1. INTRODUCTION

Our world is facing problems caused by climate change, especially in recent decades, and the main cause of this problem is greenhouse gas (GHG) emissions. The Paris Agreement, which entered into force in 2016, aims to reduce emissions as soon as possible and to offset the released and captured GHGs by the second half of the 21st century. Economic and environmental researchers are also trying to contribute to this issue by using various methods and samples on the factors affecting ghg emissions. This intensive research has not ignored the possible impact of financial indicators. In fact, it can be seen that there are two opposing views on how financial development affects emission levels. According to the first view, financial development can increase energy consumption and hence emissions by removing credit constraints (^{Shahbaz et al., 2016}; ^{Paramati, Mo, and Huang, 2021}). Although this effect will occur directly in households, it runs in parallel with the increase in output in firms. According to the second opposite view, as financial development increases, opportunities for research and development and the use of environmentally friendly technologies will increase, so positive effects on environmental quality can be expected (^{Al Mamun et al., 2018}; ^{Gill, Hassan, and Haseeb, 2019}; ^{Vo and Zaman, 2020}). The results of empirical studies examining the relationship between financial development and GHG emissions are also mixed. Therefore, there is still a need for new empirical analyses that complement the shortcomings of the existing literature.

A striking shortcoming of previous studies is that the idea that financial development affects emission levels via energy consumption, while often emphasized, has not been empirically tested for developing countries. While financial development can have a significant impact on the level and composition of energy consumption (^{Sadorsky, 2010}; ^{Al Mamun et al., 2018}; ^{Mukhtarov et al., 2020}), the level and composition of energy consumption determines the level of emissions. For this reason, it is important to show these effects using nonlinear models. ^{Zeqiraj, Sohag, and Soytas (2020)} is the only study that examines the effects of financial development considering the indirect effects due to energy consumption. However, the study focused on European Union (EU) countries and did not consider the distinction between financial markets and financial institutions. Moreover, a specific threshold for the financial development indicator was not calculated.

To address this gap in the literature, we used a panel threshold regression to test the moderating effects of financial development. In addition, the relationships were tested with both static and dynamic models and the results were compared. The second important shortcoming in the literature relates to the financial development indicator used. Many different variables such as broad money supply and credit to the private sector relative to Gross Domestic Product (GDP) have been used as indicators of financial development. However, in this study, the current financial development index published by the International Monetary Fund (IMF) was used, which addresses financial development in a comprehensive manner rather than in a single dimension. The index includes information on the depth, accessibility and efficiency of financial markets and institutions. Third, because the index breaks down the development of financial institutions and markets, we were able to examine the effects of different types of financial development separately. This allowed us to answer the question of whether the effects of different indicators are different. Fourth, we conducted second-generation stationarity tests, allowing for the cross-sectional dependence before regression. In this way, we were able to obtain more robust results. Finally, we consider the example of developing countries, which has received relatively little attention. As economies develop, the share of the financial market increases, and the composition of financial systems changes in favor of the stock market. ^{Zeqiraj, Sohag, and Soytas (2020)} pointed out that the financial system in developed countries is mainly based on the stock market, while in developing countries it is based on the banking sector. Therefore, it is important to consider these countries as a separate sample as their financial structures differ significantly from those of developed countries.

In the context of these objectives, the remainder of the study is organized as follows. Section 2 reviews the relevant literature. Section 3 presents the methodology. Data and empirical results are given in Section 4. Section 5 provides conclusions and policy implications.

2. LITERATURE REVIEW

The question of how financial development affects environmental quality has been frequently discussed in the literature. There are two opposing views on this issue. According to the first view, financial development will promote the emergence and growth of firms by removing credit and liquidity constraints, so that emissions will increase in line with the increase in production and energy demand. ^{Sadorsky (2010)} pointed out that financial development reduces financial risk and borrowing costs, increases transparency between borrowers and lenders, and thus leads to an increase in fixed investment and energy consumption. Financial development can also lead consumers to purchase energy intensive goods by facilitating borrowing, and, it can also increase energy demand by lowering the cost of starting new businesses and hiring new workers. According to the second view, financial development will increase the use of research and development and investment in clean technologies (^{Paramati, Mo, and Huang, 2021}), thereby reducing the amount of emissions.

The selection of key variables for financial development is a critical issue. It is noteworthy that various indicators of financial development have been used in the literature. Broad money supply, private sector credit, and stock market development and index performance are among the most widely used indicators. For example, ^{Shahbaz et al. (2013a}; ^2013b) used domestic credit to the private sector as an indicator of financial development. Using data from Malaysia, ^{Shahbaz et al. (2013a)} examined the impact of financial development on emission levels. They considered both the quadratic and cubic terms of financial development to examine their non-linear effects. The results of the cointegration test indicate that the relationship between financial development and environmental degradation is N-shaped. In a similar study, ^{Shahbaz et al. (2013b)} examined the nonlinear relationship between financial development and emissions in the case of Indonesia. The results of the study suggest a U-shaped relationship between the variables. Later, ^{Gill, Hassan, and Haseeb (2019)} reexamined the impact of financial development on environmental degradation in Malaysia. The results of the Autoregressive Distributed Lag (ARDL) approach revealed that financial development mitigates emissions and has a moderating role on the environmental Kuznets curve by inducing the inflexion point earlier. ^{Charfeddine and Khediri (2016)} confirmed the existence of a U-shaped relationship between financial development and emissions in the United Arab Emirates. ^{Rahman et al. (2020)} examined both the linear and nonlinear effects of financial development on emissions in Lithuania. Using the ARDL bounds testing approach, they found empirical evidence for the existence of a U-shaped relationship between financial development and emissions. ^{Tahir et al. (2021)}, on the other hand, argued that financial development stimulated the quantity of emissions in South Asian economies.

In contrast to the aforementioned studies, ^{Vo and Zaman (2020)} used broad money as an indicator of financial development. They argued that there is a negative relationship between financial development and carbon emissions for 101 developing and developed countries. The method they used was a dynamic Panel Generalized Method of Moment (GMM), but they assumed the impact of financial development is linear.

Some studies such as ^{Zhang (2011)}, ^{Abbasi and Riaz (2016)}, and ^{Zeqiraj, Sohag, and Soytas (2020)} focused on the role of stock market development. In investigating the impact of stock market development on carbon emissions in China, ^{Zhang (2011)} used both stock market capitalization and stock market efficiency as indicators of financial development. The results of the study showed that the size of the stock market has a relatively larger impact on carbon emissions than the efficiency of the stock market. ^{Abbasi and Riaz (2016)} similarly focused on the impact of stock market development. As a result of their study in Pakistan, they found that financial development reduces emissions only during periods of high liberalization. ^{Zeqiraj, Sohag, and Soytas (2020)} analyze the impact of financial development on renewable energy emissions in the context of EU member states for the first time. They used the interaction term of financial development and renewable energy to test the nonlinear effects. As a result of the cross-sectional ARDL approach, they showed that stock market development reduces emissions by increasing renewable energy consumption.

Some recent studies have derived indices as indicators of financial development. For example, ^{Shahbaz et al. (2016)} constructed a composite index of Pakistan’s financial development using M2, M3, domestic credit, stock market capitalization, traded stock market value, and stock market turnover. Their results suggest that bank-based financial development is detrimental to the environment. Similarly, ^{Rahman et al. (2019)} constructed a composite index of Pakistan’s financial development using domestic credit, broad money, commercial bank assets and liquid liabilities. As a result of the ARDL bounds test, they concluded that this index has a moderating role for the environmental Kuznets curve. ^{Al Mamun et al. (2018)} examined the relationship between financial development and energy in a panel framework using a sample of 25 OECD countries (Organisation for Economic Cooperation and Development). They constructed an index by considering the stock market, credit market, and broad money supply. The results of the study show that financial development promotes cleaner energy in OECD countries. ^{Katircioglu and Taspinar (2017)} constructed an index for Turkey using domestic credit, broad money, commercial bank assets and liquid liabilities. Their results show no significant moderating effect of financial development on the relationship between energy consumption and emissions.

It is worth noting that some recent studies use the IMF’s financial development index. For example, ^{Saud et al. (2019)} examined the role of financial development in selected Central and Eastern Europe (CEE) countries on carbon emissions. They found that environmental quality decreased in six of the countries as financial development increased, while the opposite was true in four. Later, ^{Paramati, Mo, and Huang (2021)} examined the impact of financial development using the distinction between financial markets and financial institutions. As a result of the study which looked at OECD countries, it was found that financial development increased emissions. However, it ignored the nonlinear effects of financial development. In the case of G20 countries, ^{Habiba, Xinbang, and Ahmad (2021)} used the Durbin-Hausman panel cointegration approach to test the impact of financial development on emissions. They distinguished between stock market development and financial institution development. Their results show that stock market development was found to have a negative relationship with emissions in developed countries, but a positive relationship with emissions in developing countries. In addition, financial institution development has a positive relationship with emissions in developed countries, but no significant relationship with emissions in developing countries. Thus, the study highlights the importance of considering the subcomponents of financial development when examining environmental impacts. However, energy use was considered as an exogenous variable and its moderating role on the relationship between financial development and environmental quality was not considered. ^{He, Gao, and Wang (2022)} examined the impact of financial development on GHG emissions for 162 countries and found that the financial development positively affects the GHG emissions. They also tested the mediating effect of technical progress on the process and confirmed the existence of a dilution effect.

In another study using IMF indices, ^{Li et al. (2022)} focused on the impact of financial development in Brasil, Rusia, India, China, Sudáfrica (BRICS) countries and argued that financial development increases emissions. Although the study focused on developing countries and used dynamic regression, the nonlinear effects of financial development were once again ignored.

3. METHODOLOGY

^{Hansen (1999)} developed a threshold regression method for static panels with individual-specific fixed effects. The observations are divided into two or more regimes with different regression slopes. The single threshold model can be represented as follows:

Where yit,qit,xit,1≤i≤n,1≤t≤T is an observed panel data set, I(.) is the indicator function, q _it is the threshold variable with an estimated threshold λ, μ_i is the fixed effect, and e _it is the error term.

^{Chan (1993)} and ^{Hansen (1999)} argued that the value of the threshold can be estimated by minimizing of sum of squared errors:

In the final stage, the significance of the threshold effect is tested using the likelihood ratio test. The significance of the threshold effect is tested by testing the null hypothesis that the slope coefficients do not change between regimes.

Later, ^{Seo and Shin (2016)} proposed a GMM approach based on the first difference transformation, where both the threshold variable and the regressors can be endogenous. They showed that the estimator obtained by this approach follows a normal distribution asymptotically. Moreover, unlike the least squares approach, asymptotic normality holds true irrespective of whether the regression function is continuous or not. Thus, standard inferences based on Wald statistics are possible.

A single panel threshold model following ^{Seo and Shin (2016)} is stated as:

Where β1'=1,β1,1,β1,2⋯β1,k and β2'=1,β2,1,β2,2⋯β2,k are parameter vectors in different regimes, Xit=yit-1,yit-2,⋯yit-k,xit,xit-1⋯xit-k is the set of explanatory variables including the lagged term of the dependent variable, and I(.) is the identity function. We can also re-write the regression for single threshold as follows:

The model can be extended if there is more than one threshold variable. Following ^{Hansen (1999)} and ^{Seo and Shin (2016)}, we assumed that the threshold parameters are uniform.

4. EMPIRICAL ANALYSIS

5. CONCLUSION

In this study, the relationship between financial development and ghg emissions was tested by considering the moderating role of energy consumption. The sample which includes the period 2003-2019 and 31 developing countries, was analyzed with static and dynamic panel threshold regressions. It was found that the threshold effects of the financial development index, financial institutions development index, and financial market development index cannot be rejected at a 5% level of significance. However, the magnitude of the coefficients in the lower and upper regimes varies according to the financial development indicator used.

In countries where financial market development is above the threshold (0.53), a 1% increase in energy consumption increases greenhouse gas emissions by 0.16%. Singapore is an outlier with a level of financial market development above the threshold in all sample years. Therefore, the increase in financial market development in this country has triggered the level of emissions through energy consumption more than in other countries when all years are taken as a basis. On the other hand, in 17 of 31 countries (Algeria, Argentina, Bangladesh, Chile, China, Colombia, Ecuador, Indonesia, Iran, Israel, Mexico, Morocco, Oman, Pakistan, Peru, Sri Lanka, and Trinidad and Tobago), the financial market development index is below the threshold value (0.53) in all years. In these countries, a 1% increase in energy consumption increases greenhouse gas emissions by 0.069%.

The positive effects of financial development on environmental quality emerge through financial institutions. In fact, in developing countries, an important part of the financial system is constituted by these institutions instead of financial markets. In countries where the financial institutions development index is below the threshold value (0.36), 1% increase in energy consumption increases greenhouse gas emissions by 0.26%. On the other hand, in countries where the financial institution development index is above the threshold value (0.36), a 1% increase in energy consumption increases greenhouse gas emissions by only 0.09%. In 11 of 31 countries (Brazil, Chile, China, Israel, Kuwait, Malaysia, Oman, Singapore, South Africa, Thailand, and Trinidad and Tobago), the financial institution development index is above the threshold value in all years. Therefore, it can be argued that the development of financial institutions in these countries increases the use of clean energy resources.

Nevertheless, it is clear that the findings support that financial development has a positive effect on emissions. Therefore, it would not be right to expect a solution to emerge spontaneously through financial development without implementing active policies to reduce emissions in developing countries. For this reason, it is recommended that policy makers benefit from policies such as environmental taxes, standards, and emissions trading. In particular, environmental taxation will penalize polluters and contribute to the financing of budget deficits, which is a general financial problem in developing countries.