The Impact of Energy Poverty Alleviation on Carbon Emissions in Countries along the Belt and Road Initiative

Abstract

Energy poverty and climate change are global challenges. While the Belt and Road Initiative (BRI) has fostered energy cooperation and alleviated energy poverty in BRI countries, its impact on climate change requires careful examination. This study investigates the impact of energy poverty alleviation on carbon emissions in BRI countries and analyzes the underlying mechanisms. In particular, this study constructs an integrated energy poverty assessment framework that includes three dimensions of energy accessibility, cleanliness, and affordability and utilizes the entropy-TOPSIS method to assess the status of energy poverty in 70 BRI countries. Subsequently, an empirical analysis using the Gini index as an instrumental variable was conducted to explore the impact of energy poverty alleviation on carbon emissions and the specific mechanisms. It is found that alleviating energy poverty in BRI countries will lead to an increase in per capita carbon emissions. However, improving energy cleanliness is effective in reducing per capita carbon emissions, while improving energy affordability has no significant impact on carbon emissions. This study shows that energy poverty alleviation and carbon emission reduction are two non-conflicting sustainable development goals. BRI countries can synergize energy poverty alleviation and carbon emissions reduction by focusing on clean energy development and providing energy subsidies.

1. Introduction

The Paris Agreement and the United Nations 2030 Sustainable Development Goals (SDGs) jointly agreed on the goals of climate change mitigation and sustainable development, and one of the key goals of the SDGs (SDG1) is to eradicate extreme poverty [1,2,3]. Energy poverty is one aspect of multidimensional poverty [4], which occurs when households cannot afford to consume energy or have to reduce their household energy consumption to the extent that it negatively affects their health and well-being [5]. The world today still suffers from severe energy poverty. As of 2020, only 48.5% of the population in rural areas worldwide had access to clean energy and technology for cooking [6]. In South Asia, Southeast Asia, and North Africa, energy poverty remains a serious problem [7,8]. The Belt and Road Initiative (BRI) is a global development strategy proposed by China in 2013. It encompasses infrastructure development and economic cooperation along the historic Silk Road, aiming to stimulate economic growth in a wide range of regions, including Asia, Europe, and Africa [9]. It connects countries in Asia, Africa, and Europe, promoting trade, investment, and connectivity. By the end of June 2021, China had signed 206 cooperation documents with 140 countries and 32 international organizations to implement the Belt and Road Initiative [10]. Energy poverty is also widespread in these BRI countries, such as Pakistan, Bangladesh, and Indonesia [11,12].

Many measures have been taken by countries to alleviate energy poverty, such as the development of energy technologies, increased use of clean kitchen appliances and cooking facilities, and the promotion of clean electricity, which have already yielded initial results [13]. Statistics from the International Energy Agency (IEA) show a steady upward trend in modern renewable energy since 2000, accounting for 11% of final energy use in 2018 [14]. Since 2010, there has been a gradual downward trend in the percentage of the population that relies heavily on kerosene, solid biomass, or coal. Similarly, in 2019, the global population with poor access to clean electricity had fallen to about 770 million, reaching a historically low level.

Energy poverty alleviation and climate change mitigation are inextricably linked policy objectives. Alleviating energy poverty may lead to higher levels of carbon emissions from increased residential energy consumption [15], but it may also improve energy efficiency in the residential sector by increasing the use of cleaner energy, which in turn helps to reduce carbon emissions [16]. Therefore, it is necessary to explore how the alleviation of energy poverty affects carbon emissions. In addition, the introduction of the Belt and Road Initiative has brought a large amount of infrastructure investment and energy inputs to BRI countries, thus improving their energy poverty [17], which will also have a profound impact on local carbon emissions. Meanwhile, energy poverty exists widely in BRI countries, and it is urgent to improve their energy poverty, so it is necessary to take BRI countries as the research object.

To summarize, this study constructs an integrated energy poverty index from the three dimensions of energy accessibility, affordability, and cleanliness. Thereafter, the study assesses the current status of energy poverty in 66 BRI countries using data from 2010 to 2020. Based on the above index, this study empirically examines the impact of energy poverty alleviation on carbon emissions in BRI countries by adopting the instrumental variable method, and further analyzes the impact of the improvement of energy poverty in different dimensions on the level of carbon emissions.

The contributions of this study are as follows: Firstly, this paper assesses the integrated energy poverty levels of BRI countries from the three dimensions of accessibility, cleanliness, and affordability, which more accurately locate the current status of energy poverty in each country. In addition, the year-by-year assessment indices can be used to evaluate the improvement of energy poverty in each country, which is conducive to the development of the subsequent energy policy. Secondly, this paper tests the impact of energy poverty alleviation on carbon emissions in BRI countries and uses an instrumental variable approach to address the endogeneity problem. The accuracy of the results has improved, and the instrumental variable provides a reference for subsequent research on energy poverty. Finally, the paper examines the impact of different dimensions of energy poverty on carbon emissions and identifies the key factors that lead to increased carbon emissions. Further, the paper discusses the conflict and synergy relationship between energy poverty alleviation and emission reduction, providing policy suggestions for the simultaneous realization of the two challenges.

The subsequent sections are organized as follows. Section 2 reviews the literature on energy poverty assessment and the impact of energy poverty on carbon emissions. Section 3 provides an integrated assessment of energy poverty in BRI countries in three dimensions using the entropy-TOPSIS method. Section 4 constructs the empirical model. Section 5 shows the results of the empirical tests, mechanism analysis, and robustness tests. Section 6 provides conclusions.

2. Literature Review

2.1. Related Research on Energy Poverty Assessment Methods

Currently, a variety of methods exist for assessing energy poverty, without a universally accepted assessment framework. The primary approaches can be categorized as single-indicator methods and multidimensional-indicator methods. Single-indicator methods assess energy poverty from a specific perspective using a single metric. Due to limited early attention on energy poverty and data availability constraints, single-indicator methods were commonly employed. Single-indicator methods primarily rely on income and fuel expenditures to assess energy poverty, with prominent examples including the 10% indicator, Minimum Income Standard (MIS) indicator, and Low Income High Costs (LIHC) indicator. The 10% indicator, one of the earliest methods used to assess energy poverty, identifies households as having energy poverty if they spend over 10% of their income on achieving adequate energy services [18]. This method has been adopted by several countries and combined with subjective energy poverty indicators, such as in Greece [19] and Ireland [20]. However, its sensitivity to energy prices limits its effectiveness for international comparisons [21]. The MIS indicator compares a household’s residual income with the basic cost of energy, and those unable to afford this cost are considered to be in energy poverty [22]. Compared to other single-indicator methods, MIS provides a better understanding of energy poverty depth and household vulnerability. The LIHC indicator refines the 10% indicator by focusing on the ratio of energy expenditure to income and its socio-economic implications [23]. The UK government employs both the 10% and LIHC indicators, but their dual relativity may lead to misleading results [24]. However, modifying calculation methods and threshold settings can improve assessment accuracy [25,26].

Multidimensional-indicator methods provide a comprehensive assessment of energy poverty by considering multiple dimensions of the issue. These methods utilize a range of sub-indicators, each reflecting a specific attribute of energy poverty, which are then aggregated into a single, readily interpretable index. Aristondo and Onaindia [27] employed a multidimensional assessment approach, classifying energy poverty variables into three key indicators of energy accessibility: the ability to maintain adequate household warmth, the presence of utility bill arrears, and the existence of leaks, damp walls, or broken windows. Several multidimensional indicators have gained recognition and widespread application, including the Sen-based viability index, the Energy for Development Index (EDI), and the Multidimensional Energy Poverty Index (MEPI). Notably, the Sen index, which incorporates the poverty rate, poverty gap ratio, and the Gini index, offers a straightforward yet effective approach to assessing energy poverty and has been widely adopted across diverse national contexts [28]. The Sen-based viability index assesses energy poverty through the lens of capabilities, highlighting the extent to which individuals are deprived of their freedom of choice [29]. The Energy Development Index (EDI) is a multidimensional indicator for measuring energy poverty proposed by the International Energy Agency in 2004, which combines three equally weighted indicators: commercial energy consumption per capita, the share of commercial energy in final energy use, and the share of the population using electricity [30]. The index is straightforward and easily accessible, and the index is valuable in comparing the energy development status of different countries. In addition, the index can be used to assess the energy policy needs of each country and region [31]. Multidimensional Energy Poverty Index (MEPI), proposed by Nussbaumer et al. [32], reflects the incidence of energy poverty and the average intensity of deprivation among the energy poor. In recent years, scholars have mostly used it for energy poverty analysis in specific regions [17,32,33,34]. However, MEPI is used in the absence of reliable and comprehensive data, which can affect the accuracy of the results of this index. In summary, it can be seen that since it is necessary to assess energy poverty from different directions and perspectives, and finally form a comprehensive index. Therefore, it is a challenge for the multidimensional energy poverty assessment method to assign reasonable weights.

2.2. Related Research on the Impact of Energy Poverty Alleviation Measures on Carbon Emissions

Increasing the use of clean energy, especially electricity, has long been seen as the preferred means to alleviate energy poverty [35]. Theoretically, the spread of clean energy can effectively replace traditional fuels such as firewood and coal, which can help mitigate the increasing greenhouse effect while alleviating energy poverty [36]. However, the current electricity supply is still dominated by thermal power, and low-income groups can hardly afford the high cost of electricity, which will gradually aggravate the greenhouse effect [37]. This also means that, while alleviating energy poverty, additional carbon emissions may be generated, thereby exacerbating climate change.

In addition, in some regions, governments and organizations may prefer to use traditional fossil energy sources [38], such as coal, oil, and natural gas, to alleviate energy poverty. Traditional fossil energy sources are characterized by high energy density and stability, but they also produce large amounts of greenhouse gas emissions, exacerbating the global climate change problem [15]. The standard of living and consumption capacity of people may increase when energy poverty is alleviated [39]. However, this may lead to more energy consumption and greenhouse gas emissions. Especially in the process of industrialization and urbanization, people become more dependent on energy-intensive production and lifestyles, which further increases the risk of greenhouse gas emissions.

On the other hand, the process of alleviating energy poverty is often associated with a low-carbon energy transition. Alleviating energy poverty can lead to greater use of clean energy [16], such as renewable energy sources such as solar, wind, and hydro, to replace traditional fossil energy sources such as coal, oil, and natural gas. Clean energy sources are low-carbon and non-polluting, and their use can significantly reduce greenhouse gas emissions and contribute to climate change mitigation. Alleviating energy poverty usually involves improving the infrastructure for energy supply and utilization, such as power grids, lighting equipment, and combustion appliances. By improving the energy efficiency of these facilities and equipment, it is possible to reduce energy consumption while providing sufficient energy, thereby reducing greenhouse gas emissions. In some regions, alleviating energy poverty may mean reducing reliance on biomass, such as firewood and straw, for heating, cooking, and other uses [4]. The burning of biomass releases large amounts of carbon emissions and other harmful substances, so reducing the use of biomass can help lower greenhouse gas emissions.

Overall, current research has focused on assessing energy poverty, while a limited number of studies have gradually focused on the relationship between energy poverty alleviation and carbon emission reduction. However, research on this relationship often overlooks issues of causality and endogeneity, and there is insufficient exploration of the mechanisms through which alleviating energy poverty impacts carbon emissions. Therefore, building on prior research, this study employs instrumental variable techniques to address endogeneity concerns and further investigates the mechanisms by which alleviating energy poverty affects carbon emissions.

2.3. Related Research on Energy Poverty in BRI Countries

Energy poverty is also prevalent in BRI countries. Numerous studies indicate that energy poverty is widespread in many BRI countries, particularly in South and Southeast Asia [8,40]. These studies highlight the urgent need to expand access to modern energy services and identify the lagging development of energy infrastructure and the insufficient use of clean energy as the main causes of energy poverty [8]. On the other hand, China’s Belt and Road Initiative has had a significant impact on alleviating energy poverty while promoting economic development in participating countries. Therefore, a number of studies have examined the impact of the Belt and Road Initiative on energy poverty in participating countries. Che et al. [40] found that the Belt and Road Initiative reduces energy poverty in participating countries, especially along the Maritime Silk Road, by improving energy accessibility, energy infrastructure, and energy supply levels. Zhou et al. [17] further indicated that China’s outward foreign direct investment and its spatial spillover effects can reduce energy poverty in BRI countries by improving energy accessibility, cleanliness, and modernization.

There are relatively few studies on the relationship between energy poverty and carbon emissions under the Belt and Road Initiative. Xu et al. [41] found that energy poverty in rural areas is associated with higher levels of carbon emissions and that urbanization has a mitigating effect on these emissions. Zhao et al. [42] indicated that globally, alleviating energy poverty through the promotion of electricity may exacerbate the greenhouse effect, but in non-BRI countries, alleviating energy poverty through electricity can be effective in mitigating the greenhouse effect. Meanwhile, some studies have focused on the impact of energy poverty on other social problems. Lin and Okyere [43] found that energy poverty increases the risk of mental health problems, especially in developing countries. In a follow-up study, they also explored the impact of energy poverty on subjective social status and found a quadratic (U-shaped) relationship between energy poverty and social status [44].

Overall, these studies consistently suggest that energy poverty in BRI countries exhibits significant regional disparities. Energy poverty is particularly acute in South Asia, Southeast Asia, and parts of Africa, where more international cooperation and investment are needed to improve energy infrastructure and supply. In addition, these studies advise that the promotion of renewable energy sources, the implementation of sustainable urban development policies, and energy conservation strategies are important measures to alleviate energy poverty.

2.4. Research Gap

The literature review highlights several key areas in the field of energy poverty assessment and its impact on carbon emissions. However, several research gaps can be identified. First of all, current research has focused on assessing energy poverty, while a limited number of studies have gradually focused on the relationship between energy poverty alleviation and carbon emission reduction. Current research on the relationship between energy poverty and carbon emissions often focuses on the overall impact of energy poverty alleviation on carbon emissions, neglecting the potential differences in impact across various dimensions of energy poverty. This study analyzes the impact of three dimensions of energy poverty (accessibility, cleanliness, and affordability) on carbon emissions. The results reveal that improving energy cleanliness can effectively reduce carbon emissions, while improving energy accessibility can lead to an increase in carbon emissions. This finding reveals the complex relationship between energy poverty alleviation and carbon reduction and calls for more in-depth analysis. Furthermore, existing studies often rely on OLS regression models, which can be susceptible to endogeneity issues, leading to biased results. This study employs an instrumental variable approach to address endogeneity, using the Gini index as an instrumental variable. Consequently, this approach provides a more accurate assessment of the impact of energy poverty alleviation on carbon emissions and offers a more reliable analytical method for future research. Finally, current research often overlooks the impact mechanisms of different energy poverty indicators on carbon emissions, lacking a specific analysis of how different indicators influence carbon emissions. This study conducts a separate analysis of each energy poverty indicator, revealing that electricity access rate and per capita energy consumption are the primary drivers of carbon emissions increases, while the share of renewable energy consumption contributes to a decrease in carbon emissions.

3. Assessment of the Energy Poverty Status in BRI Countries

3.1. Energy Poverty Assessment Framework

This paper provides a comprehensive and transnational analytical framework for assessing energy poverty. To assess energy poverty in BRI countries, this study builds upon existing research [27,32,45] and proposes an integrated assessment framework encompassing eight indicators across three dimensions: energy accessibility, energy cleanliness, and energy affordability. The term “clean” is used here instead of “renewable” because the focus is on the types of energy used by households. The World Health Organization’s definition of clean energy encompasses sources such as electricity and natural gas, not all of which are renewable [46].

Energy accessibility is a crucial indicator of energy poverty, as it directly reflects the ease with which households can obtain energy. In many regions, inequitable distribution of energy resources or inadequate infrastructure limits households’ access to necessary energy sources. Therefore, incorporating energy accessibility into an energy poverty assessment framework allows for an accurate measurement of the challenges households face in securing energy supplies and also provides a valuable reference for formulating effective policies and interventions. Considering electricity’s prominence in household energy consumption, this study employs both overall energy access and electricity access as indicators of accessibility. In a word, the energy accessibility dimension comprises three indicators: access to electricity, electricity consumption, and energy consumption. The rationale behind selecting the first two indicators stems from the prevalent issue of limited access to clean energy sources in BRI countries. For instance, 77% of households in Pakistan rely on biomass for their energy needs [47]. Consequently, the proportion of the population with electricity access and per capita electricity consumption reflect the degree of difficulty in accessing modern energy sources for the inhabitants of these regions. Finally, per capita energy consumption provides insights into actual household energy usage, helping determine whether households can secure sufficient energy to meet their basic needs.

Energy cleanliness is particularly important in the context of climate change and sustainable development. In this paper, access to clean fuels for cooking, the renewable energy share of total final energy consumption, and primary energy intensity are used as indicators to characterize energy cleanliness. The proportion of the population with access to clean fuels reflects the cleanliness of household energy use [32]. In addition, SDG7 proposed universal access to modern energy services to improve the efficiency and use of renewable energy [48]. Therefore, the level of renewable energy use is an important aspect reflecting the energy poverty status, and it is necessary to consider the energy structure in the indicator framework, using the renewable energy share of total final energy consumption as a characterization of the energy structure. Finally, the primary energy intensity is the amount of energy given to the economy per unit of economic output [49], reflecting the cleanliness of a country’s economic production process.

Energy affordability implies that energy costs are too high for households to pay. The firm’s cost of getting electricity as well as the number of mobile phone subscribers are used as characterization indicators. The cost of getting electricity is the median of the total cost of getting electricity access for firms in a country, and it is calculated as a percentage of per capita income. A firm’s cost of getting electricity can be used to measure the affordability of firms for energy [45], hence it is used as an indicator in this paper. In addition, with reference to the study by Nussbaumer, Bazilian, and Modi [32], the energy needs of households for communication should also be taken into account when assessing energy poverty. Moreover, meeting the basic energy needs, the communication expenditure of a household reflects its further affordable household energy expenditures, therefore, this study adds the number of mobile phone subscribers as a characterization indicator of energy affordability to represent the further energy needs of the household.

In summary, this paper assesses energy poverty from three dimensions: energy accessibility, cleanliness, and affordability. These dimensions are correlated with each other and are important in providing an integrated framework for assessing a country’s energy poverty status. The indicator framework is shown in Table 1.

3.2. Data Sources

As shown in Table 1, the data used in this paper are mainly from World development indicators [6], the Doing Business (DB) dataset [50], the Energy Information Administration [51], and the Sustainable Development Goals (SDGs) [49]. Due to data availability, a total of 70 RBI countries were selected as a sample for this paper. Among them, the indicators of “primary energy intensity” and “firm’s cost of getting electricity” have negative impacts, and this paper normalizes them by using 1 minus the original indicator. The data source for each indicator is shown in Table 1.

Table 1. Indicator framework for energy poverty assessment in BRI countries.

Dimension	Indicators	Reference	Data Source	Impact
Energy accessibility	Access to electricity (% of population)	Sadath and Acharya [34]	WDI	+
	Electricity consumption (kWh per capita)	Nussbaumer, Bazilian and Modi [32]	UNSD	+
	Energy consumption (MMBtu per capita)	Gonzalez-Eguino [4]	EIA	+
Energy cleanliness	Access to clean fuels for cooking (% of population)	Wang et al. [52]	WDI	+
	Renewable energy share of total final energy consumption (%)	Bhide and Monroy [53]	WDI	+
	Primary energy intensity (MJ/GDP)	Che et al. [54]	WDI	−
Energy affordability	Firm’s cost of getting electricity (% of income per capita)	Ayodele, Ogunjuyigbe and Opebiyi [45]	DB	−
	Mobile phone subscribers (per 100 people)	Nussbaumer, Bazilian and Modi [32]	WDI	+

Note: “+” represents the indicator of positive impact, and “−” represents the indicator of negative impact.

Table 1. Indicator framework for energy poverty assessment in BRI countries.

Dimension	Indicators	Reference	Data Source	Impact
Energy accessibility	Access to electricity (% of population)	Sadath and Acharya [34]	WDI	+
	Electricity consumption (kWh per capita)	Nussbaumer, Bazilian and Modi [32]	UNSD	+
	Energy consumption (MMBtu per capita)	Gonzalez-Eguino [4]	EIA	+
Energy cleanliness	Access to clean fuels for cooking (% of population)	Wang et al. [52]	WDI	+
	Renewable energy share of total final energy consumption (%)	Bhide and Monroy [53]	WDI	+
	Primary energy intensity (MJ/GDP)	Che et al. [54]	WDI	−
Energy affordability	Firm’s cost of getting electricity (% of income per capita)	Ayodele, Ogunjuyigbe and Opebiyi [45]	DB	−
	Mobile phone subscribers (per 100 people)	Nussbaumer, Bazilian and Modi [32]	WDI	+

Note: “+” represents the indicator of positive impact, and “−” represents the indicator of negative impact.

3.3. Integrated Energy Poverty Index for BRI Countries

This study examines energy poverty in 70 BRI countries from 2010 to 2020. We utilize the entropy-TOPSIS method to calculate an integrated energy poverty index for each country. This method objectively assigns weights to various indicators, reducing subjectivity and enabling a more accurate assessment of energy poverty levels. The entropy-TOPSIS approach considers both the best and worst scenarios, allowing us to rank countries based on their energy poverty status. This method also facilitates comparisons of energy poverty levels over time and across countries, providing valuable insights for understanding energy poverty trends and developing strategies for alleviation. The integrated energy poverty index calculated from the above method, ranging from 0 to 1, reflects the severity of energy poverty, with higher values indicating a greater level of energy poverty.

Figure 1 shows the spatial distribution of average energy poverty status in BRI countries from 2010 to 2020. As shown, gray areas represent non-BRI countries or missing data. Green areas indicate lower levels of energy poverty, while red areas indicate more severe levels of energy poverty. The countries with the highest levels of energy poverty are mainly in Southeast Asia and sub-Saharan Africa, such as Timor-Leste, Djibouti, and Mauritania. The main obstacle to reducing energy poverty in these countries is energy cleanliness. In 2020, Djibouti’s access to clean fuels and cooking technologies will only be limited to 9.25 percent of the population, far below the average of 64.79 percent for that year. Some Southeast Asian countries, such as Timor-Leste, have a renewable energy share of only 11.42 percent, well below the average of 33.0 percent for that year. The BRI countries with the lowest levels of energy poverty are mainly located in West Asia, Southeast Asia, and Europe, such as Qatar, Bahrain, Singapore, the United Arab Emirates, and Kuwait. These countries have a significant advantage in terms of energy accessibility, with the proportion of the population with access to electricity remaining at 100 percent. In 2020, per capita energy consumption in Qatar, a country located in West Asia, will reach 778.41 MMBtu, ten times higher than the average. At the same time, the energy affordability of these regions is also better than that of BRI countries as a whole. For example, Singapore’s firm’s cost of getting electricity in 2020 will be only 22% of per capita income, much lower than the average of 1268.16%. However, these less energy-poor BRI countries lag far behind most of the BRI countries in terms of cleanliness, which is mainly due to the low proportion of renewable energy and high primary energy intensity, so the use of clean energy will become the main initiative to improve energy poverty in these countries.

In addition, we selected a few representative countries to further demonstrate the results of our assessment of energy poverty status in BRI countries, as shown in Figure 2. These include the United Arab Emirates, Luxembourg, and Equatorial Guinea, representing the countries with the most significant energy poverty alleviation, the most severe deterioration, and the most severe energy poverty, respectively.

In Figure 2, the dashed line represents the average energy poverty level in BRI countries. The United Arab Emirates, located in West Asia, has a rapidly developing economy dominated by oil production and petrochemical industries. The country experiences relatively low energy poverty, ranking 6th among BRI countries in 2010. By 2020, its energy poverty situation would have further improved, rising to 4th place. While the United Arab Emirates demonstrates high levels of energy accessibility and affordability, its clean energy performance falls significantly below average. Consequently, the country has recently increased its focus on research and development of renewable energy sources. Similar to the overall trend among BRI countries, the United Arab Emirates’ primary pathway for alleviating energy poverty has been through improved affordability, with a substantial 74.53% improvement in this area, the highest among all BRI countries. This progress is evident in the increase in mobile phone subscriptions per 100 people, rising from 128.82 in 2010 to 197.84 in 2020, compared to the BRI average of 107.99 in 2020. In addition, the country’s energy cleanliness improved by 0.30%, yet its ranking dropped by four places among the BRI countries, suggesting that improvements in energy cleanliness lagged behind those of other countries.

Luxembourg has experienced a significant deterioration in energy poverty, despite the fact that the country has always ranked high among the BRI countries as one of the countries with a low level of energy poverty. Between 2010 and 2020, energy poverty in Luxembourg worsened by 20.93%, primarily due to reduced energy accessibility. This decline is evident in the decrease in electricity consumption per capita from 13.11 million kWh in 2010 to 9.85 million kWh in 2020. Additionally, energy consumption per capita decreased by 32.03%.

Equatorial Guinea is the most energy-poor country, and its economic development is largely dependent on the export of fossil energy, which has certain limitations. As shown in Figure 2, it can be seen that energy cleanliness and affordability are the main obstacles to alleviating energy poverty in Equatorial Guinea. In 2020, the percentage of the population with access to clean fuels for cooking will be only 23.9%, which is significantly lower than the average of 64.79% of the BRI countries, and so there is much room for improvement in its clean energy. In terms of energy affordability, the country’s mobile phone subscriptions per 100 people are 46.62, which is lower than the average of 107.99.

The three cases we have selected are typical of countries where energy poverty has been alleviated or deteriorated, as well as countries where energy poverty is very severe, reflecting to a certain extent the overall status of energy poverty in BRI countries.

4. Model

4.1. Baseline Model

Estimating the impact of energy poverty alleviation on carbon emissions in BRI countries directly using the least squares (OLS) method may be endogenous due to missing variables and measurement errors. This study utilizes eight indicators to construct an assessment framework for status and focuses on the integrated energy poverty index as the explanatory variable. These indicators include energy consumption-related indicators such as per capita energy consumption. Since increases in per capita energy consumption directly impact per capita carbon emissions, endogeneity problems are introduced. Therefore, directly employing the OLS regression model without addressing these endogeneity concerns may lead to inaccurate results. Alternative methods are necessary to correct for these biases and provide more reliable estimations.

The endogeneity problem is a common challenge in causal inference and refers to the existence of a correlation between the explanatory variable and the error term, which leads to biased OLS estimates. The application of the Instrumental Variable (IV) method in causal inference is based on its ability to address the endogeneity problem. The IV method solves the endogeneity problem by introducing an exogenous variable (i.e., an instrumental variable) that is correlated with the explanatory variable but not with the error term.

To address endogeneity, this paper uses the level of income inequality across countries as an instrumental variable and employs the Gini index to measure the level of income inequality. Income inequality is significantly associated with energy poverty [55], and it is one of the most important causes of households falling into energy poverty. Higher levels of income inequality reduce the energy affordability of the population and increase energy poverty. Affordability is worst in countries with medium levels of economic development and higher income inequality [56]. In addition, previous studies have shown that income inequality similarly affects energy accessibility and cleanliness [57]. Therefore, the use of the Gini index as an instrumental variable for energy poverty level satisfies its relevance. On the other hand, existing studies show that there is no clear correlation between income inequality and carbon emissions. The relationship between income inequality and per capita carbon emissions varies across income levels [58], and the Gini index has little effect on CO₂ emissions [59]. These evidences suggest that income inequality does not affect carbon emissions and satisfies exclusion. The baseline regression model of this study is set as follows:

$${CarbonEmissions}_{i,t}={\beta }_0+\widehat{{\beta }_{1}EnergyPovert{y}_{i,t}}+{\beta }_2{X}_{i,t}+{\gamma }_i+{\lambda }_t+{ϵ}_{i,t}$$

(1)

$$\widehat{EnergyPovert{y}_{i,t}}={\beta }_0^{\prime }+{\beta }_1^{\prime }Gin{i}_{i,t}+{{\beta }^{\prime }}_2{X}_{i,t}+{{\gamma }^{\prime }}_i+{{\lambda }^{\prime }}_t+{{ϵ}^{\prime }}_{i,t}$$

(2)

where the dependent variable ${CarbonEmissions}_{i,t}$ represents the per capita carbon emissions, $EnergyPovert{y}_{i,t}$ denotes the integrated energy poverty index, $Gin{i}_{i,t}$ is the Gini index as an instrumental variable, $X_{i,t}$ is a list of control variables, ${\gamma }_i$ and ${\lambda }_t$ reflect country fixed effects and year fixed effects, respectively, and ${ϵ}_{i,t}$ is the error term.

4.2. Measurement

This study analyzes unbalanced panel data from 70 BRI countries from 2010 to 2020. The Gini index data in this study, sourced from the World Bank, is based on primary household survey data obtained from government statistical offices and World Bank country departments. Due to missing data across countries and years, constructing a balanced panel was not feasible. Consequently, this study uses unbalanced panel data for estimation.

4.2.1. Dependent Variable

This paper uses per capita CO₂ emissions as the explanatory variable, which is obtained from the World Bank’s Word Development Indicators (WDI) database. CO₂ emissions per capita are obtained by dividing the country’s total CO₂ emissions by its total population. Figure 3 illustrates the CO₂ emissions per capita for all countries in the world in 2020. It can be seen that there are obvious regional differences in CO₂ emissions per capita among countries in the world, with developed countries such as the United States, Russia, Canada, and Australia having high CO₂ emissions per capita, while regions such as Africa, South Asia, and South America have low CO₂ emissions per capita.

4.2.2. Explanatory Variable

In Section 3, this paper constructs an integrated energy poverty index in terms of the three dimensions of energy accessibility, cleanliness, and affordability and assesses the current state of energy poverty in BRI countries. A higher energy poverty index indicates a higher level of energy poverty in the country. In the empirical analysis, we use the 2010–2020 integrated energy poverty index of 70 BRI countries as the explanatory variables.

4.2.3. Instrumental Variable

In this study, we leverage the Gini index as an instrumental variable to investigate the impact of energy poverty alleviation on carbon emissions in BRI countries. The Gini index serves as a robust measure of income distribution within a population, capturing the degree of deviation from perfect equality. Its values range from 0 (representing perfect equality) to 100 (indicating perfect inequality). Our Gini index data is primarily sourced from the World Bank, which collates and calculates it based on nationally representative household surveys conducted by government statistical agencies and the World Bank itself.

4.2.4. Control Variables

Economic growth. The effect of economic growth on CO₂ has been verified by a large number of studies [60]. Therefore, economic growth measured by GDP per capita (2017 U.S. invariant) was included in the empirical model.

Industrial upgrading. The current process of shifting the industrial structure from the highly polluting secondary industry to the tertiary industry is critical to CO₂ emissions [61]. Industrial structure upgrading, evaluated as the ratio of secondary industry output to GDP, is incorporated into the model.

Labor Force. Demographic factors, particularly labor force size, significantly influence CO₂ emissions, as highlighted by Wang et al. [62]. While incorporating population dynamics is crucial, using the total population would lead to double-counting its effect on emissions, producing misleading results. Therefore, we use the population growth rate to represent changes in the labor force. This approach captures labor force dynamics while mitigating potential biases from total population measurement.

Level of technology. The impact of the level of technology on carbon emissions is obvious, as a higher level of technology will increase the efficiency of energy use and produce fewer carbon emissions at the same level of economic development. Meanwhile, a higher level of technology can also support the development of renewable energy, which can help reduce emissions [63,64]. Therefore, we use the level of technology as a control variable, which is indicated by the number of scientific and technical journal articles, an indicator that measures a country’s scientific production and academic influence.

International cooperation. Increasingly frequent international trade and foreign direct investment exchanges will strengthen the business and production activities of enterprises, which will have a non-negligible impact on CO₂ emissions [65]. Therefore, trade openness is included in the empirical model, which is assessed by the ratio of total import and export trade to total GDP. Meanwhile, the level of foreign investment is also included in the model, which is measured by the total investment of foreign-invested enterprises as a proportion of GDP.

4.3. Data Summary and Source

The data used in this paper are mainly from the World Bank’s World Development Indicators database (WDI), the U.S. Energy Information Administration (EIA), and the United Nations Statistics Division (UNSD).

Table 2. Data summary.

	(1)	(2)	(3)	(4)	(5)
Variables	Obs	Mean	Std.	Min	Max
Integrated energy poverty index	517	0.847	0.0583	0.494	0.945
Per capita CO2 emissions	517	0.00441	0.00367	3.61 × 10−5	0.0218
Gini index	517	36.49	7.364	23.20	63.40
Per capita GDP	517	12,287	16,535	305.3	108,351
The ratio of secondary industry	517	25.43	7.575	5.270	65.87
Population growth rate	517	0.708	1.127	−2.258	4.488
Level of technology	517	6702	13,239	1.510	85,419
Trade openness	517	51.97	25.66	13.24	191.5

presents the data summary of the variables used in this paper.

5. The Impact of Energy Poverty on Carbon Emissions in BRI Countries

5.1. Results of the Baseline Model

Table 3. Results of the baseline model.

	(1)	(2)	(3)
	OLS	First Stage	Second Stage
Variables	Per Capita CO2 Emissions	Integrated Energy Poverty Index	Per Capita CO2 Emissions
Integrated energy poverty index	−0.0631 ***		−0.101 ***
	(0.00861)		(0.0211)
Gini index		0.000774 ***
		(0.000235)
Per capita GDP	−2.14 × 10−7 ***	−3.56 × 10−6 ***	−3.55 × 10−7 ***
	(6.63 × 10−8)	(6.08 × 10−7)	(1.18 × 10−7)
The ratio of secondary industry	4.12 × 10−5 ***	0.000245	5.13 × 10−5 ***
	(1.54 × 10−5)	(0.000188)	(1.85 × 10−5)
Population growth rate	−0.000165 *	−0.00309 **	−0.000254 **
	(9.10 × 10−5)	(0.00129)	(0.000108)
Level of technology	−1.80 × 10−8	−5.37 × 10−7 **	−3.21 × 10−8 **
	(1.31 × 10−8)	(2.09 × 10−7)	(1.40 × 10−8)
Trade openness	−1.53 × 10−6	0.000391 ***	1.38 × 10−5
	(5.02 × 10−6)	(0.000134)	(9.51 × 10−6)
Constant	0.0599 ***
	(0.00720)
Observations	491	491	491
R-squared	0.978		0.233
Country FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
K-P F			10.84

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

shows the basic regression results of the impact of the integrated energy poverty index on per capita carbon emissions in BRI countries. Column (1) of Table 3 shows the OLS regression results, indicating that there is a significant negative correlation between the integrated energy poverty index and per capita carbon emissions of BRI countries, controlling for two-way fixed effects. To address the possible endogeneity of the model, this paper uses the Gini index of the BRI countries as an instrumental variable, and employs the two-stage least squares (2SLS) regression method for the regression analysis. Column (2) reports the first-stage regression results of 2SLS. The results show that the Gini index of BRI countries has a significantly positive effect on the integrated energy poverty index. The result suggests that the higher the income gap, the higher the integrated energy poverty index and the more severe the status of energy poverty in BRI countries. Column (3) reports the results of the second stage of the 2SLS regression, which show that the integrated energy poverty index of BRI countries has a significantly negative effect on per capita carbon emissions. It indicates that the alleviation of the integrated energy poverty index in BRI countries will significantly increase per capita carbon emissions. Comparing the OLS estimates in column (1) and the 2SLS estimates in column (3), the coefficients of the instrumental variables are slightly higher than those of the OLS estimates, which may be due to the measurement error of the integrated energy poverty index in BRI countries.

The results show that the alleviation of energy poverty leads to an increase in per capita carbon emissions in BRI countries. This finding differs from some of the previous studies. For example, Zhao et al. [66] found a positive correlation between energy poverty and carbon emissions, indicating that regions with higher levels of energy poverty tend to have higher carbon emissions. This may be due to the fact that, as the level of energy poverty is alleviated, people have access to more convenient and abundant energy supplies, and they may therefore use more energy. In this case, while energy poverty is alleviated, the ensuing increase in energy consumption may lead to an increase in carbon emissions. In addition, the alleviation of energy poverty may be achieved by making energy consumption more affordable, which may lead to an increase in energy consumption and hence carbon emissions. However, increased energy cleanliness implies that residents switch from biomass or fossil energy sources to cleaner renewable energy sources, which may also reduce carbon emissions but is inconsistent with our empirical results. Furthermore, Zhao et al. [66] indicated that the relationship between energy poverty and carbon emissions can vary across regions. Their study focused specifically on Chinese provinces, which differs from the scope of our research. Additionally, a study by Li et al. [67] found an inverted U-shaped relationship between energy poverty levels and carbon emissions. This suggests that alleviating energy poverty initially increases carbon emissions until a certain threshold is reached. This finding aligns with our results and provides a potential explanation for the discrepancy with previous studies. Our research and that of Zhao et al. [66] likely examine regions at different stages of development, leading to contrasting impacts of energy poverty on carbon emissions. In order to further explore specific mechanisms, subsequent sections will discuss the impact of each dimension of energy poverty on carbon emissions.

5.2. Robustness Test

5.2.1. Replacing the Explanatory Variable

In the empirical tests, this paper uses the integrated energy poverty index calculated from Section 3 as the explanatory variable. Therefore, the estimation results of this paper are affected by the method of energy poverty assessment. Many studies have used the weighted average method rather than the TOPSIS method used in Section 3 to construct a multidimensional energy poverty index [32]. Therefore, this paper recalculates the multidimensional energy poverty index of BRI countries using the weighted average method and takes it as an alternative explanatory variable for the robustness test. The multidimensional energy poverty index is calculated by standardizing, normalizing, and weighting the indicators. The standardization and normalization methods are the same as those in Section 3, and the weights are calculated using the entropy weight method. In contrast to the integrated energy poverty index, a lower multidimensional energy poverty index indicates a more severe level of energy poverty.

Table 4. Multidimensional energy poverty index and per capita CO₂ emissions.

	(1)	(2)	(3)
	OLS	First Stage	Second Stage
Variables	Per Capita CO2 Emissions	Multidimensional Energy Poverty Index	Per Capita CO2 Emissions
Multidimensional energy poverty index	0.0375 ***		0.0666 ***
	(0.00662)		(0.0139)
Gini index		−0.00117 ***
		(0.000369)
Per capita GDP	−2.30 × 10−7 ***	6.42 × 10−6 ***	−4.23 × 10−7 ***
	(6.34 × 10−8)	(9.72 × 10−7)	(1.31 × 10−7)
The ratio of secondary industry	3.50 × 10−5 **	−0.000250	4.32 × 10−5 **
	(1.55 × 10−5)	(0.000284)	(1.87 × 10−5)
Population growth rate	−0.000171 *	0.00525 **	−0.000292 ***
	(9.34 × 10−5)	(0.00203)	(0.000112)
Level of technology	−1.83 × 10−8	8.83 × 10−7 ***	−3.68 × 10−8 **
	(1.37 × 10−8)	(3.34 × 10−7)	(1.47 × 10−8)
Trade openness	−4.00 × 10−6	−0.000595 ***	1.39 × 10−5
	(5.15 × 10−6)	(0.000204)	(9.61 × 10−6)
Constant	0.00630 ***
	(0.000779)
Observations	491	491	491
R-squared	0.978		0.152
Country FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
K-P F			10.10

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

shows the results of the robustness tests. Column (1) shows the results of the OLS regression, indicating that there is a significant positive relationship between the multidimensional energy poverty index and per capita carbon emissions in BRI countries. As in the previous section, the Gini index of BRI countries is also used as an instrumental variable in the robustness test, and the 2SLS method is used for the regression analysis. Column (2) reports the first-stage regression results of the 2SLS, which show that the Gini index of BRI countries has a significantly negative effect on the multidimensional energy poverty index. The results indicate that the higher the income gap, the lower the multidimensional energy poverty index and the more severe the energy poverty status in BRI countries. Column (3) reports the regression results of the second stage of the 2SLS, which show that the multidimensional energy poverty index of BRI countries has a significantly positive effect on per capita carbon emissions. The results indicate that the alleviation of energy poverty in these countries will significantly increase per capita carbon emissions. The results of using the multidimensional energy poverty index as an explanatory variable are the same as those of the baseline regression, indicating that the findings of this paper are robust.

5.2.2. Replacing Samples

In the previous study, this paper focuses on the impact of energy poverty alleviation on carbon emissions, using BRI countries as the study sample. One of the main steps in the integrated assessment method proposed in Section 3 is the use of entropy weights to determine the weights, and changes in the alternative samples may lead to changes in the weights of the indicators. In addition, different alternative samples may also make the assessment results change based on the TOPSIS method. In a word, these two concerns imply that changes in the sample may have an impact on the final assessment results. Therefore, this paper tests the robustness of the research method by changing the alternative samples. Specifically, the paper expands the study sample from BRI countries to all countries for which data are available.

Table 5. Integrated energy poverty index and per capita CO₂ emissions.

	(1)	(2)	(3)
	OLS	First Stage	Second Stage
Variables	Per Capita CO2 Emissions	Integrated Energy Poverty Index	Per Capita CO2 Emissions
Integrated energy poverty index	−0.0523 ***		−0.0856 ***
	(0.00910)		(0.0149)
Gini index		0.00103 ***
		(0.000229)
Per capita GDP	−5.97 × 10−8 ***	−8.39 × 10−7 ***	−8.91 × 10−8 ***
	(1.55 × 10−8)	(2.46 × 10−7)	(2.34 × 10−8)
The ratio of secondary industry	3.90 × 10−5 ***	0.000307 *	4.92 × 10−5 ***
	(1.48 × 10−5)	(0.000180)	(1.71 × 10−5)
Population growth rate	−8.04 × 10−5	−0.00354 ***	−0.000182 **
	(7.32 × 10−5)	(0.00114)	(9.06 × 10−5)
Level of technology	−1.24 × 10−8	−2.08 × 10−7	−1.29 × 10−8
	(1.12 × 10−8)	(1.89 × 10−7)	(1.21 × 10−8)
Trade openness	−2.23 × 10−6	0.000375 ***	1.11 × 10−5
	(5.30 × 10−6)	(0.000114)	(7.57 × 10−6)
Constant	0.0487 ***
	(0.00762)
Observations	668	668	668
R-squared	0.981		0.184
Country FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
K-P F			20.40

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

shows the results of the robustness tests with the study sample replaced. Column (1) presents the results of the OLS regressions, which show that there is a significant negative relationship between the integrated energy poverty index and per capita carbon emissions for all countries in the world. Column (2) reports the first-stage regression results of the 2SLS, demonstrating a significantly positive effect of the Gini index on the integrated energy poverty index for all countries in the world. The results indicate that the higher the income inequality, the higher the integrated energy poverty index, and the more severe the status of energy poverty in all countries of the world. Column (3) reports the results of the second stage of the 2SLS regression, which show that the effect of the integrated energy poverty index on per capita carbon emissions is significantly negative for all countries in the world. The results indicate that the alleviation of energy poverty in all countries of the world significantly increases per capita carbon emissions. The results are consistent with those of the baseline regression, confirming the robustness of this study.

5.3. Results of Mechanisms

The baseline model confirms that alleviating energy poverty in BRI countries leads to an increase in per capita carbon emissions. Section 3 collectively reflects the level of energy poverty in three dimensions: energy accessibility, cleanliness, and affordability. In this section, we further investigate the impact of each dimension on per capita carbon emissions, so as to explore in depth the mechanisms of energy poverty alleviation on carbon emissions in BRI countries.

In this section, we modify the model to include three dimensions of energy poverty ${{EnergyPoverty}^n}_{i,t}$, namely energy accessibility, cleanliness, and affordability, together as explanatory variables. These three sub-indices were calculated from Section 3. Since these sub-indices are also used to assess the status of energy poverty, they perform the same function as the integrated energy poverty index, with higher values reflecting a more severe level of energy poverty in the country for that dimension. In this study, a linear regression model consistent with the previous section is set up to examine the impact of the three dimensions of energy poverty on per capita carbon emissions.

$${CarbonEmissions}_{i,t}={\beta }_0+\widehat{{\beta }_{1}EnergyPovert{y}_{i,t}^n}+{\beta }_2{X}_{i,t}+{\gamma }_i+{\lambda }_t+{ϵ}_{i,t}$$

(3)

Table 6. Impact of the level of energy poverty in each dimension on CO₂ emissions.

	(1)	(2)	(3)	(4)	(5)
	OLS	OLS	FE	FE	FE
Variables	Per Capita CO2 Emissions	Per Capita CO2 Emissions	Per Capita CO2 Emissions	Per Capita CO2 Emissions	Per Capita CO2 Emissions
Energy accessibility	−0.0440 ***	−0.0503 ***	−0.0499 ***	−0.0610 ***	−0.0585 ***
	(0.00111)	(0.00223)	(0.00299)	(0.00637)	(0.00593)
Energy cleanliness	0.00503 ***	0.00374 ***	0.00360 ***	0.0135 ***	0.0122 **
	(0.000786)	(0.000840)	(0.000806)	(0.00519)	(0.00474)
Energy affordability	0.00223 ***	0.00133 **	0.00103 *	0.000337	0.000129
	(0.000615)	(0.000619)	(0.000582)	(0.000390)	(0.000434)
Per capita GDP		−3.15 × 10−8 ***	−3.04 × 10−8 ***	−2.08 × 10−7 ***	−1.43 × 10−7 ***
		(9.92 × 10−9)	(1.13 × 10−8)	(3.99 × 10−8)	(5.34 × 10−8)
The ratio of secondary industry		1.15 × 10−5	9.09 × 10−6	3.02 × 10−5 **	1.80 × 10−5
		(1.07 × 10−5)	(7.49 × 10−6)	(1.22 × 10−5)	(1.31 × 10−5)
Population growth rate		0.000471 ***	0.000458 ***	−0.000138	−0.000167 *
		(7.74 × 10−5)	(8.95 × 10−5)	(8.57 × 10−5)	(9.21 × 10−5)
Level of technology		2.83 × 10−9	3.80 × 10−9	−3.51 × 10−8 ***	−1.27 × 10−8
		(5.87 × 10−9)	(5.29 × 10−9)	(9.83 × 10−9)	(1.02 × 10−8)
Trade openness		7.75 × 10−6 **	7.40 × 10−6 **	1.87 × 10−6	−1.81 × 10−6
		(3.48 × 10−6)	(3.64 × 10−6)	(4.08 × 10−6)	(4.35 × 10−6)
Constant	0.0389 ***	0.0448 ***	0.0448 ***	0.0518 ***	0.0500***
	(0.00107)	(0.00226)	(0.00304)	(0.00815)	(0.00745)
Observations	517	517	517	491	491
R-squared	0.981	0.973	0.841	0.982	0.97
Country FE	Yes	Yes	No	Yes	Yes
Year FE	Yes	Yes	Yes	No	Yes

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

shows the impact of the alleviation of three dimensions of energy poverty on local per capita carbon emissions in BRI countries. Columns (1) and (2) show the results of OLS regressions without and with control variables, respectively. Columns (3) and (4) are the results of regressions with a year-fixed effect only and a country-fixed effect only, respectively. Column (5) presents regression results with two-way fixed effects.

Initially, for the energy accessibility dimension, the results show that the sub-index of energy accessibility has a negative effect on per capita CO₂ emissions, suggesting that an improvement in energy accessibility increases per capita CO₂ emissions. Secondly, for the dimension of energy cleanliness, the results show that the effect of the sub-index of energy cleanliness on per capita CO₂ emissions is positive, which reveals a positive correlation between improving energy cleanliness and emission reduction. The result also suggests that improvements in energy cleanliness can be achieved at the same time as reductions in per capita CO₂ emissions.

Finally, as for energy affordability, the results of OLS and year-fixed effect tests show a positive effect of energy affordability on per capita CO₂ emissions. However, the results of the country fixed effect and two-way fixed effects tests are not significant, which shows that the positive relationship between improving energy affordability and the level of per capita carbon emissions may be the result of the unobserved characteristics of the sample (countries). The results of the two-way fixed effects test are therefore more plausible, suggesting that improving energy affordability does not directly affect the level of carbon emissions.

Improving energy accessibility evidently increases residents’ per capita energy consumption, inevitably leading to an increase in carbon emission levels. On the other hand, the improvement in energy cleanliness is manifested by residents’ increased use of clean energy sources such as natural gas and electricity. Moreover, in many countries, there is a gradual promotion of renewable energy generation, which results in lower carbon emissions compared to traditional fossil fuels. In summary, alleviating energy poverty and mitigating climate change are not conflicting sustainable development goals, but policymakers still need to take into account their impact on carbon emissions when designating policies to alleviate energy poverty. Therefore, policymakers should pay more attention to improving the cleanliness and affordability of energy in order to synergize energy poverty alleviation and carbon emissions reduction.

5.4. Analysis of Indicators

This study uses three dimensions of sub-indices to construct a framework to assess the level of integrated energy poverty, and uses the integrated energy poverty index calculated as the explanatory variable of the study. We hope to further explore the mechanism by which energy poverty alleviation affects carbon emissions. Therefore, we examine each of the eight indicators in the integrated energy poverty assessment framework and empirically test the impact of the improvement of each indicator on the per capita carbon emissions in BRI countries.

Table 7. Energy poverty indicators and per capita carbon emissions.

	(1)	(2)	(3)	(4)	(5)
	OLS	OLS	FE	FE	FE
Variables	Per Capita CO2 Emissions	Per Capita CO2 Emissions	Per Capita CO2 Emissions	Per Capita CO2 Emissions	Per Capita CO2 Emissions
Access to electricity	−1.43 × 10−5 *	−1.08 × 10−5	−9.69 × 10−6	2.03 × 10−5 **	1.89 × 10−5 **
	(7.31 × 10−6)	(7.38 × 10−6)	(6.16 × 10−6)	(9.96 × 10−6)	(9.01 × 10−6)
Electricity consumption	0.000905 ***	0.00100 ***	0.00102 ***	−7.87 × 10−5	−5.73 × 10−5
	(7.03 × 10−5)	(7.97 × 10−5)	(0.000176)	(0.000240)	(0.000220)
Energy consumption	1.78 × 10−5 ***	1.12 × 10−5 ***	1.05 × 10−5	4.73 × 10−5***	4.59 × 10−5 ***
	(2.56 × 10−6)	(2.78 × 10−6)	(6.90 × 10−6)	(7.89 × 10−6)	(8.86 × 10−6)
Access to clean fuels	−1.09 × 10−5 **	−1.36 × 10−5 ***	−1.37 × 10−5 ***	1.11 × 10−6	2.01 × 10−6
	(4.26 × 10−6)	(4.61 × 10−6)	(3.65 × 10−6)	(6.64 × 10−6)	(7.48 × 10−6)
Renewable energy share	−4.61 × 10−5 ***	−5.21 × 10−5 ***	−5.12 × 10−5 ***	−7.39 × 10−5 **	−7.37 × 10−5 ***
	(4.94 × 10−6)	(5.35 × 10−6)	(4.95e-06)	(2.89 × 10−5)	(2.67 × 10−5)
Primary energy intensity	0.000319 ***	0.000355 ***	0.000350 ***	0.000538 ***	0.000565 ***
	(3.31 × 10−5)	(3.77 × 10−5)	(5.46 × 10−5)	(0.000181)	(0.000202)
Firm’s cost of getting electricity	−2.70 × 10−8	−2.11 × 10−8	−2.55 × 10−8	2.89 × 10−8	3.50 × 10−8
	(4.21 × 10−8)	(4.08 × 10−8)	(2.39 × 10−8)	(4.12 × 10−8)	(4.51 × 10−8)
Mobile phone subscribers (per 100 people)	3.22 × 10−6	4.05 × 10−6	4.46 × 10−6 *	1.15 × 10−6	3.23 × 10−7
	(2.62 × 10−6)	(2.56 × 10−6)	(2.54 × 10−6)	(1.68 × 10−6)	(1.87 × 10−6)
Observations	517	517	517	491	491
R-squared	0.878	0.886	0.888	0.986	0.986
Control variables	No	Yes	Yes	Yes	Yes
Country FE	No	No	No	Yes	Yes
Year FE	No	No	Yes	No	Yes

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

reports the relationship between the energy poverty indicators and per capita carbon emissions, with all indicators together as explanatory variables and per capita CO₂ emissions as the dependent variable. Columns (1) and (2) show the results of OLS regressions without and with control variables, respectively. Columns (3) and (4) show the results of tests with year-fixed effect only and country-fixed effect only, respectively. Column (5) presents regression results for two-way fixed effects.

In Table 7, the results show improvements in the proportion of the population with access to electricity and in per capita energy consumption increase per capita CO₂ emissions. However, improvements in the consumption of renewable energy decrease per capita CO₂ emissions, while higher primary energy intensity increases per capita CO₂ emissions. Moreover, the effect of per capita electricity consumption on per capita carbon emissions is significantly positive in the OLS and year fixed effects tests, but not significant in the country and two-way fixed effects tests, which could also be due to the presence of unobserved sample characteristics. Similarly, the non-significant results for the proportion of the population using clean fuels in the two-way fixed effects similarly indicate the presence of unobserved sample characteristics in the other tests.

These findings also explain the results of the mechanism analysis. The regression results for each indicator show that, while overall energy poverty alleviation leads to an increase in CO₂ per capita, the improvement of different energy poverty indicators has distinct impacts on CO₂ per capita. Improvements in the proportion of the population with access to electricity and per capita energy consumption are the main drivers of the increase in per capita CO₂ emissions, while an improvement in the level of renewable energy consumption mitigates per capita CO₂ emissions. This result also informs policymakers that governments can alleviate energy poverty by developing incentives to promote the development and utilization of renewable energy.

6. Conclusions

The main purpose of this paper is to verify whether the alleviation of energy poverty in the Belt and Road Initiative (BRI) countries will lead to an increase in carbon emissions. Therefore, this paper uses per capita carbon emissions as the explanatory variable and the comprehensive energy poverty index as the explanatory variable to construct a regression model. In addition, this paper uses the data of 70 BRI countries from 2010 to 2020 for empirical testing. In order to address possible endogeneity, this paper uses the Gini index, which characterizes the level of income inequality as an instrumental variable, and empirically tests it using the two-stage least squares method. The results suggest that alleviating energy poverty in BRI countries will lead to an increase in carbon emissions. The paper also conducts a robustness test by replacing the explanatory variables and the study sample separately and finds that the estimates remain robust.

In order to further explore the mechanisms of energy poverty alleviation on carbon emissions, this paper examines the effect of improvements in energy poverty levels in different dimensions on carbon emissions. The results suggest that alleviating energy poverty leads to an increase in per capita CO₂ emissions by increasing energy accessibility. However, alleviation of the energy cleanliness dimension can also achieve a reduction in per capita CO₂ emissions. In summary, the above results suggest that energy poverty alleviation and emission reduction are two non-conflicting sustainable development goals.

This study offers several policy implications for BRI countries. Our findings demonstrate that alleviating energy poverty and reducing carbon emissions are not mutually exclusive goals. Therefore, policymakers should adopt a comprehensive approach when designing energy poverty alleviation programs, incorporating carbon reduction targets, and considering various influencing factors. The mechanism analysis highlights the importance of energy cleanliness in reducing carbon emissions. BRI countries should prioritize the utilization of clean energy sources, such as solar and wind power, to address energy poverty. This strategy can achieve a win-win scenario by simultaneously alleviating energy poverty and reducing carbon emissions. Governments should strengthen their support for renewable energy through subsidies and incentives, encouraging businesses and individuals to transition away from traditional energy sources. Additionally, our study finds that enhancing energy affordability does not necessarily lead to increased carbon emissions. Therefore, governments can also focus on improving energy affordability, for example, by establishing targeted energy subsidy programs to help low-income households afford energy services. Furthermore, it’s also crucial to encourage energy companies to develop and promote affordable, clean energy technologies.