Keywords

JEL Classification

14.1 Introduction

Over the last decade, energy economics literature has provided augmented energy demand functions, which add a set of socio-economic variables onto the basic energy model. Financial development is one of the promising ones, which has been built by referring to the finance-growth theory. Even though the literature has not yet as expanded as energy-growth literature, the number of these studies has been steadily increasing.

Theoretical settings of finance-energy literature have basically inspired from finance-led growth hypothesis. Therefore, previous literature has addressed that financial development can affect energy demand via three channels according to the previous literature. Direct effect channel implies that consumers can find easier and cheaper borrowing opportunity as financial system develops in order to purchase durable goods which consume energy a lot. Business effect channel implies that business sector can also find opportunity to borrow easily and less costly as financial markets improve which in turn affect energy demand via investment and production decisions. Finally, wealth effect channel implies that increasing financial activities can affect economic agents’ confidence by creating a wealth effect which can promote economic activity and demand for energy. However, Sadorsky (2010) argues that energy demand might be irresponsive to financial development given the validity of growth-led finance hypothesis, and this relationship can only be resolved through empirical analyses.

Most of the previous studies looking into the impact of financial development on energy consumption have measured financial development using bank-related variables and as a matter of fact, indeed, investigated the impact of banking sector development (see, for example: Tang and Tan 2014; Zeren and Koc 2014; Aslan et al. 2014; Altay and Topcu 2015; among others). However, relatively little research has been done with the stock market variables on the impact that financial development has on energy consumption (see: Sadorsky 2011; Coban and Topcu 2013; Topcu and Altay 2017; Topcu and Payne 2017; Altay and Topcu 2017).

Unlike finance-energy literature, the number of studies in finance-environment literature is relatively limited (see, for example: Tamazian and Rao 2010; Jalil and Feridun 2011; Al-mulali and Sab 2012a, b; Lee 2013; Omri 2013; Ozturk and Acaravci 2013; Shahbaz 2013; Shahbaz et al. 2013a, b, c, d, 2016; Boutabba 2014; Al-mulali et al. 2015a, b; Mugableh 2015; Salahuddin et al. 2015; Ziaei 2015; Dogan and Seker 2016; Dogan and Turkekul 2016; Farhani and Ozturk 2015; Rafindadi 2016). Specifically, the impact of stock market on environment is an area which is almost untouched. To best of our knowledge, the literature includes only a few number of studies. For instance, Paramati et al. (2018) stock market indicators have a positive impact on carbon emissions in a panel of 20 developing countries, whereas the impact turns out negative in a panel of 23 developed countries. Likewise, Paramati et al. (2017)’s stock market development has a negative impact on emissions in developed G20 nations, whereas it has a positive impact in developing G20 nations. Lanoie et al. (1998) report for two developed nations (the US and Canada) that efficient capital markets spur environmental quality. Dasgupta et al. (2001) find for a four number of developing countries (Argentina, Chile, Mexico, and the Phillipines) that stock market development improves environmental performance. Tamazian et al. (2009) report that stock market development decreases carbon emissions in BRIC countries. Zhang (2011) finds for China that the stock market is not yet well-developed to significantly decrease carbon emissions. Abbasi and Riaz (2016) report that stock market development dramatically increases emissions in Pakistan. In the case Malaysia, Iatridis (2013) finds that carbon emissions increase with the stock market development.

Overall, the aforementioned literature roughly suggests that stock market development increases carbon emissions and deteriorates the environment. However, this evidence is not very robust as previous attempts include limited number of developing countries. Therefore, the aim of this study is to extend the analysis with the inclusion of more developing countries to provide better insights for policy makers. Given this motivation, this study intends to be the encompassing one in the literature. Unlike previous attempts, in addition, this study not only investigates the long-run relationship, but also provides short-run evidences.

Rest of the study is structured as follows: Sect. 14.2 describes model and data, Sect. 14.3 presents empirical approaches and findings, and finally, Sect. 14.4 discusses policy implications and gives concluding remarks.

14.2 Model and Data

As a proxy of environmental degradation, per capita carbon emissions (CO2) are described as a function of per capita energy consumption (e), per capita income (y), its squared term (ysq), and stock market development (s). The analysis includes 60 developing countries and is based on annual observations spanning from 1990 to 2014. Table 14.1 lists these countries.

Table 14.1 Sample of countries
Full size table

Environmental degradation is represented by carbon emissions measured by metric tons per capita, energy consumption is represented by energy use kg of oil equivalent per capita, income is represented by GDP per capita measured using constant 2010 US$, and stock market development is represented by stock market capitalization measured as a share of GDP. All data are sourced from World Bank World Development Indicators Database (2018), with the exception of the stock market data which is sourced from World Bank Global Financial Development Database (2018). To interpret the results in terms of elasticities, all variables are transformed into natural logarithms.

Table 14.2 presents the descriptive statistics of the data. When we review the mean values of the variables, carbon emissions are 14.72, energy consumption is quite closer to 7, income is 8.38, and stock market indicator is slightly less than 3. Notice that, the variable that has the highest standard derivation is the stock market development proxy, which is closely followed by carbon emissions.

Table 14.2 Descriptive statistics
Full size table

14.3 Methods and Findings

14.3.1 Unit Root Testing

Necessary precondition for implementing an Engle-Granger-based panel cointegration analysis is to provide that the variables in consideration are integrated of order one. Besides, prior to a panel ARDL estimation, it is necessary to ensure that the variables in interest are level or first-difference stationary. In this context, panel unit root tests developed by Im et al. (2003, IPS) were utilized, and findings were reported in Table 14.3. Accordingly, there seems no restriction for conducting the related analyses.

Table 14.3 Unit root results
Full size table

14.3.2 Cointegration

Since the variables in consideration are integrated of order one, this study employs an Engle-Granger-based panel cointegration analysis which was recently developed by Pedroni (1999, 2004) for the investigation of a possible cointegration relationship.

Within dimension tests

1.

Panel-ν stat:

\(Z_{\nu } = T^{2} N^{3/2} \left( {\sum\limits_{i = 1}^{N} {\sum\limits_{t = 1}^{T} {\hat{L}_{11i}^{ - 2} \hat{e}_{i,t - 1}^{2} } } } \right)^{ - 1}\)

2.

Panel-rho stat:

\(Z_{\rho } = T\sqrt N \left( {\sum\limits_{i = 1}^{N} {\sum\limits_{t = 1}^{T} {\hat{L}_{11i}^{ - 2} \hat{e}_{i,t - 1}^{2} } } } \right)^{ - 1} \sum\limits_{i = 1}^{N} {\sum\limits_{t = 1}^{T} {\hat{L}_{11i}^{ - 2} (\hat{e}_{i,t - 1} \Delta \hat{e}_{i,t} - \hat{\lambda }_{i} } )}\)

3.

Panel-pp stat:

\(Z_{t} = \left( {\hat{\sigma }_{N,T}^{2} \sum\limits_{i = 1}^{N} {\sum\limits_{t = 1}^{T} {\hat{L}_{11i}^{ - 2} \hat{e}_{i,t - 1}^{2} } } } \right)^{ - 1/2} \sum\limits_{i = 1}^{N} {\sum\limits_{t = 1}^{T} {\hat{L}_{11i}^{ - 2} (\hat{e}_{i,t - 1} \Delta \hat{e}_{i,t} - \hat{\lambda }_{i} } )}\)

4.

Panel-adf stat:

\(Z_{t}^{*} = \left( {\tilde{s}_{N,T}^{*2} \sum\limits_{i = 1}^{N} {\sum\limits_{t = 1}^{T} {\hat{L}_{11i}^{ - 2} \hat{e}_{i,t - 1}^{*2} } } } \right)^{ - 1/2} \sum\limits_{i = 1}^{N} {\sum\limits_{t = 1}^{T} {\hat{L}_{11i}^{ - 2} (\hat{e}_{i,t - 1}^{*} \Delta \hat{e}_{i,t}^{*} )} }\)

Between dimension tests

5.

Group-rho stat:

\(\tilde{Z}_{\rho } = TN^{ - 1/2} \sum\limits_{i = 1}^{N} {\left( {\sum\limits_{t = 1}^{T} {\hat{e}_{i,t - 1}^{2} } } \right)^{ - 1} } \sum\limits_{t = 1}^{T} {(\hat{e}_{i,t - 1} \Delta \hat{e}_{i,t} - \hat{\lambda }_{i} } )\)

6.

Group-pp stat:

\(\tilde{Z}_{t} = N^{ - 1/2} \sum\limits_{i = 1}^{N} {\left( {\hat{\sigma }_{i}^{2} \sum\limits_{t = 1}^{T} {\hat{e}_{i,t - 1}^{2} } } \right)^{ - 1/2} } \sum\limits_{t = 1}^{T} {(\hat{e}_{i,t - 1} \Delta \hat{e}_{i,t} - \hat{\lambda }_{i} } )\)

7.

Group-adf stat:

\(\tilde{Z}_{t}^{*} = N^{ - 1/2} \sum\limits_{i = 1}^{N} {\left( {\sum\limits_{t = 1}^{T} {\hat{s}_{i}^{*2} \hat{e}_{i,t - 1}^{*2} } } \right)^{ - 1/2} } \sum\limits_{t = 1}^{T} {\hat{e}_{i,t - 1}^{*} \Delta \hat{e}_{{_{i,t} }}^{*} }\)

Pedroni (1999, 2004) has proposed seven test statistics, which are shown above. These statistics assume that the variables are not level-stationary, and the cointegration vector is heterogeneous across the cross-section units. In this sense, the null of no cointegration was tested against the alternative hypothesis of cointegration by using the tests, four of which are termed as “panel statistics” and the others as “group statistics”. Findings presented in Table 14.4 support the cointegration relationship.

Table 14.4 Cointegration results
Full size table

14.3.3 Estimation

The present study employs a panel ARDL model for investigating the impact of financial development on environmental degradation. Our model incorporates with the pooled mean group estimator (PMG) that was developed by Pesaran et al. (1999). The considered model is formulated in the following manner:

$$\begin{aligned} \ln \,{\text{CO}}_{2} \,_{it} & = \alpha_{i} + \sum\limits_{j = 1}^{ki} {\beta_{ij} \ln {\text{CO}}_{2} \,_{i,t - j} } + \sum\limits_{j = 0}^{fi} {\delta_{ij} \ln {\text{FD}}_{i,t - j} } + \sum\limits_{j = 0}^{hi} {\phi_{ij} \ln {\text{EC}}_{i,t - j} } \\ & \quad + \sum\limits_{j = 0}^{ri} {\partial_{ij} \ln {\text{GDPPC}}_{i,t - j} } + \sum\limits_{j = 0}^{di} {\tau_{ij} \ln {\text{GDPPC}}2_{i,t - j} + } \varepsilon_{it} \\ \end{aligned}$$
(14.1)

In order to see the separate effects (i.e., short and long) of financial development on environmental degradation, Eq. (14.1) can be parameterized as follows:

$$\begin{aligned} \Delta \ln {\text{CO}}_{2} \,_{it} & = \alpha_{i} + \varpi_{i} \ln {\text{CO}}_{2} \,_{i,t - 1} + \delta_{i}^{*} \ln {\text{FD}}_{it} + \phi_{i}^{*} \ln {\text{EC}}_{it} + \partial_{i}^{*} \ln {\text{GDPPC}}_{it} \\ & \quad + \tau_{i}^{*} \ln {\text{GDPPC}}2_{it} + \sum\limits_{j = 1}^{ki - 1} {\beta_{ij}^{**} \Delta \ln {\text{CO}}_{2} \,_{i,t - j} } + \sum\limits_{j = 0}^{fi} {\delta_{ij}^{**} \Delta \ln {\text{FD}}_{i,t - j} } \\ & \quad + \sum\limits_{j = 0}^{hi} {\phi_{ij}^{**} \Delta \ln {\text{EC}}_{i,t - j} } + \sum\limits_{j = 0}^{ri} {\partial_{ij}^{**} \Delta \ln {\text{GDPPC}}_{i,t - j} } \\ & \quad + \sum\limits_{j = 0}^{di} {\tau_{ij}^{**} \Delta \ln {\text{GDPPC}}2_{i,t - j} + } \varepsilon_{it} \\ \end{aligned}$$
(14.2)

where ϖ represents error correction coefficient, the notations \(\delta^{*} ,\phi^{*} ,\partial^{*} ,\tau^{*}\) and \(\delta^{**} ,\phi^{**} ,\partial^{**} ,\tau^{**}\) illustrate the long- and short-run coefficients, respectively.

Pesaran et al. (1999) developed two estimators, namely the mean group (MG) and the pooled mean group (PMG) which both can be utilized to estimate Eq. (14.2). However, since the MG does not allow certain parameters to be distributed homogenously across cross-section units, this study utilizes the PMG for the estimation of Eq. (14.2).

As both pooling and averaging, the PMG estimator allows the intercepts, short-run coefficients, and error variances to differ freely across groups, but constraints the long-run coefficients to be the same (Pesaran et al. 1999). Because of initial conditions or some structural factors that have a possibility to influence all groups in a similar way, utilizing the PMG estimator seems to be appropriate for the considered purpose.

According to findings illustrated in Table 14.5, the model that we try to solve has a stable equilibrium. It is proved by negative and statistically significant error correction coefficient. Besides, estimates reveal that stock market development decreases environmental degradation in the short-run, while the impact turns to positive in the long-run. As expected, energy consumption is the major factor that raises carbon dioxide emissions either in the long or in the short-run. Finally, carbon dioxide emission is positively affected by per capita income in the long-run, whereas the link is statistically insignificant in the short-run. Despite of the desired signs, the environmental Kuznets curve hypothesis is not satisfied given the insignificant coefficients provided either from the short- or long-run estimations.

Table 14.5 Panel ARDL estimation results
Full size table

14.4 Conclusion

In recent years, the number of studies that investigate the impact of global warming and climate change on environmental quality has increased. A great number of these studies have employed urbanization, financial development, energy consumption, and trade into the function. The results of these studies, however, are volatile across the income level of related countries.

Unlike previous studies in the area, this study considers the simultaneous use of energy consumption, income, and stock market development in order to estimate the separate impact of each (i.e., short and long) variable on environmental degradation. For this purpose, panel ARDL model is utilized for 60 developing countries over the period 1990–2014. Due to the production structure of developing countries, the relationship between economic activity and environmental degradation could be nonlinear, which has been called the environmental Kuznets hypothesis. The environmental Kuznets curve hypothesis argues that in the initial stages of development, environmental degradation raises and then it decreases as incomes increase.

The results of this study show that stock market development decreases environmental degradation in the short-run; however, environmental degradation rises with stock market development in the long-run. As discussed earlier, the existing literature documents that stock market development positively affects carbon emissions in developing countries. This is very consistent with the long-run results of this study. However, the short-run results of this study show that stock market development is not harmful to the environment in the short-run. This split reveals that the impact of stock market development on environment can vary over time. This is probably due to the underdeveloped capital market structure of developing countries. Because developing countries are not better able to transform stock market development into the production, this development does not strictly lead to an environmental degradation in the short-run.