Financial inclusion and carbon emissions in Asia: Implications for environmental sustainability

Abstract This study explores how carbon emissions are affected by financial inclusion. Using a balanced panel data set of 26 Asian countries, we compute a composite index, through the principal component analysis (PCA) technique, of financial inclusion based on a set of attributes related to financial inclusion. Our main analysis also delineates the subsamples of developed and developing Asian economies. The results reveal a long (short)-run positive (negative) impact of financial inclusion on carbon emissions across the Asian countries. This finding is also true for the developed country subsample, implying nonlinearity in short- and long-run relationships. For the developing countries, a more pronounced long-run positive impact compared to developed countries is found. Furthermore, the pairwise causality test results indicate the existence of bi-directional causality between financial inclusion and carbon emissions. These findings have important policy implications, especially in the context of the strategic integration of financial inclusion and climate change strategies.


Introduction
Climate change has resulted in devastating consequences for human life and proved to be one of the main impediments for a sustainable and stable environment across the globe, therefore garnering much attention from researchers worldwide (Hussain, Gul, et al. 2023).The enormous disposal of pollutants and harmful gases due to the utilisation of fossil fuels and massive industrial discharge are the primary contributors to global warming (Hussain, Akbar, et al. 2023;Wawrzyniak and Dory� n 2020).Consequently, the worldwide temperature rise may surpass the threshold of 2 � C; and if so, the world may experience disastrous consequences of climate change, affecting all aspects of life in the form of a rapid rise in sea levels, mass extinction, super droughts, water contamination and health problems (Hussain, Gul, and Ullah 2023;Kayani, Ashfaq, and Siddique 2020;Lu 2018).Carbon dioxide (CO 2 ) is the main cause of environmental degradation among the greenhouse gases (hereafter GHGs) because, in 2019, it contributed 64% of GHG emissions. 1 The deterioration of environment due to environmental pollutants has shifted the attention of researchers to explore the nexus between macroeconomic variables and the harmful pollutants.For instance, previous studies ascertain the association between carbon emissions and financial development (Jiang and Ma 2019;Kayani, Ashfaq, and Siddique 2020), economic growth (Le, Chuc, and Taghizadeh-Hesary 2019), energy consumption (Charfeddine and Kahia 2019), trade openness (Shahbaz et al. 2017), urbanisation (Raghutla and Chittedi 2021), and population growth (Yeh and Liao 2017).However, empirical studies on the association between carbon emissions and the emerging concept of financial inclusion are scant.Financial inclusion helps in developing the financial sector of a country and is considered vital in achieving economic development.Financial inclusion has also become an important attribute of the financial sector development (Dahiya and Kumar 2020).Le, Le, and Taghizadeh-Hesary (2020) argue that the literature provides contradictory evidence about the nexus between environmental sustainability and financial inclusion.On one hand, financial inclusion enables individuals and businesses to access credit schemes at lower cost, which makes the investment in green technology more affordable (G€ ok 2020).This way, an inclusive financial system positively contributes to environmental sustainability by encouraging individuals and businesses to use green technology and adopt better environmental practices, which in turn lead to a reduction in GHG emissions (Hussain, Gul, and Ullah 2023).On the other hand, an accessible financial system may also damage environmental quality by boosting manufacturing and industrial activities through affordable financing, which in turn increases CO 2 emissions (Charfeddine and Kahia 2019;Hussain, Akbar, et al. 2023).Financial inclusion also motivates individual consumers to use energy-intensive electric appliances (Hussain, Akbar, et al. 2023).The use of these energy-intensive goods poses a serious threat to environmental sustainability.
Given the existing contradictory empirical evidence on the nexus between financial inclusion and carbon emissions, this study re-examines this link using a strongly balanced panel data set of 26 Asian countries.Le, Chuc, and Taghizadeh-Hesary (2019) argue that despite economic growth in Asia, more significant progress has not been achieved due to the low access of the masses to the financial sector.According to Bhardwaj, Hedrick-Wong, and Thomas (2018), 2/3 of the poor people reside in Asia and almost a billion people have neither a formal bank account nor the access to any financial service.About 27% of the adults hold a formal bank account, and only about 33% have utilised either a loan or credit facility in the developing countries of the Asian region.Much effort has been put to include the masses into the financial sector; however, it is still very challenging to achieve financial inclusion at a maximum level mainly due to contextual differences across countries, which vary along political, economic, cultural, ethnic and religious lines.The climate deterioration is another challenge for sustainable and comprehensive economic growth that involves a broader segment of population.Asia contributed about a half of the world's GHG emissions during 2021 (see Figure 1), making it the biggest contributor to global warming.Le, Le, and Taghizadeh-Hesary (2020) claim that Asia is more vulnerable to the hazardous climate change in comparison to the other continents of the world.Hussain, Akbar, et al. (2023) advocate a hetergenous impact of financial inclusion on carbon emissions which vary with the level of economic development and technological advancement across different countries.Developing countries are more inclined to achieve a higher level of economic development, thereby expanding the production capacity via available credit facilities rather than investing in energyefficient technology for environmental protection (Jiang and Ma 2019).On the contrary, developed countries are usually characterised by a strong environmental protection mechanism which compels these countries to strive for more technological advancement rather than mass scale industrial expansion (Acheampong, Amponsah, and Boateng 2020;Qin et al. 2021).Thus, we conduct a comparative evaluation of the financial inclusion-carbon emissions nexus for developed and developing countries of Asia.The existing empirical literature lacks such a comparative analysis in the context of Asia.
Notable studies in this strand of literature such as Le, Le, and Taghizadeh-Hesary (2020), Renzhi andBaek (2020), andHussain, Akbar, et al. (2023) use unbalanced panel data, while we use strongly balanced data and the pooled mean group (PMG) estimation technique to examine the short-and long-run impact of financial inclusion on carbon emissions, which is important to ascertain the existence of nonlinearity.The causal effect is determined using the Dumitrescu and Hurlin (2012) (hereafter D-H) panel causality approach.We find that financial inclusion has a significant positive (negative) impact on carbon emissions in the long run (short run), implying a nonlinear U-shaped relationship.This result also holds for the developed country subsample; whereas in developing countries, financial inclusion only deteriorates the environmental quality.We also find bi-directional short-run Granger causality between financial inclusion and carbon emissions.The overall results support the notions that an inclusive financial system in developed countries might enhance green investment in the short run and accumulate capital and enhance investment opportunities (Hussain, Gul, et al. 2023), and more intensive production and consumption activities in turn increase carbon emissions in the early stages of financial inclusion until a certain turning point, beyond which the level of carbon emissions starts to decrease.
The rest of the paper is organised as follows.We first provide a review of the existing literature.In the following section, we present the details of data and methodology.The findings of econometric estimations are presented and discussed in the next section.The final section concludes the study.

Literature review
Theoretically, there exist divergent views among researchers regarding the impact of inclusive financial development on environmental sustainability (Raghutla and Chittedi 2021).Some scholars suggest that financial development mitigates GHG emissions by facilitating the energy supply sector to upgrade production technology and equipment through alleviating financial constraints with lower borrowing cost (Renzhi and Baek 2020;Hussain, Gul, et al. 2023).The development related to financial systems, and the availability of capital and funding channels through banks and stock markets, can positively contribute to environmental sustainability by reducing GHG emissions (G€ ok 2020).
Following the reasoning that financial development reduces environmental deterioration (Koshta, Bashir, and Samad 2021), the existing literature acknowledges the constructive role of the financial sector development in combating climate change and achieving inclusive and sustainable economic growth (Charfeddine and Kahia 2019).Analysing a data set for 19 emerging economies over the period from 1990 to 2013, Saidi and Mbarek (2017) empirically show the negative impact of the financial system development on GHG emissions.Dogan and Seker (2016) confirm the cointegration of financial systems-related development with CO 2 emissions through an array of advanced techniques such as the fully modified OLS (FMOLS) and dynamic OLS (DOLS), with an extensive data set of 23 renewable energy-using economies over a longer period from 1985 to 2011.They also support the notion that financial development reduces GHG emissions.Zaidi et al. (2019) analyse the connection between the development related to financial systems and GHG emissions based on advanced statistical tests such as continuously updated bias-corrected and continuously updated fully modified estimators in 17 APEC economies from 1990 to 2016 using the environmental Kuznets curve (EKC) framework.The findings also show that the development of financial systems is negatively associated with CO 2 emissions.
However, some others posit that the financial sector development exacerbates environmental deterioration via an upsurge in CO 2 emissions.A well-functioning financial system encourages businesses to obtain low-cost capital required for expanding production capacity, resulting in the deterioration of environmental quality (Raghutla and Chittedi 2021).Financial development dramatically promotes social consumption by providing better credit utilisation, which could facilitate individual consumers' purchase of more energy-intensive electric appliances (G€ ok 2020), automobiles among many others, which in turn degrades air quality (Koshta, Bashir, and Samad 2021).Capital markets are considered important indicators of economic development.The persistent performance of the stock market attracts individual and institutional investors and stimulates both the production and consumption of fossil fuels, which in turn increase CO 2 emissions (Rajpurohit and Sharma 2021).From this perspective, environmental deterioration increases with financial development.
A plethora of empirical studies, such as those of Al-Mulali, Ozturk, and Lean (2015), Kayani, Ashfaq, and Siddique (2020), and Jiang and Ma (2019), support the theoretical claim that the financial system development deteriorates environmental quality.Using the data of 12 Asian economies during 1993-2013, Lu (2018) ) reveals that financial development contributes to environmental degradation.Using the ARDL approach, Ali et al. (2019) report on the long-and short-run association between the financial sector development and carbon emissions with an extended sample over a period from 1971 to 2010.Their findings reveal that the development of financial systems leads to further environmental degradation.Considering financial inclusion to be an inextricable part of financial development, Le, Le, and Taghizadeh-Hesary (2020) investigate the causal impact of financial inclusion on CO 2 emissions by taking a broader panel of 31 Asian economies from 2004 to 2014 and using Driscoll-Kraay standard errors models.The empirical findings reveal that financial inclusion contributes to environmental deterioration, thereby increasing CO 2 emissions.
The third school of thought supports nonlinearity in the relationship between the financial sector development and climate change.For instance, considering 25 OECD countries as a sample over the period of 1971-2007, Hung, Li, and Shen (2018) ascertain the nonlinear association of financial systems-related development with GHG emissions.Renzhi and Baek (2020), while analysing a data set of 103 economies, examine the connection of climate change (CO 2 emissions) with financial development in the EKC framework, and support the inverted U-shaped relationship.Hussain, Akbar, et al. (2023) examine the nonlinear relationship between financial inclusion and CO 2 emissions based on 74 countries and establish the inverted Ushaped relationship between financial inclusion and CO 2 emissions.We have summarised some existing empirical works in Table 1.
The existing empirical evidence highlights at least two areas for improvement.First, the nonlinear relationship between financial inclusion and carbon emissions in Asia could be explored.This could be done with balanced panel data and by examining the magnitude of short-and long-run coefficients of the impact of financial inclusion on carbon emissions.Second, a comparative analysis is also needed to ascertain if there is a difference in this relationship across developing and developed Asian economies.

Data
Our sample consists of 26 Asian countries, namely Brunei Darussalam, Iraq, Israel, Japan, Kazakhstan, the Republic of Korea, Sri Lanka, Tajikistan, Vietnam, Kuwait, Lebanon, Malaysia, Bhutan, Cambodia, India, Qatar, Saudi Arabia, Singapore, Indonesia, Jordan, the Kyrgyz Republic, Mongolia, Pakistan, the Philippines, the United Arab Emirates, and Bangladesh.These countries are considered based on the availability of complete data for all variables from 2004 to 2014.We collect the data related to financial inclusion and macroeconomic indicators from the Global Findex database (https:// globalfindex.worldbank.org/) and the World Development Indicators (WDI), respectively.
To compare between developed and developing countries, the sample is further divided into two groups.Following Essandoh, Islam, and Kakinaka (2020) and Omar and Inaba (2020), 14 countries that fall in the category of high-income and uppermiddle-income countries are grouped as developed countries.The developed countries consist of Brunei Darussalam, Iraq, Israel, Japan, Jordan, Kazakhstan, the Republic of Korea, Kuwait, Lebanon, Malaysia, Qatar, Saudi Arabia, Singapore, and the United Arab Emirates; and 12 lower-income and lower-middle-income countries are grouped as developing countries, i.e.Bangladesh, Bhutan, Cambodia, India, Indonesia, the Kyrgyz Republic, Mongolia, Pakistan, the Philippines, Sri Lanka, Tajikistan and Vietnam.Shahzad et al. (2017) claim that carbon emissions have a greater cataclysmic impact on environmental degradation among anthropogenic GHG emissions.Hence, we consider CO 2 emissions as an indicator of environmental degradation.
Financial inclusion refers to the access of individuals and businesses to bank deposits, credit, insurance and other financial services in order to reduce poverty and income inequality (Kim, Yu, and Hassan 2018).The inclusive financial system reduces the cost of borrowing, which allows all participants of an economy to achieve financial stability (Le, Chuc, and Taghizadeh-Hesary 2019).
More precisely, financial inclusion has three aspects, i.e. obtainability, usage and the penetration of financial facilities (Sarma 2016).Renzhi and Baek (2020) support the notion that the obtainability and utilisation of financial services are key pillars for a financially inclusive system.In this study, we consider four different proxies for the construction of a composite measure of financial inclusion, i.e. the number of branches of commercial banks per 100,000 adults, bank deposit amount, the number of ATMs per 100,000 adults, and the amount of bank credit (Chatterjee 2020; Le, Le, and Taghizadeh-Hesary 2020).These proxies for financial inclusion correspond to two important aspects: the former two indicators demonstrate the availability of the banking system while the latter two demonstrate the use of the banking system (Hussain, Gul, et al. 2023).
We use principal component analysis (PCA) to construct the financial inclusion index because it helps in creating a simple standalone indicator by reducing the dimensionality of data.To test the suitability of the selected proxies for PCA, we apply the Bartlett test and the Kaiser-Meyer-Olkin (KMO) test.The results are reported in Table A1 in Appendix.The value of the KMO test is greater than 0.5, indicating the adequacy of the proxies for PCA (Le, Chuc, and Taghizadeh-Hesary 2019).
PCA is a two-step process.In the first step, different factors are identified which account for the maximum variation in the original variables, with the lowest correlation between two components.Then, in line with Hussain, Akbar, et al. (2023), an index is estimated based on the factors having an eigenvalue > 1.The cumulative variance of each component and the pattern matrix of PCA are reported in Tables A2-A3 in Appendix.
Table 2 provides the details of the variables and summary statistics for the 26 selected Asian countries.The mean and standard deviation of carbon emissions are 8.55 and 11.11, respectively.The summary of the financial inclusion proxies reveals that the mean values of the branches of commercial banks per 100,000 adults and ATMs per 100,000 adults are close to the mean values of these proxies given by Le, Le, and Taghizadeh-Hesary (2020), while outstanding deposits (% of GDP) and loans (% of GDP) are in conformity with those in Le, Chuc, and Taghizadeh-Hesary (2019).The scatter plot of financial inclusion (on x-axis) and carbon emissions (on yaxis) shown in Figure 2 indicates an upward sloping curve, showing a possible positive long-run relationship between the variables.

Methodology
Before estimating our final model, we first check for cross-sectional dependence through the Pesaran (2015) cross-sectional dependence test.The unit root properties are determined using the cross-sectional augmented IPS (CIPS) unit root test proposed by Pesaran (2007).The cointegration among the variables is ascertained through the panel cointegration tests following Pedroni (2004), Kao (1999) and Westerlund (2005).Further, we apply the panel ARDL approach to establish the long-and short-run relationships.The D-H panel causality approach is employed to explore the direction of causality.For brevity, we only provide the details of the ARDL and D-H methods.
We estimate the following general form of regression to examine the impact of financial inclusion on CO 2 emissions: where CO 2 represents the natural log of carbon emissions and FI stands for the financial inclusion index.In light of the existing literature and given the availability of the balanced panel data, we include EC, the natural log of energy consumption (Charfeddine and Kahia 2019), TO, the natural log of trade openness (Hussain, Akbar, et al. 2023), and IND, the natural log of industrialisation (Zhou and Liu 2016) as control variables; e i, t is the error term.The above function is estimated using the panel ARDL approach with the PMG estimation to examine the long-and short-run impact because this approach simultaneously caters to serial correlation and endogeneity issues. 2The PMG approach constrains the long-run coefficients to be identical but allows the short-run and error correction coefficients to differ across countries.Hence, the following equation is estimated to empirically test the relationship between financial inclusion and CO 2 emissions: where y i, t represents CO 2 emissions, ; i is the error correction coefficient enabling us to measure the speed of the adjustment to the long-run equilibrium, and u i, t−1 is the deviation from the long-run state.h is the vector of common long-run coefficients measuring the impact of exogenous variables, such as the financial inclusion index, trade openness, energy consumption and industrialisation as represented by x i, t : W and d are coefficients of the short-term dynamics, p and q are the lag lengths selected using the Schwarz Bayesian criterion.l i represents the vector of coefficients for the country-specific effects, � i, t is the deterministic vector of regressors, i.e. constants, the time trend, and/or dummy variables, and e i, t is the disturbance term with different variances across countries.Further, we employ the D-H panel causality test in order to examine the causality direction, which is considered as a useful tool for regulators and policy makers to design a comprehensive policy.The D-H test is considered superior to traditional causality tests since it allows for cross-sectional dependence.It is as follows: (4) where d, g and b are coefficients.The D-H panel causality test is based on the individual Wald statistics of Granger (1969) causality tests averaged across the crosssectional units.The auto generated Wald statistics through the D-H test can be written as: where W i,T represents the individual Wald statistic for each cross-sectional unit.

Findings
We first apply the Pesaran (2015) test to check for cross-sectional dependence in the full sample and two subsamples.The test is suitable for a larger cross-sectional and smaller time-series data set as used in the current research (N ¼ 26 > T ¼ 11).Table 3 reports the result of the test with the null hypothesis that there is no cross-sectional dependence.Consistent with Le, Le, and Taghizadeh-Hesary (2020), the result of the cross-sectional dependence test rejects the null hypothesis and concludes that crosssectional dependence is present across the economies in Asia.
The unit root properties for the level and first difference series are tested using the CIPS unit root test of Pesaran (2007), with intercept as well as with intercept and trend.The results shown in Table 4 conclude that all the variables are I(I), except for weak evidence that IND might be stationary in level, when both intercept and trend are included in the unit root test.Thus, we further test the cointegration among the variables using the Pedroni (2004), Kao (1999) and Westerlund (2005) approaches.
The Pedroni (2004) approach is more suitable when N > T. The results reported in Table 5 collectively conclude the presence of a long-run relationship among the selected variables in the Asian region.We test the long-and short-run impact of financial inclusion on carbon emissions using the panel ARDL technique based on the PMG estimation.The estimation results for the full sample as well as for the developed and developing subsamples are presented in Table 6.The upper panel of the table reports the coefficients and standard errors for the long-run impact of selected independent variables on carbon emissions.The short-run impact, along with error correction terms, is shown in the lower panel.All the error correction terms are negative and significant at the 1% level of significance, implying a long-run equilibrium relationship between the selected variables.The speed of adjustment to the long-run equilibrium is −0.350 for the whole sample, and −0.246 for the developed countries and −0.667 for the  developing countries in Asia.The higher speed of adjustment to the long-run equilibrium in the developing countries implies that the selected variables have a lower impact on carbon emissions in the short run in these countries.
It is important to mention that we follow the framework proposed by Narayan and Narayan (2010), which suggests that the change in the magnitude and in the signs of short-and long-run coefficients can help ascertain the inherent nonlinearity in the relationship.If long-run coefficients are higher than short-run coefficients, then the countries fail to improve environmental quality, and vice versa.For the full sample, we find that financial inclusion has a positive (1.123) and significant, at the 1% level of significance, impact on carbon emissions in the long run; whereas the short-term impact of financial inclusion is negative (−0.401) and significant at the 10% level.The difference in signs between the short-and long-run impacts implies that the relationship between carbon emissions and financial inclusion is nonlinear.The findings are consistent with those of Hussain, Akbar, et al. (2023).Furthermore, although financial inclusion may reduce carbon emissions in the short run, it deteriorates the environment in the long run.The finding that financial inclusion has led to higher emissions of CO 2 in the long run is in line with that of Le, Le, and Taghizadeh-Hesary (2020).However, we posit that the impact is negative in the short run, which implies the existence of nonlinearity.Energy consumption, trade and industrialisation also increase carbon emissions in the long run.
The results for the developed country subsample in Table 6 are similar to full sample findings where we see the positive (negative) impact of financial inclusion on carbon emissions in the long (short) run.Our finding that the short-run impact of financial inclusion is negative for developed countries seems consistent with the notion that the presence of better industrial systems and strict environmental regulations encourages businesses to make investment in environmentally friendly endeavours (Jiang and Ma 2019;Acheampong 2019).However, we posit this impact is short lived in Asian developed economies.
For the developing subsample, we find a significant and positive impact of financial inclusion only in the long run.While Khan and Ozturk (2021) support the pollution inhibiting role of financial development in developing countries, we argue that the higher disastrous impact of financial inclusion on climate change is attributable to the fact that Asian developing countries are under pressure to achieve faster economic development by using limited available credit for large scale expansion, which further exacerbates environmental quality (Qin et al. 2021).
Finally, we test for the pairwise causality between financial innovation and carbon emissions in a panel setting through the D-H test.The results for pairwise causality tests are provided in Table 7. Keeping in view the short span of our data, we use lag order one in our analysis ( � Zikovi� c, Tomas � Zikovi� c, and Vlahini� c Lenz 2020).The null hypothesis of no causality from financial inclusion to carbon emissions and vice versa is rejected for the full sample and for both developed and developing subsamples.

Conclusion
Massive unmodulated development has created challenges including environmental deterioration and climate change.Global warming is a serious threat to the existence of human beings and has thus been considered a primary concern over the last couple of decades.This study empirically investigates the impact of financial inclusion on carbon emissions based on the balanced panel data of 26 Asian countries from 2004 till 2014, and analyses the difference in this impact between developed and developing countries.
Our findings reveal a positive (negative) impact of financial inclusion on CO 2 emissions in the long (short) run for the full sample and this conclusion remains valid for developed countries.The results are consistent with the idea that financial inclusion, which facilitates the easy access of individuals and businesses to financial services, motivates businesses to expand the scale of production and increases the ability of individual consumers to purchase energy-intensive electric appliances, which triggers the use of energy from fossil fuels, resulting in higher CO 2 emissions in the long run.
As regards the more salient adverse effect of financial inclusion on climate change in developing economies, it is attributed to the notion that businesses tend to expand their production through financial loans instead of investing in green technology projects, since developing countries are under extreme pressure to develop economically, which leads to further environmental deterioration.Future studies could explore the role of governance in the relationship between financial inclusion and the various proxies of climate change.
Policy makers should integrate their climate change policies with the financial sector development and optimise the inclusivity of financial resources as per climate change adaptation strategies.For instance, under the widely acknowledged green credit guidelines of China, the financial system provides adequate support through green credit financial services to reduce carbon emissions and contribute to green economy and sustainable development.Besides, regulatory authorities may align initiatives for financial inclusion with environmental protection policies.

Figure 2 .
Figure 2. The scatter plot of annual cross-sectional averages of financial inclusion (x-axis) and (the log of) carbon emissions (y-axis).

Table 1 .
Summaries of some existing panel data studies on the relationship between financial inclusion (FI)/financial development (FD) and CO 2 emissions.

Table 2 .
Details of the variables and basic statistics.

Table 3 .
The result of the Pesaran (2015) cross-sectional dependence test.
Notes: �� and ��� denote statistical significance at the 5% and the 1% level, respectively.CO 2 stands for the natural log of carbon emissions, FI represents the financial inclusion index, TO is the natural log of trade openness, EC represents the natural log of energy consumption, and IND is the natural log of industrialisation; the same applies to the following tables.

Table 4 .
The CIPS unit root test.�� and ��� denote statistical significance at the 10%, 5% and 1% levels, respectively.All the series, except the series for FI, are in the logarithmic form.

Table 7 .
Results for D-H Granger causality tests.Notes: If the p value is less than 0.05, it means that the independent variable Granger causes the dependent variable and this relationship is statistically significant.