ARDL Bound Testing Approach for a Green Low-Carbon Circular Economy in Turkey

Abstract

This study analyzes Turkey’s development toward a green economy between 1990 and 2022 within the framework of certain green economic indicators. The data consist of secondary data from the official databases of the World Bank and the Turkish Statistical Institute (TURKSTAT). In the study, the total amount of carbon emissions was chosen as an indicator of green growth, while gross domestic product per capita (GDP) represents economic growth, domestic loans granted by banks to the private sector (as a percentage of GDP) and foreign direct investment represent financial development, and electricity generation represents pollution. To determine whether the variables are cointegrated and to determine the direction and strength of the relationship between the variables, the ARDL bounds test and the FMOLS and DOLS long-run estimators were used. Finally, Toda Yamamoto (TY)–Granger tests were performed to determine causality. The long-term relationship between the variables was confirmed by the results of the ARDL bounds test. The error correction coefficient (CointEq(−1)) was estimated to be statistically significant and negative (−0.757) when the short-term analysis was performed. This result shows that the short-term imbalances will be corrected in less than a year, and the system will approach the long-term equilibrium. In the long-term analysis of the model, all variables selected to explain the dependent variable were found to have a statistically significant impact on the dependent variable. The GDP per capita variable, the indicator of economic growth, has a negative effect on the dependent variable, while the other independent variables have a positive effect. The results of the causality analysis indicate that the dependent variable carbon emissions (CO₂) has a unidirectional causality relationship with domestic credit provided to the private sector by banks (DC), which represents financial development, and with total electricity production (EP), which serves as an indicator of pollutants.

1. Introduction

Sustainable development is a paradigm shift toward an integrated approach to development that aims to strike a balance between economic growth, environmental protection and social justice [1], even if it remains a topic of debate in academia. The findings from this historical perspective are of great importance for policy-making processes as the global community strives for a sustainable future [2,3]. The effectiveness of environmental policy in ensuring sustainable development has emerged as one of the most important issues of our time [4]. Ensuring economic growth while improving social equity and preventing environmental degradation are key factors for policy success [5,6]. Efforts to strike a balance between the demand for “more development” and the need to “protect the environment” continue to be a central theme of research [7].

The balance between economic growth and environmental sustainability depends on the success of global environmental policies [8,9,10,11]. These policies have a significant impact on green trade, address global challenges and shape economic perceptions, underlining the complexity of achieving sustainable development in a globalized world [12]. As countries face burning issues such as economic inequalities, climate change and environmental degradation, environmental policies are becoming increasingly important for securing a sustainable future [13]. Findings from recent research provide valuable guidance for the development of policies that balance economic goals with environmental sustainability, serving policy makers, stakeholders and researchers [14]. In the face of global environmental challenges, the urgency for integrated policymaking has never been greater [15,16]. Ensuring sustainable development is highly complex [17,18].

The green economy is a system that prioritizes nature and the environment, promotes sustainable and balanced production, consumption and exchange processes and strives for an equal distribution of economic and social benefits [19]. At the same time, it represents a transition from the traditional industrial approach to development to an inclusive development paradigm that focuses on sustainability and restoring the ecological balance that has been disrupted.

The global green economy should not only aim to increase GDP through green motives, but it must also make the necessary structural changes [20]. Cato et al. [21] stated that the literature on green economy focuses on various issues such as green economy and green growth and that economic growth should not only be assessed in terms of GDP but should also include an assessment of welfare and environmental quality. Studies on this topic suggest that the development of strategies that consider all relevant factors is limited [22].

Governments can use various incentives to promote sustainable economic and social growth, such as green procurement, encouraging the consumption of goods and services produced for the benefit of the environment, and green taxes [23,24]. Green jobs can be described as jobs that are committed to reducing the consumption of natural resources and energy, reducing the amount of waste generated by industry and households and reducing pollution and greenhouse gas emissions. These jobs also focus on preserving ecosystems and protecting biodiversity [25,26]. Creating a sustainable green economy is the responsibility of states, international organizations, public authorities, local and regional communities, businesses, households and citizens. While all these concepts are well defined, the main question is to what extent they are implemented economically or socially. Not every concept can offer universal solutions; some may have shortcomings or be inapplicable under certain conditions.

This study assesses Turkey’s development toward a green economy from 1990 to 2022 using selected green economic indicators. The relationships between green growth, ecological development, financial development and pollutants are analyzed.

The Contradiction Between Environmental Protection and Economic Growth

In the current literature, the relationship between environmental protection and economic growth is usually seen as a conflicting set of goals. However, recent studies suggest that sustainable development policies can bridge this gap and offer opportunities to achieve economic growth without jeopardizing the environment. Alnsour et al. [27] analyzed carbon dioxide emissions and their impact on economic indicators in Jordan. It was found that these indicators have a positive effect on CO₂ emissions in both the short term (SR) and the long term (LR) and lead to a deterioration of the environmental situation. Zhang et al. [28] investigated the interactions between economic growth and environmental degradation in the context of China’s sustainable development policy. Using a vector autoregression model, the authors established a relationship between GDP growth in Henan Province and the increase in environmental pollutants. However, they also pointed out that effective environmental regulations could mitigate these effects and emphasized that sustainable economic practices and environmental protection can coexist. Dogaru [29] examined the benefits offered by green economic initiatives and green growth strategies and emphasized their role in promoting sustainable development. These strategies serve as practical tools for achieving sustainable development goals by reducing the environmental impact of economic activities. Khan et al. [30] examined this relationship in Pakistan and concluded that while tourism and economic growth contribute to an increase in CO₂ emissions, integrated policy approaches that incorporate economic, environmental and energy considerations can promote sustainable tourism practices. This study highlights that increasing economic competitiveness while minimizing pollution can help achieve broader sustainable development goals. Ali et al. [31] showed that economic policy uncertainty and financial development can increase CO₂ emissions, while environmentally friendly technological innovations and the quality of institutions play a crucial role in reducing environmental pollution. This study emphasizes that the integration of green technological innovations and sound institutional frameworks with environmental policies is of great importance for achieving sustainability. Idris et al. [32] investigated the opportunities for a green economy in the Middle East and North Africa (MENA) region by examining key indicators: access to clean fuels, GDP and CO₂ emissions between 2000 and 2018. The study used the ARDL bounds test with seasonally adjusted data to determine the cointegration between the indicators used. The researchers concluded that access to clean fuels has a positive effect on economic growth, while CO₂ emissions are negatively impacted. Saleem et al. [33] examined the relationship between GDP, domestic credit to the private sector, renewable energy consumption, environmental taxes, green growth and carbon emissions for 12 countries in Asia over the period 1990–2018. The results suggest that green growth, renewable energy consumption, clean technology and innovation, environmental taxes and domestic credit reduce carbon emissions.

Studies that analyze green economic growth using time-series analysis often use environmental factors (such as carbon emissions, ecological footprint and natural resource depletion), economic factors (such as GDP as an indicator of economic growth, domestic credit to private firms as a measure of financial development, foreign direct investment and trade openness) and pollutants (such as data on energy production or consumption). The number of studies using carbon emissions to represent green economic growth is quite high. GDP is usually used as a benchmark for assessing the economy and economic growth. When it comes to defining economic and social development, many different factors are used to show the characteristics of government policies. In the transition to a green economy, other environmental indicators are now also used, and economic security factors are used alongside prosperity to determine green growth [34]. When evaluating green growth and transitioning an economy to a green economy, it is important to include new or alternative criteria. This would help researchers understand the conditions in the study area in terms of economic, social and environmental conditions and their relationship to green growth. These criteria are particularly important in the implementation of a new Green Deal that facilitates the transformational processes required to achieve sustainable development goals. The selection of factors is crucial to recognizing whether or not an economy is making progress toward sustainable development and a new green economic deal. Therefore, the following points are important to achieve the goals for a successful transition to a green economy: setting a timeline for achieving these goals, determining the factors to be considered and how these factors should be measured and finding data to help develop relevant strategies and policies.

The aim of this study is to evaluate Turkey’s development toward a green economy between 1990 and 2022 within the framework of selected green economic indicators. For this purpose, the long-term and causal relationships between the variables of green growth (total carbon emissions), economic growth (GDP per capita in USD), financial development (domestic lending to the private sector as a percentage of GDP), foreign direct investment (as a percentage of GDP) and total electricity generation (GWh) as pollutants were analyzed.

2. Materials and Methods

2.1. Materials

Equation (1) was formed in the study to analyze Turkey’s development toward a green economy within the framework of selected green economic indicators. In the construction of the model, variables that have been empirically proven to influence the green economy in previous studies and variables that have the potential to influence carbon emissions were considered. In the study, carbon emissions were chosen as the dependent variable, and the independent variables consisted of GDP per capita, foreign direct investment, the amount of domestic credit extended by banks to the private sector and total annual electricity production.

$${CO}_2={\beta }_0+{\beta }_1{EG}_t+{\beta }_2{DC}_t+{\beta }_3{FDI}_t+{{\beta }_{4}EP}_t+{\epsilon }_t$$

(1)

Here, CO₂ represents total carbon emissions (MI), EG represents GDP per capita (USD), DC represents domestic bank lending to the private sector (% GDP), FDI represents foreign direct investment (% GDP), EP represents total electricity generation (GWh), and ε_t represents the error term. Annual data were used for these variables between 1990 and 2022. Natural logarithms were used to analyze the relationships between the variables in terms of elasticity. The descriptive statistics can be found in

Table 1. Descriptive statistics.

	lnCO2	lnEG	lnDC	lnFDI	lnEP
Mean	5.618320	8.728089	3.394403	−0.038408	11.98654
Median	5.651189	8.987674	3.269590	0.226965	12.07994
Maximum	6.120843	9.439719	4.261222	1.287408	12.72106
Minimum	5.037866	7.714503	2.639818	−1.186175	10.96029
S.D.	0.339957	0.585697	0.584711	0.725447	0.537133
Ske.	−0.139722	−0.320274	0.208735	−0.032781	−0.358882
Kur.	1.696908	1.438339	1.406305	1.710706	1.937331
Jarque–Bera	2.442190	3.917497	3.731950	2.291544	2.261120
Sum	185.4046	288.0269	112.0153	−1.267471	395.5559
Sum Sq. Dev.	3.698255	10.97729	10.94037	16.84074	9.232390
Observations	33	33	33	33	33

, and their graphs can also be seen in Figure 1.

2.2. Method

2.2.1. Unit Root Tests

In order to analyze Turkey’s transition to a green economy in the context of selected green economic indicators, the ARDL bounds test approach, DOLS, FMOLS and the Granger causality test, which are widely used methods of time-series analysis in the literature, were applied. In the cointegration analysis, the unit roots of the variables and their degrees of stationarity must be determined, as the cointegration analysis includes assumptions about the degrees of stationarity of the series. If the time series of the variables in question remain statistically the same over time, this means that they are stationary. In other words, the basic statistics of the variables, such as mean, variance and covariance, remain constant. Therefore, it is crucial for the study that the series is stationary. The variables in this study were tested with ADF and PP to determine their stationarity. In both the ADF and PP tests, H₀ means that the variable has a unit root and is non-stationary, while H₁ says the opposite. The ADF unit root test requires the choice of a lag length, so AIC was chosen to determine the lag length.

2.2.2. ARDL Bounds Test

The ARDL approach is a method for determining cointegration between variables and has several significant advantages over alternative cointegration approaches. ARDL can provide robust results in cases where the sample size is considered small [35,36], the variables can be stationary at different levels, unlike many other cointegration techniques [37], and different lag lengths can be used for different variables [38]. Pesaran et al. [39] proposed different conditional error correction models that differ depending on the inclusion of the constants and trend terms in the ARDL model. For this study, Case III was chosen which includes an unconstrained constant and a trend. The model for the ARDL test, which was determined using Case III in Equation (1), was modified as follows:

$$\begin{array}{c}{∆\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_{2\mathrm{t}}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1{\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_{2\mathrm{t}-1}+{\mathrm{\beta }}_2{\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}}_{\mathrm{t}-1}+{\mathrm{\beta }}_3{\mathrm{l}\mathrm{n}\mathrm{D}\mathrm{C}}_{\mathrm{t}-1}+{\mathrm{\beta }}_4{\mathrm{l}\mathrm{n}\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{t}-1}+{\mathrm{\beta }}_5{\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{P}}_{\mathrm{t}-1}+{\sum }_{\mathrm{i}=1}^{\mathrm{p}}{\mathrm{\beta }}_{6\mathrm{i}}{∆\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}}_{\mathrm{t}-\mathrm{i}}+\\ {\sum }_{\mathrm{i}=0}^{{\mathrm{q}}_1}{\mathrm{\beta }}_{7\mathrm{i}}{∆\mathrm{l}\mathrm{n}\mathrm{D}\mathrm{C}}_{\mathrm{t}-\mathrm{i}}+{\sum }_{\mathrm{i}=0}^{{\mathrm{q}}_2}{\mathrm{\beta }}_{8\mathrm{i}}{∆\mathrm{l}\mathrm{n}\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{t}-\mathrm{i}}+{\sum }_{\mathrm{i}=0}^{{\mathrm{q}}_3}{\mathrm{\beta }}_{9\mathrm{i}}{∆\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{P}}_{\mathrm{t}-\mathrm{i}}+{\mathrm{\epsilon }}_{\mathrm{t}}\end{array}$$

(2)

In the study, the F-bound test was applied to examine whether cointegration exists between the variables used in Equation (2). The null hypothesis (H₀) states that there is no long-run relationship among the selected variables (H₀: β₁ = β₂ = β₃ = β₄ = β₅ = 0). To reject the H₀ and confirm the presence of a long-run relationship, the F-statistic must exceed the upper bound of critical values [40]. Narayan [35] reported the critical values used in the study for the F-bound test which is known to produce reliable results even in cases where the sample size is relatively small.

Since the study aims to estimate both short-run and long-run relationships, the error correction model (ECM) was employed. Using the ECM, Equation (1) was reformulated as Equation (3):

$$\begin{array}{c}{∆\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_{2\mathrm{t}}={\mathrm{\beta }}_0+{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{p}}}{\mathrm{\beta }}_{1\mathrm{i}}{∆\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_{2\mathrm{t}-\mathrm{i}}+{\displaystyle \sum _{\mathrm{i}=0}^{{\mathrm{q}}_1}}{\mathrm{\beta }}_{2\mathrm{i}}{∆\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}}_{\mathrm{t}-\mathrm{i}}+{\displaystyle \sum _{\mathrm{i}=0}^{{\mathrm{q}}_2}}{\mathrm{\beta }}_{3\mathrm{i}}{∆\mathrm{l}\mathrm{n}\mathrm{D}\mathrm{C}}_{\mathrm{t}-\mathrm{i}}+{\displaystyle \sum _{\mathrm{i}=0}^{{\mathrm{q}}_3}}{\mathrm{\beta }}_{4\mathrm{i}}{∆\mathrm{l}\mathrm{n}\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{t}-\mathrm{i}}+{\displaystyle \sum _{\mathrm{i}=0}^{{\mathrm{q}}_4}}{\mathrm{\beta }}_{5\mathrm{i}}{∆\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{P}}_{\mathrm{t}-\mathrm{i}}\\ +\mathrm{\omega }{\mathrm{E}\mathrm{C}\mathrm{T}}_{\mathrm{t}-1}+{\mathrm{\epsilon }}_{\mathrm{t}}\hfill \end{array}$$

(3)

Here, the ECT shows whether the short-term deviations are approaching the long-term equilibrium. If the term is negative, this condition is fulfilled.

2.2.3. Causality Test

The causality analysis is important because it shows both the short-term relationships and the direction of causality, indicating possible feedback effects. A causality test was also conducted in this study to determine these relationships. Numerous studies in this area use the causality test by Granger causality test [41], which was the first to introduce causality into econometric analysis. The TY [42] test allows the series to have different stationarities. In this approach based on VAR (vector autoregression), the maximum degree of stationarity (dmax) is first determined for the variables, followed by the selection of the lag length (k) for the VAR model. The variables are then included in the model with a lag equal to k + dmax. The Wald test is used to estimate causality in the model. Equation (4) shows the TY [42] test used in this study.

$$\begin{array}{c}\left[\begin{array}{c}{\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_2\\ \mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}\\ \mathrm{l}\mathrm{n}\mathrm{D}\mathrm{C}\\ \mathrm{l}\mathrm{n}\mathrm{F}\mathrm{D}\mathrm{I}\\ \mathrm{l}\mathrm{n}\mathrm{E}\mathrm{P}\end{array}\right]=\left[\begin{array}{c}\mathrm{\alpha }\\ \mathrm{\beta }\\ \mathrm{\delta }\\ \mathrm{\phi }\\ \mathrm{ԛ}\end{array}\right]+{\displaystyle \sum _{\mathit{i}=1}^{\mathit{k}}}\left[\begin{array}{cccccc}{\mathrm{b}}_{11\mathrm{i}}& {\mathrm{b}}_{12\mathrm{i}}& {\mathrm{b}}_{13\mathrm{i}}& {\mathrm{b}}_{14\mathrm{i}}& {\mathrm{b}}_{15\mathrm{i}}& {\mathrm{b}}_{16\mathrm{i}}\\ {\mathrm{b}}_{21\mathrm{i}}& {\mathrm{b}}_{22\mathrm{i}}& {\mathrm{b}}_{23\mathrm{i}}& {\mathrm{b}}_{24\mathrm{i}}& {\mathrm{b}}_{25\mathrm{i}}& {\mathrm{b}}_{26\mathrm{i}}\\ {\mathrm{b}}_{31\mathrm{i}}& {\mathrm{b}}_{32\mathrm{i}}& {\mathrm{b}}_{33\mathrm{i}}& {\mathrm{b}}_{34\mathrm{i}}& {\mathrm{b}}_{35\mathrm{i}}& {\mathrm{b}}_{36\mathrm{i}}\\ {\mathrm{b}}_{41\mathrm{i}}& {\mathrm{b}}_{42\mathrm{i}}& {\mathrm{b}}_{43\mathrm{i}}& {\mathrm{b}}_{44\mathrm{i}}& {\mathrm{b}}_{45\mathrm{i}}& {\mathrm{b}}_{46\mathrm{i}}\\ {\mathrm{b}}_{51\mathrm{i}}& {\mathrm{b}}_{52\mathrm{i}}& {\mathrm{b}}_{53\mathrm{i}}& {\mathrm{b}}_{54\mathrm{i}}& {\mathrm{b}}_{55\mathrm{i}}& {\mathrm{b}}_{56\mathrm{i}}\end{array}\right]\times \left[\begin{array}{c}{\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_{2\mathrm{t}-\mathrm{i}}\\ {\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}}_{\mathrm{t}-\mathrm{i}}\\ {\mathrm{l}\mathrm{n}\mathrm{D}\mathrm{C}}_{\mathrm{t}-\mathrm{i}}\\ {\mathrm{l}\mathrm{n}\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{t}-\mathrm{i}}\\ {\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{P}}_{\mathrm{t}-\mathrm{i}}\end{array}\right]\hfill \\ \hfill +{\displaystyle \sum _{\mathrm{j}=\mathrm{k}+1}^{{\mathrm{d}}_{\mathrm{m}\mathrm{a}\mathrm{x}}}}\left[\begin{array}{cccccc}{\mathrm{b}}_{11\mathrm{j}}& {\mathrm{b}}_{12\mathrm{j}}& {\mathrm{b}}_{13\mathrm{j}}& {\mathrm{b}}_{14\mathrm{j}}& {\mathrm{b}}_{15\mathrm{j}}& {\mathrm{b}}_{16\mathrm{j}}\\ {\mathrm{b}}_{21\mathrm{j}}& {\mathrm{b}}_{22\mathrm{j}}& {\mathrm{b}}_{23\mathrm{j}}& {\mathrm{b}}_{24\mathrm{j}}& {\mathrm{b}}_{25\mathrm{j}}& {\mathrm{b}}_{26\mathrm{j}}\\ {\mathrm{b}}_{31\mathrm{j}}& {\mathrm{b}}_{32\mathrm{j}}& {\mathrm{b}}_{33\mathrm{j}}& {\mathrm{b}}_{34\mathrm{j}}& {\mathrm{b}}_{35\mathrm{j}}& {\mathrm{b}}_{36\mathrm{j}}\\ {\mathrm{b}}_{41\mathrm{j}}& {\mathrm{b}}_{42\mathrm{j}}& {\mathrm{b}}_{43\mathrm{j}}& {\mathrm{b}}_{44\mathrm{j}}& {\mathrm{b}}_{45\mathrm{j}}& {\mathrm{b}}_{46\mathrm{j}}\\ {\mathrm{b}}_{51\mathrm{j}}& {\mathrm{b}}_{52\mathrm{j}}& {\mathrm{b}}_{53\mathrm{j}}& {\mathrm{b}}_{54\mathrm{j}}& {\mathrm{b}}_{55\mathrm{j}}& {\mathrm{b}}_{56\mathrm{j}}\end{array}\right]\times \left[\begin{array}{c}{\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_{2\mathrm{t}-\mathrm{j}}\\ {\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}}_{\mathrm{t}-\mathrm{j}}\\ {\mathrm{l}\mathrm{n}\mathrm{D}\mathrm{C}}_{\mathrm{t}-\mathrm{j}}\\ {\mathrm{l}\mathrm{n}\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{t}-\mathrm{j}}\\ {\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{P}}_{\mathrm{t}-\mathrm{j}}\end{array}\right]+\left[\begin{array}{c}{\mathrm{\epsilon }}_{1\mathrm{t}}\\ {\mathrm{\epsilon }}_{2\mathrm{t}}\\ {\mathrm{\epsilon }}_{3\mathrm{t}}\\ {\mathrm{\epsilon }}_{4\mathrm{t}}\\ {\mathrm{\epsilon }}_{5\mathrm{t}}\end{array}\right]\end{array}$$

(4)

3. Results

3.1. Unit Root Test and Unit Root with Break Test

The results of the ADF and PP unit root tests can be found in

Table 2. ADF and PP unit root test results (constant).

Variables	ADF		PP
	Level	1st Difference	Level	1st Difference
lnCO2	−1.077	−6.045 ***	−1.568	−7.123 ***
lnEG	−1.451	−5.865 ***	−1.072	−5.877 ***
lnDC	0.568	−3.618 ***	1.00	−3.668 ***
lnFDI	−2.090 **	−6.0498 ***	−1.969 **	−12.076 ***
lnEP	−3.146 **	−4.205 ***	−8.415 ***	−4.225 ***

According to the Schwarz information criterion (SIC), the maximum lag length was chosen as 8 for ADF. For the PP test “Barlett Kernel” and “Newey West Bandwidth” methods were used for the kernel method. The p value is presented in parentheses. *** significance at the 1% level, ** significance at the 5% level.

. Both the ADF and PP tests show that all variables are stationary at level I(1). The assumption that the variables are stationary at I(0) or I(1) was fulfilled, and the data set was found to be suitable for ARDL.

The Break Test results are presented in

Table 3. Unit root with Break Test result.

Variables	Schwarz Information Criterion (SIC)			F-Statistic
	Level	1st Difference	Break Date	Level	1st Difference	Break Date
lnCO2		−6.514 ***	2019		−6.463 ***	2007
lnEG		−6.248 ***	2001		−6.248 ***	2001
lnDC		−6.064 ***	2021	−6.383 ***		2005
lnFDI	−4.937 ***		2004	−4.937 ***		2004
lnEP		−5.527 ***	2009		−5.527 ***	2009

According to the Schwarz information criterion (SIC) and F-statistic, the maximum lag length was chosen as 8 for the Breakpoint Unit Root Test. The p value is presented in parentheses. *** significance at the 1% level.

. Structural breaks in the data have no impact on the results of this research.

3.2. ARDL Bounds Test Results

Next, the lag length for the ARDL model must be determined in the analysis. For this purpose, a VAR model is used, considering the values of AIC, SIC and HQC (Hannan–Quinn Information Criterion). The lag length is chosen based on the one that gives the lowest critical value (

Table 4. Lag length test results.

Lag	LogL	LR	FPE	AIC	SIC	HQ
0	33.254	NA	1.05 × 10−7	−1.883	−1.650	−1.808
1	173.695	224.705 *	4.89 × 10−11 *	−9.579	−8.178 *	−9.131 *
2	195.7120	27.887	6.94 × 10−11	−9.380	−6.811	−8.558
3	223.871	26.281	8.69 × 10−11	−9.591 *	−5.854	−8396

* The lowest value that shows the most appropriate lag length for the relevant criterion.

). The model created with the selected lag length should not have any problems with the automatic selection.

The maximum lag length was set at 3, as the study does not have a large number of observations. Diagnostic tests were performed, and stability was taken into account when selecting the most appropriate model. Figure 2 lists 20 models that fulfill the minimum requirement of the AIC. Figure 2 lists 20 models that meet the minimum AIC requirement.

When determining the appropriate lag length for model estimation, the model with the lowest AIC value was selected. This model was selected from the models that met the following criteria: normal distribution, no heteroscedasticity, no autocorrelation and stability. The results for the ARDL model (3, 3, 3, 2, 3) which met these criteria are shown in

Table 5. ARDL bounds test.

Cointegration	Significance	F-Value	F-Bounds Test		t-Value	T-Bounds Test
			I(0)	I(1)		I(0)	I(1)
Yes	6.344 ***	10%	2.45	3.52	−6.576 ***	−2.57	−3.66
		5%	2.86	4.01		−2.86	−3.99
		1%	3.74	5.06		−3.43	−4.60

*** significant at 1%.

.

When examining the results of the F-bound test, the calculated F-value exceeds the upper critical threshold values at the 1% significance level (Table 4). Thus, the null hypothesis is rejected, and the hypothesis indicating a cointegration relationship is accepted. Since an unrestricted conditional ECM (case III) was used, the t-bound test was also applied. As the results for this test were also statistically significant at the 1% level, the cointegration between the variables was confirmed. As cointegration was demonstrated, the SR and LR coefficients were estimated (

Table 6. ARDL Cointegration Long and Short Run Coefficients.

LR Analysis
Var.	Coef.	T	Prob.
lnEG	−0.054 ***	−2.245	0.046
lnDC	0.114 ***	5.595	0.000
lnFDI	0.041 ***	2.833	0.016
lnEP	0.552 ***	17.657	0.000
SR Analysis
Var.	Coef.	T	Prob.
C ***	−1.285	−6.511	0.000
D(lnCO2(−1)) *	0.309	2.112	0.058
D(lnCO2(−2)) *	0.269	1.879	0.073
D(lnEG)	−0.039	−1.655	0.126
D(lnEG(−1)) *	0.046	1.946	0.077
D(lnEG(−2)) ***	0.103	4.065	0.001
D(lnDC)	0.027	1.190	0.258
D(lnDC(−1))	−0.038	−1.164	0.268
D(lnDC(−2)) ***	−0.127	−4.285	0.001
D(lnFDI)	0.002	0.340	0.739
D(lnFDI(−1)) **	−0.021	−2.489	0.030
D(lnEP) ***	0.774	7.023	0.000
D(lnEP(−1))	0.157	1.063	0.310
D(lnEP(−2)) ***	−0.470	−3.469	0.005
CointEq(−2) ***	−1.380 ***	−6.576	0.000
Sensitivity analysis
R2	0.948
Adjusted R2	0.900
F statistic	19.755
Prob (F statistic) ***	0.000000

*** significant at 1%, ** significant at 5%, * significant at 10%.

).

When examining the results of the short-run estimate, it is found that the error correction coefficient (CointEq(-1)) is negative and statistically significant. Therefore, it is expected that the model will converge to equilibrium in the long run. The results of the long-run estimation also show that all variables in the model have a statistically significant influence on the dependent variable in the LR.

3.3. Diagnostic Tests

Diagnostic tests were performed for this study to determine the functionality of the model. The test statistics for autocorrelation, normality, heteroscedasticity and model specification errors were acceptable for the predicted model (

Table 7. ARDL diagnostic test results.

Test	F-Stat	Prob.	Result
Breusch–Godfrey serial correlation LM test	0.454	0.089	No problem of serial correlations
Breusch–Pagan–Godfrey heteroscedasticity test	1.523	1.000	No problem of serial correlations
Jarque–Bera test	0.308	0.856	Estimated residual is normal
Ramsey test	0.008	0.770	Model is specified correctly

).

The CUSUM and CUSUMSQ tests were chosen to assess the structural stability of the estimated long-run model. The CUSUM (Cumulative Sum) test plots for the ARDL model are shown in Figure 3a, while Figure 3b shows the CUSUMSQ (Cumulative Sum of Squares) test plots. Since the CUSUM and CUSUMSQ values remain within the 5% critical value, it is assumed that the coefficients are stable in the long run and that there are no structural breaks in the model.

3.4. FMOLS and DOLS Test Results

The study examined FMOLS and DOLS estimators to improve and validate the reliability of the ARDL results. These estimators identify long-run relationships between variables and interpret coefficients while addressing endogeneity issues and providing reliable results in small samples. FMOLS and DOLS are widely used methods that complement each other.

The DOLS estimator, first introduced by Stock and Watson [43], has been further refined in subsequent studies [44,45]. This method can help to estimate a long-run constant relationship when there is cointegration between economic variables.

The FMOLS method addresses diagnostic problems associated with standard estimators. In addition, FMOLS provides accurate and unbiased results even with small sample sizes [46,47].

FMOLS and DOLS tests were used to evaluate the accuracy and power of the ARDL boundary test results.

Table 8. DOLS test.

Var.	Coef.	S.E.	t	Prob.
EG	−0.093 ***	0.026	−3.483	0.003
DC	0.172 ***	0.014	11.765	0.000
FDI	0.076 ***	0.018	4.239	0.000
EP	0.489 ***	0.020	24.406	0.000

*** significance level of 1%.

shows the results of the DOLS test, while

Table 9. FMOLS test.

Var.	Coef.	S.E.	t	Prob.
EG	−0.412	0.027	−1.509	0.142
DC	0.150 ***	0.013	10.870	0.000
FDI	0.080 ***	0.010	7.802	0.000
EP	0.456 ***	0.019	23.686	0.000

*** significance level of 1%.

shows the results of the FMOLS test.

The coefficient, sign and significance levels resulting from the ARDL bounds test and from the FMOLS and DOLS tests for the indicator variables of the green economy are remarkably similar.

The economic growth (EG) variable has a negative effect on carbon emissions (CO₂), which represent green growth, in all models. In the DOLS models, this negative effect is statistically significant at the 1% level, while it is not significant in the FMOLS model. All other variables have a positive effect on carbon emissions (CO₂) in all models and are statistically significant at the 1% level. The DOLS and FMOLS test results for green economic indicators are consistent with the ARDL results.

3.5. Granger Causality Test

The TY test was used to determine the influence of the variables on each other (

Table 10. Granger causality analysis.

Variables	lnCO2	lnEG	lnDC	lnFDI	lnEP
lnCO2	-	-	20.155 (0.000) ***	-	19.566 (0.000) ***
LnEG	-	-	-	-	-
LnDC	-	-	-	-	-
LnFDI	-	1.014 (0.000) ***	5.635 (0.059) *	-	-
lnEP	-	-	6.876 (0.032) **	-	-

***, ** and * show the rejected H₀ hypothesis at 1%, 5% and 10% significance levels. (p) values can be found in parentheses. The most appropriate lag length was chosen as 2 according to Akaike information criteria (AIC).

).

The evaluation of the results in Table 10 shows that CO₂ (dependent variable) has a unidirectional causal relationship with the variables DC and EP. It is found that ADI has a unidirectional causality relationship with EG and DC. Finally, it is found that there is a unidirectional causal relationship between EP and DC.

4. Discussion

Carbon emissions are significantly influenced by economic growth and development [48]. However, previous research does not show complete consensus on the direction of the relationship between economic growth and carbon emissions. The results of the long-run estimation of the ARDL model show that economic growth (EG) has a negative relationship with carbon emissions (CO₂). This negative effect is statistically significant at the 5% level and shows that a 1% increase in economic growth (EG) leads to a 0.05% decrease in carbon emissions (CO₂). These results are consistent with some studies [49,50,51,52,53,54,55,56,57] but contradict others [33,58,59,60,61,62]. This discrepancy may be due to the shift in Turkey’s economic structure from a carbon-intensive heavy industry to a low-carbon service sector in recent years.

Increased consumption drives production, while potential investment opportunities boost employment and energy consumption, potentially exacerbating environmental degradation. There are conflicting views on how financial development affects carbon emissions. Some scholars argue that more accessible and affordable financial instruments can increase investment and consumption, which has a negative impact on the environment. While increased consumption boosts production and creates new investment opportunities, it also creates more jobs and higher energy consumption, which can harm the environment. However, if foreign direct investment is directed toward environmentally friendly projects and a greater proportion of economic growth and foreign investment is channeled into the renewable energy sector, financial development can help improve environmental quality. In addition, the development of a financial system can provide the opportunity to finance a renewable energy industry. Low financing costs in financially developed countries will facilitate the financing of environmentally friendly initiatives and help reduce energy consumption.

The long-run coefficient of domestic bank lending to the private sector (DC), which represents financial development, shows that it has a positive effect on carbon emissions (CO₂). At the 1% significance level, this positive effect is also statistically significant. A 1% change in domestic credit (DC) granted by banks to the private sector leads to a 0.11% increase in carbon emissions (CO₂) if the other variables remain constant. These results are consistent with [27,56,63,64].

The study shows a statistically significant relationship between the second variable representing financial development, foreign direct investment (FDI) and carbon emissions (CO₂) at a significance level of 1. The relationship shows a positive effect on carbon emissions (CO₂). Specifically, a 1% increase in FDI leads to a 0.04% increase in carbon emissions (CO₂), while the other variables remain constant. These results are consistent with [27,59,65,66].

This effect on the variables of financial development can be explained by the “pollution-haven hypothesis”, the “scale effect” and the “technology effect”, which are widely discussed in the literature. Developing countries, which generally have less stringent environmental regulations, attract foreign companies that want to reduce their production costs. This leads to a shift of carbon-intensive production activities from industrialized countries to developing countries, which increases pollution. Turkey, in particular, has been struggling with economic problems in recent years. In countries experiencing an economic downturn, unemployment rises, purchasing power falls, aggregate demand declines, interest rates rise, inflation escalates, and it becomes difficult for companies to invest in green energy. In times of economic hardship, governments often relax environmental regulations to encourage investor confidence and economic activity, which favors the use of fossil fuels or other cheap energy sources and leads to higher carbon emissions [67]. Finally, foreign direct investment increases energy demand by stimulating economic growth, which, when energy sources are based on fossil fuels, often leads to higher carbon emissions. The expansion of production capacity, the exploitation of natural resources and the growth of energy-intensive sectors are the main mechanisms driving this effect.

Although Turkey is facing economic challenges, no country can afford the luxury of ignoring environmental damage. Even during an economic crisis, Turkey should be able to strive for sustainable development not only in economic but also in environmental terms. Instead of relaxing environmental regulations, businesses and consumers should be offered other incentives and encouraged to limit their environmental damage. After all, when the economic crisis that a country is facing comes to an end, the environmental damage that has already been done could be irreversible. As the increase in carbon emissions and the looming problems of climate change seem to be inevitable, a balanced environmental and economic strategy is of great importance for countries facing economic problems like Turkey. Reducing taxes on environmentally friendly products while increasing taxes on others and granting tax exemptions to foreign investors who produce or use renewable energy in their production can be some of the solutions to Turkey’s dilemma between economic downturn and environmental protection.

High carbon emissions are the biggest obstacle to green growth in Turkey, as the country promotes the use of coal [68]. A statistically significant relationship is found between total electric power production (TEP) and carbon emissions (CO₂) at a 1% significance level. The relationship shows a positive effect on carbon emissions (CO₂). If the other variables are held constant, a 1% increase in total electricity production (TEP) from energy sources increases carbon emissions (CO₂) by 0.55%. Power plants fuelled by fossil fuels (coal, natural gas and oil) dominate total electricity generation in Turkey and account for about 58% of total electricity generation [69].

This result is consistent with the findings of Kadioglu and Gurbuz [70]. They examined the relationship between total annual electricity production from renewable sources and total annual electricity production from fossil sources for Turkey between 1986 and 2020. The ARDL model and Granger causality tests were used to investigate this relationship. The research results show that annual electricity production from coal, oil and natural gas has a positive impact on carbon emissions. From 1986 to 2020, 69.31% of the electricity generated in Turkey came from fossil fuels. Given the high share of fossil fuels in electricity generation compared to other energy sources, it is logical that the consumption of coal, oil and natural gas has a positive impact on carbon emissions. Studies showing that fossil fuel consumption has a positive long-term impact on carbon emissions underline this view [61,71,72,73]. On the other hand, renewable energy sources have a negative relationship with carbon emissions. Between 1986 and 2020, only 30.69% of the electricity generated in Turkey came from renewable energy sources. There are numerous studies in the literature that indicate that renewable energy sources have a negative impact on carbon emissions [32,49,59,61,67,71,72,74].

5. Conclusions

Reducing greenhouse gas emissions is one of the most important steps to safeguard not only the present but also the lives of future generations. This situation requires urgent and collective action for environmental sustainability. Therefore, countries must first turn to renewable energy sources to meet their energy needs. Thanks to its geopolitical location, Turkey has great potential for renewable energy sources. This potential is crucial not only for providing a clean, economical and secure energy supply under the Turkish Action Plan but also for supporting the country’s economy and achieving green transformation goals.

The energy generated from investments in renewable energy should be used effectively. R&D services should be promoted at all levels to ensure the development of new concepts to increase energy efficiency, and all parts of society should be educated and informed about efficient energy usage.

The integration of sustainable planning of industry and agriculture in Turkey is crucial for the future of the country in terms of ensuring a green economy. There are many issues regarding the sustainable use of resources in both industry and agriculture. For example, industrial zones are created to support ındustrial development without the necessary environmental controls which causes problems in the surrounding agricultural areas. In this context, industrial development should be ensured while considering both the environment and agriculture for a green and sustainable future.

In Turkey, to achieve a green economy, both industrial and agricultural production should be planned considering sustainable methods. Sustainability reporting should become mandatory for companies above a certain size. For sectors that produce particularly high carbon emissions, such as steel, chemicals and energy, annual emission ceilings should be set and tightened each year. Penalties should be imposed if the carbon emission cap is exceeded. A trading platform should be established so that businesses with surplus emission permits can sell these permits to other businesses so that the carbon market in Turkey functions. However, the cost of purchasing surplus emissions should not be more attractive than the penalty costs for businesses that exceed the cap. Companies that exceed the cap on excess emissions should be subject to additional taxes or receive certain tax exemptions. Special verification mechanisms should be established to monitor sustainability performance and ensure the reliability of these data. Special support should be provided for research and development to improve environmental performance, and these projects should be supported with low-interest loans and grant programs. In addition, tax benefits and incentives should be offered for investment in sustainable energy and energy efficiency projects. Appropriate software solutions and technologies should be used to collect and report emission data and the traceability of data in the supply chain should be ensured. Sustainability criteria should also be added to contracts with suppliers, and these targets should be monitored and audited. Training programs, especially for SMEs, on sustainability reporting, environmental management systems and carbon reduction should be established.

Deepening scientific and technological reforms is important to support green growth. Investment in high-quality, environmentally sound scientific and technological innovation should be increased and resources mobilized for social innovation. Emphasis should be placed on training professionals who can provide innovative solutions to environmental problems. And a combination of education, research and production should be achieved. It is crucial that economic expansion is combined with environmental protection by introducing a green budgeting approach. This will help to achieve the goals of economic development and environmental sustainability.

In this study, Turkey’s transition to a green economy between 1990 and 2022 was examined in the context of various economic and environmental indicators. However, the study has some limitations. First, the analysis was limited to certain macroeconomic indicators, which may have limited the ability to obtain more comprehensive results. A broader perspective could be achieved, for example, by including different dimensions such as social indicators or sectoral data. Despite the accuracy and continuity of the data sources used to increase the reliability of the study, possible shortcomings or limitations in the data sets may affect the generalizability of the results.

Future research should aim to analyze the links between the green economy and sustainable development more comprehensively. In this context, it is of great importance to investigate new factors that may influence the course of the green transformation process, especially the impact of variables such as policy, technological innovation and social awareness. In addition to quantitative research, qualitative studies conducted with entrepreneurs, public institutions and civil society organizations—the main actors in this process—can also make an important contribution to the existing literature. Such a multidisciplinary approach will add depth to both the theoretical framework and the policy recommendations.

In conclusion, this study provides important insights into a green economy and sustainable development policy and serves as a starting point for more comprehensive research. It can be emphasized that policy makers should pay attention to the relationship between financial development and energy production in their goals to reduce carbon emissions and support sustainable economic growth.