The Impact of Oil Price on Carbon Dioxide Emissions in the Transport Sector: The Threshold Effect of Environmental Policy Stringency

Abstract

Carbon dioxide emissions from the transport sector make a significant contribution to global greenhouse gases, and understanding the factors that influence these emissions is beneficial for devising effective emission reduction policies. Oil prices are an important influencing factor since the fuel used in the transport sector is primarily based on oil, and fluctuations in oil prices directly impact the sector’s CO₂ emissions. Additionally, environmental policies, as a key means of controlling CO₂ emissions, can affect the relationship between oil prices and CO₂ emissions in the transport sector. Therefore, this study aims to examine the impact of oil prices on CO₂ emissions in the transport sector and explore the nonlinear role of environmental policy stringency in this relationship. Based on data from 27 OECD member countries and 6 non-member countries from 1990 to 2019, we used the environmental policy stringency index as a threshold variable to construct a panel threshold regression model. The analysis results indicate a double-threshold effect: when the environmental policy stringency index is low, the impact of oil prices on CO₂ emissions in the transport sector is not significant. However, when the index reaches the first threshold, the impact of oil prices significantly increases; upon reaching the second threshold, the effect is further intensified. This paper also analyzes the three subindicators—market-based policies, non-market-based policies, and technology support policies—to clarify the distinct impact mechanisms of different types of environmental policies. Finally, based on the research findings, we propose policy recommendations to achieve carbon dioxide emission reduction targets in the transport sector.

1. Introduction

In recent years, the issue of global warming caused by significant greenhouse gas emissions has received widespread attention worldwide. The transport sector’s carbon dioxide (CO₂) emissions are one of the major contributors to greenhouse gases [1]. Data from the International Energy Agency (IEA) show that in 2021, CO₂ emissions from the transport sector accounted for 23% of total CO₂ emissions [2]. With the accelerating pace of industrialization and urbanization, and the growing demand for convenient transportation and freight in modern society, CO₂ emissions from this sector are continuously increasing [3,4]. Therefore, CO₂ emission reduction in the transport sector has become a key focus for governments, and this sector is considered the most challenging in terms of CO₂ reduction efforts [5,6,7]. Taking the United States, China, and the European Union as examples, these three are the main contributors to CO₂ emissions from the transport sector. They have implemented several carbon reduction policies targeting this sector. In the United States, through the Corporate Average Fuel Economy (CAFE) standards mandated by the Energy Independence and Security Act (EISA), goals have been set to increase annually from 2010 to 2030, compelling automakers to continually improve vehicle fuel efficiency to meet these standards [8]. Additionally, the U.S. has implemented complementary policies including fuel taxes and new vehicle purchase tax credits [9]. In China, the government vigorously promotes electric vehicles through the new energy vehicle industry policy, reducing CO₂ emissions in the transport sector with purchase subsidies and investments in electric vehicle-related infrastructure [10,11,12]. Meanwhile, the European Union focuses on advancing the use of renewable energy in the transport sector. According to its 2020 Climate Change Package, targets have been set to increase the share of liquid biofuels in the transport sector to at least 10% by 2020, thus significantly reducing CO₂ emissions from the transport sector [1].

Greenhouse gas emissions from the transport sector primarily originate from the fossil fuels burned by cars, trucks, ships, trains, and airplanes [13]. Over 94% of transportation fuels are oil-based, including gasoline and diesel, and burning these fuels results in direct emissions [14]. Therefore, changes in oil prices directly affect fuel usage in the transport sector, underscoring their importance. We believe it is essential to empirically examine and identify the impact of oil prices on CO₂ emission in the transport sector using real data. However, the impact of oil price fluctuations on CO₂ emissions in the transport sector is currently unclear [15]. Intuitively, an increase in oil prices would reduce fuel demand in the transport sector, leading to lower CO₂ emissions [16,17]. However, some argue that the demand for fuel in the transport sector is inelastic, making it difficult for higher oil prices to curb demand, thus having little effect on CO₂ reduction [18,19]. We believe this discrepancy may be due to overlooking the nonlinear relationship between oil prices and CO₂ emissions in the transport sector caused by environmental policy stringency. Therefore, this paper not only focuses on the impact of oil prices on CO₂ emissions in the transport sector but also examines whether environmental policy stringency affects this relationship. Specifically, we conducted the following analysis: First, we applied a nonlinear panel threshold model to panel data from 27 OECD member countries and 6 non-member countries from 1990 to 2019. Then, we used the environmental policy stringency (EPS) index as a threshold variable, which is widely used in research as a proxy for environmental measures [20]. Additionally, we further analyzed the subcomponents of the EPS index, including market-based policies, non-market policies, and technology support policies, to determine the mechanisms and heterogeneity of environmental policy stringency in the relationship between oil prices and CO₂ emissions in the transport sector.

This paper makes the following contributions: First, empirical research on the impact of oil prices on CO₂ emissions in the transport sector is relatively scarce. We investigate the nonlinear relationship between oil prices and carbon emissions in the transport sector, identifying that this nonlinearity is related to environmental policy stringency, providing new insights in this field. Second, our analysis effectively addresses the discrepancies in the literature, reconciling views that oil prices have both minimal and significant impacts on carbon emissions in the transport sector. Third, by examining the multidimensionality of the environmental policy stringency index, we clarify the mechanisms and heterogeneity of different types of environmental policies. Fourth, the policy implications of this study provide specific guidance for policymakers on how to effectively implement carbon emission reduction policies in the transport sector. By revealing the nonlinear relationship between oil prices and the stringency of environmental policies, and its impact on CO₂ emissions, this research emphasizes the need to consider this dynamic relationship when formulating and adjusting emission reduction strategies in the transport sector. This means that policymakers, when designing emission reduction measures, should not only focus on changes in oil prices but also take into account the strictness of environmental policy implementation, to ensure these policies can effectively respond to the continuously changing economic and market conditions. Furthermore, by analyzing the multidimensionality of environmental policy, this study provides insights for policymakers on how to comprehensively utilize market mechanisms, direct regulations, and technological support as various policy tools.

The remaining sections are as follows: Section 2 reviews and summarizes the research on the determinants of CO₂ emissions in the transportation sector. Section 3 describes the methods and data used in this study. Section 4 presents and discusses the results of the empirical analysis. Finally, Section 5 concludes the study and offers relevant policy recommendations.

2. Literature Review

Currently, the determinants of CO₂ emissions in the transport sector have attracted considerable attention from scholars. Based on their research objectives, studies can be simply divided into two categories. The first category aims to comprehensively consider the driving factors of CO₂ emissions in the transport sector. The second category focuses on the impact of one or several core variables on CO₂ emissions in the transport sector.

In the first category of studies, examining the driving factors behind CO₂ emissions in China’s transport sector is currently mainstream. Lin and Benjamin [21] utilized time series data from 1980 to 2010 and found that per capita GDP, energy intensity, and carbon intensity have a greater impact on carbon emissions compared to urbanization. Lin et al. [22] employed a panel model to quantify the impact of socio-economic factors and their regional differences on carbon emissions in China’s provincial transport sectors from 2000 to 2015, revealing that the level of secondary industry development is the most significant factor. Sun et al. [23] used the three-dimensional gray correlation analysis model to analyze and found that the five most influential factors on carbon emissions in China’s provincial transport sectors are energy structure, logistics scale, population, GDP, and tertiary industry factors. Cai et al. [24] applied the log-average weight decomposition method to analyze the driving factors of CO₂ emissions in China’s transport sector and discovered that per capita GDP, transportation energy intensity, and population size all contribute to the increase in CO₂ emissions, while the energy usage per unit turnover and transportation intensity reduce the growth of CO₂ emissions.

In addition to studies that focus on total carbon emissions as the dependent variable, some research targets the carbon emission intensity of China’s provincial transport sector, considering its driving factors. Huang et al. [25] used a dynamic panel quantile regression method to analyze the main drivers of carbon emission intensity in China’s provincial transportation industry from 2000 to 2016. This study took into account the heterogeneous impacts of factors such as per capita GDP, energy intensity, urbanization, private vehicles, and freight turnover across different carbon emission intensity quantiles. Similarly, Liu et al. [26] used data from the transport sectors of 30 Chinese provinces to investigate regional differences in carbon emission intensity and identified the impact of urbanization levels, energy structure, population size, and industrial structure on carbon emission intensity.

In addition to focusing on research in China, some of the literature also examines the transport sectors of specific countries or multiple countries other than China. Timilsina and Shrestha [27] analyzed the potential factors contributing to the growth of CO₂ emissions in the transport sectors of selected Asian countries between 1980 and 2005. The study found that in countries such as China, India, and South Korea, per capita income and population growth were the main factors driving the increase in CO₂ emissions in the transport sector. Solaymani [1] utilized the Logarithmic Mean Divisia Index (LMDI) to decompose the carbon emissions of the transport sectors in the USA, China, India, Canada, Russia, and Brazil. The findings indicated that the main factor reducing CO₂ emissions is carbon intensity, while the primary drivers for the increase in CO₂ emissions are the electricity structure and the economic output effect. Timilsina and Shrestha [28] used the same method to analyze CO₂ emissions in the transportation sectors of Latin American and Caribbean countries. The results revealed that economic growth and changes in the energy intensity of transportation were the main drivers of the increase in CO₂ emissions. Rasool et al. [29] concentrated on analyzing the determinants of carbon emissions in Pakistan’s transport sector, finding that rising oil prices and economic growth help reduce CO₂ emissions, whereas increases in energy intensity, population concentration, and road infrastructure lead to higher CO₂ emissions. Georgatzi et al. [30] used panel data to examine the relationship between the stringency of environmental policies, green transportation technologies, and investments in transportation infrastructure and CO₂ emissions in 12 European countries. The results showed that transportation infrastructure investment did not affect CO₂ emissions, whereas the stringency index of environmental policies and green transportation technologies had significant emission reduction effects. In contrast, Alataş [31] examined the impact of environmental technological advancements on CO₂ emissions in the transport sector of 15 EU countries and found that the effect was not statistically significant. Bhowmik et al. [32] also investigated the impact of green transportation technologies on CO₂ emissions in the transport sectors of China, India, Russia, the United States, and Brazil. The results showed that in low-emission countries, the impact of green transportation technologies was not significant, whereas in medium- and high-emission countries, green transportation technologies had a substantial emission reduction effect. Through these studies, we can gain a more comprehensive understanding of the drivers and complexities of CO₂ emissions in the transport sector.

The second category of studies focuses on the impact of core variables on CO₂ emissions in the transport sector, primarily concentrating on technological innovation and energy efficiency. In terms of energy efficiency, Li et al. [33] found that its effect on reducing carbon emissions in China’s transport sector varies across different stages, but overall, it has a significant positive impact on reducing emissions. Irfan et al. [34] discovered similar results using an Indian sample, noting that a 1% increase in energy efficiency leads to a reduction of carbon emissions by 1.343% to 1.665% in India’s transport sector in the long term. Regarding technological innovation, Demircan Çakar, Gedikli, Erdoğan, and Yıldırım [13] studied the impact of innovative technologies on carbon emissions in the transport sector of Mediterranean countries. After establishing a long-term relationship using panel cointegration methods, long-term coefficients were obtained using PMG and DFE methods, showing that the emission reduction effect of technological innovations is stronger in developed countries. Similarly, Chen et al. [35] and Yang et al. [36] found that technological innovation is the most crucial factor in reducing carbon emissions in China’s transportation industry. Lee et al. [37] further considered the impact of the digital economy on carbon emissions in China’s provincial transport sector, with results indicating that the digital economy has a carbon reduction effect.

Overall, through our review of the relevant literature, we found that existing studies rarely address the impact of oil prices on CO₂ emissions in the transport sector. However, over 94% of transportation fuels in this sector are oil-based, primarily including gasoline and diesel, which are direct sources of carbon dioxide emissions [14]. Therefore, the role of oil prices in this area is crucial. Our study fills this gap. Compared to previous research that mainly focused on a few countries like China, our sample includes 33 countries, making the results more generalizable. Our research findings indicate that environmental policy stringency is a key factor in the relationship between oil prices and CO₂ emissions in the transport sector. Furthermore, we conducted a detailed analysis of three subindices: market-based policies, non-market-based policies, and technological support policies, clarifying the differentiated impact mechanisms of various types of environmental policies. Specifically, the type and environmental policy stringency play a decisive role in regulating the relationship between oil prices and CO₂ emissions. These findings are important to environmental policymakers, not only providing directions for policy adjustments but also offering specific recommendations on how to optimize the environmental performance of the oil-dependent transport sector through different combinations of policies.

3. Empirical Strategy, Data, and Variable Definitions

3.1. Panel Threshold Regression Model

This paper posits that the impact of oil prices on CO₂ emissions in the transport sector is heterogeneous under different levels of environmental policy stringency, necessitating consideration of threshold effects. Therefore, we utilize the threshold panel model developed by Hansen. This model indicates that the regression coefficients can change with structural breaks or threshold variables, which are endogenously determined [38,39]. By endogenously dividing based on the characteristics of the environmental policy stringency index data, we can define different levels of environmental policy stringency intervals. Within these intervals, we determine how oil prices influence CO₂ emissions in the transport sector.

The setup for a single-threshold model with fixed effects is as follows:

$$\begin{array}{c}{Y}_{i,t}={\delta }_i+{\beta }^{\prime }{X}_{i,t}+{\theta }_1{g}_{i,t}I\left(d_{i,t}\le \gamma \right)+{\theta }_2{g}_{i,t}I\left(d_{i,t}>\gamma \right)+{\omega }_{i,t}\end{array}$$

(1)

where $Y_{i,t}$ represents the dependent variable (CO₂ emissions in the transport sector), and $g_{i,t}$ is the core explanatory variable (oil prices). $X_{i,t}$ includes a set of control variables with ${\beta }^{\prime }$ as their corresponding coefficient. $d_{i,t}$ is the threshold variable (environmental policy stringency index), and $\gamma$ is a specific threshold parameter (a specific value of the environmental policy stringency index) that divides the equation into two states with two separate threshold coefficients, ${\theta }_1$ and ${\theta }_2$. $I$ is an indicator function. ${\delta }_i$ reflects the individual effects of each country. The random disturbance term ${\omega }_{i,t}$ is assumed to be independently and identically distributed.

The principle of estimating threshold regression models primarily follows the principle of minimizing the sum of squared residuals [40]. The estimator for the threshold parameter γ is the value that minimizes the RSS for the single-threshold model [41]:

$$\begin{array}{c}\widehat{\gamma }=arg\underset{\gamma }{\min }{S}_1\left(\gamma \right)\end{array}$$

(2)

where $S_1$ represents the RSS of the single-threshold model. $\widehat{\gamma }$ is a consistent estimator for $\gamma$, and confidence intervals can be determined by testing the hypothesis $H_0:\gamma ={\gamma }_0$ using the likelihood ratio (LR) statistic, as follows [42]:

$$\begin{array}{c}LR\left(\gamma \right)={\displaystyle \frac{S_1\left(\gamma \right)-S_1\left(\widehat{\gamma }\right)}{{\widehat{\sigma }}^2}}\end{array}$$

(3)

When the threshold parameter $\gamma$ is given, the threshold coefficients ${\theta }_1$ and ${\theta }_2$ can be estimated using ordinary least squares [43].

Determining whether the threshold effect is statistically significant is also an important step, which involves testing the null hypothesis $H_0:{\theta }_1={\theta }_2$ (linear model) against the alternative hypothesis $H_a:{\theta }_1\ne {\theta }_2$ (single-threshold model). If the null hypothesis holds, then there is no threshold effect, and the model is a linear model [44]. The test statistic used in this process is the F-statistic:

$$\begin{array}{c}F={\displaystyle \frac{\left(S_0-S_1\right)}{{\widehat{\sigma }}^2}}\end{array}$$

(4)

where $S_0$ and $S_1$ represent the RSS for the linear model and the single-threshold model, respectively. Under $H_0$, the threshold $\gamma$ is not identified, and the F-statistic has a non-standard asymptotic distribution. Therefore, we use the bootstrap method to determine the critical values of the F-statistic to test the significance of the threshold effect [45].

The computation of a single-threshold model can similarly be extended to multiple-threshold models. Specifically, after rejecting the null hypothesis of the single-threshold model, we need to further test the double-threshold model, whose null hypothesis assumes a single-threshold effect and the alternative hypothesis assumes a double-threshold effect. Similarly, when the null hypothesis of the double-threshold model is rejected, we need to test the triple-threshold model, where the null hypothesis is the double-threshold effect and the alternative hypothesis is the triple-threshold effect [38].

3.2. Data and Variable Definitions

To examine the relationship between oil prices, CO₂ emissions in the transport sector, and environmental policy stringency, this study utilizes a dataset covering the period from 1990 to 2019 for 33 countries (27 OECD countries and 6 non-OECD countries). The list of sample countries is shown in

Table 1. List of sample countries.

OECD Countries	Non-OECD Countries
Australia, Austria, Belgium, Canada, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Japan, Netherlands, Norway, Poland, Portugal, Slovakia, Slovenia, South Korea, Spain, Sweden, Switzerland, Turkey, United Kingdom, United States	Brazil, China, India, Indonesia, Russia, South Africa

.

Firstly, the dependent variable is the per capita CO₂ emissions in the transport sector. We use CO₂ emissions data for the transport sector compiled by Ritchie et al. [46], with the original data sourced from the International Energy Agency (IEA).

Secondly, our core explanatory variable is the price of oil, derived from the OECD’s imported crude oil price data. Some missing values are filled using the global oil prices (WTI) from the U.S. Energy Information Administration.

Thirdly, the threshold variables in this paper are the environmental policy stringency (EPS) index and its subcomponents. Over the past few decades, as environmental degradation issues have intensified and awareness of environmental issues has increased, countries have developed a series of environmental policies [47]. Although the effects of these policies could be considered individually, using a more comprehensive index allows for a better assessment of the effectiveness of environmental measures in curbing emissions [48]. In this context, the concept of the EPS index was initially proposed by Botta and Koźluk [49] and later revised and updated by Kruse et al. [50]. The current EPS index primarily focuses on policies mitigating climate change and air pollution, representing the strictness of environmental policies on a scale from 0 to 6, with higher numbers indicating stricter environmental regulation [51]. Its advantage lies in making the EPS measurement standards comparable across different countries/regions. This index consists of three subindices, each with equal weight, representing different aspects of environmental policy measures. These subindices include market-based policies (such as carbon taxes and pollution pricing schemes like trading programs), non-market-based policies (emission limits and standards policies), and technology support policies (policies supporting the innovation and adoption of clean technologies) [50].

Fourth, in order to isolate the impact of oil prices on CO₂ emissions in the transport sector as accurately as possible, we also controlled for a range of factors that could affect CO₂ emissions. Economic growth has driven an increase in transportation activities, which in turn has led to a rise in CO₂ emissions from the transport sector [27]. Trade facilitates the flow of goods produced in one country for consumption or further processing [52], potentially driving an increase in CO₂ emissions from transportation. The transportation industry is the world’s second-largest energy consumer, accounting for over 30% of global total energy consumption, and energy consumption is directly linked to CO₂ emissions [53]. Technology transfers conducted through FDI may impact CO₂ emissions in the transport sector [53]. Government spending can help curb the increase in CO₂ emissions. For instance, investments in technology and education encourage the innovation of energy-saving technologies and low-carbon consumption patterns, thereby reducing CO₂ emissions [54]. Based on the above reasons, we have chosen control variables consistent with the existing literature (see [52,55,56]), which are foreign direct investment (FDI), international trade (TRADE), government expenditure (GOV), per capita real GDP (GDP), and primary energy consumption per capita (PT). The definitions of the variables and data sources are presented in

Table 2. Variable definitions and data sources.

Symbol	Variable	Unit	Data Source
Explained variable
CO2	Transport sector CO2 emissions	Metric tons per capita	Our world in Data
Explanatory variable
OIL	Crude oil import prices	USD/barrel	OECD
Threshold variables
EPS	Environmental policy stringency index	Index (from 0 to 6)	OECD
MBI	Market-based instrument index	Index (from 0 to 6)	OECD
NMBI	Non-market-based instrument index	Index (from 0 to 6)	OECD
TS	Technology support policy index	Index (from 0 to 6)	OECD
Control variables
GDP	Real GDP per capita	Constant 2015 USD	World bank
FDI	Foreign direct investment	As % of GDP	World bank
TRADE	Trade	As % of GDP	World bank
GOV	Government spending	As % of GDP	World bank
PT	Primary energy consumption per capita	Million tons	OECD

.

For the variables of per capita CO₂ emissions in the transport sector (CO₂), oil price (OIL), and per capita primary energy consumption (PT), we have applied logarithmic transformations. This means that the coefficient for the oil price variable we are concerned with is an elasticity coefficient, indicating the percentage change in per capita CO₂ emissions in the transport sector for every 1% increase in oil prices. Descriptive statistics for the variables used in the regression analysis are shown in

Table 3. Descriptive statistics of variables.

Variable	Obs	Mean	Standard Deviation	Min	Max
$lnC{O}_2$	990	0.430	0.817	−2.579	1.841
$lnOIL$	990	3.669	0.667	2.456	4.769
$EPS$	990	1.885	1.137	0	4.722
$MBI$	990	1.066	0.801	0	4.167
$NMBI$	990	3.099	1.945	0	6
$TS$	990	1.490	1.253	0	6
$lnGDP$	975	9.870	1.050	6.271	11.375
$FDI$	976	3.641	7.913	−40.087	86.479
$TRADE$	979	73.612	40.115	15.156	252.250
$GOV$	979	18.478	4.161	5.694	30.324
$lnPT$	990	−12.689	0.637	−14.949	−11.676

.

4. Results

4.1. Panel Threshold Model Results

To prevent spurious regression results caused by non-stationary time series, we first conducted panel unit root tests. Since the panel data we used were unbalanced, we utilized the Im–Pesaran–Shin test. The results of the panel unit root tests are shown in

Table 4. Unit root tests.

Variable	Im–Pesaran–Shin Test (Level)
	C	C&T
$lnC{O}_2$	$3.537$	${-1.662}^{**}$
$lnOIL$	${-1.900}^{**}$	${-8.994}^{***}$
$EPS$	${-1.869}^{**}$	${-2.452}^{***}$
$MBI$	$0.250$	${-3.430}^{***}$
$NMBI$	${-3.574}^{***}$	$-0.551$
$TS$	${-3.058}^{**}$	${-2.308}^{***}$
$lnGDP$	$5.581$	${-6.108}^{***}$
$FDI$	${-13.331}^{***}$	${-10.760}^{***}$
$TRADE$	$1.716$	${-5.053}^{***}$
$GOV$	${-2.463}^{***}$	${-2.872}^{***}$
$lnPT$	$0.751$	${-4.963}^{***}$

Note: *** p < 0.01, ** p < 0.05.

. As seen from Table 4, the level sequences of all variables passed the significance tests, indicating that all variable series were stationary.

Next, we constructed the LR statistic and determined the number of thresholds in the model based on 300 bootstrap repetitions. The results of the threshold effect tests are shown in

Table 5. Threshold effect tests (threshold variables: EPS).

Threshold	Hypothesis	F-Value	p-Value	Threshold Estimated Value	95% Confidence Interval
Threshold variables: Environmental policy stringency index (EPS)
Single	H0: no threshold H1: one threshold	${28.19}^{***}$	$0.000$	$2.94$	$[2.82,2.97]$
Double	H0: one threshold H1: double threshold	${11.37}^{**}$	$0.040$	$3.56$	$[3.33,3.61]$
Triple	H0: double threshold H1: triple threshold	$8.29$	$0.417$

Note: *** p < 0.01, ** p < 0.05.

, where both the F-test of the threshold effect and the bootstrapped p-value indicate that the hypotheses of single and double thresholds are rejected at the 1% significance level. This demonstrates a significant double-threshold effect in the impact of oil prices on CO₂ emissions in the transport sector when the environmental policy stringency (EPS) index is used as the threshold variable. Therefore, a double-threshold model should be adopted.

We conducted regression estimates using a double-threshold model, and the results are presented in

Table 6. Baseline regression results.

Variables	Threshold Model
$lnOIL(EPS\le 2.94)$	$-0.025$ $\left(0.019\right)$
$lnOIL(2.94<EPS\le 3.56)$	${-0.038}^{**}$ $\left(0.016\right)$
$lnOIL(EPS>3.56)$	${-0.050}^{***}$ $\left(0.018\right)$
$EPS$	$0.009$ $\left(0.019\right)$
$lnGDP$	${0.547}^{***}$ $\left(0.102\right)$
$FDI$	$-0.0001$ $\left(0.0002\right)$
$TRADE$	$0.0009$ $\left(0.001\right)$
$GOV$	$-0.012$ $\left(0.009\right)$
$lnPT$	${0.607}^{***}$ $\left(0.150\right)$
$N$	$964$
$F$	$50.03$
$R^2$	$0.902$

Note: Robust standard errors in parentheses, *** p < 0.01, ** p < 0.05.

. The results indicate that when the EPS is below 2.94, the impact of oil prices on CO₂ emissions in the transport sector is minimal, with a coefficient of −0.025, which is not statistically significant. When the EPS is between 2.94 and 3.56, the impact of oil prices on CO₂ emissions in the transport sector increases, with a coefficient of −0.038, and is statistically significant at the 5% level. When the EPS exceeds 3.56, the impact of oil prices on CO₂ emissions in the transport sector further increases, with a coefficient of −0.050, and is statistically significant at the 1% level.

4.2. Robustness Test

In our analysis using the panel threshold model, we found that the negative impact of oil prices on CO₂ emissions in the transport sector strengthens with increasing environmental policy stringency, indicating that environmental policy stringency plays a positive moderating role. Therefore, we further conducted a robustness test using a moderation effect model. The formula for the moderation effect model is as follows:

$$\begin{array}{c}{lnC{O}_2}_{i,t}={\alpha }_0+{\alpha }_1{lnOIL}_{i,t}+{\alpha }_2{EPS}_{i,t}+{\alpha }_3{lnOIL}_{i,t} \ast {EPS}_{i,t}+{\alpha }_4{Control}_{i,t}+{\mu }_{i,t}\end{array}$$

(5)

where ${lnC{O}_2}_{i,t}$ is the outcome variable, ${lnOIL}_{i,t}$ is the treatment variable, ${EPS}_{i,t}$ is the moderating variable, and ${lnOIL}_{i,t} \ast {EPS}_{i,t}$ is the interaction term. The results of the moderation effect model are illustrated in Figure 1, which plots the marginal effects of the interaction term. Specifically, Figure 1 shows how the environmental policy stringency index, ${EPS}_{i,t}$, affects the relationship between crude oil prices, ${lnOIL}_{i,t}$, and CO₂ emissions in the transport sector, ${lnCO2}_{i,t}$. From Figure 1, it can be observed that as ${EPS}_{i,t}$ increases, the marginal effect of ${lnOIL}_{i,t}$ on ${lnC{O}_2}_{i,t}$ gradually increases and becomes statistically significant. This result is consistent with our baseline regression findings and similarly indicates that at low levels of EPS, an increase in oil prices does not lead to a decrease in CO₂ emissions in the transport sector; only when EPS exceeds a certain threshold do oil prices impact CO₂ emissions.

To further ascertain the robustness of the empirical results, we arbitrarily divided the entire sample into two subsamples using an EPS threshold of 2.94: a high-ESP-level sample and a low-ESP-level sample. Previous panel threshold regression results indicated that when EPS is below 2.94, oil prices do not impact the transport sector’s CO₂ emissions; however, when EPS exceeds 2.94, the effect of oil prices on CO₂ emissions in the transport sector is statistically significant. Therefore, if the empirical results are robust, the same outcomes should be observed in subsample regressions. The results of these regressions, as shown in

Table 7. Results of the robustness test.

Variables	Sample (EPS < 2.94)	Sample (EPS > 2.94)
$lnOIL$	$-0.004$ $\left(0.019\right)$	${-0.037}^{**}$ $\left(0.016\right)$
$Control$	$Y$	$Y$
$FE$	$Y$	$Y$
$N$	$741$	$223$
$F$	$35.01$	$26.08$
$R^2$	$0.82$	$0.41$

Note: Robust standard errors in parentheses, ** p < 0.05.

, reveal that $lnOIL$ is not significant in the low-ESP-level sample but is significantly negative in the high-ESP-level sample, further verifying the robustness of the empirical findings.

4.3. Mechanism Analysis

Environmental policy is composed of three types of policies, which may represent three different impact mechanisms. To conduct a mechanism analysis, we use three subindicators of the environmental policy stringency index: 1. market-based instruments (MBIs) (such as carbon taxes and trading schemes, and other pollution pricing policies), 2. non-market-based instruments (NMBIs) (emission limits and standards policies), and 3. technology support (TS) policies (policies supporting the innovation and adoption of clean technologies) [50].

First, we determined the number of thresholds and their estimates for the models corresponding to the three threshold variables by conducting threshold effect tests. The results are shown in

Table 8. Threshold effect tests (subindices of the environmental policy stringency index).

Threshold	Hypothesis	F-Value	p-Value	Threshold Estimated Value	95% Confidence Interval
Threshold variables: Market-based instrument index (MBI)
Single	H0: no threshold H1: one threshold	${27.41}^{***}$	$0.000$	$2.17$	$[1.92,2.33]$
Double	H0: one threshold H1: double threshold	${7.768}^{**}$	$0.043$	$0.17$	$[0.00,0.33]$
Triple	H0: double threshold H1: triple threshold	$13.04$	$0.333$
Threshold variables: Non-market-based instrument index (NMBI)
Single	H0: no threshold H1: one threshold	${36.38}^{***}$	$0.000$	$5.25$	$[4.86,5.50]$
Double	H0: one threshold H1: double threshold	$8.22$	$0.323$
Triple	H0: double threshold H1: triple threshold	$5.67$	$0.533$
Threshold variables: Technology support policy index (TS)
Single	H0: no threshold H1: one threshold	${40.24}^{***}$	$0.000$	$0.25$	$[0.00,0.50]$
Double	H0: one threshold H1: double threshold	${14.19}^{***}$	$0.000$	$3.25$	$[2.62,3.50]$
Triple	H0: double threshold H1: triple threshold	$5.94$	$0.290$

Note: *** p < 0.01, ** p < 0.05.

. The findings indicate that models with MBIs and TS as threshold variables exhibit double-threshold effects and, thus, a double-threshold model should be used for estimation. Meanwhile, the model with NMBIs as the threshold variable shows only a single-threshold effect, suggesting that a single-threshold model should be used for estimation.

The regression results for the mechanism analysis are shown in

Table 9. Mechanism analysis results.

Variables	Threshold Model (MBI)	Variables	Threshold Model (NMBI)	Variables	Threshold Model (TS)
$lnOIL(MBI\le 0.17)$	$-0.019$ $\left(0.022\right)$	$lnOIL(NMBI\le 5.25)$	${-0.037}^{**}$ $\left(0.015\right)$	$lnOIL(TS\le 0.25)$	${-0.056}^{**}$ $\left(0.023\right)$
$lnOIL(0.17<MBI\le 2.17)$	${-0.037}^{**}$ $\left(0.016\right)$	$lnOIL(NMBI>5.25)$	${-0.054}^{***}$ $\left(0.015\right)$	$lnOIL(0.25<TS\le 3.25)$	${-0.032}^{*}$ $\left(0.017\right)$
$lnOIL(MBI>2.17)$	$-0.017$ $\left(0.018\right)$			$lnOIL(TS>3.25)$	${-0.045}^{**}$ $\left(0.017\right)$
$MBI$	${-0.065}^{***}$ $\left(0.023\right)$	$NMBI$	$-0.005$ $\left(0.009\right)$	$TS$	$0.014$ $\left(0.012\right)$
$lnGDP$	${0.635}^{***}$ $\left(0.085\right)$	$lnGDP$	${0.624}^{***}$ $\left(0.101\right)$	$lnGDP$	${0.515}^{***}$ $\left(0.087\right)$
$FDI$	$0.00002$ $\left(0.0002\right)$	$FDI$	$-0.00005$ $\left(0.0003\right)$	$FDI$	$0.0003$ $\left(0.0003\right)$
$TRADE$	$0.0008$ $\left(0.001\right)$	$TRADE$	$0.001$ $\left(0.001\right)$	$TRADE$	$0.0002$ $\left(0.001\right)$
$GOV$	$-0.011$ $\left(0.008\right)$	$GOV$	$-0.011$ $\left(0.008\right)$	$GOV$	${-0.017}^{*}$ $\left(0.01\right)$
$lnPT$	$0.609$ $\left(0.134\right)$	$lnPT$	$0.556$ $\left(0.157\right)$	$lnPT$	${0.624}^{***}$ $\left(0.149\right)$
$N$	$964$	$N$	$964$	$N$	$936$
$F$	$47.75$	$F$	$78.28$	$F$	$61.44$
$R^2$	$0.906$	$R^2$	$0.893$	$R^2$	$0.901$

Note: Robust standard errors in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.

. The results indicate the following: (1) There is an optimal range for MBIs from 0.17 to 2.17, within which the impact of oil prices on CO₂ emissions in the transport sector is negative and statistically significant. (2) Across all ranges of NMBIs, the impact of oil prices on CO₂ emissions in the transport sector is consistently negative and significant. The impact intensifies when NMBIs exceed 5.25. (3) Across all ranges of TS, the impact of oil prices on CO₂ emissions in the transport sector is negative and significant, exhibiting a pattern where the effect is stronger at the extremes and weaker in the middle.

4.4. Discussion of Results

We used a forest plot to summarize the results of the panel threshold analysis and the subsequent mechanism analysis results, as shown in Figure 2.

Based on the results of the panel threshold analysis, we can conclude that the stricter the environmental policies of a country, the more effective the reduction in CO₂ emissions from the transport sector due to increases in oil prices. According to the results, when the environmental policy stringency (EPS) is at a low level, the rise in oil prices has little impact on CO₂ emissions. This indicates that although higher oil prices lead to reduced oil consumption and thus decreased CO₂ emissions from oil, the rise in energy prices reflects the scarcity of oil, which prompts businesses to switch to cheaper alternatives, such as coal, ultimately leading to an increase in CO₂ emissions [57,58,59]. We found that when the EPS increases to the first threshold of 2.94, the rise in oil prices significantly reduces CO₂ emissions in the transport sector. This may be because strict environmental policies discourage businesses from choosing coal as an alternative energy source, instead prompting them to opt for renewable energy. Wang et al.’s [60] research indicates that strict environmental policies will drive the transformation of economies towards renewable energy, thereby reducing CO₂ emissions. Overall, our empirical results effectively reveal the importance of environmental policy in the relationship between oil prices and CO₂ emissions in the transport sector.

Through mechanism analysis, we further clarified how environmental policy stringency influences the relationship between oil prices and CO₂ emissions in the transport sector through three different types of environmental policies, each operating through distinct mechanisms.

The analysis results using market-based instruments (MBIs) as the threshold variable indicate that there is an optimal range for MBIs (0.17 < MBI ≤ 2.17). MBIs consist of taxes and trading schemes. Environmental tax policy, a common market economic tool that follows the “polluter pays” principle [61], directly increases the cost of CO₂ emissions. The literature shows that reasonable environmental taxes can reduce CO₂ emissions and promote the use of renewable energy, thereby lowering CO₂ emissions [61,62]. Trading schemes are also a common market-based environmental policy that uses economic incentives to reward reductions in CO₂ emissions, exemplified by the European Union Emissions Trading System launched in 2005 [63]. Heinrichs et al. [64] found that incorporating the transport sector into the European Union’s carbon emissions trading scheme can effectively reduce CO₂ emissions. According to our research, environmental taxes (direct costs) and trading schemes (opportunity costs) both directly affect the cost of petroleum use in the transport sector. When the costs induced by market-based instruments (MBIs) are low, increases in oil prices may lead the transport sector to opt for alternatives like coal, resulting in increased CO₂ emissions. However, when the costs induced by MBIs are high, rising oil prices encourage enterprises to reduce their use of petroleum products and switch to renewable energy sources, thereby effectively lowering CO₂ emissions in the transport sector [65]. Notably, as the stringency of MBIs further increases, the impact of rising oil prices on CO₂ emissions in the transport sector becomes negligible. This may be because, under high market incentives, enterprises have minimized the proportion of petroleum products in their energy consumption, making the impact of oil price fluctuations on costs no longer significant.

The analysis using non-market-based instrument (NMBI) policies as the threshold variable shows that, across all ranges, the impact of oil prices on CO₂ emissions in the transport sector is significantly negative and strengthens with the increased stringency of NMBI. This suggests that NMBI policies, represented by tools such as emission limits and the maximum allowable sulfur content in diesel, impose direct costs on CO₂ emissions in the transport sector [66]. Compared to MBI policies, the effects of NMBI policies are evidently stronger. This indicates that NMBIs, through stringent standards and restrictions, effectively reduce CO₂ emissions in the transport sector. As policies tighten, enterprises respond more actively in reducing emissions, further consolidating the suppression of CO₂.

The analysis results show that when technology support (TS) policies are used as the threshold variable, the impact of oil prices on CO₂ emissions in the transport sector is significantly negative across all intervals (with significance at the 10% level within the 0.25 < TS ≤ 3.25 range). These findings indicate that reducing CO₂ emissions in the transport sector cannot rely solely on traditional environmental policy changes; it necessitates technology support policies to drive technological innovation [67]. Technological innovation can be divided into two aspects: technological innovation in transportation components (such as electric vehicles) and technological innovation in alternative fuels (especially renewable energy). In terms of transportation components, the widespread adoption of electric vehicles is a key measure for reducing CO₂ emissions, making the promotion of “zero-emission” and low-energy-consuming electric vehicles particularly important [26,68]. Additionally, there is evidence that improvements in vehicle technology have reduced greenhouse gas emissions in the transport sector of European countries [69,70]. In the area of alternative fuels, technological innovation can expand the production of renewable energy and reduce its costs, making the transportation industry more likely to choose green energy over high-emission fuels such as coal when oil prices rise [71,72,73]. Therefore, technology support policies, by driving these innovations, can significantly reduce CO₂ emissions in the transport sector.

5. Conclusions

The main results indicate that the impact of oil prices on CO₂ emissions in the transport sector varies with the stringency of environmental policies. When environmental policies are lenient, the impact of oil prices is not significant; however, as the stringency of the policies increases, the suppressive effect of oil prices on CO₂ emissions strengthens. Mechanistic analysis shows that different types of environmental policies affect the relationship between oil prices and CO₂ emissions differently. Market-based policies, such as carbon taxes and trading schemes, significantly enhance the suppressive effect of oil prices on CO₂ emissions within certain ranges, whereas non-market policies, such as emission limits, demonstrate a suppressive effect on CO₂ emissions across all ranges. Technology support policies also significantly strengthen the negative impact of oil prices on CO₂ emissions across all ranges.

In light of the above conclusions, when formulating environmental policies for the transport sector, governments should pay particular attention to how the environmental policy stringency influences the impact of oil prices on CO₂ emissions in the transport sector. Based on our empirical analysis results, governments can develop targeted CO₂ emission reduction policy combinations for the transport sector, aligned with the threshold stages of the environmental policy stringency index. For instance, in countries where environmental policy stringency is below the threshold, such as Brazil and the United States, governments should raise emission standards and strengthen regulation and introduce or enhance market-based policies such as carbon taxes and trading schemes. Additionally, enhancing technological support policies, encouraging enterprises to increase investments in green energy technologies, and fostering innovation and application in clean technologies are also critical. These measures will ensure that enterprises can opt for clean energy alternatives instead of oil when oil prices rise, ultimately reducing CO₂ emissions in the transport sector.

This study has certain limitations. First, due to the availability of data, the sample in this study primarily comprises developed countries, such as OECD member states, and while it also includes major developing countries like China and India, the consideration of developing countries remains insufficient. These countries often face more complex environmental challenges; thus, their actual conditions are not fully reflected in the study. Secondly, this study primarily focuses on the threshold effects of environmental policy stringency, providing key insights into the impact of environmental policies. However, we acknowledge that there may be other potential impact channels that have not been explored due to the constraints of this paper’s length. These channels have not been thoroughly analyzed in this study. Future research will aim to identify and evaluate these yet-to-be-covered impact channels to provide a more comprehensive understanding of how environmental policies influence the relationship between oil prices and carbon emissions in the transport sector.