The Impact of Fossil Energy Prices on Carbon Emissions: The Dual Mediation of Energy Efficiency and Renewable Energy

Abstract

This paper empirically examines the nexus between fossil energy prices and carbon emissions using a balanced panel of 119 economies spanning the period from 1990 to 2023. The baseline regression results indicate that a 1% rise in fossil energy prices results in a 0.009% reduction in CO₂ emissions, equivalent to approximately 3.1 million tons of CO₂. Further analysis reveals two key mechanisms. First, energy efficiency partially mediates the price–emission relationship: higher prices significantly improve efficiency, which in turn reduces CO₂ emissions, although a rebound effect of 13.6% offsets part of the expected savings. Second, renewable energy penetration serves as an additional pathway, with higher prices accelerating renewable adoption and thereby contributing to carbon mitigation. Overall, the findings confirm the direct and indirect impacts of fossil energy prices on emissions, underscoring their role as an effective lever for achieving global sustainability targets. Policy implications include the need to align fossil energy prices with true economic and environmental costs, while complementing price mechanisms with efficiency standards and renewable incentives to counterbalance hirebound effects.

1. Introduction

Over the past decades, progressive industrialization and globalization have been mostly fueled, in large part, by enormous energy consumption resulting in significant environmental pollution, particularly in the form of carbon dioxide (CO₂) emissions. According to British Petroleum, global CO₂ emissions increased from 22,680.4 Mt in 1990 to 39,023.9 Mt in 2023 (see Figure 1), with 1.7% growth on average per year. To address environmental sustainability, countries committed to the Paris Agreement aim to restrict global warming to 2 degrees Celsius over pre-industrial levels, with a 1.5-degree Celsius target in mind.

A key aspect of carbon emission mitigation addresses energy consumption [1]. Fossil energy prices have a significant impact on the dynamics of energy supply and consumption [2]. Therefore, integrating fossil energy prices into carbon reduction decisions is vital for sustainable development. The impact of fossil energy prices on CO₂ emissions mainly includes the “cost pressure effect” and the “technology dynamics effect” [3,4]. Increased energy price lessens the elasticity of energy supply and reduces energy consumption [5,6]. Shephard’s Lemma theory suggests that when fossil energy prices are high, firms are more willing to switch to fossil energy alternatives or improve efficiency to maximize profits [3]. Consistent with this mechanism, recent cross-country evidence shows that reduced emissions of CO₂ are linked to increased carbon pricing, with elasticities varying by sector and policy design [7].

As research by [8,9,10] suggested, improving energy efficiency is essential to global carbon reduction. Therefore, a substantial amount of literature has emerged to study the link between energy efficiency and energy consumption; [11], for example, concluded that raising fossil energy prices might greatly enhance energy efficiency. Energy efficiency and market electricity prices were found to be positively correlated in the short and long term by [12].

However, extant research has suggested that the impact of energy efficiency on energy consumption depends on the energy rebound effects [13,14]. Energy rebound effects are twofold. Energy efficiency, on the one hand, results in lower energy consumption. On the other hand, while improved energy efficiency reduces energy consumption directly, the associated economic growth can induce additional energy use, thereby offsetting part of the expected savings [15,16]. Therefore, it is important to take the rebound effect into consideration when examining the effects of energy efficiency on energy consumption and carbon reduction.

Existing research on CO₂ emissions mostly focuses on individual regions, such as China, the G7, Pakistan, and BRICS [17,18,19,20,21], and lacks comprehensive surveys covering long time series and multi-country samples. These findings require validation based on richer samples. Complementing this gap, a recent global STIRPAT study covering 136 countries documents the role of structural drivers—particularly urbanization—in shaping CO₂ outcomes with heterogeneous effects across income groups [22]. In addition, prior studies have conducted preliminary examinations of the direct effects of changes in fossil energy prices on CO₂ emissions [10,23,24]. However, whether and when energy efficiency serves as an influential pathway remain to be fully explored.

There are two contributions in this article. First, our analysis adds to the relevant literature on CO₂ emissions by revealing the underpinning relationship among energy price, energy efficiency, and CO₂ emissions. Prior literature has examined the relationships between fossil energy prices, energy efficiency, and CO₂ emissions separately [11]. However, these studies failed to examine energy efficiency as the mechanism in shaping the relationship between fossil energy prices and carbon reduction. Second, it introduces energy efficiency as a mediator, offering a systematic framework to assess indirect effects. Incorporating the rebound effects theory, it demonstrates how efficiency impacts the energy price-CO₂ relationship, thereby enriching discussions on CO₂ emission factors across 119 countries.

2. Literature Review and Hypothesis Development

2.1. Direct Effects of Fossil Energy Prices on CO₂ Emissions

Several academics have recognized that fossil energy prices can significantly lower CO₂ emissions [4,10,23,25]. At the firm level, evidence from Sweden indicates that carbon pricing results in a significant reduction in the emission intensity of manufacturing firms, corroborating the price-induced abatement channel [26]. Extant studies on the relationship between fossil energy prices and carbon reduction are listed in

Table A1. Current research of fossil energy prices-carbon emissions.

Authors	Region	Year	Correlation
Li et al. (2020)	China	2002–2016	Negative
Umar et al. (2021)	Thirteen African nations	1990–2017	Negative
Wu et al. (2023)	EU	2000–2018	Negative
Hammoudeh et al. (2022)	USA	2006–2013	inverted U-shape
Al-Mulali et al. (2023)	27 developed economies	1990–2012	Negative
Mujtaba et al. (2021)	India	1986–2014	two-way impact
Mensah et al. (2019)	22 African countries	1990–2015	Negative
Yuan et al. (2010)	China’s industrial sector	1993–2007	Negative
Mukhtarov et al. (2022)	Iran	1980–2019	Negative

. For instance, ref. [27] found that the rise in fossil energy prices was an effective means of reducing CO₂ emissions, based on the Chinese sample data. Ref. [28] reached the same findings on the beneficial role of fossil energy prices in reducing CO₂ emissions in thirteen African nations. Ref. [29] found a positive correlation between fossil energy prices and CO₂ emissions in EU countries. From a regional perspective, recent European evidence highlights how economic activity and fossil-fuel use jointly shape fossil CO₂ trajectories, informing the design of complementary structural policies [30].

Some researchers have confirmed that higher fossil energy prices dramatically reduce energy consumption in 22 African countries, the United States, and China [31,32,33]. The increased fossil energy prices raise costs and reduce profits, which force companies to boost the energy efficiency of their production through technological innovation, especially for energy-intensive petrochemical enterprises [25].

Furthermore, increased fossil energy prices pressure energy users to consider the marginal productivity of input factors, which, in turn, reduces energy use [6]. Many scholars support this view from an economic perspective that energy demand will decline in the face of rising fossil energy prices, resulting in carbon reduction effects [4,10]. Specifically, ref. [34] revealed that coal and natural gas prices impacted EUA prices through dark and spark spreads, with a positive correlation between the dark spread and EUA prices. Ref. [35] found that the deployment of renewable energy sources significantly reduced CO₂ emissions and decreased the demand for EUA. Ref. [36] demonstrated that a coal price shock had a negative impact on EUA prices, resulting from fuel switching to natural gas. Ref. [37] revealed that higher prices for carbon-intensive fuels tend to induce a shift toward low-carbon energy, thereby mitigating CO₂ emissions and reducing the price of carbon credits. Sectoral evidence further shows that higher oil prices reduce transport CO₂ when environmental policy stringency exceeds estimated thresholds, implying complementarity between prices and regulation [38].

Therefore, rising fossil energy prices and decreased energy supply elasticity caused by tighter supply-side costs will restrain energy consumption and ultimately mitigate CO₂ emissions [6]. In this regard, this article proposes the following hypothesis:

Hypothesis 1.

The rise in fossil energy prices can effectively reduce carbon emissions.

2.2. Indirect Effects of Fossil Energy Prices on Carbon Emissions

The positive effects of fossil energy prices on energy efficiency have been well-documented [11,39,40]. The association between energy efficiency improvements and carbon reduction, however, is highly controversial. One stream of studies has suggested that energy efficiency is an effective path to alleviate CO₂ emissions [21]. Ref. [41], for example, proposed that improved energy efficiency in China’s pulp and paper industry saved energy and slowed the rate of CO₂ emissions. Ref. [9] also emphasized that by 2060, one-fifth of Saudi Arabia’s decarbonization might be attributed to advances in energy efficiency.

By contrast, other scholars argue that improved energy efficiency may produce a certain “rebound effect” or “carbon-increasing effect” [13,14]. Energy efficiency improvements may stimulate new energy demand, which makes energy savings often less effective than expected in reducing CO₂ emissions [42].

Table A2. Current research on rebound effect of energy efficiency.

Authors	Specification	Technique	Region	Year	Rebound Effect
Shao et al. (2019)	Cobb–Douglas		China	1991–2016	93.96%
Adha et al. (2021)	KHREM model		Indonesia	2002–2018	87.2%/45.5%
Kong et al. (2023)	Cobb–Douglas		Beijing	2015–2019	69.64%
Steren et al. (2022)		2SLS	Israel	2007–2016	62%
Omondi et al. (2023)	Cobb–Douglas		Kenya	2009–2019	15.64%
Baležentis et al. (2021)		OLS	EU	2011–2015	2.55%
Bhringer and Rivers (2018)	CGE		US	2010	63.3%
Berner et al. (2021)	S-FAVAR		Italy, UK, USA	2008–2019	78–101%
Zheng et al. (2022)		2SLS	China	2003–2017	123%

categorizes rebound effects into partial rebound (or zero-rebound or super-conservation) with values < 1, resulting in lower carbon emissions [43,44,45]. High- and middle-income countries are less prone to increased energy use with improved efficiency [42,46,47]. The backfire effect (≥1) may offset CO₂ reduction from efficiency gains [48,49]. Therefore, the magnitude of the rebound effect becomes critical to whether fossil energy prices affect CO₂ emissions through energy efficiency. Building on this framework, we put forth the following hypothesis:

Hypothesis 2.

When rebound efficiency values are less than 1, fossil energy prices can reduce CO₂ emissions through energy efficiency.

The specific theoretical mechanism is shown in Figure 2.

3. Data and Method

3.1. Model Establishment

The impact of human activity on the environment is commonly described by the IPAT model, which was first put forth by Ehrlich and Holdren in 1971 [50]. This model, which represents the environmental impact as a product of population, wealth, and technology, is considered a vital tool in environmental and social science research for assessing the impact of different factors on the environment and formulating corresponding policies. The expression is as follows:

$$I=PAT$$

(1)

Equation (1) provides only a simple framework and assumes strict ratios between the factor coefficients, which makes it difficult to effectively and comprehensively capture how the environment is affected by changes in the three variables over time. Subsequently, refs. [51] developed the STIRPAT model with stochastic effects based on the IPAT model for population, per capita gross domestic product (GDP), and technology, which shows the following equation:

$$I_{it}=a_{it}{P^b}_{it}{A^c}_{it}{T^d}_{it}{e}_{it}$$

(2)

where $a$ indicates the constant term, and $b$, $c,$ and $d$ are the parameters of P, A, and T, respectively. $e$ represents a random error item. The subscript $i$ ($i$ = 1, 2, …, n) refers to the country, and the subscript $t$ ($t$ = 1, 2, …, T) refers to the time period. Because the STIRPAT model is nonlinear, it is necessary to take the logarithm of both sides of the equation to reduce heteroscedasticity. The model expression after taking logarithms is as follows:

$${\mathrm{l}\mathrm{n}I}_{it}=a_{it}+b\left(\mathrm{l}\mathrm{n}{P}_{it}\right)+c\left(\mathrm{l}\mathrm{n}{A}_{it}\right)+d\left(\mathrm{l}\mathrm{n}{T}_{it}\right)+e_{it}$$

(3)

To more comprehensively identify causal disturbances, some studies extend the STIRPAT model by including additional explanatory variables. For instance, refs. [27,52] incorporated urbanization and energy structure into the STIRPAT model, respectively, highlighting the importance of these factors in influencing CO₂ emissions. Urbanization and energy consumption structure are therefore incorporated into the STIRPAT model by drawing on the above approach, and the model is modified as follows.

$$\mathrm{l}\mathrm{n}{CO}_{2it}=\alpha +{\beta }_1\mathrm{l}\mathrm{n}{FEP}_{it}+{\beta }_2\mathrm{l}\mathrm{n}{GDP}_{it}+{\beta }_3\mathrm{l}\mathrm{n}{POP}_{it}+{\beta }_4\mathrm{l}\mathrm{n}{TEC}_{it}+{\beta }_5\mathrm{l}\mathrm{n}{ES}_{it}+{\beta }_6\mathrm{l}\mathrm{n}{UR}_{it}+e_{it}$$

(4)

where $\mathsf{\alpha }$ is the constant term, ${\beta }_1-{\beta }_6$ refer to the parameters that need to be tested. $e_{it}$ denote the error term. ln is the logarithmic form of variables. CO₂ represents the carbon emissions of country i at time period t. EP represents the energy price. GDP, POP, TEC, ES, and UR indicate the GDP, total population, technological progress, energy structure, and urbanization of each country, respectively.

In the theoretical hypothesis section, this article points out that fossil energy prices not only directly affect CO₂ emissions but also influence CO₂ emissions via energy efficiency. To verify whether energy efficiency can weaken CO₂ emissions, this article draws on [53] definition of the energy efficiency rebound effect (R), and constructs the following equation: 1 plus energy consumption elasticity in relation to energy efficiency [54] as follows:

$$R=1+\xi =1+{\displaystyle \frac{\partial \mathrm{l}\mathrm{n}AEC}{\partial \mathrm{l}\mathrm{n}EEI}}$$

(5)

where $\xi$ refers to the elasticity of actual energy consumption efficiency, EEI denotes energy efficiency, and AEC denotes actual energy consumption.

The elasticity of actual energy consumption efficiency in Equation (5) can be simulated in advance through Equation (6), as follows:

$$\mathrm{I}\mathrm{n}{AEC}_{it}=\gamma +\xi \mathrm{I}\mathrm{n}{EEI}_{it}+{\mathsf{\gamma }}_1\mathrm{I}\mathrm{n}{FEP}_{it}+{\gamma }_2\mathrm{I}\mathrm{n}{GDP}_{it}+{\gamma }_3\mathrm{I}\mathrm{n}{POP}_{it}+{\gamma }_4\mathrm{I}\mathrm{n}{TEC}_{it}+{\gamma }_5\mathrm{I}\mathrm{n}{ES}_{it}+{\gamma }_6\mathrm{I}\mathrm{n}{UR}_{it}+e_{it}$$

(6)

In this equation, $\mathsf{\gamma }$ indicates the constant term, and ${\gamma }_1-{\gamma }_6$ denote the estimated coefficients. The additional variables and symbols are identical to those in Equation (4).

Table A3. The types of rebound effects and their corresponding elasticity values.

Rebound Efficient	Elasticity	Description
R > 1	ξ > 0	Backfire
R = 1	ξ = 0	Full rebound
0 < R < 1	−1 < ξ < 0	Partial rebound
R = 0	ξ = −1	Zero rebound
R < 0	ξ < −1	Super-conservation

describes various types of rebound effects and their corresponding elasticity values. It includes reverse effects (R > 1, ξ > 0), complete rebound (R = 1, ξ = 0), partial rebound (0 < R < 1, −1 < ξ < 0), zero rebound (R = 0, ξ = −1), and super energy-saving (R < 0, ξ < −1) [55]. The reverse effects cause CO₂ emissions and energy consumption to rise even when efficiency is improved. A complete rebound has no substantial impact on energy savings, while a partial rebound produces less savings than expected. Zero rebound matches expected savings, and super energy-saving exceeds expectations.

This article incorporates energy efficiency as an intermediate variable into Equation (4) to verify whether a mediating effect exists. The existence of a mediating effect must satisfy three conditions: (1) it is clear that the core explanatory variable influences the dependent variable before the intermediate variable is included; (2) the core explanatory variable significantly affects the intermediate variable; and (3) after the intermediate variable is included, the intermediate variable significantly influences the dependent variable, but the influence of the core explanatory variable on the dependent variable decreases, or even becomes insignificant. Therefore, this article constructs Equation (7) to verify the second condition for the existence of a mediating effect, and the specific model construction is as follows:

$$EEI=\alpha +{\beta }_1\mathrm{I}\mathrm{n}{FEP}_{it}+{\beta }_2\mathrm{I}\mathrm{n}{GDP}_{it}+{\beta }_3\mathrm{I}\mathrm{n}{POP}_{it}+{\beta }_4\mathrm{I}\mathrm{n}{TEC}_{it}+{\beta }_5\mathrm{I}\mathrm{n}{ES}_{it}+{\beta }_6\mathrm{I}\mathrm{n}{UR}_{it}+e_{it}$$

(7)

Among them, EEI denotes the energy efficiency, which is an intermediate variable. If the coefficient β₁ is significant, it means that the core explanatory variable significantly affects the intermediate variable, thus satisfying the second condition for the mediating effect. Based on the verification of the first and second conditions for the existence of a mediating effect, we further construct Equation (8) to verify the third condition for the mediating effect:

$$In{CO}_{2it}=\alpha +{\beta }_{1}InF{EP}_{it}+{\beta }_{2}In{GDP}_{it}+{\beta }_{3}In{POP}_{it}+{\beta }_{4}In{TEC}_{it}+{\beta }_{5}In{ES}_{it}+{\beta }_{6}In{UR}_{it}+{{\beta }_{7}In{EEI}_{it}+e}_{it}$$

(8)

In this equation, if the coefficient β₁ significantly decreases or becomes insignificant, but the significance test is passed by the coefficient β₇, it satisfies the third condition for the existence of a mediating effect.

3.2. Variable Descriptions and Data Sources

Considering the significant fluctuations in fossil energy prices since 1990, and the relatively complete sample data available, the research sample encompasses data for 119 countries from 1990 to 2023, with the aim of exploring the relationships among fossil energy prices, CO₂ emissions, and energy efficiency. The sample is selected based on geographical distribution, level of economic development, energy structure, population size, and the diversity of energy policies and regulations, with a priority given to countries with reliable long-term data.

Dependent variable: Total amount of CO₂ emissions (denoted as CO₂).

The issue of CO₂ emissions is a global concern and closely linked to global climate change. The unit of measurement for this indicator is kilotons. To meet the growing demand, fossil fuels, including coal, oil, and natural gas, are extensively utilized by people, which release a large amount of CO₂ emissions during combustion. With the acceleration of global industrialization and population growth, the demand for energy is also increasing, resulting in an annual rise in CO₂ emissions.

Core explanatory variable: fossil energy prices (denoted as FEP).

This article adopts the calculation method of [53], which selects the price of Brent crude oil as the proxy variable for fossil energy prices, and multiplies it by a country’s annual average actual exchange rate against the US dollar after being adjusted for inflation based on the US actual consumer price index (with 2010 as the reference year) to obtain an energy price index, which reflects the actual price a country pays for a barrel of crude oil in US dollars. This specification is consistent with recent empirical practice that employs Brent as an explanatory variable [56]. For robustness, we re-estimate the models using a Relative Energy Price (REP); detailed results are reported in Section 4.2.1.

Intermediary variable: Energy efficiency (expressed by EEI).

The index is applied as an intermediary variable, gauged by the ratio of total energy supply to GDP (in billion USD, 2015 prices and exchange rates), and employed to quantify the rebound effects of actual energy consumption (AEC).

Control variables:

The control variables include economic activity (expressed by GDP) [57,58], with each country’s 2015 GDP at constant prices as the proxy indicator for regional economic activity; population size (expressed by POP), which is measured by each country’s total population [46,59]; urbanization level (expressed by UR), which is gauged by the ratio of urban population to total population [60,61]; and energy structure (denoted as ES), which is calculated by the share of fossil total energy supply [62,63]. In the process of carbon reduction, technological progress (expressed by TEC) plays an irreplaceable role; this article, accordingly, introduces charges for the use of intellectual property (royalties and license fees) as a measure of technological progress [64].

Among them, the statistics of CO₂, GDP, POP, and TEC are obtained from the World Bank database. The data of both ES and EEI are from the International Energy Agency (IEA).

Table A4. Descriptive statistics of selected variables.

Variable	Definition	Unite	Mean	Std. Dev.	Min	Max
ln CO2	The amount of CO2 emissions	Kiloton	10.245	1.996	5.763	16.400
lnFEP	Fossil energy prices	$	6.702	3.020	0	27.215
lntec	Intellectual property	Thousand $	18.754	3.019	0.634	30.289
lnGDP	Total GDP	Trillion $	4.229	1.910	0.343	9.988
lnPOP	Population size	108	16.356	1.558	12.448	21.086
lnES	Energy structure	%	0.661	0.280	0.020	1
lnUR	Urbanization level	%	0.586	0.217	0.054	1
lnEEI	The energy efficiency index	%	2.250	0.472	0.502	3.742

presents the descriptive statistics and detailed descriptions (i.e., observations, average value, standard deviation, minimum value, and maximum value) after logarithmic application.

The country sample encompasses all major world regions and spans all four income classes (low, lower-middle, upper-middle, high). Income categories are held fixed over time. This breadth reduces regional-composition bias and provides the cross-country variation needed to identify the price–emissions elasticity and its heterogeneity. Figure 3 visualizes per capita CO₂ at four benchmark years, illustrating persistent spatial heterogeneity over time.

4. Results

4.1. Baseline Regression

This article employs pooled ordinary least squares (OLS) [34], fixed effects, and random effects methods to estimate factors influencing CO₂ emissions [65]. The results are presented in

Table 1. Basic result and endogeneity testing.

	(1)	(2)	(3)	(4)	(5)	(6)
Methods	OLS	Fixed Effect	Random Effect	Lag 1	Lag 2	Lag 3
lnFEP	−0.005	−0.009 ***	−0.008 ***
	(0.003)	(0.002)	(0.002)
L.lnFEP				−0.009 ***
				(0.002)
L2.lnFEP					−0.008 ***
					(0.002)
L3.lnFEP						−0.007 ***
						(0.002)
lnTEC	0.012 **	−0.012 ***	−0.015 ***	−0.012 ***	−0.013 ***	−0.013 ***
	(0.005)	(0.003)	(0.003)	(0.003)	(0.003)	(0.003)
lnGDP	0.638 ***	0.657 ***	0.600 ***	0.657 ***	0.658 ***	0.659 ***
	(0.013)	(0.018)	(0.016)	(0.019)	(0.019)	(0.019)
lnPOP	0.342 ***	0.873 ***	0.583 ***	0.878 ***	0.873 ***	0.870 ***
	(0.011)	(0.033)	(0.022)	(0.034)	(0.034)	(0.035)
lnES	2.323 ***	1.437 ***	1.693 ***	1.424 ***	1.410 ***	1.398 ***
	(0.037)	(0.052)	(0.049)	(0.053)	(0.053)	(0.053)
lnUR	−0.224 ***	1.268 ***	1.373 ***	1.257 ***	1.253 ***	1.275 ***
	(0.068)	(0.104)	(0.094)	(0.106)	(0.109)	(0.111)
Constant	0.361 *	−8.014 ***	−3.268 ***	−8.105 ***	−8.004 ***	−7.976 ***
	(0.202)	(0.513)	(0.348)	(0.523)	(0.533)	(0.543)
Observations	3293	3293	3293	3243	3191	3137
R-squared	0.943	0.788		0.782	0.775	0.768

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

; specifically, columns [1,2,3] provide the basic findings from these models. The F-test rejects pooled OLS in favor of specifications with unit effects. Similarly, the Hausman test indicates a significant correlation between regressors and country-specific effects. Accordingly, a fixed-effects model is adopted as the baseline. In column [2] of Table 1, the regression coefficient of fossil energy prices is reported as −0.009 (0.002) [66]. This coefficient suggests that a 1% increase in fossil energy prices leads to a 0.009% decrease in CO₂ emissions. These estimates indicate a statistically significant yet economically modest association: a 1% rise in fossil energy prices is linked to a 0.009% decline in CO₂ emissions. Using the 2023 CO₂ total of our 119-economy sample (34,453.01 Mt) as the baseline, a one-percent increase in the fossil energy price index (proxied by Brent) corresponds to approximately 3.101 Mt of CO₂, and to 1.244 Mt of standard coal, using a conversion factor of 2.493 t CO₂ per tce [21]. Hence, an increase in fossil energy prices can effectively suppress CO₂ emissions. To address potential endogeneity, we estimate a fixed-effects two-stage least squares (FE-2SLS) model using the 1–3-period lags of the fossil energy price as instruments; with country-clustered standard errors, the IV estimate remains negative and statistically significant, and diagnostics indicate strong instruments and satisfactory overidentification (

Table A6. FE-2SLS estimates.

	(1)
Variables	IV (FE-2SLS L1–L3 as IV)
lnFEP	−0.014 **
	(0.006)
lnTEC	0.005 *
	(0.003)
lnGDP	0.426 ***
	(0.090)
lnPOP	0.729 ***
	(0.144)
lnES	2.671 ***
	(0.279)
lnUR	1.223 ***
	(0.381)
Observations	3689
Number of id	119
R-squared	0.789
Country FE	Yes
Year FE	Yes
K-P rk Wald F	19,891
Under-id LM p	0.000200
Hansen J p	0.662

Notes: Standard errors (in parentheses) are clustered at the country level. Instruments: L1 lnEP, L2 lnEP, L3 lnEP. Reported diagnostics include Kleibergen–Paap rk Wald F, underidentification LM p-value, and Hansen J p-value. *** p < 0.01, ** p < 0.05, * p < 0.1.

).

The reason for this could be that rising fossil energy prices can lead to high costs for businesses or individuals, and the dual pressures of shrinking costs and profits can force businesses to reduce energy waste and consumption [3,4]. In addition, fluctuations in fossil energy prices will underscore the importance of innovation in business [6]. Businesses need to find more efficient and environmentally friendly ways to use energy, increasing productivity and reducing costs [5,6]. In summary, higher fossil energy prices are associated with lower CO₂ emissions [46]. We view price signals as one of several complementary tools to be paired with efficiency and structural policies.

Considering the control variables, the coefficients of economic development level, population size, fossil energy consumption share, and urbanization are found to be significantly positive. These results suggest that economic growth, population size, and urbanization play adverse roles in alleviating CO₂ emissions. Furthermore, the share of fossil energy consumption is positively correlated with CO₂ formation, indicating that higher consumption of traditional fossil energy sources leads to the growth of CO₂ emissions. A possible explanation is that fossil energy sources like coal, oil and natural gas emit large amounts of CO₂ during the combustion process. Additionally, there is a significant positive association between CO₂ emissions and economic progress. As the economy grows, residents’ demand and consumption levels will increase accordingly, leading to a rise in energy consumption and thus promoting the emissions of CO₂. Furthermore, population size is also a significant factor affecting CO₂ emissions. As the population grows, the clothing, food, housing, transportation, and socio-economic activities of residents will inevitably promote an increase in CO₂ emissions. Additionally, the acceleration of urbanization is also positively related to CO₂ emissions. Urbanization brings more construction and transportation activities, which consume substantial amounts of energy, resulting in a rise in CO₂ emissions [67]. Nevertheless, we also discovered an adverse link between CO₂ emissions and technological advancement [68]. Technological progress can enhance the efficiency of energy utilization and reduce energy waste, thereby lowering CO₂ emissions. This suggests that through technological innovation and improvement, we can effectively control CO₂ emissions.

The results in Table 1 (4–6) indicate that increases in fossil energy prices can reduce CO₂ emissions, and that this effect remains significant when fossil energy prices are lagged by one and two periods, both significant at the 1% level. The coefficients of the regression results decrease over time (0.009–0.009–0.008–0.007), suggesting that the impact of fossil energy prices on CO₂ emissions diminishes over time. This may be due to the fact that other factors (e.g., technological advances) have gradually increased their impact on CO₂ emissions, thus offsetting some of the effects of fossil energy prices. This sets the stage for the next step in studying the rebound effect of energy efficiency.

4.2. Robustness Tests

4.2.1. Core Explanatory Variable Substitution

In this research, the Brent crude oil price index (expressed by REP) is utilized as one of the indicators to measure international oil market prices, reflecting changes in crude oil prices. To address potential heterogeneity issues associated with indicator selection, relative energy prices are chosen for robustness testing. A novel method proposed by [69] is referenced for measuring relative energy prices. The results, as presented in

Table A5. Results of robust tests.

	(1)	(2)	(3)
Variables	lnREP	D-K	lnGreen_gas
lnFEP		−0.009 **	−0.007 ***
		(0.003)	(0.002)
lnREP	−0.172 ***
	(0.050)
lnTEC	−0.007 ***	−0.012 **	−0.008 ***
	(0.003)	(0.006)	(0.002)
lnGDP	0.706 ***	0.657 ***	0.617 ***
	(0.019)	(0.042)	(0.016)
lnPOP	0.897 ***	0.873 ***	0.656 ***
	(0.035)	(0.108)	(0.028)
lnES	1.268 ***	1.437 ***	1.407 ***
	(0.053)	(0.380)	(0.044)
lnUR	1.140 ***	1.268 ***	1.157 ***
	(0.109)	(0.116)	(0.089)
Constant	−8.564 ***	−8.014 ***	−11.110 ***
	(0.539)	(1.634)	(0.438)
Country FE	Yes	Yes	Yes
Time FE	Yes	Yes	Yes
Observations	3051	3293	3293
R-squared	0.793	0.788	0.803

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

(1), demonstrate statistical significance at the 1% level. This indicates that an increase in relative energy prices has a significant suppressive effect on CO₂ emissions, contributing to a carbon reduction rate of 0.172%. Importantly, these findings align with the conclusions of [7,70], thereby reaffirming the robustness of the results reported in this article. The results continue to be negative and significant, indicating that decreased CO₂ emissions are linked to higher relative energy prices.

4.2.2. Experimental Method Substitution

To effectively address the issues of serial correlation, heteroskedasticity, and cross-sectional dependence in panel data, the study also used Driscoll and Kraay’s concepts to build a fixed effects regression model using Driscoll-Kraay standard errors. The results in Table A5 (2) show that the fossil energy prices reject the null hypothesis and are negative at the 5% significance level. This consistency with expectations reinforces the robustness of the analysis and supports the notion that increasing fossil energy prices can effectively contribute to carbon reduction efforts [19,24].

4.2.3. Dependent Variable Substitution

Greenhouse gas, denoted by Green_gas, is used in place of the dependent variable in this article to further confirm its robustness. The results, as shown in column 3 of Table A5, indicate that the upward adjustment of global fossil energy prices from 1990 to 2023 continues to have a significant inhibitory effect on greenhouse gas emissions. The regression coefficient of fossil energy prices on greenhouse gas emissions is −0.007 (0.002), indicating that a 1% increase in fossil energy prices leads to a 0.007% decrease in greenhouse gas emissions. This result is basically consistent with the benchmark regression results. Due to the large base of global greenhouse gas emissions, even slight changes will have a huge impact on the environment, which further demonstrates the empirical robustness of this article.

4.3. Heterogeneity Analyses

4.3.1. Income-Group Heterogeneity

Income groups follow the World Bank 2024 four-category classification and are treated as time-invariant. Specifically, “High” refers to upper-middle- and high-income economies; “Low” refers to low- and lower-middle-income economies. In

Table A7. Heterogeneity analysis.

	(1)	(2)
Variables	Income Groups	NEI
lnFEP	−0.019 ***	−0.035 ***
	(0.002)	(0.003)
High#c.lnFEP	0.010 ***
	(0.004)
Importer#c.lnFEP		0.027 ***
		(0.004)
lnTEC	0.004 ***	0.003 ***
	(0.001)	(0.001)
lnGDP	0.408 ***	0.421 ***
	(0.016)	(0.015)
lnPOP	0.845 ***	0.756 ***
	(0.030)	(0.028)
lnES	2.771 ***	2.851 ***
	(0.058)	(0.057)
lnUR	1.123 ***	0.807 ***
	(0.101)	(0.099)
Constant	−14.519 ***	−12.971 ***
	(0.451)	(0.423)
Observations	4046	3842
Number of id	119	113
R-squared	0.805	0.813
Country FE	Yes	Yes
Year FE	Yes	Yes
Low slope	−0.0193
High slope	−0.00973
Exporters slope		−0.0348
Importers slope		−0.00770
Difference p-value	0.00984	<0.001

Notes: *** p < 0.01, standard errors in parentheses.

, column (1) shows that the interaction with the high-income indicator is positive and statistically significant. The implied price–emissions elasticities are approximately −0.019 for low- and lower-middle economies and −0.010 for upper-middle- and high-income economies, with a statistically significant difference at about the one-percent level.

A smaller absolute elasticity in upper-middle- and high-income economies is consistent with higher baseline efficiency and more diversified energy mixes. These economies also use broader complementary instruments, such as performance standards, ETS, and price corridors, which dampen marginal responses. This pattern aligns with the efficiency/structure channels highlighted in the discussion section. It also aligns with cross-country evidence that emission responses to pricing vary by development stage and policy architecture. Recent multi-country studies have documented significant heterogeneity in ex-post carbon-pricing effects and elasticities across contexts [35].

4.3.2. Heterogeneity by Net Energy Imports

Net energy imports (NEIs), defined as the share of net imports in total energy use, are used to classify countries into two time-invariant groups: those with a positive sample-average NEI are categorized as net importers, while those with a non-positive average are classified as net exporters. Column (2) of Table A7 shows that the mitigation response is substantially weaker among importers. Specifically, the estimated elasticity is approximately −0.008 for importers, compared to −0.035 for exporters, and the difference between the two groups is highly statistically significant.

Importers often use price smoothing or broad subsidies [71]. Pass-through from international prices to retail prices is lower. Energy use is more diversified, and efficiency is higher. These features mute the short-run impact of price increases on emissions. Exporters tend to have more energy-intensive production and stronger cost pass-through. The pattern is consistent with the efficiency and structural channels discussed in the paper [72].

5. Further Discussion on the Rebound and Mediating Effects

5.1. Rebound Effect

Before verifying whether a mediating pathway exists for fossil energy prices to affect CO₂ emissions through energy efficiency, this article refers to the work of [53] to calculate the rebound effect of energy efficiency using the elasticity of actual energy use with respect to energy efficiency. The findings are displayed in column 1 of

Table 2. The rebound effect.

	(1)
Variables	In AEC
lnEEI	−0.864 ***
	(0.010)
lnFEP	−0.001
	(0.001)
lnTEC	0.002 **
	(0.001)
lnGDP	0.839 ***
	(0.009)
lnPOP	−0.073 ***
	(0.015)
lnES	−0.085 ***
	(0.022)
lnUR	−0.205 ***
	(0.044)
Constant	2.341 ***
	(0.232)
Country FE	Yes
Time FE	Yes
Observations	3293
R-squared	0.912

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

. The rate at which energy efficiency contributes to actual energy use is −0.864 (0.010), indicating that a 100% increase in energy efficiency will result in an 86.4% decrease in lnAEC. Therefore, the rebound effect of energy efficiency can be calculated as 13.6%, which also verifies Hypothesis 2. The above quantitative test shows that the energy efficiency has partial rebound effects on CO₂ emissions, which is in line with the findings of [44,48]. This partial rebound suggests that while efficiency improvements yield significant energy savings, a portion of the gains is offset by induced energy demand, highlighting the behavioral and structural feedbacks embedded in real-world energy systems [73].

5.2. Mediating Effect

5.2.1. Energy Efficiency Mediation

The previous empirical results strongly indicate that fossil energy prices have a significant effect on carbon emission reduction (see

Table 3. Intermediary effect of energy efficiency.

	(1)	(2)	(3)
Variables	ln CO2	ln EEI	ln CO2
lnEEI			−0.917 ***
			(0.019)
lnFEP	−0.009 ***	0.009 ***	−0.001
	(0.002)	(0.001)	(0.001)
lnTEC	−0.012 ***	0.002	−0.010 ***
	(0.003)	(0.002)	(0.002)
lnGDP	0.657 ***	0.445 ***	1.064 ***
	(0.018)	(0.013)	(0.016)
lnPOP	0.873 ***	−0.516 ***	0.401 ***
	(0.033)	(0.024)	(0.027)
lnES	1.437 ***	−0.325 ***	1.139 ***
	(0.052)	(0.038)	(0.040)
lnUR	1.268 ***	−1.080 ***	0.278 ***
	(0.104)	(0.075)	(0.081)
Constant	−8.014 ***	9.482 ***	0.677
	(0.513)	(0.370)	(0.424)
Country FE	Yes	Yes	Yes
Time FE	Yes	Yes	Yes
Observations	3293	3393	3293
R-squared	0.788	0.689	0.881

Notes: *** p < 0.01, standard errors in parentheses.

(1)). To test the mechanism and pathway between fossil energy prices and carbon emission reduction, we introduce energy efficiency as a mediating variable. Based on this, our Model 7 verifies the effect of fossil energy prices on the mediating variable, and the results are shown in column 2 of Table 3, with a regression coefficient of 0.009 (0.002). Through a significant test of 1%, it demonstrates that the increase in fossil energy prices can significantly encourage energy efficiency improvement, verifying that energy efficiency is one of the important steps in using fossil energy prices as a lever to reduce carbon emissions.

Next, the third condition for the existence of a mediating effect is verified: the inclusion of mediating variables has a significant effect on the explained variables, while the core explanatory variables have a reduced or even insignificant effect on the explained variables. The column of Table 3 (3) shows that the improvement in energy efficiency significantly reduces CO₂ emissions with a regression coefficient of −0.917 (0.019), which rejects the null hypothesis and is negative at the 1% level of significance. As expected, the effect of lnEP on CO₂ emissions is not significant. This verifies that energy efficiency is a partial mediator of energy price leverage in reducing carbon emissions.

5.2.2. Renewable Energy Share Mediation

With the energy efficiency pathway established and the negative total effect persisting in

Table 4. Intermediary effect of renewable energy share.

	(1)	(2)	(3)
Variables	ln CO2	ln RES	ln CO2
lnRES			−1.766 ***
			(0.075)
lnFEP	−0.009 ***	0.002 ***	−0.006 ***
	(0.002)	(0.000)	(0.002)
lnTEC	−0.012 ***	0.003 ***	−0.007 ***
	(0.003)	(0.001)	(0.002)
lnGDP	0.657 ***	−0.066 ***	0.549 ***
	(0.018)	(0.004)	(0.018)
lnPOP	0.873 ***	−0.087 ***	0.777 ***
	(0.033)	(0.008)	(0.033)
lnES	1.437 ***	−0.564 ***	0.337 ***
	(0.052)	(0.012)	(0.065)
lnUR	1.268 ***	−0.408 ***	0.486 ***
	(0.104)	(0.025)	(0.105)
Constant	−8.014 ***	2.533 ***	−4.426 ***
	(0.513)	(0.123)	(0.532)
Country FE	Yes	Yes	Yes
Time FE	Yes	Yes	Yes
Observations	3293	3090	3090
R-squared	0.788	0.675	0.821

Notes: *** p < 0.01, standard errors in parentheses.

(1), we examine a parallel mechanism via the renewable energy share (RES). The coefficient from lnFEP to the renewable share is 0.002 (0.000) and is significant at the 1% level (Table 4 (2)). This result suggests that renewable energy penetration is stronger under higher prices. This finding is consistent with recent evidence on the positive relationship between fossil-fuel prices and renewable energy uptake [74,75]. Rising fossil energy prices not only raise the cost of conventional fuels but also enhance the relative price advantage and investment incentives of renewable technologies, motivating structural substitution toward cleaner energy sources [17].

When the mediator is included (Table 4 (3)), the renewable share reduces CO₂ emissions with a coefficient of −1.766 (0.075) at the 1% level [76]. Meanwhile, lnFEP remains negative and significant at −0.006 (0.002). This pattern indicates partial mediation: while the renewable pathway is material, a direct price effect persists. The renewable pathway complements the energy efficiency channel by conveying structural decarbonization alongside efficiency gains [77].

6. Conclusions and Policy Implications

6.1. Conclusions

This study provides comprehensive scientific proof about the relationship between fossil energy prices and carbon emissions by employing panel data from 119 economies over the period 1990–2023. The results consistently indicate that higher fossil energy prices are negatively associated with CO₂ emissions, underscoring their role as a crucial lever for carbon mitigation. Rising energy costs impose significant adjustment pressures on both producers and consumers, which, in turn, stimulate shifts in consumption behavior, production investment, and technological choices toward more efficient and cleaner energy use.

Mechanism testing further demonstrates that energy efficiency functions as a partial mediator in this relationship. Specifically, a 1% increase in fossil energy prices is estimated to reduce CO₂ emissions by approximately 0.009%, of which about 86.4% of the expected energy savings are realized, while rebound effects offset the remaining 13.6%. The elasticity of actual energy consumption confirms the existence of a partial rebound rather than a complete backfire in terms of efficiency (−0.864). In other words, efficiency gains are substantial, but not absolute, since a portion of the savings is reabsorbed into new energy demand.

Moreover, a parallel mechanism is observed through the share of renewable energy. Rising fossil energy prices are found to significantly accelerate the deployment of renewable energy, which exerts an independent and negative effect on CO₂ emissions [75,78]. This finding highlights a dual pathway by which fossil energy prices contribute to decarbonization: through both efficiency improvements and structural energy substitution.

Taken together, the evidence suggests that price-based instruments are effective yet complex tools for reducing carbon emissions. They generate direct mitigation outcomes, indirect efficiency-mediated reductions, and complementary renewable expansion. At the same time, rebound effects partially erode the expected benefits, which makes it essential for policymakers to explicitly account for behavioral and systemic feedback when designing energy and climate strategies.

6.2. Policy Implications

First, the evidence suggests that aligning fossil energy prices more closely with their true economic and environmental costs can serve as a powerful mechanism to reduce CO₂ emissions. Governments—particularly those in high-emission economies—should consider comprehensive pricing reforms, including the gradual reduction in fossil-fuel subsidies and the incorporation of carbon costs into energy markets. Such reforms not only incentivize energy efficiency but also reallocate resources toward cleaner technologies.

Second, the documented 13.6% rebound effect implies that efficiency policies alone are insufficient to fully achieve carbon reduction targets. While efficiency improvements clearly contribute to mitigation, part of the gains will be offset unless complemented by measures that manage demand-side responses. Policymakers should therefore implement balanced frameworks that combine price adjustments with behavioral interventions, stricter efficiency standards, and regulatory caps to limit rebound effects.

Third, the analysis of renewable energy as a parallel mediation pathway reveals that higher fossil-energy prices can foster structural transformation of the energy mix. Policy design should therefore go beyond efficiency incentives and actively promote renewable deployment through fiscal instruments (e.g., feed-in tariffs, tax credits), infrastructure investment, and long-term innovation support. The interaction between higher energy costs and renewable expansion creates synergies that amplify the overall decarbonization impact. Hence, integrating price signals with stringent efficiency standards and behavioral measures is essential to contain rebound effects and ensure sustained carbon mitigation.

Fourth, the international scope of this study (119 countries, three decades) underlines the need for heterogeneous policy design. High-income economies with advanced renewable sectors may prioritize managing rebound effects and strengthening efficiency standards, while low- and middle-income countries should emphasize energy affordability, access to clean energy, and gradual transition mechanisms to avoid adverse social impacts. Such differentiation enhances the operability and applicability of our recommendations across development stages. Coordinated international frameworks, such as carbon clubs or price corridors, can help align national efforts and prevent carbon leakage.

Finally, the results suggest that fossil energy pricing should not be viewed in isolation but rather as part of a broader sustainable development agenda. Complementary strategies—such as promoting the circular economy, technological innovation, and sustainable urbanization—are necessary to reinforce the long-term effectiveness of energy price policies. By integrating efficiency, renewable expansion, and demand management, governments can maximize the environmental benefits while minimizing unintended rebound effects.

In conclusion, this study demonstrates that energy price adjustments are a critical yet nuanced instrument for global carbon reduction. Their effectiveness lies not only in direct suppression of emissions but also in the dual mechanisms of efficiency and renewables, moderated by rebound effects. Designing policies that internalize these dynamics will be essential for meeting international climate targets and achieving sustainable development in the coming decades. In practice, policy packages that combine price signals with efficiency standards and market-design reforms are more likely to deliver durable abatement.

6.3. Limitations

This study focuses on average cross-country effects using international price proxies and fixed effects. Although robustness tests mitigate concerns about endogeneity and dependence, some structural channels and data heterogeneity may remain unobserved. Future work could integrate retail price pass-through, policy stringency, and sectoral data for finer identification.

Notably, the construction of the energy structure indicator—defined as the fossil share in total energy supply—relies on internationally harmonized data that aggregates diverse fuels into a common physical unit, tonnes of oil equivalent (toe), based on their lower heating values. As [79] cautions, this heating-value-only aggregation implicitly treats all energy forms as functionally equivalent. When the energy mix shifts significantly—for instance, from coal toward natural gas or electricity—ignoring differences in energy quality (e.g., useful work potential or exergy) may bias rebound effect estimates upward. Ref. [15] empirically observed a similar pattern, noting that unadjusted physical aggregation tends to overstate rebound effects in settings undergoing rapid decarbonization. Interpreted in this light, our central estimate of 13.6% likely represents an upper bound. Future work could explore quality-adjusted aggregation schemes—such as exergy-based or price-weighted indices—to better capture efficiency gains embedded in structural energy transitions.