Agricultural Value Added, Renewable Energy, and the Environmental Kuznets Curve: Evidence from Turkey

Abstract

In this study, the relationship between economic growth and carbon emissions for the period 1968–2022 in Turkey was evaluated within the framework of the EKC (Environmental Kuznets Curve) hypothesis. In addition, the impacts of renewable energy consumption and agricultural value added on carbon emissions were analyzed using the ARDL bounds testing approach. The validity of the results was also tested using the FMOLS and DOLS methods. The findings confirmed the existence of a cointegration relationship between carbon emissions and per capita income, renewable energy consumption, and agricultural value added. Long-term analyses indicate that renewable energy consumption reduces carbon emissions, whereas growth in agricultural value added leads to an increase in emissions. In addition, it has been determined that the EKC hypothesis is valid in both the long and short terms and that increases in per capita income raise emissions up to a certain threshold and have a mitigating effect when this threshold is exceeded. The results of the short-term analysis showed that the effects of renewable energy consumption vary across periods, and that agricultural value added increases emissions in the short term. This study provides empirical evidence for Turkey by incorporating sectoral variables within the EKC framework and offers meaningful insights for policymakers regarding the environmental impacts of agricultural value added and renewable energy use in the context of a developing country. Accordingly, fiscal policy instruments such as green taxation, carbon credit trading mechanisms, and financial and agricultural subsidies should be more effectively utilized in Turkey to support structural transformation in agriculture and promote the use of clean energy, in line with the findings that suggest the need for targeted agricultural and energy policies aligned with Turkey’s SDG commitments.

1. Introduction

The Environmental Kuznets Curve (EKC) hypothesis provides a fundamental framework for understanding the relationship between economic growth and the environment, positing that degradation increases during the early stages of growth but declines after a certain income threshold is surpassed [1]. Many studies do not include sectoral factors that cause emissions in their analysis. This study addresses this gap by examining the environmental impacts of renewable energy consumption and agricultural value added within the EKC framework, focusing on Turkey as a developing economy that is heavily dependent on fossil fuels and traditional agricultural practices. These issues are central to current global sustainability debates, particularly those framed by the United Nations Sustainable Development Goals (SDGs), including SDG 7 (Affordable and Clean Energy), SDG 12 (Responsible Consumption and Production), and SDG 13 (Climate Action).

In addition to macro-level sustainability goals, recent approaches underscore the importance of coherent, sector-specific strategies that are both actionable and aligned with long-term environmental objectives. Within this evolving discourse, the concept of pragmatic sustainability has gained prominence by promoting feasible, innovation-driven frameworks that integrate economic and ecological priorities. Second-hand clothing initiatives, the integration of natural capital into circular economy models, the adoption of hybrid renewable energy systems, and the sustainable management of forests all play important roles in reducing environmental impacts, enhancing agricultural and rural sustainability, and advancing sustainable development [2,3,4,5]. This underscores the value of sectoral, integrative, and practice-oriented strategies within the pragmatic sustainability paradigm. Given this context, sector-specific and integrative strategies are essential for advancing environmental sustainability in Turkey, where this study contributes by empirically testing the EKC hypothesis through an analysis of long-term dynamics.

According to the EKC hypothesis, at the first stage of economic growth, environmental degradation increases due to an increase in production volume (scale effect). As growth continues, the services sector, where growth effects are less, begins to grow (composition effect). With economic growth, reducing emissions and increasing waste management allows for investment in technologies (technological impact). But this trend should not mean that environmental degradation is a natural consequence of economic growth and will correct itself over time. On the contrary, in the face of the risk of exceeding environmental thresholds and irreversible damage, the necessity of implementing structural policies arises [6]. Emissions changes cannot be explained by carbon intensity or growth rates alone; more factors affecting carbon emissions should be included in the EKC research framework. In particular, sector-based analysis is crucial for effectively assessing policy monitoring and achieving carbon neutrality targets.

In Turkey, energy consumption plays a decisive role in growth dynamics and the link between energy and growth [7,8]. Given Turkey’s dependency on imported fossil fuels, renewable energy holds strategic importance, not only environmentally but also economically. For instance, [9] found that a 1% increase in renewable energy consumption leads to a 0.10% rise in the GDP per capita, underscoring the macroeconomic benefits of renewable energy beyond environmental considerations. Moreover, the use of renewable energy is not limited to reducing environmental impacts; it also provides multidimensional contributions such as supporting rural development, reducing income inequality, and modernizing the agricultural sector, thereby playing an important role in achieving sustainable development goals, especially in developing countries.

The agricultural sector is another critical area that significantly affects both economic and environmental outcomes. Although the share of the agricultural sector in Turkey’s GDP declined from 25.8% in 1980 to 4.8% in 2022, it remains strategically important, as it still employs 15.8% of the total workforce [10]. Due to increasing food demand, its role as a fundamental driver of rural development, and its high vulnerability to climate change, the agricultural sector can be considered one of the most critical sectors in terms of environmental sustainability.

Most studies linking agriculture and emissions focus on broad indicators such as agricultural activities or the volume of production. However, this study includes the variable of agricultural value added, which accounted for approximately 4.1% of global national income as of 2021. Notably, the top 20 economies—generating 78% of the global income—also produce 56% of the global agricultural output [10]. Agricultural value added represents the net output derived from the difference between agricultural inputs and outputs; in this respect, it is a key indicator that directly reflects productivity growth and economic contribution in the sector while also allowing for the testing of the EKC theory in the context of agriculture.

The increase in agricultural added value can contribute to environmental sustainability with lower emissions, as opposed to energy-intensive production structures. High technology integration in developed countries makes it possible to achieve high agricultural added value at lower environmental costs. In developing countries, the increase in agricultural value added is usually achieved through traditional and resource-intensive forms of production rather than through modern technologies or sustainable practices. Access to clean technologies in these countries is restricted due to high costs and limited financing opportunities. Since economic growth is prioritized, environmental regulations and sanctions are not sufficiently strict. Therefore, in developing countries, the increase in agricultural value added may lead to short-term economic gains but long-term environmental costs. As a middle-income country with limited technology integration, Turkey is likely to face similar environmental risks.

In light of these observations, the aim of this study is to test the validity of the Environmental Kuznets Curve hypothesis in the Turkish economy and to empirically analyze the effects of agricultural added value and renewable energy use on carbon emissions, identified as complementary indicators for evaluating the effectiveness of climate policies. In this study, agricultural value added as a share of the GDP is used instead of absolute agricultural output to capture the structural composition effect of agriculture within the economy, which provides deeper insights into sectoral environmental dynamics. Similarly, renewable energy consumption is included as a key explanatory variable, not only due to its emission-reducing potential but also because of its broader macroeconomic contributions and strategic relevance in fossil fuel-dependent countries like Turkey. Despite a growing body of EKC-related research, sectoral applications—especially those focusing on agricultural value added as a share of GDP—remain underexplored. In the context of Turkey, empirical studies that isolate the environmental effects of agriculture within the EKC framework are particularly scarce. This study addresses this gap by offering a sector-specific analysis using agricultural value added and renewable energy within the EKC framework for the Turkish economy.

The key contributions of the study include the following: First, it is one of the few studies to jointly analyze the effects of renewable energy consumption and agricultural value added on carbon emissions in Turkey, a developing country with a dual dependency on fossil fuels and traditional agricultural methods. Second, the study incorporates agricultural value added as a share of GDP to reflect sector-specific environmental impacts, an approach that provides more insight than absolute production metrics. Third, it applies a comprehensive econometric methodology combining ARDL bounds testing with FMOLS and DOLS estimators, enhancing the robustness and reliability of the long-run findings. Lastly, it evaluates both short-run and long-run dynamics, offering a more complete view of the EKC hypothesis over time.

2. Theoretical Background

The EKC has emerged as a synthesis of approaches that argue that increasing income levels can raise environmental quality by increasing demand for environmental protection, as opposed to approaches that consider economic growth as the root cause of environmental degradation. Different results are presented depending on the analysis and variables regarding the validity of the EKC hypothesis.

Many studies on the validity of the EKC hypothesis have presented findings with supportive evidence for the EKC hypothesis [11,12,13]. Conversely, refs. [14,15,16] reported that the EKC hypothesis is not valid. However, ref. [17] found that the impact of economic growth on environmental degradation varies depending on a certain threshold level in per capita income and that the EKC hypothesis is partially valid. Accordingly, while below-the-threshold GDP growth increases resource depletion and erosion, above the threshold, these effects weaken, and the impact of growth on some environmental indicators becomes limited or meaningless. These findings suggest that the validity of the EKC hypothesis may differ depending on the country group, periodic scope, variables used, and econometric methods.

Although renewable energy consumption in general has been found to have a reducing effect on carbon emissions in the literature [18,19], studies have also found that renewable energy sources have ineffective or limited effects. As shown in [20], the effects of renewable energy sources on carbon emissions vary: hydropower and wind energy contribute to emission reductions, but solar energy shows no meaningful impact. The study in [21] found no statistically significant relationship between renewable energy consumption and environmental indicators such as ecological footprint and load capacity factor. According to [22], renewable energy improves environmental quality, but the EKC hypothesis is not valid. Also, ref. [12] states that the effects of renewable energy consumption and human capital-reducing carbon emissions differ according to the countries’ EKC return points.

There is a mutual interaction between the environment and agriculture. Climatic conditions have impacts on the amount and diversity of agricultural production. For example, ref. [23] revealed that the increase in carbon emissions has negative effects on both economic growth and agricultural added value. Similarly, refs. [24,25,26,27] found that climatic conditions such as temperature increases and drought lead to decreases in agricultural productivity and income; in contrast, favorable climatic conditions can have positive effects on agricultural production and yields.

On the other hand, agricultural production also has effects on the environment. Agricultural added value can be considered equivalent to production growth and productivity in the agricultural sector. Further production increases environmental damage directly or indirectly by increasing resource use. Previous studies addressing the impact of agricultural activities on carbon emissions differ depending on the period studied, the country’s agricultural policies, modes of practice, and use of technology. Findings from previous studies have revealed that added agricultural value can have effects on carbon emissions in different directions.

Reference [28] found that agricultural added value improves environmental quality in Vietnam, whereas [29] noted that the growth of agricultural added value contributes to the increase in ecological footprint. Similarly, [30] in OECD countries, [31] in Pakistan, and [32] in Brazil found that agricultural added value leads to higher CO₂ emissions. In contrast, ref. [33] concluded that agricultural added value in Somalia reduces environmental quality, while [34,35] reported no statistically significant relationship.

In some studies, it is stated that the increase in agricultural added value increases the environmental impact first, but the environmental cost decreases after a certain income or technology threshold is crossed. Similar to the EKC hypothesis, rising agricultural income levels make it possible to invest in cleaner technologies and move towards sustainable practices. According to [36], agricultural growth in developing countries contributes to increased emissions in the short term, while the long-term effects appear to be less significant.

In the literature, studies that directly use the variable ‘agricultural value added’ are quite limited; in particular, there is a significant gap in analyses examining its relationship with emissions in the context of Turkey. This gap necessitates a more in-depth investigation and the implementation of original empirical studies tailored to the Turkish economy. Including agricultural value added in such analyses is particularly meaningful for assessing environmentally sustainable growth. References [37,38] stated that agricultural value-added growth in Turkey reduces emissions in the long term. This finding suggests that higher added value and lower emissions may be achieved through improved agricultural technologies, renewable energy, and sustainable farming practices. On the other hand, ref. [39] concluded that agricultural value added increases carbon emissions in Turkey. Similarly, refs. [40,41] found that agricultural value added increases CO₂ emissions in both the short and long runs, whereas renewable energy consumption reduces emissions only in the short run; its long-term effect was found to be statistically insignificant. These mixed results indicate that the long-term impact of renewable energy and agricultural practices in Turkey should be examined more thoroughly.

3. Research Methodology

In the studies that measure the validity of the EKC hypothesis, it is seen that different variables such as income level, foreign direct investments, energy consumption, commercial deficit, urbanization, renewable energy, and industrialization level are included in the analysis. In this study, income, agricultural added value, and renewable energy consumption were used as variables affecting carbon emission. Only the share of agricultural value added in the GDP was used to include, in the analysis, the relative weight of the economy, not the absolute level of agricultural production. This variable reflects the structural composition of the economy rather than the direct scale effect and is therefore suitable for analyzing the relative size of environmental impacts. Although alternative proxies such as direct agricultural emissions or input–output-based environmental indicators could be used, these are often subject to limitations in consistency, comparability, and long-term availability, especially in developing countries. In contrast, agricultural value added as a share of the GDP is a widely available and standardized macroeconomic indicator. This approach is particularly appropriate in the context of the EKC framework, which emphasizes relative sectoral dynamics over absolute outputs. Moreover, using AGRI as a share of the GDP aligns with previous studies [28,29] that employ this measure to explore the sector’s contribution to environmental degradation in an aggregated economic context.

In this study, the impact of renewable energy consumption and agricultural value added on carbon emissions is examined within the scope of the Environmental Kuznets Curve (EKC) hypothesis. For this purpose, data from the period 1968–2022 pertaining to Turkey were analyzed. A total of 55 observations were considered. Various trials were conducted to determine the most suitable functional form for the model, and it was concluded that the double-logarithmic form is the most appropriate. The empirical model is expressed as follows:

$${\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_2={\mathsf{\theta }}_0+{\mathsf{\theta }}_1{\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{t}}+{{\mathsf{\theta }}_2{\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{D}\mathrm{P}}^2}_{\mathrm{t}}+{\mathsf{\theta }}_3{\mathrm{l}\mathrm{n}\mathrm{R}\mathrm{E}\mathrm{W}}_{\mathrm{t}}+{{\mathsf{\theta }}_4\mathrm{l}\mathrm{n}\mathrm{A}\mathrm{G}\mathrm{R}\mathrm{I}}_{\mathrm{t}}$$

(1)

In the equation, CO₂ represents carbon emissions; GDP represents gross domestic product; REW represents renewable energy consumption; and AGRI represents agricultural value added. In addition, t denotes the time period, the parameters from θ₁ to θ₄ represent the long-term elasticity coefficients, and ln indicates the natural logarithmic transformation. As assumed in the EKC hypothesis, an inverted U-shaped relationship is expected between economic growth and carbon emissions in the model. The validity of the EKC hypothesis depends on the coefficient θ₁ being positive and the coefficient θ₂ being negative. This implies that as the gross domestic product increases, carbon emissions also increase, but after a certain turning point, carbon emissions begin to decrease. The relevant turning point is calculated using the formula ${\mathrm{G}\mathrm{D}\mathrm{P}}^{\mathrm{*}}=-{\mathsf{\theta }}_1/2{\mathsf{\theta }}_2$, and exp(GDP)* provides the monetary value of the turning point [42] (p. 7743). However, in developing countries, it is anticipated that environmental pressures continue to increase alongside economic growth, and the calculated turning point has not yet been reached [43] (p. 446).

The sign of the coefficient θ₃ is expected to be positive, while the sign of θ₄ may be positive or negative depending on agricultural practices in the country. Comprehensive explanations of the variables in the model are presented in

Table 1. Data definitions.

The Symbol	Variable Name	Definition	Source
CO2	Emissions of carbon	Per capita (metric tons)	Global Footprint Network [44]
GDP	Gross domestic product	Per capita (constant 2015 USD)	World Bank (WDI) [45]
REW	Renewable energy consumption	Per capita (kWh)	Our World in Data [46]
AGRI	Agricultural added value	At constant prices (percentage of GDP)	World Bank (WDI) [45]

.

3.1. Stationary Test

In this study, the ARDL bounds testing approach was used to determine the impact of economic growth, renewable energy consumption, and agricultural value added on carbon emissions. In the ARDL bounds testing approach, it is first necessary to test the stationarity of the variables. In this approach, the dependent variable must become stationary at the first difference, while the independent variables must be stationary at level or at the first difference. Unit root tests are used to test the stationarity of the variables. In this study, the Augmented Dickey–Fuller (ADF) and Phillips–Perron (PP) unit root tests, which are frequently used in the literature, were applied. The ADF approach is conducted by adding a sufficient number of lagged values of the dependent variable to the regression equation in order to eliminate autocorrelation, assuming that the error terms exhibit autocorrelation [47]. In the ADF test, it is assumed that the distribution of the error term is independent and homoscedastic [48] (p. 10). Unlike this approach, the PP test allows for a non-parametric structure with weak dependence and heteroscedastic distribution of the error term [49]. In both approaches, the null hypothesis suggests that there is a unit root (non-stationary) [50] (p. 264).

3.2. ARDL Bounds Testing Approach

In time-series analyses, there are different approaches for examining the cointegration relationship among non-stationary series. The ARDL approach developed by Pesaran and Shin is a method that can be applied to series that become stationary at level and first difference [51]. In this approach, short- and long-run relationships between the dependent and independent variables can be estimated simultaneously, addressing problems due to omitted variables and autocorrelation [52]. Moreover, it enables reliable results in studies with small samples [53]. In this study, the ARDL bounds testing approach, which offers several advantages due to its greater flexibility and practicality compared to other methods, has been employed. The ARDL bounds testing approach consists of three stages. First, an unrestricted error correction model (UECM) is established to estimate short- and long-run parameters. The model in which the dependent variable used in the analysis is included is presented in the following equation [54,55]:

$$\begin{array}{cc}\hfill {△\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_{2\mathrm{t}}={\mathsf{\alpha }}_0+& {\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}}{\mathsf{\alpha }}_{\mathrm{i}1}△{\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_{2\mathrm{t}-\mathrm{i}}+{\displaystyle \sum _{\mathrm{k}=0}^{\mathrm{n}}}{\mathsf{\alpha }}_{1\mathrm{k}}△{\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{t}-\mathrm{k}}\hfill \\ & +{\displaystyle \sum _{\mathrm{k}=0}^{\mathrm{n}}}{\mathsf{\alpha }}_{2\mathrm{k}}{△{\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{D}\mathrm{P}}^2}_{\mathrm{t}-\mathrm{k}}\hfill \\ & +{\displaystyle \sum _{\mathrm{k}=0}^{\mathrm{n}}}{\mathsf{\alpha }}_{3\mathrm{k}}{△\mathrm{l}\mathrm{n}\mathrm{R}\mathrm{E}\mathrm{W}}_{\mathrm{t}-\mathrm{k}}\hfill \\ & +{\displaystyle \sum _{\mathrm{k}=0}^{\mathrm{n}}}{\mathsf{\alpha }}_{4\mathrm{k}}{△\mathrm{l}\mathrm{n}\mathrm{A}\mathrm{G}\mathrm{R}\mathrm{I}}_{\mathrm{t}-\mathrm{k}}+{{\mathsf{\lambda }}_1\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_{2\mathrm{t}-1}+{\mathsf{\lambda }}_2{\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{t}-1}\hfill \\ & +{\mathsf{\lambda }}_3\mathrm{l}\mathrm{n}{{\mathrm{G}\mathrm{D}\mathrm{P}}^2}_{\mathrm{t}-1}+{\mathsf{\lambda }}_4{\mathrm{l}\mathrm{n}\mathrm{R}\mathrm{E}\mathrm{W}}_{\mathrm{t}-1}+{\mathsf{\lambda }}_5{\mathrm{l}\mathrm{n}\mathrm{A}\mathrm{G}\mathrm{R}\mathrm{I}}_{\mathrm{t}-1}+{\mathrm{e}}_{1\mathrm{t}}\hfill \end{array}$$

(2)

In Equation (2), △ denotes the difference operator, α₀ the constant term, e₁t the error term, and α_1–5 and λ_1–5 the coefficients. n shows the appropriate lag length. The cointegration relationship is tested using the F-test. The null hypothesis (λ₁ = λ₂ = λ₃ = λ₄ = λ₅ = 0) states that there is no cointegration. If the F-statistic is lower than the critical bounds, the null is accepted; if higher, it is rejected; if in between, no conclusion can be drawn [53]. If cointegration exists, lag lengths are determined using the AIC or SIC, and the model is estimated via OLS [52,54]. In the second stage, long-run coefficients are estimated [52]:

$$\begin{array}{cc}\hfill {\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_2={\mathsf{\alpha }}_0+{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}1}}& {\mathsf{\alpha }}_{1\mathrm{i}}{△\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{O}}_{2\mathrm{t}-\mathrm{i}}\hfill \\ & +{\displaystyle \sum _{\mathrm{k}=0}^{\mathrm{n}2}}{\mathsf{\alpha }}_{2\mathrm{k}}{△\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{t}-\mathrm{k}}\hfill \\ & +{\displaystyle \sum _{\mathrm{k}=0}^{\mathrm{n}3}}{\mathsf{\alpha }}_{3\mathrm{k}}{△{\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{D}\mathrm{P}}^2}_{\mathrm{t}-\mathrm{k}}+{\displaystyle \sum _{\mathrm{k}=0}^{\mathrm{n}4}}{\mathsf{\alpha }}_{4\mathrm{k}}{△\mathrm{l}\mathrm{n}\mathrm{R}\mathrm{E}\mathrm{W}}_{\mathrm{t}-\mathrm{k}}\hfill \\ & +{\displaystyle \sum _{\mathrm{k}=0}^{\mathrm{n}5}}{\mathsf{\alpha }}_{5\mathrm{k}}{△\mathrm{l}\mathrm{n}\mathrm{A}\mathrm{G}\mathrm{R}\mathrm{I}}_{\mathrm{t}-\mathrm{k}}+{\mathrm{e}}_{\mathrm{t}}\hfill \end{array}$$

(3)

The long-run coefficients help identify the direction and magnitude of the effect between variables [56,57]. The third stage involves estimating the short-run coefficients through the error correction model [52]:

$$\begin{array}{cc}△{lnCO}_2={\alpha }_0+\hfill & {\displaystyle \sum _{i=1}^{n1}}{\alpha }_{1i}{△lnCO}_{2t-i}\hfill \\ & +{\displaystyle \sum _{k=0}^{n2}}{\alpha }_{2k}{△lnGDP}_{t-k}\hfill \\ & +{\displaystyle \sum _{k=0}^{n3}}{\alpha }_{3k}{△{GDP}^2}_{t-k}+{\displaystyle \sum _{k=0}^{n4}}{\alpha }_{4k}{△lnREW}_{t-k}\hfill \\ & +{\displaystyle \sum _{k=0}^{n5}}{\alpha }_{5k}{△lnAGRI}_{t-k}+{\Psi COINTEQ}_{t-k}+e_t\hfill \end{array}$$

(4)

In this equation, △ is the first-difference operator, α₀ the constant, and eₜ the error term. The coefficients α_1–5 refer to the short-term dynamics. COINTEQ denotes the error correction term, and Ψ represents the speed of adjustment. Ψ is expected to be negative and significant. If between 0 and −1, shocks converge monotonically; between −1 and −2, convergence occurs with decreasing fluctuations; values beyond −2 imply divergence [58,59].

3.3. Robustness Check

In this study, the FMOLS (Fully Modified Ordinary Least Squares) and DOLS (Dynamic Ordinary Least Squares) methods were employed to assess the reliability of the results obtained from the ARDL model. Various studies in the literature have suggested that the FMOLS and DOLS approaches are appropriate for evaluating the robustness of long-run ARDL estimates [60]. Developed by [61], FMOLS is a semi-parametric regression method designed to produce consistent estimates in time series that are integrated of order one (I(1)) and cointegrated. This method corrects for common issues in time-series data, such as autocorrelation and endogeneity, thus providing more reliable results. Compared to the conventional OLS method, FMOLS yields more consistent estimates and facilitates inference due to its asymptotic distribution properties. Therefore, it is widely used in multivariate cointegration analyses [61]. The DOLS method, on the other hand, is a cointegration estimation technique that provides reliable inferences despite uncertainties regarding the integration order of the series and stands out for its asymptotic efficiency [62]. In other words, DOLS addresses the asymptotic inefficiency of OLS by incorporating short-run dynamics into the model when estimating long-run cointegration relationships. As a result, the distribution of parameter estimates becomes more concentrated around the true values, allowing for more reliable outcomes [63]. Additionally, DOLS effectively handles small-sample bias, autocorrelation, and deviations from normality [60].

4. Results

In the ARDL model, the descriptive statistics of the variables included in the analysis are presented in

Table 2. Descriptive statistics.

	lnCO2	lnGDP	lnREW	lnAGRI
Mean	0.106258	8.712519	7.089166	2.657173
Median	0.158690	8.655083	7.289702	2.724388
Maximum	0.702991	9.550741	8.359624	3.727795
Minimum	−0.892514	8.069710	5.388940	1.710866
Std. Dev.	0.445196	0.423007	0.783811	0.632076
Skewness	−0.392169	0.350538	−0.576747	0.173185
Kurtosis	2.112708	2.005981	2.631408	1.710175
Jarque–Bera	3.214005	3.390710	3.360517	4.087464
Jarque–Bera probability	0.200488	0.183534	0.186326	0.129544
Observations	55	55	55	55

. The variable with the highest mean value is GDP per capita followed, respectively, by renewable energy consumption, agricultural value added, and carbon emissions. It was observed that the variable with the highest volatility is renewable energy consumption, while the variable with the lowest volatility is GDP per capita. According to the Jarque–Bera statistics, all variables were found to exhibit a normal distribution.

In the ARDL model, the stationarity of the variables included in the analysis was first tested. For this purpose, the ADF and PP unit root tests, which are widely used in the literature, were applied. In both unit root tests, differencing was applied to the non-stationary series, and the differencing process continued until stationarity was achieved. According to the results obtained from the unit root tests, all variables were found to be non-stationary at level but stationary at the first difference (

Table 3. ADF and PP unit root results.

Model	Variables	ADF	PP	ADF	PP	Decision
		I (0)		I (1)
Fixed Model	lnCO2	−2.13 [0] (0.23)	−0.23 [1] (0.19)	−8.14 [0] (0.00) *	−8.14 [2] (0.00) *	For I (1) $H_0$ refuse
	lnGDP (lnGDP2)	0.89 [0] (0.99)	1.15 [4] (0.99)	−6.98 [0] (0.00) *	−6.98 [3] (0.00) *
	lnREW	−1.13 [0] (0.69)	−0.94 [5] (0.76)	−8.27 [0] (0.00) *	−8.74 [5] (0.00) *
	lnAGRI	−1.20 [0] (0.66)	−1.40 [7] (0.57)	−7.28 [0] (0,00) *	−7.46 [6] (0.00) *
Fixed and Trendy Model	lnCO2	−3.12 [0] (0.11)	−3.06 [2] (0.12)	−8.44 [0] (0.00) *	−8.44 [1] (0.00) *
	lnGDP (lnGDP2)	−1.66 [0] (0.75)	−1.74 [1] (0.71)	−7.09 [0] (0.00) *	−7.14 [4] (0.00) *
	lnREW	−2.71 [0] (0.23)	−2.71 [0] (0.23)	−8.20 [0] (0.00) *	−8.67 [5] (0.00) *
	lnAGRI	−2.37 [0] (0.38)	−2.37 [0] (0.38)	−7.31 [0] (0.00) *	−7.74 [7] (0.00) *

Note: The values in the table represent the test statistics, [ ] the maximum lag length and bandwidth, and ( ) the probability value of the test statistic. *, indicates significance at the 1% level.

). These results indicate that the variables do not become stationary at the second difference, meaning that the ARDL model is applicable.

Following the stationarity tests of the variables, the model estimation was carried out. Before estimating the model, it is necessary to determine the appropriate lag lengths of the variables by taking into account the maximum lag length. In cases where the variables in the ARDL model consist of annual series, it is suggested that the maximum lag length be taken as two [52,55]. Accordingly, the maximum lag length was set to two, and the Akaike Information Criterion (AIC) was taken into consideration to determine the optimal lag length of the variables in the model. As a result of the evaluations, it was concluded that the most appropriate model is ARDL (1, 2, 1, 2, 2). In the ARDL model, the F bounds test was applied to examine the cointegration relationship between the variables. According to the findings obtained from the bounds test, the F-statistic value was found to be higher than the lower and upper bounds of the critical values in the table, indicating the existence of a cointegration relationship between the variables at the 1% significance level (

Table 4. Bounds testing for long-run relationship.

Test Statistic	Value	Signif.	I (0)	I (1)
F-statistic	7.581204	10%	2.345	3.28
K	4	5%	2.763	3.813
		1%	3.738	4.947

Note: Case II restricted intercept and no trend model was used, and the critical values were taken from [20].

).

Following the determination of the cointegration relationship among the variables, the long-run coefficients of the model were estimated and are presented in

Table 5. Long-run coefficients based on ARDL model.

Variable	Coefficient	Std. Error	t-Statistic	Prob.
lnGDP	19.206	1.711	11.221	0.000 *
lnGDP2	−1.011	0.091	−11.097	0.000 *
lnREW	−0.163	0.034	−4.807	0.000 *
lnAGRI	0.153	0.075	2.038	0.048 **
C	−89.533	8.011	−11.175	0.000 *
Turning point	$13,396
Diagnostic Tests	Test Statistic		p Values
BG-LM	1.108		0.340
BPG	1.197		0.318
Ramsey reset	0.011		0.990
Jarque–Bera	0.264		0.876
F-statistic	445.548		0.000 *
Durbin–Watson	1.923
R2	0.992
Adjusted R2	0.990

* and ** indicate that the coefficients are statistically significant at the 1% and 5% significance levels, respectively.

. The estimation results of the model revealed that the GDP per capita, its squared term, and renewable energy consumption were statistically significant at the 1% significance level. According to the long-run estimation results, a 1% increase in the GDP per capita increases carbon emissions by approximately 19%, while renewable energy consumption reduces carbon emissions by 0.16%. The findings obtained are largely consistent with theoretical expectations. Indeed, renewable energy is considered one of the important and sustainable solutions to environmental problems caused by fossil fuels. The finding of this study that renewable energy consumption reduces CO₂ emissions in the long run is consistent with the results reported in [39,64]. In addition, the study in [37] also identifies renewable energy as a causal determinant of CO₂ emissions. On the other hand, the findings reported by [40] contradict the current results by stating that the long-run effect of renewable energy on the environment is statistically insignificant.

In addition, a 1% increase in agricultural value added was found to increase carbon emissions by 0.15%, and the coefficient of this variable was found to be statistically significant at the 5% level. This finding can be explained by the increasing dependence of the agricultural sector in Turkey on carbon-intensive production inputs such as mechanization, the use of chemical fertilizers and pesticides, and fossil fuel-based irrigation systems in the long term. In particular, the transition to modern agricultural techniques increases agricultural productivity while also leading to higher energy consumption and carbon emissions. This situation indicates that the growth of the agricultural sector may increase environmental costs in the long run. The findings obtained from this study are consistent with the results reported in [39,41] within the Turkish context but differ from those reported in [37,38].

In the long-run analysis of the ARDL model for Turkey, when the signs of the GDP per capita variables are examined, it is observed that the GDP per capita has a positive coefficient, while its squared term has a negative coefficient. This indicates that the EKC hypothesis is valid in the long run. Accordingly, it is predicted that an increase in the GDP per capita will reduce carbon emissions after reaching a certain threshold. This study’s findings are generally consistent with those of [40]. Both studies confirm the validity of the EKC hypothesis for Turkey and show that agricultural value added increases CO₂ emissions. While this study finds a long-term emission-reducing effect of renewable energy consumption, [40] reports a significant reduction only in the short term and states that the effect is not statistically significant in the long term.

According to the model estimation results, the turning point predicted by the EKC hypothesis is estimated to occur when per capita income reaches approximately USD 13,396. Based on World Bank data, Turkey’s real per capita income for the year 2022 was reported as USD 14,713 [45]. This suggests that the transition from increasing to decreasing environmental pressure may have already occurred in Turkey or is likely to occur in the very near future. The finding is largely consistent with various studies in the literature. For instance, in studies focusing on Turkey, [65] reported the EKC turning point at USD 9920, while [42] estimated it to be in the range of USD 13,523–14,077. Additionally, the study in [41], which used ecological footprint as the dependent variable, the turning point was calculated as USD 8184, and it was noted that environmental degradation began to decline during the analysis period. In this context, it is concluded that environmental policies in Turkey should be maintained decisively, and a balance between economic growth and environmental sustainability must be ensured.

In the estimated ARDL model, the BG-LM autocorrelation test and the Breusch–Pagan–Godfrey (BPG) heteroscedasticity test were applied. The test results indicated that there were no issues of autocorrelation or heteroscedasticity. Furthermore, it was determined that the Durbin–Watson statistic had a value close to two, confirming the absence of autocorrelation. The Jarque–Bera test revealed that the error terms were normally distributed. The Ramsey RESET test statistics also indicated that the model has the correct functional form. According to the R² values obtained from the model, it was determined that the model has high explanatory power. The probability value of the F-statistic showed that all independent variables in the model significantly affect the dependent variable.

Following the estimation of the long-run coefficients, an error correction model (ECM) based on the ARDL model was established within the scope of the third step in order to estimate the short-run coefficients. As seen in

Table 6. Error correction for the selected ARDL model.

Variable	Coefficient	Std. Error	t-Statistic	p Values
D (lnGDP)	6.132	2.751	2.229	0.031 **
D (lnGDP (−1))	−0.395	0.138	−2.852	0.006 *
D (lnGDP2)	−0.268	0.155	−1.726	0.091 ***
D (lnREW)	−0.110	0.029	−3.743	0.000 *
D (lnREW (−1))	0.083	0.029	2.887	0.006 *
D (lnAGRI)	0.179	0.071	2.491	0.017 **
D (lnAGRI (−1))	−0.152	0.069	−2.194	0.034 **
CointEq (−1) *	−0.752	0.105	−7.153	0.000 *
R-squared	0.781
Adjusted R-squared	0.747
Log likelihood	102.563
Durbin–Watson stat	1.923

*, **, and *** indicate that the coefficients of the variables are statistically significant at the significance levels of 1%, 5%, and 10%, respectively.

, all variables included in the model were found to be statistically significant at different significance levels. The results obtained from the error correction model show that the GDP per capita in the current period has a positive effect on carbon emissions, while the lagged GDP per capita has a reducing effect on carbon emissions. On the other hand, it was determined that the squared term of the GDP per capita in the current period has a negative effect on carbon emissions. This indicates that the EKC hypothesis is valid in the short term.

In addition, while carbon emissions increase with the lagged renewable energy consumption, they decrease with current renewable energy consumption. This finding suggests that renewable energy consumption may have temporary negative effects in the short term and may cause transition costs. It was found that increases in agricultural value added in the current period increase carbon emissions, whereas increases in the previous period have a reducing effect. This suggests that due to the nature of agricultural activities, they may trigger both short-term emission-generating and delayed emission-reducing processes. In particular, intensive agricultural activities increase emissions in the short term through energy use and input dependency; however, in the following periods, they may reduce emissions through mechanisms such as decreased activity volume after harvest, an increase in biomass, and carbon interaction with the soil.

The coefficient of the error correction term was found to be statistically significant at the 1% level. The fact that the coefficient lies between −1 and 0 indicates that short-term shocks can be eliminated in a monotonic manner. According to the value obtained, approximately 75% of the imbalances that occurred in the previous period can be corrected within one year; in other words, it is concluded that the short-term shocks can fully return to the long-term equilibrium level in about 1.3 years.

CUSUM and CUSUMQ tests, as proposed by [66], were applied to determine whether the long- and short-term coefficients obtained from the ARDL model were stable over time. In the CUSUM and CUSUMQ tests, the cumulative sums of error terms are iteratively examined on the graph, depending on time. In this way, it can be determined whether the structural breaks that may occur in the coefficients of the model and the predictive power of the model maintain their validity in different periods [66]. If the graphs obtained under the CUSUM and CUSUMQ tests lie within the critical value zone of 5% significance, it appears that the coefficients in the model are stationary and have no structural fracture [67]. The results of both tests in Figure 1 show that the null hypothesis is accepted, the graph is among the critical values, and the coefficients are stable.

The robust model was estimated for the validity of the results obtained from the ARDL model. For this purpose, the FMOLS and DOLS methods were used, and the results obtained are presented in

Table 7. FMOLS and DOLS results for robustness check.

FMOLS	Coefficient	Std. Error	t-Statistic	Prob.
lnGDP	19.183	1.235	15.521	0.000 *
lnGDP2	−1.007	0.066	−15.235	0.000 *
lnREW	−0.137	0.024	−5.634	0.000 *
lnAGRI	0.173	0.050	3.453	0.001 *
C	−89.876	5.746	−15.639	0.000 *
Turning point	$13,853
R-squared	0.988
Adjusted R-squared	0.987
DOLS	Coefficient	Std. Error	t-Statistic	Prob.
lnGDP	20.259	1.501	13.492	0.000 *
lnGDP2	−1.064	0.079	−13.331	0.000 *
lnREW	−0.178	0.027	−6.486	0.000 *
lnAGRI	0.176	0.066	2.650	0.012 **
C	−94.594	7.045	−13.425	0.000 *
Turning point	$13,687
R-squared	0.993
Adjusted R-squared	0.990

* and ** indicate that the coefficients for variables are significant at the significance levels of 1% and 5%, respectively.

. The differences in the values of ARDL, FMOLS, and DOLS coefficients are due to the structural characteristics of the methods. The results from the FMOLS and DOLS regressions have been determined to be largely consistent with the results from the ARDL long-term model. As the regression results reveal, the EKC hypothesis is valid for Turkey. It has been found that renewable energy reduces carbon emissions, while agricultural added value has the effect of increasing carbon emissions. On the other hand, the turning point in per capita income predicted by the EKC hypothesis appears to be relatively consistent with the ARDL long-run results. The turning point is estimated at USD 13,583 in the FMOLS model and USD 13,687 in the DOLS model. Considering that Turkey’s real per capita income is USD 14,713 [45], these values indicate that Turkey is either very close to or has just surpassed the turning point at which environmental pressures begin to decline. The results support the EKC hypothesis, which suggests that environmental degradation may decrease as income increases, highlighting the necessity of aligning economic growth with sustainability principles.

5. Discussion

Differences in national production and consumption structures play a critical role in identifying sectoral sources of greenhouse gas emissions and in assessing the effectiveness of country-specific policy instruments. In emerging economies such as Turkey, sectoral analysis enables the formulation of targeted policy proposals for environmental sustainability. To empirically assess these sectoral dynamics in the context of Turkey, this study analyzes the validity of the EKC hypothesis over the period 1968–2022. In addition, it examines the impact of renewable energy consumption and agricultural value added on carbon emissions using the ARDL bounds testing approach. To ensure robustness, the results were also tested using the FMOLS and DOLS estimation methods. The empirical findings confirm a long-run cointegration relationship among carbon emissions, GDP per capita, renewable energy consumption, and agricultural value added.

In the long run, renewable energy consumption reduces emissions, while growth in agricultural value added contributes to increased emissions. The EKC hypothesis is found to be valid for Turkey in both the short and long runs. Short-run results indicate that the effects of renewable energy and agricultural value added on emissions vary over time, which may be explained by the costs of energy transition and agriculture’s ongoing dependence on fossil fuels and chemical inputs. The error correction model shows that about 74% of short-term deviations are corrected annually, indicating a convergence to long-run equilibrium within approximately 1.3 years. Similarly, recent studies conducted in the European Union suggest that the impact of renewable energy on sustainability may be negative at lower levels and positive at higher levels of adoption, indicating that such effects can vary depending on scale and context [68].

The emission-increasing effect of agricultural value added is consistent with previous studies such as [29,30,31]. On the other hand, studies that report different results, such as [69], show that crop production and fisheries contribute to reducing CO₂ emissions, while the effect of livestock is statistically insignificant. Similarly, the long-term emission-reducing impact of renewable energy consumption reflects a widely observed trend in the literature [69,70,71]. However, divergent findings reported by studies such as [20,21] suggest that results in this field may vary depending on country-specific conditions and contextual factors. For instance, ref. [72] reported a negative effect of renewable energy consumption, suggesting that its current use in the agricultural sector may not be fully efficient or adequately integrated into productive activities. In line with the present study’s findings, ref. [73] also reported that renewable energy consumption reduces the ecological footprint in Denmark, whereas value added from agriculture, forestry, and fisheries—as well as economic growth—increases environmental pressure.

According to the analysis findings, since Turkey has either reached or is very close to the EKC turning point, it is of great importance to maintain sustainable environmental policies with determination. The results from the study suggest the need to maintain a delicate balance between economic growth and environmental sustainability and emphasize the importance of supporting renewable energy policies with stronger incentives, particularly in Turkey, where reducing dependence on foreign fossil fuel sources could significantly benefit national economic growth.

In the case of Turkey, an increase in the real share of agricultural added value in the GDP, coupled with the low technology density of the production structure and traditional forms of production that create environmental pressure, leads to an increase in emissions. This shows that not only the volume of production, but also the mode of production and the technologies used, are decisive in terms of environmental impacts. In this context, it is necessary to accelerate structural transformation in the agricultural sector, adopt low-impact technologies, and activate financial and economic incentives to increase the use of clean energy. The adoption of technologies that enhance productivity—such as precision farming, low-emission equipment, and smart irrigation—can help alleviate the environmental burden of agricultural activities and should be accelerated through targeted subsidies and technical support.

Environmentally friendly and sustainable farming practices offer a promising path toward reducing emissions from the agricultural sector, particularly when supported by national policies aimed at increasing the sector’s added value. In this respect, strategies such as improved livestock management, agroforestry, forest conservation, reforestation, sustainable fishing practices, and the adoption of energy-efficient technologies may be prioritized [21,74].

The use of renewable energy sources, which can reduce environmental impacts while lowering energy costs in agricultural production processes, should be expanded. Supporting energy efficiency and renewable energy investments will contribute to reducing environmental pressures while enabling Turkey to sustain its economic growth in a more environmentally sustainable manner. To this end, the adoption of renewable energy should be encouraged through various instruments such as green taxes, emissions trading schemes, R&D support, and efforts to increase public awareness. The steps taken towards a carbon credit trading platform in Turkey are promising in terms of developments in this field [75]. Establishing the necessary infrastructure for monitoring and reporting emissions in the agricultural sector and including this sector within the scope of the Emissions Trading System (ETS) may contribute to the reduction of carbon emissions. In addition to supporting the agricultural sector by providing additional income to farmers, such mechanisms can also contribute to the dissemination of sustainable practices such as reducing the use of chemical fertilizers. Existing agricultural subsidies should be restructured to discourage environmentally harmful practices and support sustainable production methods while complementing these reforms with green taxation instruments such as carbon taxes and border carbon adjustments to internalize environmental costs.

In order to reduce the restrictions encountered in the export of agricultural products to the EU, Turkey should enhance the compliance of its agricultural products with the EU’s health and safety standards. As a matter of fact, in line with the European Green Deal, important steps are being taken to integrate agricultural production in Turkey to fight climate change and meet emission reduction targets. In particular, limiting pesticide and fertilizer use, increasing organic farmland, and promoting agroecological methods are among the key strategies supporting both environmental sustainability and rural development [76].

6. Conclusions

In conclusion, the development of holistic and inclusive policies that aim to reduce environmental impacts while sustaining economic growth is of paramount importance. This study, based on national-level macroeconomic data, focuses on broad structural trends but does not account for micro-level diversity in agricultural practices or regional disparities in energy use. Additionally, certain structural factors—such as industrial policies, technological advancements, and urbanization—were excluded due to data limitations. Although this study offers sector-specific insights into the environmental impacts of agriculture and renewable energy, it does not account for institutional and policy-related uncertainties. As highlighted by [77], ESG-based sustainability uncertainties—such as those arising from fluctuating environmental, social, and governance conditions—can significantly influence both policy effectiveness and investment decisions.

Accordingly, it is recommended that future studies expand the analytical scope to include not only the renewable energy and agriculture sectors but also other emission-intensive sectors such as industry, transportation, and services. In addition, incorporating structural determinants—such as industrial policies, technological innovation, and energy consumption patterns—could enrich the explanatory capacity of future models. Evaluating the impact of market-based environmental instruments—such as carbon taxation, emissions trading schemes, and green finance—may also lead to more comprehensive findings. Micro-level analyses and institutional indicators could further support a multidimensional assessment of Turkey’s emissions landscape.

In light of recent empirical studies, several promising directions emerge for future research, particularly regarding methodological diversity and variable selection in the analysis of the agriculture–energy–emissions nexus. Future studies aiming to deepen the Environmental Kuznets Curve (EKC) framework in the context of agriculture and renewable energy can be enriched through broader variable sets and methodological improvements. Firstly, the inclusion of green innovation indicators such as research and development (R&D) expenditures and the number of patents in the model may reveal the technological aspect of environmental improvements [78,79]. In addition, analyses can be conducted based on component indicators such as energy efficiency (e.g., energy intensity or energy productivity) and resource efficiency (e.g., productivity ratios based on DMC) [80]. On the regulatory dimension, it is possible to use indicators such as environmental governance quality and eco-efficiency scores [81]. Similarly, the effects of policy uncertainty on sustainability highlight the value of including indicators of institutional instability into the model [82]. From a financial perspective, green public expenditures and financial development indicators may help analyze how public investments and credit systems contribute to environmental goals [78,83]. Finally, the application of advanced econometric methods such as Fourier ARDL, NARDL, and wavelet coherence may allow for more effective identification of structural breaks, asymmetric effects, and time-varying relationships [73,79,84]. Such extensions will contribute to a more holistic and policy-oriented understanding of emission dynamics in the context of Turkey’s evolving energy–agriculture structure.