy play? Resour
Policy 85:103914. https:// doi. org/ 10. 1016/j. resou rpol. 2023. 103914
Publisher's Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
Springer Nature or its licensor (e.g. a society or other partner) holds
exclusive rights to this article under a publishing agreement with the
author(s) or other rightsholder(s); author self-archiving of the accepted
manuscript version of this article is solely governed by the terms of
such publishing agreement and applicable law.
A preview of this full-text is provided by Springer Nature.
Content available from Environmental Science and Pollution Research
This content is subject to copyright. Terms and conditions apply.
