Environmental Policy Stringency and Carbon Dioxide Emission: Asymmetric Causality Analysis

Abstract

The effectiveness of policy tools used to combat global environmental degradation and climate change is gaining importance. The role of environmental policies in reducing carbon dioxide emissions and whether rising emissions levels lead to tightening of environmental policies are the main research questions of this study. Using panel causality methods, this study examines the relationship between the Environmental Policy Stringency Index (EPS) and carbon dioxide (CO₂) emissions across 23 OECD countries from 1991 to 2020. The results of the Dumitrescu–Hurlin panel causality test show bidirectional causality between EPS and CO₂. According to the asymmetric causality test, we find causality from EPS to CO₂ during positive shocks. Furthermore, when examining negative shocks, we report a unidirectional causality originating only from EPS in the US. Our results demonstrate that the bidirectional relationship between EPS and CO₂ can contribute to the development of effective and balanced environmental policies if policymakers and society consider this interaction.

1. Introduction

One of the most common tools for governments is the environmental policy stringency (EPS) index developed by the OECD to monitor environmental quality changes. The EPS indicator is a composite index. The index is obtained by collecting information on specific environmental policy instruments related primarily to climate and air pollution. This indicator focuses on upstream sectors such as energy and transportation because these sectors, which have similar importance among countries, are also of primary importance from an environmental perspective. Within this framework, strict environmental rules and regulations mainly rely on environmental taxes, renewable energy, and support for energy efficiency [1]. Figure 1 shows the components that generate the environmental policy stringency index. The aggregation procedure for both energy and broader indicators follows two steps. First, instrument-specific indicators (e.g., taxes on SOx, NOx, and CO₂) are combined into mid-level indicators by type (e.g., environmental taxes). Second, these are grouped into market-based and non-market instruments. Subcomponents can also be aggregated into “stick” (penalizing) and “carrot” (rewarding) policies. At each level of aggregation, equal weights are applied [2]. The EPS index takes a value between 0 and 6. Policies become stricter from 0 to 6.

The structure of the EPS index indicates that CO₂ emissions contribute as a variable in taxation and trading plans. Therefore, the variation in CO₂ emissions could be directly related to the stringency and effectiveness of environmental policies. Monitoring and analyzing these emissions are critical in determining whether environmental policies are successfully implemented and affect the evaluation of EPS. While the common wisdom is that stringent environmental policies lead to a decrease in CO₂ emissions, inadequate policies can also result in augmenting emissions. It should be noted that reducing CO₂ emissions does not solely depend on the policies’ success but also on a well-operated implementation. Thus, a legally stringent environmental policy requires support from the implementation stage [3].

In environmental policies, the most common issue that has been stressed is the reduction of fueling carbon emissions. A milestone policy document, The Paris Climate Agreement, was signed by 197 countries in 2015 as a precautionary or proactive policy paper addressing the global climate crisis. Through the agreement, governments have willingly made novel domestic regulations regarding carbon emissions in line with the targets specified in the agreement. The anticipated impact of those regulations is pioneering the enhancement of environmental quality, especially in reducing CO₂ emissions.

Reducing carbon emissions not only improves countries’ environmental quality but is also crucial for economic growth and development. Governments have the opportunity to strike an optimal balance between economic growth, development, and carbon emissions. The Paris Climate Agreement supports determining this balance based on each country’s priorities and characteristics [4]. The expectations from the Paris Agreement’s regulations indeed caught the attention of researchers worldwide, which led to the rising numbers of studies on CO₂ emissions and EPS. However, few studies investigate whether there is a relationship between CO₂ emissions and the stringency of environmental policies. Ref. [5] contributes to this scarce literature by examining the asymmetric causality between CO₂ and EPS. However, their sample covering the top five carbon emission producer countries provides limited results in this global case. There is also other research that focuses on analyzing the symmetric relationship between the stringency of environmental policies and CO₂. In this strand of the literature, there are few studies covering OECD countries in their data [6,7,8,9,10]. The vast majority of the current research [11,12,13], however, primarily considers China due to its high carbon emissions levels.

This study, with its sample spanning numerous countries, directly contributes to overcoming the generalizability problem encountered in previous studies with limited sample sizes. Furthermore, this study’s asymmetry aspect contributes to the global EPS literature. Specifically, OECD data show that EPS has increased over time in many countries and that differences persist across countries. Therefore, the policy–emission relationship may not respond in the same way to both negative and positive shocks. Consequently, examining positive and negative shocks separately strengthens the methodological approach to uncovering hidden relationships emphasized in the literature.

This research investigates whether there is causality between CO₂ and EPS in OECD countries. In this respect, our research’s main motivation point is that we identify a gap in the literature regarding the global impact of environmental degradation and attempt to fill this gap with a new sample in our analysis. This paper attempts to contribute to the current studies by diverging itself by methodology and sample selection. Utilizing an asymmetrical analysis on a larger scale, we aim to thoroughly explain the relationship between the stringency of environmental policies and carbon emissions. The key contribution of this study lies in its combined use of both symmetric and asymmetric causality approaches. Furthermore, the Dumitrescu–Hurlin and Yılancı–Aydın tests are applied to differentiate between positive and negative shocks. This methodological approach adds depth to the analysis by revealing not only average relationships but also different effects depending on the direction of the shocks. In addition to the distinction between symmetric and asymmetric causality, the country-specific findings presented in this study enable a clearer and more comparative assessment of the heterogeneous effects of environmental policies. In these respects, this study contributes to the literature through its global scope, methodological novelty, and asymmetric analysis.

Our research investigates the period from 1991 to 2020, covering 23 OECD countries. We utilize panel causality and asymmetric causality tests in the analysis. Furthermore, we also test the positive and negative shocks of the stringency of environmental policies on CO₂ emissions to improve the robustness of policy recommendations on enhanced environmental quality. Our results are twofold. On the one hand, the symmetric causality test documents a bidirectional relationship between EPS and CO₂. On the other hand, the asymmetric causality test reports significant one-way causality from EPS to CO₂. This impact is observed only for the USA under a negative shock. We then test the positive and negative shocks for the asymmetric causality between EPS and CO₂. In this analysis, we document one-way and bidirectional causality for different countries.

2. Literature Review

While extracting the current literature, we identified that the vast majority of the studies have used the EPS index in their analysis. Yet, the overall number of studies that use EPS and CO₂ as variables is limited. The findings of the studies in the literature are summarized below. The literature review suggests three strands of research. In the first group, there are studies documenting the negative impact of EPS on environmental degradation. The second group consists of reporting positive effects. The last group, however, indicates ambiguous results.

A recent study by [14] explores the effect of the stringency of environmental policies on renewable energy consumption. In the analysis for the period covering 1990–2019 and for 32 OECD countries, they employ the GMM estimator and Driscoll–Kraay standard errors technique. The study reveals that increasing environmental policy stringency encourages renewable energy. Hence, this result could be pivotal in achieving the net-zero CO₂ emission target of OECD governments by 2050. Ref. [15] assessed the impact of environmental policy practices on carbon emissions in the industrial sector by panel PMG and QARDL models. Considering the Eurozone, China, South Korea, Japan, the UK, and the US from 1997 to 2020, they present that the stringency of environmental policies has a significant negative impact on carbon emissions.

Ref. [16] examined the regulatory effect of environmental policy stringency between technological innovation and CO₂ emissions using DOLS, FMOLS, CCR, and MMQR estimation approaches. The study covered the period between 1990 and 2020 and G7 countries. According to the findings, technological innovation leads to an increase in environmental quality by reducing CO₂ emissions, and the implementation of stricter environmental rules increases the positive effects of green technologies, such as using environmentally friendly technologies to achieve the carbon neutrality target.

There are also wider perspectives on research on this topic. The relationship between green innovations, imports, exports, environmental policy stringency, and consumption-based CO₂ emissions in high-income OECD economies for the period between 1990 and 2020 is analyzed by [9] using the MMQR method. According to the findings, a one-unit increase in the stringency of environmental policies reduces CO₂ emissions by 0.32–0.39 units. Examining the asymmetric effects of stringency of environmental policies on CO₂ emissions in high-emitting countries (China, USA, India, Russia, and Japan), ref. [5] utilizes a nonlinear panel ARDL approach using 1990–2019 data. Their results indicate that increases in the stringency of environmental policies improve environmental quality by reducing CO₂ emissions in the long run and document a one-way symmetric causality from EPS to CO₂ emissions. Regarding asymmetric panel causality, a positive shock to EPS does not affect CO₂ emissions, but CO₂ emissions cause a negative shock to EPS.

Ref. [17] investigated the effectiveness of environmental taxes and the stringency of environmental policies in reducing CO₂ emissions. They utilized the AMG test for a panel of seven developing economies from 1994 to 2015 and documented an inverted U-shaped relationship between CO₂ emissions and the stringency of environmental policies. Indeed, they state that the stringency of environmental policies requires time to become effective. Their robustness checks by FMOLS also document the stringency of environmental policies and taxes that could effectively reduce CO₂ emissions. In another approach, ref. [18] analyzed the long-run impact of environmental policies and human development on CO₂ emissions in Group of Seven and BRICS economies from 1995 to 2015 through panel cointegration and causality tests. The causality analysis presents a bidirectional causality between environmental tightening policies (EPS) and CO₂ emissions for Germany, Japan, the United Kingdom, and the United States and unidirectional causality from CO₂ emissions to EPS for Canada, China, and France. Furthermore, according to the cointegration analysis, both environmental tightening policies and human development have a decreasing impact on CO₂ emissions.

Widening the sample, ref. [19] scrutinized the relationship between the stringency of environmental policy and CO₂ emissions in BRIICTS (Brazil, Russia, India, Indonesia, China, Turkey, and South Africa) from 1993 to 2014. By applying the PMG-ARDL estimator, they documented an inverted U-shaped relationship between environmental policy stringency and CO₂ emissions. The outcomes indicate that stringent ecological policy initially does not lead to environmental improvements, but stringent environmental policy still improves environmental quality after a threshold. In their PMG analysis, ref. [8] examined the role of strict environmental regulations on environmentally friendly technological innovation, carbon emissions, GDP, exports, and imports for 20 OECD countries between 1999 and 2015. They document strict environmental policies, and imports reduce CO₂ emissions in the long run.

Accordingly, ref. [7] utilized the fixed effects model and investigated the relationship between economic growth and environmental quality in the context of the Kuznets curve for 32 countries in the period covering 1992 and 2012. Their findings show that environmental policies are effective in reducing environmental damage associated with economic growth. Ref. [11] conducted panel data analysis (fixed effects–random effects and OLS) using data from 35 industrial sectors in China from 2006 to 2014. Their results yield a negative relationship between environmental regulations and carbon efficiency. Their explanation relies on the argument that environmental regulations on carbon efficiency led to inconsistency and inefficiency of the measures related to the regulations.

The relationship between the EPS and CO₂ is generally negative. However, rare studies have also found a positive impact between the two variables. The research by [10] inquired into the relationship between EPS and CO₂ using the dose–response function econometric approach for 37 OECD countries and the years 1990–2020. Their results suggest that environmental policies implemented at low levels initially increased carbon intensity, while environmental policies implemented at stricter levels reduced the amount of carbon emissions per capita. Ref. [13] scrutinized the effects of the shadow economy and strictness of environmental policy on economic growth and energy consumption in China during the period from 1993 to 2019. Using the nonlinear ARDL approach, their outcomes yielded a boost in energy consumption despite the increase in the strictness of environmental policies. This situation was interpreted as EPS not being as effective as it should be in China, where non-renewable energy is used to a large extent. According to the asymmetric Granger causality test results, there is one-way causality from energy consumption to EPS.

The number of studies that yield inconclusive results is also limited. In their research on the USA, ref. [20] documents that much of the reduction in air pollution (NO) emissions experienced by the country over the period between 1990 and 2008 was due to changes in environmental policies. Investigating the effects of environmental regulations on environmental performance, ref. [12] utilized the GMM method for 283 Chinese cities from 2003 to 2010 and four different indicators for environmental quality. Their estimations show that the current environmental control measures and regulations do not achieve the goal of controlling and reducing pollution. There is also research focusing on nonlinear impacts in this matter. Ref. [6], examining the nonlinear effects of environmental regulations and economic growth on PM2.5 in 30 OECD countries over 1998–2015, revealed an initial boost in PM2.5 emissions in line with environmental policy stringency, whereas there is no significant correlation otherwise.

3. Materials and Methods

3.1. Data

In this research, we investigate the relationship between the Environmental Policy Stringency Index (EPS) and carbon emissions (per kg) for 1991–2020 for 23 OECD countries through two-way interaction via symmetric and asymmetric causality tests. In order to ensure the continuity of the data for EPS and CO₂ variables, the research sample consisted of 23 OECD countries in the years 1991–2020. We obtained EPS Index data from the OECD database and CO₂ from the World Bank database. All data are annual. Our sample covers 23 OECD countries. In the analyses, we used EViews 12.0, Stata 18.0, and Gauss 12.0. Descriptive statistics of the variables are given in

Table 1. Main characteristics of the series.

Variables	Obs.	Mean.	Std. Dev.	Min.	Max.
EPS	690	2.289533	1.036466	0.1666667	4.888889
CO2	690	0.2799767	0.1339037	0.0477577	0.8198865

Source: Prepared by authors.

.

3.2. Methodology

Dumitrescu and Hurlin’s causality test of the null hypothesis checks the homogeneity of the result, whereas the alternative hypothesis controls the heterogeneity in the model with a heterogeneous structure. In addition, this test requires a stationary series. Finally, the Dumitrescu and Hurlin causality test provides relatively more reliable results in the case of T > N. The causality test model is as follows [21]. In this equation, Y and x demonstrate stationary values:

$${\gamma }_{i,t}={\alpha }_i +\sum _{k=1}^K{\mathsf{{\rm Y}}}_i^{\left(k\right)} {\mathsf{{\rm Y}}}_{i, t-k} + \sum _{k=1}^K{\beta }_i^{\left(k\right)} x_{i,t-k}+{\mathrm{ℇ}}_{i,t}$$

(1)

For country i = 1, … N and years t = 1, … T.

There are studies in the literature that test symmetric causality between EPS and CO₂. However, the number of studies using asymmetric causality tests is quite limited. For this reason, we used the asymmetric causality test in addition to the symmetric causality test. The rationale for using the asymmetric causality test in the study is to detect the hidden relationship between two variables. We tested the causality between the EPS and CO₂ by the Dumitrescu and Hurlin causality test and the Yılancı and Aydın asymmetric causality test. Ref. [22] developed an asymmetric causality test for panel data. Symmetric causality analysis provides a widely used standard framework in the literature, revealing the average effect and overall directional relationship between variables. However, the relationship between environmental policies and emissions often exhibits a structure in which policy tightening and loosening can have different effects, meaning it can vary with the direction of shocks. Therefore, using an asymmetric causality approach, the aim was to separate positive and negative shocks and test whether they produce distinct causal dynamics. Symmetric causality tests, by their nature, do not consider the direction of shocks and may therefore disregard possible asymmetric effects. In this context, we aimed to increase the robustness of the findings by using both methods together in this study. Asymmetry in variables indicates that the data set responds to negative and positive shocks. Variables used in the analysis may show different reactions to shocks. In the case of ignoring these variations, it will not be possible to document the existing relationship between variables. Hence, an analysis considering the asymmetric relationship reveals the hidden relationship between variables and strengthens the reliability of the analysis [23].

Before proceeding with the Dumitrescu and Hurlin causality test, it is essential to conduct preliminary tests to ensure that the basic assumptions of panel data analysis are met. First, the presence of cross-sectional dependence in the panel must be tested. This is because inter-country economic and environmental interactions (e.g., global shocks or common policy trends) can lead to interdependence between units. In this case, classical panel causality approaches risk, producing misleading results. Therefore, in this study, the CD test developed by Pesaran (2004) [24] and the Breusch–Pagan–LM test were used to test for the presence of cross-sectional dependence. The two tests are complementary, and using them together allows for a more robust assessment of different types of addiction. Additionally, the CIPS test, a second-generation panel unit root test that accounts for this dependence, was performed. Furthermore, testing whether the slope coefficients in the panel are homogeneous is an important preliminary step. The Dumitrescu and Hurlin (2012) [21] causality test is a causality test developed for heterogeneous panels. For this purpose, the delta test developed by Pesaran and Yamagata (2008) [25] was used. In conclusion, these preliminary tests were used in this study to prevent biases that might arise from incorrect model selection and to increase the reliability of causality findings.

4. Results

To document our results, we first perform the cross-sectional dependency, homogeneity, and unit root tests. We utilized Breusch–Pagan’s (1980) LM [26] and Pesaran’s (2004) [24] CD tests to check the existence of cross-sectional dependency. The Breusch–Pagan–LM test is used in cases where the unit size (N) of the panel is small and the time dimension (T) is large (N < T). Breusch–Pagan–LM test statistics are obtained by the following equation [25]:

$${CD}_{lm }= T\sum _{i=1}^{N-1}\sum _{j=i+1}^N{\widehat{\rho }}_{ij}^2$$

(2)

Pesaran developed the CD test for both the case where the unit dimension is larger than the time dimension (N > T) and the case where the time dimension is larger than the unit dimension (T > N). The equation is as follows [24]:

$${CD}_{lm}=\sqrt{{\displaystyle \frac{1}{N\left(N-1\right)}} } \sum _{i=1}^{N-1}\sum _{j=i+1}^N(T{\widehat{\rho }}_{ij}^2-1)$$

(3)

The cross-sectional dependency test results are given in

Table 2. Cross-sectional dependence tests results.

Variables	Statistics	BP (LM	Pesaran CD
EPS	Test statistic	6405.543	79.904
	p value	0.0000	0.0000
CO2	Test statistic	6336.271	79.327
	p value	0.0000	0.0000

Source: Prepared by authors.

. According to Table 2, the null hypothesis is rejected since the probability value for both tests was less than 0.05 (0.000). These results allow us to state that there is cross-sectional dependency in the series. The fact that the results indicate a significant level of cross-sectional dependence is important for interpreting causality analyses. The presence of cross-sectional dependence suggests common shocks across countries (global economic developments, environmental policies, energy prices, etc.). This situation may require careful interpretation of causality test results.

In the second step, we use the homogeneity test developed by Pesaran and Yamagata (2008) [25]. This test checks whether the slope coefficients are homogeneous or not. The Pesaran and Yamagata (2008) [25] test is based on the Swamy (1970) [27] homogeneity test model as an improved version. The Delta test (Δ) is calculated with the following formula [25]:

$$\widetilde{∆}=\sqrt{N}{\displaystyle \frac{N^{-1}\tilde{S}-k}{\sqrt{2k}}}$$

(4)

The test results are given in

Table 3. Homogeneity test results.

Test	Test Statistic	p Value
D	29.267	0.0000
D adj	30.851	0.0000

Source: Prepared by authors.

. According to the Delta test results, the H₀ hypothesis was rejected at the 1% significance level, and heterogeneity was revealed.

We then utilize the second-generation unit root test due to the cross-sectional dependency between the series. The rationale behind using the second-generation tests derives from providing more reliable and consistent results than the unit root tests to be applied with panel data. As there is cross-sectional dependency, it is more appropriate to use the second-generation unit root tests that consider the cross-sectional dependency. The CIPS statistic (Cross-sectionally Augmented Version of IPS), proposed by Pesaran (2007) [28] and based on the test of Im, Pesaran, and Shin (2003) [29], controls the stationarity of the cross-sectional units as a whole. The CIPS statistic is calculated using the following equation [28]:

$$CIPS(N, T)= t- bar= N^{-1}\sum _{i=1}^N{t}_i(N,T)$$

(5)

The test results are given in

Table 4. CIPS panel unit root test results.

Variables	Constant	Constant + Trend
EPS	−1.61657 [>=0.10]	−0.04677 [>=0.10]
d(EPS)	−5.87419 [<0.01] *	−5.84291 [<0.01] *
CO2	−1.87518 [>=0.10]	−0.01501 [>=0.10]
d(CO2)	−5.60227 [<0.01] *	−5.75685 [<0.01] *

Source: Prepared by authors. *: Indicates statistical significance at the 1% level.

. Table 4 shows EPS and CO₂ as the variables for which no difference was taken; d(EPS) and d(CO₂) are the variables for which the first difference was taken. Examining the table, it is seen that both EPS and CO₂ variables are not statistically significant at the level values (constant and constant + trend below) (p ≥ 0.10), meaning that the series contain a unit root and are not stationary. In contrast, the series for which the first differences were taken (d(EPS) and d(CO₂)) are statistically significant at the 1% significance level. This finding shows that both variables have an I(1) process, meaning they become stationary when their first differences are taken. Therefore, using series for which first differences were taken in the causality analysis yields more reliable and consistent findings.

Additionally, we first test the causality between CO₂ and EPS for 1991–2020 in 23 OECD countries utilizing the Dumitrescu and Hurlin causality test. The results are reported in

Table 5. Results of Dumitrescu & Hurlin’s (2012) [21] Granger non-causality test.

Test Statistic	EPS → CO2	p Value	Test Statistic	CO2 → EPS	p Value
Wbar	3.0853	0.0000	Wbar	2.6449	0.0000
Zbar	7.0715	0.0000	Zbar	5.5782	0.0000
Zbartilde	5.8787	0.0000	Zbartilde	4.5856	0.0000
H0: CO2 does not Granger-cause EPS.			H0: EPS does not Granger-cause CO2.
H1: CO2 does Granger-cause EPS			H1: EPS does Granger-cause CO2

Source: Prepared by authors.

. According to the results of the analysis, we document bidirectional causality between EPS and CO₂.

After the Dumitrescu and Hurlin causality test, we investigate the causality between CO₂ and EPS with the Yılancı and Aydın asymmetric causality test. The test results are presented in

Table 6. Asymmetric panel causality tests results (−,−).

Countries	EPS ↛ CO2 (−)					CO2 ↛ EPS (−)
	Wald. Stat.	Bootstrap Critical Values			Wald. Stat.	Bootstrap Critical Values
		1%	5%	10%		1%	5%	10%
Australia	265.3753	1050.8002	504.4487	383.4833	30.3404	360.2648	244.4408	185.0352
Austria	288.7244	720.7988	412.1528	312.8138	30.9201	431.8531	242.4291	176.5366
Belgium	288.5686	778.2454	445.6478	356.7252	68.1431	386.1018	250.7996	192.8259
Canada	171.9888	1316.3990	792.6163	603.6011	63.8072	488.0741	305.2307	239.4408
Denmark	349.0714	1766.6433	1137.3127	911.0572	68.2962	1177.2840	564.8522	441.1843
Finland	407.8512	1710.9739	1059.5336	886.4469	49.1779	1373.2141	612.6872	472.1212
France	208.4269	1160.7180	853.5947	696.8659	60.8085	1172.9750	567.8932	429.4048
Germany	302.9404	1366.1987	893.3482	736.3786	88.7661	1496.4190	704.3210	564.0602
Greece	413.2200	1408.6697	769.9117	629.1056	96.4963	1318.1740	734.7771	555.8637
Hungary	404.4917	1341.7270	919.6880	687.8243	135.2001	1297.3550	873.3864	645.6376
Ireland	141.3675	834.6102	509.9762	395.4734	117.3014	1077.8646	628.4925	490.9271
Italy	176.3528	821.7688	484.9061	394.9862	63.1580	1414.3178	688.5691	553.4617
Japan	247.2522	982.3327	613.7726	486.7917	89.2331	1314.1260	858.6429	661.6927
Korea	252.1729	1178.1571	715.6704	540.9968	58.1845	1262.2499	827.3528	603.3462
Netherlands	145.7649	677.7928	466.4946	374.0452	40.9360	1485.9189	748.7380	529.4511
Norway	126.9987	588.5316	436.9693	337.0802	45.0256	1515.0831	903.8341	580.4181
Portugal	99.7952	453.1394	287.6365	222.2494	21.3230	1091.6373	574.3378	466.0713
Spain	113.7975	466.0743	300.8336	245.4342	14.2849	597.4844	328.5780	240.5257
Sweden	220.3431	734.5786	534.3524	427.8095	36.0263	738.8177	473.9547	386.2452
Switzerland	237.7150	907.0917	672.7999	500.6898	14.0268	524.5166	324.3223	246.2029
Türkiye	327.9621	1366.2814	750.9120	563.7110	32.7384	510.3416	324.0011	224.4773
United Kingdom	320.2567	629.0698	402.5549	320.9554	39.8514	275.0451	179.9824	141.9401
United States *	488.6366	560.3013	331.5473	252.0789	18.2789	297.6297	164.4749	127.0044

Source: Prepared by authors. *: One-way causal relationship.

. Looking at the findings on negative shocks, only in the USA is there significant, unilateral causality from EPS to CO₂. In the other countries included in the analysis, there is no significant causal relationship between EPS and CO₂ or between CO₂ and EPS. We can express this in two ways. Firstly, when examining OECD policy documents, it is stated that the stringency of environmental policies generally increases. Therefore, the negative shock channel includes a limited number of observations in many countries. This reduces statistical power and may lead to spurious causality being detected in fewer countries. Secondly, the fact that environmental policies are mostly sticky, or in other words, permanent, suggests that this finding is consistent with the increasing stringency of environmental policies [1]. Another notable finding in Table 6 is specific to the USA, where there is significant unidirectional causality from EPS to CO₂. In asymmetric causality tests, significant negative shocks for the independent variable indicate that a decrease in the independent variable will affect the dependent variable. Accordingly, a decrease (relaxation) in environmental policies in the United States affects carbon emission levels—either the decrease stops or carbon emissions drop to a low level. This effect may stem from the effective implementation of environmental policies in the United States and the binding nature of environmental regulations. The United States Environmental Protection Agency (EPA) is an independent agency of the US government that addresses environmental protection issues [30]. It is responsible for maintaining and enforcing national environmental standards under various US laws (such as the Clean Air Act). It has the authority to impose fines, sanctions, and other measures. The existence of an independent agency responsible for environmental protection, rather than relying solely on existing environmental laws in the US, contributes to the effective implementation of environmental policies. Therefore, individual and corporate behavior is sensitive to environmental policies in the United States. The fact that similar results are not obtained in other countries outside the US in our empirical findings may be due to different environmental policies adopted by those countries or the lack of sensitivity in the behavior of individuals and companies.

Later, we also tested the causality between EPS and CO₂ with positive shocks. The test results are given in

Table 7. Asymmetric panel causality test results (+,+).

Countries	EPS ↛ CO2 (+)				CO2 ↛ EPS (+)
	Wald. Stat.	Bootstrap Critical Values			Wald. Stat.	Bootstrap Critical Values
		1%	5%	10%		1%	5%	10%
Australia *	59.9466	986.1210	495.5973	325.6062	338.4587	297.3410	209.9330	167.6777
Austria	122.0846	735.3470	399.5112	309.6320	96.8561	288.6628	199.3904	155.9273
Belgium **	164.7911	319.2590	160.1994	127.4089	235.5591	246.0031	156.2068	125.5740
Canada *	137.1131	945.5100	592.6890	438.3684	433.8142	295.8390	207.3538	168.3691
Denmark	252.2334	1075.9230	581.7795	437.8127	145.0117	333.9066	224.1142	178.5252
Finland	393.3282	1194.3728	679.3320	518.5630	113.9259	441.9607	253.6853	207.4991
France *	525.7898	1272.5401	692.1359	543.3413	354.8153	461.3001	268.9735	214.4510
Germany **	708.0312	1407.8317	502.4387	409.2956	420.1518	365.0640	273.2365	222.8300
Greece *	1676.9260	1442.9789	844.7632	590.9618	182.5411	468.7142	255.1312	220.4193
Hungary *	1436.8655	1121.7174	575.4949	467.4162	130.1287	306.2055	199.7484	160.3214
Ireland **	849.5464	910.4108	490.3514	363.7816	252.7342	352.2620	238.7444	194.8468
Italy *	709.9269	728.6869	432.0843	314.9298	112.3696	355.6760	224.1415	195.6590
Japan *	456.0430	797.3385	397.6485	292.1621	159.0091	351.5868	236.7121	178.8117
Korea **	484.6309	817.1952	369.9326	289.5567	216.2388	397.6542	252.0953	211.5466
Netherlands *	731.5390	593.5238	269.9163	196.7793	262.4056	739.9748	351.2877	273.8355
Norway *	793.0060	454.4849	281.2360	194.0231	362.1338	1136.6430	531.0194	411.7248
Portugal **	323.7447	244.0672	126.4457	82.7980	616.4656	604.1134	279.2957	213.4642
Spain **	113.8180	229.6786	114.4278	72.7429	461.4107	906.1917	395.2901	299.7262
Sweden **	284.6928	883.2769	379.2044	283.5164	291.4311	389.5863	255.2447	196.1089
Switzerland **	394.2524	1138.2782	517.7492	373.8945	314.7701	446.1230	306.0584	247.6090
Türkiye **	478.5465	532.0424	257.6913	199.9019	727.5888	438.2368	265.4217	228.1319
United Kingdom *	88.5483	238.9444	126.0686	93.8593	248.9373	307.8066	196.0647	164.6778
United States *	30.9343	272.8173	135.5620	107.7286	489.6025	484.6365	273.5996	192.9466

Source: Prepared by authors. *: One-way causal relationship. **: Two-way causal relationship.

. We present bidirectional causality for Belgium, Germany, Ireland, Korea Rep., Portugal, Spain, Sweden, Switzerland, and Türkiye, whereas there is unidirectional causality for Greece, Hungary, Italy, Japan, Netherlands, and Norway. When we investigated the reverse causality from CO₂ to EPS, our outcomes presented a unidirectional relationship for Australia, Canada, France, the United Kingdom, and the United States.

5. Conclusions

This study investigates the interaction between EPS and CO₂ through symmetric and asymmetric causality. First, we analyzed the symmetric causality between EPS and CO₂ for 23 OECD countries for 1991–2020 by utilizing the Dumitrescu–Hurlin panel causality test. In this context, the use of both symmetric and asymmetric causality tests and the sole scrutinizing of the interaction between EPS and CO₂ with positive and negative shocks distinguish our study from the current literature.

The findings of the symmetric causality test reveal that there is bidirectional causality between EPS and CO₂. However, as we suspected a hidden causality, we further checked for positive and negative shocks. Therefore, in estimating an asymmetric causality test, we document the following results: According to the negative shocks of the asymmetric panel causality, there is significant one-way causality between EPS and CO₂ only in the USA, although no significant causality is reported between CO₂ and EPS. However, we do not report any significant causality from CO₂ to EPS. This is an expected outcome. Even if carbon emissions decrease, it is not rational to expect countries to relax their environmental policies. This is because the fundamental aim of global environmental policies is not only to maintain current emission levels but also to ensure the continuous protection and improvement of all environmental factors, especially air quality.

Secondly, we tested the causality between EPS and CO₂ with positive shocks. We then presented bidirectional causality for Belgium, Germany, Ireland, Korea Rep., Portugal, Spain, Sweden, Switzerland, and Türkiye, whereas the causality is unidirectional for Greece, Hungary, Italy, Japan, Netherlands, and Norway. The results of the reverse causality from CO₂ to EPS yield a unidirectional relationship for Australia, Canada, France, the United Kingdom, and the USA.

We evaluated the symmetric and asymmetric results separately with other studies in the literature. Our results for the symmetric causality test are in line with the results of [18], while they are contradictory to [5]’s outcomes. However, in terms of asymmetric causality, as the USA is the only country affected by the negative shocks, our results are in compliance with [5]. Yet, there are other research documented various results. While ref. [11] found a negative relationship between environmental regulations and carbon efficiency, refs. [12,13] revealed that the stringency in environmental policies could be more effective.

The mutual interaction between EPS and CO₂ is a remarkable finding. Therefore, based on the empirical findings, effective implementation and tightening of environmental policies are critical in reducing CO₂ emissions. The findings of this study are directly related to SDG 13, particularly in terms of integrating climate policies into national policies and increasing resilience to climate risks. Specifically, the findings highlight the importance of reducing policy lag in countries where emissions are triggered by expanded polymer emissions, as well as increasing the reliability of policy signals in countries where EPS drives emissions. This two-pronged approach aligns with the implementation goals under SDG 13 [31,32]. Strict environmental policies should not be limited to legal regulations; economic incentives, awareness campaigns, and clean technology investments should also support them. In addition, using carbon capture and storage technologies in industrial facilities and power plants, encouraging the use of electric vehicles to replace fossil fuel vehicles, and increasing green bond issuances to finance environmental projects can reduce the amount of CO₂ released into the atmosphere. Finally, this study focuses only on the relationship between EPS and CO₂. Future research should focus on new variables, countries, and methods to enrich the literature and provide new policy recommendations.