Would narrowing the income gap help mitigate the greenhouse effect? Fresh insights from spatial and mediating effects analysis

Abstract

Fully considering income inequality (ICIE) and its potential environmental effects can help achieve the goal of reducing carbon emissions. Based on this, to effectively explore the nexus between ICIE and carbon dioxide (CO₂) emissions, we first deduce the theoretical framework and then empirically check the spatial impact of ICIE on greenhouse effect based on a global dataset from 1995 to 2016. In addition, this study also conducts the regional heterogeneous analysis and further detects the causal mediating route (i.e., indirect route) between variables. The findings suggest that: (i) Rising ICIE affects CO₂ emissions positively; in other words, narrowing income gap can effectively mitigate greenhouse effect across the globe; (ii) In countries with high gross domestic product (GDP) level, expanded ICIE will exacerbate the greenhouse effect, while in low-GDP countries, ICIE is negatively correlated with CO₂ emissions; and (iii) ICIE can indirectly influence greenhouse effect through substantial economic and technical routes, while the composition route is insignificant. Finally, this paper highlights policy suggestions for carbon reduction by adjusting income distribution, formulating targeted strategies in various countries, and promoting technical innovation.

1 Introduction

In the past few decades, with the continuous advancement of urbanization and industrialization, the economies of all countries in the world have entered a stage of rapid growth, which has significantly promoted the expansion of economic scale (Alam et al. 2016; Sarkodie and Strezov 2019; Song et al. 2018). Although economic growth has reduced absolute poverty to some extent, it has inevitably widened the income gap (Dollar and Kraay 2002). As Rubin and Segal (2015) suggested, incomes at the top were more sensitive to economic growth than those at the bottom. They concluded that economic growth in a country or area was positively correlated with widening income gap; in other words, rapid economic growth would gradually exacerbate income inequality (ICIE). Notably, uneven income distribution will not only promote the widening of the regional and urban–rural ICIE, but also increase residents’ resentment toward the rich, and aggravate social contradictions. To a large extent, this may lead to political and social instability and affect harmonious and sustainable development nationally (Meschi and Vivarelli 2009; Sulemana and Kpienbaareh 2018; Zhao et al. 2020). Thus, all countries should take measures to narrow the gap of residents’ income.

In addition to narrowing ICIE, mitigating climate change has become another major challenge facing mankind in the twenty-first century (Dong et al. 2022c; Grunewald et al. 2017; Ren et al. 2023a; Yan et al. 2023; Zhang et al. 2023; Zhao et al. 2021). In the past few decades, the development of industrialization to promote economic growth has consumed large amounts of fossil energy, which emits a lot of ecological pollutants (Li et al. 2023; Khan et al. 2021a, b; Ren et al. 2023c; Zhao and Dong 2023). Besides, the consequence of massive fossil fuels use is the increasingly growing carbon dioxide (CO₂) emissions (Duan et al. 2021; Khan et al. 2021a, b). Following the statistical data of former British Petroleum (BP 2020), overall global CO₂ emissions have increased significantly, with a growth rate of 55.43% from 1995 to 2019. This has exacerbated global warming, accelerated the rise of sea levels and the melting of glaciers, substantially intensified the greenhouse effect, and led to abnormal climate change, which produces a catastrophic impact on the ecological environment and the evolution of human society (Chen et al. 2023a; Xiang et al. 2023; Yang et al. 2017). In recent years, the signing of agreements such as the Kyoto Protocol and the Paris Agreement, and the recent vigorous advocacy of carbon neutrality, effectively reflect the confidence and determination of governments in various countries to accelerate carbon reduction (Dou et al. 2021; Zou et al. 2023).

In light of the above background, the ICIE-CO₂ nexus has attracted great and high attention from scholars. Since the environmental Kuznets curve (EKC) which refers to the work of Kuznets (1955) stresses the obvious causal nexus between income and environment, many scholars have begun to investigate whether narrowing the income gap can significantly influence ecological environment quality with the increasing attention to ICIE (Baek and Gweisah 2013; Heerink et al. 2001). Concerning CO₂ emissions, relevant literatures on the impact of ICIE on greenhouse effect have not obtained a consistent and uniform conclusion by using various econometric models, which mainly consist of positive correlation (Baloch et al. 2020; Uzar and Eyuboglu 2019), negative correlation (Khan et al. 2018), and an inverted U-shaped linkage (Liu et al. 2019a). And it has been verified that unbalanced income distribution between high- and low-income groups has a significant impact on the environment through personal preference choices (Baek and Gweisah 2013). However, few studies have explored the ICIE-CO₂ nexus from the spatial dimension. Also, the heterogeneous effect and causal mediating effect between ICIE and CO₂ are worth considering. In this regard, we first use a theoretical model to derive the expected ICIE-CO₂ nexus, and then apply the global sample data to systematically examine the spatial impact of narrowing income gap on greenhouse gas emissions across the globe. Furthermore, this study analyzes the regional heterogeneity and conducts a causal mediating effect analysis through three major mediators, mainly including economic, technical, and composition routes.

Accordingly, our study contributes to the extant literature from three ways: (1) Based on the detailed theoretical framework and proposed hypothesis, this study empirically explores the spatial impact of ICIE on the greenhouse effect by using a spatial econometric method. This not only effectively fills the academic gap of the ICIE-CO₂ nexus from the spatial econometrics dimension, but also offers theoretical support for national policymakers to promote carbon reduction in the context of income distribution; (2) By dividing the whole panel into two sub-panels, this study creatively checks the different effect across various areas. Such an analysis can help local governments adjust their policies to balance income distribution and reduce CO₂ emissions; and (3) To test the internal impact mechanism, we conduct causal mediating analysis through three typical major mediators, which is particularly useful for exploring the transmission pathways of ICIE on greenhouse effect, thus providing an effective reference for formulating carbon emission reduction policies.

We organize the rest of this paper in the following structure. The next section summarizes the related literature and proposes the literature gaps, followed by the theoretical framework between the variables deduced in Sect. 3. In Sect. 4, the spatial methodology and data are described. Section 5 analyzes the empirical findings. Section 6 further explores the transmission pathways, and the last section concludes the entire paper.

2 Literature review

2.1 Relevant studies on the income-CO₂ nexus

In recent years, a growing body of scholars has investigated the causal relationship between income and CO₂ emissions, which can be traced back to the EKC hypothesis proposed by Grossman and Krueger (1995) and drawn inspiration from the work of Kuznets (1955). This hypothesis validates the inverted U-shaped relationship between economic growth and environmental degradation, which is also supported by Alam et al. (2016), Churchill et al. (2018), Galeotti et al. (2006), Iwata et al. (2010), Lin et al. (2017), and Saboori et al. (2012). However, some scholars’ findings are not consistent with the EKC hypothesis. Specifically, Kang et al. (2016) empirically examined the CO₂ EKC hypothesis in China by using a spatial panel model covering the period 1997–2012. The results confirmed that a spatial spillover effect could affect the shape of CO₂ EKC, and both economic growth and CO₂ emissions showed an inverted N-shaped trajectory. Also, by conducting an empirical analysis, He and Richard (2010) found very little proof to support the EKC premise. Mikayilov et al. (2018) investigated the economic growth-CO₂ nexus in Azerbaijan over the period 1992–2013, and concluded that economic growth had a statistically positive impact on CO₂ emissions. Chen et al. (2023b), Jahanger (2022), and Ren et al. (2023b) obtained the same conclusion.

2.2 Relevant studies on the income inequality-CO₂ nexus

At present, there have been global concerns about whether ICIE will influence greenhouse effect, and some scholars have dedicated considerable time and effort to examine the ICIE-CO₂ nexus. To be specific, Zhang and Zhao (2014) explored the ICIE-CO₂ nexus at the national and regional levels for the period 1995–2010 in China, indicating that a positive relationship existed between the variables, and more equitable income distribution might be conducive to promoting carbon reduction. Furthermore, using panel data of China’s 403 prefecture-level cities for the period 1996–2014, Liu et al. (2019b) emphasized that the widening ICIE could deteriorate environmental quality and aggravate the greenhouse effect. The positive nexus between ICIE and CO₂ was also supported by Baek and Gweisah (2013), Baloch and Danish (2022), Hao et al. (2016), Uzar and Eyuboglu (2019), Yang et al. (2022), and Zhu et al. (2018). In contrast to these findings, Khan et al. (2018), using panel data during the period 1980–2014, revealed that ICIE in India and Pakistan could significantly reduce CO₂ emissions. Wan et al. (2022) applied the cross-country panel data (i.e., 217 nations) to test the transmission mechanism between ICIE and CO₂ emissions; they found that high ICIE boosted the increase in scientific expenditure, and thus reduced CO₂. Furthermore, considering the nonlinear characteristics, Liu et al. (2019a) discovered that ICIE could exacerbate greenhouse effect in the short term, but promote carbon reduction in the long run. By employing panel data of G7 countries between 1870 and 2014, Uddin et al. (2020) implied that ICIE showed a significant positive impact on CO₂ emissions from 1870 to 1880 and a negative effect from 1950 to 2000, which suggested a significant nonlinear relationship. The nonlinear nexus was also verified by Ozturk et al. (2022).

Many scholars have also studied the heterogeneous relationship impact of ICIE on CO₂ emissions (Safar 2022). For instance, Jorgenson et al. (2017) selected two measures of ICIE, and indicated that the share of income of the top 10% of income earners would promote CO₂ emissions, while the impact of the Gini coefficient on CO₂ emissions was insignificant. The findings of Grunewald et al. (2017) showed that widening ICIE could mitigate CO₂ emissions for low- and middle-income economies, while in high-income economies, ICIE was positively associated with CO₂ emissions. Furthermore, based on a panel dataset for 78 countries from 1990 to 2017, Wu and Xie (2020) revealed that higher ICIE could promote carbon emission reduction in countries belonging to the Organization for Economic Co-Operation and Development (OECD) and high-income non-OECD countries. Following an extended EKC model, Chen et al. (2020) insisted that, for developing countries, ICIE had a detrimental impact on CO₂ emissions, while the impact of ICIE on CO₂ emissions was not significant for most developed countries. Although many scholars were committed to studying the causal ICIE-CO₂ nexus, few studies had examined the spatial effect and causal mediating routes between ICIE and CO₂, which could help supplement relevant literature.

2.3 Literature gaps

By reviewing and summarizing the existing literature related to ICIE-CO₂ nexus, we find that although numerous scholars have comprehensively investigated the positive, negative, and nonlinear ICIE-CO₂ nexus, few scholars have focused on the spatial effect of ICIE on the greenhouse effect. A better understanding of the spatial correlation characteristics of CO₂ emissions is of great value for a clear study of the ICIE-CO₂ nexus. Furthermore, due to the differences in economic level, geographical location, natural resource endowment, and population of various countries, the ICIE-CO₂ nexus may have regional heterogeneity. Exploring the regional heterogeneous impact of narrowing ICIE on greenhouse effect can provide sufficient reference and theoretical basis for policymakers of national governments to formulate relevant measures and policies according to local conditions. Thus, this regional heterogeneity deserves further discussion. In addition, at present, few scholars have examined how ICIE affects greenhouse effect, that is, the ways in which ICIE influences carbon reduction. However, this impact mechanism, especially the spatial causal mediating effect in the ICIE-CO₂ nexus is often ignored. Discussing the specific impact channels from ICIE to CO₂ emissions can be particularly useful for policymakers to develop suitable and effective measures to mitigate the greenhouse effect.

3 Theoretical framework

To construct the theoretical model between ICIE and greenhouse gas emissions, this section first assumes that the utility of individual i relies on the impacts of both personal consumption (C) and environmental quality (Q); the utility function of individual i can be assumed as ${U}_{i}={C}_{i}+{\theta }_{i}Q$, where ${\theta }_{i}$ represents the degree of preference of individual i for environmental quality. Notably, energy consumption (E) and environmental quality show a significant negative linkage (Zhao et al. 2022). Thus, ${Q}_{E}=\partial Q/\partial E<0$.

According to the utility function assumed above, we can find that, in addition to increased personal consumption, residents can also enhance their utility by improving the environmental quality. Thus, the government can control energy consumption by levying environmental taxes and fees (expressed as S) on polluting enterprises, thereby improving environmental quality. Accordingly, ${Q}_{S}=\partial Q/\partial S>0,{E}_{S}=\partial E/\partial S<0$. Moreover, we assume that the rate of government taxes and fees is $\uptau$, which refers to the proportion of individual i’s income used to improve environmental quality of the total income (denoted as Y_i). Thus, ${C}_{i}=\left(1-\tau \right){Y}_{i}, S=(\tau -{\tau }^{2}/2)Y$, where Y denotes per capita income.

By referring to the work of Magnani (2000), we suppose that the indirect utility function of individual i is presented in the following equation:

$$U_{i} = \left( {1 - \tau } \right)Y_{i} + \theta_{i} \left( {\tau - \frac{{\tau^{2} }}{2}} \right)Y$$

(1)

Taking the partial derivative of $\uptau$ in Eq. (1) and setting it to 0, we can obtain the following equation:

$$\partial U_{i} /\partial \tau = - Y_{i} + \theta_{i} Y\left( {1 - \tau } \right) = 0$$

(2)

According to Eq. (2), we can obtain the $\tau_{i}$, as follows:

$$\tau_{i} = 1 - G_{i} /\theta_{i} , {\text{where}} G_{i} = Y_{i} /Y$$

(3)

We further assume that Y_m represents the median income, thus, ${G}_{m}={Y}_{m}/Y$. Notably, the smaller the G_m, the more severe the ICIE (Magnani 2000). Accordingly, the maximum tax rate (i.e., ${\tau }^{*}$) and government taxes and fees (i.e., ${S}^{*}$) are obtained as follows:

$$\tau^{*} = 1 - G_{m} /\theta_{m}$$

(4)

$$S^{*} = \left[ {\tau^{*} - \frac{{\left( {\tau^{*} } \right)^{2} }}{2}} \right]Y$$

(5)

Thus, the marginal effect of individual i’s relative income G_i on the maximum tax rate ${\tau }_{i}^{*}$ is as follows:

$$\frac{{\partial \tau_{i}^{*} }}{{\partial G_{i} }} = \frac{1}{{\theta_{i} }}\left( { - 1 + \frac{{\partial \theta_{i} }}{{\partial G_{i} }} \cdot \frac{{G_{i} }}{{\theta_{i} }}} \right)$$

(6)

It has been confirmed that preference for environmental quality (${\theta }_{i}$) is positively correlated with personal relative income (G_i); thus, when relative income increases, residents are willing to pay a higher part of their income to improve environmental quality, that is,$\partial {\tau }_{i}^{*}/\partial {G}_{i}>0$, and $\frac{{\partial \theta_{i} }}{{\partial G_{i} }} \cdot \frac{{G_{i} }}{{\theta_{i} }} > 1$. Such a finding is also verified by Drabo (2011). Thus, the impact of ICIE on environmental quality can be gauged as follows:

$$\frac{\partial Q}{{\partial G_{m} }} = \frac{\partial Q}{{\partial E}} \cdot \frac{\partial E}{{\partial S^{*} }} \cdot \frac{{\partial S^{*} }}{{\partial \tau^{*} }} \cdot \frac{{\partial \tau^{*} }}{{\partial G_{m} }} = \frac{{Q_{E} E_{s} \cdot \left( {1 - \tau^{*} } \right)Y}}{{\theta_{m} }}\left( { - 1 + \frac{{\partial \theta_{m} }}{{\partial G_{m} }} \cdot \frac{{G_{m} }}{{\theta_{m} }}} \right)$$

(7)

According to the above discussion, $Q_{E} < 0, E_{{S^{*} }} < 0$, and $\frac{{\partial \theta_{m} }}{{\partial G_{m} }} \cdot \frac{{G_{m} }}{{\theta_{m} }} > 1$; thus, we can obtain $\partial {\text{Q}}/\partial G_{m} > 0$. This implies that a balanced income distribution can help improve environmental quality; in other words, rising ICIE can facilitate CO₂ emissions. Such a finding is confirmed by numerous scholars (Baek and Gweisah 2013; Baloch et al. 2020; Liu et al. 2019b; Uddin et al. 2020; Zhang and Zhao 2014): the imbalance of income distribution will affect residents’ carbon emission reduction behavior by influencing household energy consumption preference, residents’ ability to purchase environmentally friendly products, and their willingness to protect the environment. Therefore, this section puts forward the hypothesis, as follows:

Hypothesis 1

Widening ICIE will aggravate the greenhouse gas emissions; in other words, equal income distribution can effectively promote carbon reduction.

4 Methodology and data

4.1 Empirical model

As Dietz and Rosa (1997) noted, the widely used stochastic impacts by regression on population, affluence, and technology (STIRPAT) model emphasized that population, affluence, and technology affected CO₂ emissions significantly, expressed by population density, economic growth, and technological innovation, respectively. Thus, the model can be built as follows:

$${\text{CO}}_{{2{\text{it}}}} = {\text{f}}({\text{Den}}_{{{\text{it}}}} ,{\text{ GDP}}_{{{\text{it}}}} , {\text{Tec}}_{{{\text{it}}}} )$$

(8)

where CO₂ indicates the amount of per capita CO₂ emissions of each country; Den indicates population density; GDP denotes economic growth, which can be expressed by the per capita gross domestic product (GDP); Tec indicates technological innovation, which is assessed by energy efficiency, i.e., the ratio of total output value to energy use.

To explore the ICIE-CO₂ nexus, this study focuses on adding ICIE into Eq. (8). In addition, the STIRPAT model including ICIE is also extended by introducing a series of variables (e.g., trade structure, industrial structure adjustment, urbanization level, and labor force) into the model. Accordingly, the extended multivariate framework is presented as follows:

$${\text{CO}}_{{2{\text{it}}}} = {\text{f}}\left( {Gini_{{{\text{it}}}} ,Den_{{{\text{it}}}} ,GDP_{{{\text{it}}}} ,Tec_{{{\text{it}}}} ,Tra_{{{\text{it}}}} ,Ind_{{{\text{it}}}} ,Urb_{{{\text{it}}}} ,Lab_{{{\text{it}}}} } \right)$$

(9)

where Gini indicates income inequality, which is evaluated by the Gini coefficient; Tra denotes trade structure, which can be calculated by the ratio of total import and export trade in total output value; Ind refers to industrial adjustment, which is calculated by the proportion of the output value of secondary industry in total output value; Urb describes urbanization level measured by the proportion of the urban population in the total population; and Lab represents labor force. The specific description of all the selected variables is displayed in Table A1 in the Appendix.

Table 1 Descriptive statistics of the variables (after logarithm)

To further eliminate the risk of potential heteroscedasticity and the influence of data fluctuation, this study applies natural logarithm processing to Eq. (9); the model can be presented as follows:

$${\text{LnCO}}_{{2{\text{it}}}} = \alpha_{0} + \alpha_{1} {\text{LnGini}}_{{{\text{it}}}} + \mathop \sum \limits_{k = 2}^{8} \alpha_{k} LnX_{{{\text{it}}}} + \varepsilon_{{{\text{it}}}}$$

(10)

where subscripts i and t represent country and year, respectively. ${\alpha }_{0}$ and ${\varepsilon }_{it}$ indicate the intercept and error terms, respectively. ${\alpha }_{1}-{\alpha }_{8}$ are the parameters to be estimated. X denotes a vector that contains control variables (i.e., economic growth, technological innovation, population density, urbanization level, industrial structure adjustment, labor force, and trade structure). Following the analysis in Sect. 3, we expect the regression parameter of income inequality (i.e., ${\alpha }_{1}$) to be positive.

4.2 Estimation strategy

4.2.1 Exploratory spatial data analysis

Before we quantitatively explore the causal ICIE-CO₂ nexus, it is indispensable to check whether there is a spatial effect on greenhouse gas emissions, which is crucial for selecting an appropriate econometric regression method. Thus, we select exploratory spatial data analysis (ESDA) to test the spatial correlation of CO₂ emissions, which consists mainly of global and local spatial autocorrelation analysis. Specifically, global spatial autocorrelation analysis aims to describe the spatial distribution of CO₂ emissions in the whole region, which is measured by the global Moran’s index (Moran 1950); the calculation equation is as follows:

$${\text{Moran}}^{\prime}{\text{s}} I = \frac{{\mathop \sum \nolimits_{i} \mathop \sum \nolimits_{j} W_{ij} \left( {x_{i} - \overline{x}} \right)\left( {x_{j} - \overline{x}} \right)}}{{\mathop \sum \nolimits_{i} \left( {x_{i} - \overline{x}} \right)}}$$

(11)

where $\overline{x} = \frac{1}{N}\mathop \sum \limits_{i = 1}^{N} x_{i}$, and x_i is the logarithmic form of CO₂ emissions in country i; N indicates the sample size of the study (i.e., 45 countries across the globe). Besides, the value of Z determines the power of spatial autocorrelation, and ${\text{Z}}=\frac{I-E(I)}{SD(I)}$. Notably, the larger the Z value (absolute value), the stronger the spatial autocorrelation.

Moreover, local spatial autocorrelation analysis is to evaluate the characteristics of regional spatial agglomeration. The conventional evaluation method is the Moran scatter plot, which usually contains four quadrants and is used to intuitively reflect the specific form of the spatial correlation between a certain area and its adjacent areas. Notably, the first to fourth quadrants represent high-high (HH), low–high (LH), low-low (LL), and high-low (HL) agglomeration, respectively.

4.2.2 Panel spatial autoregressive model

After verifying the spatial correlation of CO₂ emissions, the next step is to evaluate the specific spatial relationship in the ICIE-CO₂ nexus. Due to the possibility of spatial correlation in CO₂ emissions, we further apply the spatial autoregressive (SAR) model to detect the underlying effect between the two variables (i.e., Gini and CO₂). This model can enable us to identify the disequilibrium shocks of some driving factors on CO₂ emissions by considering the spatial lag of CO₂ emissions. As Kelejian and Prucha (2010) stressed, SAR model can apply instrumental variables to perform regression estimation through the generalized method of moments (GMM) technique, which can tackle the possible endogeneity problems in the model with a high probability. Accordingly, the SAR model is applied as the benchmark regression model in our empirical estimation. The specific SAR model is presented as follows:

$${\text{LnCO}}_{{2{\text{it}}}} = \varphi_{1} + \lambda \mathop \sum \limits_{j = 1}^{N} W_{{{\text{ij}}}} {\text{LnCO}}_{{2{\text{it}}}} + \beta_{1} {\text{LnGini}}_{{{\text{it}}}} + \mathop \sum \limits_{k = 2}^{8} \beta_{k} {\text{Ln}}X_{{{\text{it}}}} + \varepsilon_{{{\text{it}}}}$$

(12)

where Eq. (12) presented above contains the spatial impact feature in the endogenous spatial impact ($\sum_{j=1}^{N}{W}_{ij}{{\text{LnCO}}}_{2{\text{it}}}$). W_ij represents a spatial weight matrix, and ${W}_{{\text{ij}}}=1/{d}_{{\text{ij}}}$, where d_ij is the Euclidean distance between the centroids of two countries s_i and s_j, and ${W}_{{\text{ij}}}=0$ (LeSage and Pace 2009). ${\varphi }_{i}$ indicates the fixed-effect parameter. $\uplambda$ is the coefficient for the spatial impact. The estimated residual terms (${\varepsilon }_{it}\sim N(0,{\sigma }_{\varepsilon }^{2})$) are identically and independently distributed (i.i.d.).

4.3 Data

To empirically detect the spatial effect of ICIE-CO₂ nexus in various countries, we utilize a panel dataset from 1995 to 2016 of 45 economies. Our sample does not consist of other countries because the inaccessibility of data. Additionally, the data on per capita CO₂ emissions (denoted by CO₂), income inequality (denoted by Gini), economic growth (expressed as GDP), technological innovation (denoted by Tec), industrial structure adjustment (denoted by Ind), urbanization level (denoted by Urb), the labor force (denoted by Lab), population density (denoted by Den), and trade structure (denoted by Tra), have been collected from the World Development Indicators (WDI) published by the World Bank (2017). The specific descriptive statistics of all selected variables are reported in Table 1.

5 Estimation results and discussion

5.1 Results of the spatial autocorrelation test

Before applying the SAR model, we employ the annual Moran’s I to check the potential existence of the spatial effects of the CO₂ emissions for 45 countries; the specific results are reported in Table 2. It is obvious that all statistics values in this table are positive and statistically significant at the 1% statistical significance level, which indicates that CO₂ emissions exhibit a positive spatial spillover effect in our sample countries during the period 2004 to 2016. However, Moran’s I test can only describe the spatial autocorrelation characteristics of the overall sample in our study. Therefore, to explore the spatial autocorrelation between each individual in the overall sample, we display a Moran scatter plot for the years 2004, 2008, 2012, and 2016, respectively, which is displayed in Fig. 1. From this figure, it is clear most countries are distributed mainly in the first and third quadrants; in other words, most countries are in a state of HH and LL agglomeration, which also apparently presents a strong positive significant spatial autocorrelation. The above analysis implies that the spatial model is preferred to the traditional regression model.

Table 2 Moran’s I statistics for CO₂ emissions (2004–2016)

Fig. 1

Moran’ I scatter plot of CO₂ emissions

5.2 Results of the SAR estimation model

To further check the existence of spatial effect and whether the spatial model estimation can be more effective than the traditional regression model, this study conducts the estimation of the panel data model with fixed effects (PD-FE) and the SAR model simultaneously; the empirical results are posted in Table 3. Notably, the spatial coefficient (i.e., lambda) is significantly positive at the 1% level. This implies that significant spatial effect in the ICIE-CO₂ nexus is built. Moreover, the coefficients of ICIE in the PD-FE and SAR models are negative and positive, respectively, which indicates that it is necessary to utilize the results of the SAR as the baseline estimation (see Table 3).

Table 3 Comparison of the estimates between the panel data model with fixed effects and SAR

The finding indicates that a growth of ICIE by 1% will facilitate CO₂ emissions by 0.0374%; by implication, widening ICIE cannot contribute to achieve the goal of reducing carbon emissions. This conclusion coincides with the estimated findings of Meng et al. (2011) and Zhang and Zhao (2014), who emphasized that equal income distribution significantly helped reduce CO₂ emissions, which effectively verified Hypothesis 1. In our opinion, the specific reasons can be interpreted from the following two aspects. On the one hand, as Boyce (1994) indicated, when the income gap was too large, the poor tended to over-exploit natural resources and destroyed the environment to improve their living standards and meet basic living needs. At the same time, the education level of the poor inhibits environmental awareness, which increases CO₂ emissions to some extent. In addition, although rich people pay more attention to environmental quality, those who are wealthy do not necessarily increase investments to improve the environment (Scruggs 1998). To protect their own interests, wealthy people tend to transfer their property to other countries with lower political and environmental risks. This implies that widening ICIE can promote CO₂ emissions. On the other hand, the go-between theory indicates that as income continues to increase, the distribution of residents’ income gradually becomes equal, which significantly increases the relative income of go-betweens. Thus, they tend to pay more for environment-friendly products that will improve environmental quality (Magnani 2000). Consequently, when the income gap gradually widens, the relative income of the go-betweens decreases, and the preference for environment-friendly products weakens. This will increase residents’ consumption of polluted products, thereby positively affecting the greenhouse effect.

For the control variables, the regression parameters of economic growth, trade structure, and population density are significantly positive, while the labor force has an insignificant effect on aggravating global greenhouse effect. To be specific, economic growth relies heavily on massive energy consumption, and frequent factors flows and production activities will inevitably produce massive CO₂ emissions (Dong et al. 2022a; Duan et al. 2023). Moreover, higher population density means a country is densely populated, and residents’ demand for housing and transportation results in the consumption of more energy to meet the requirements of industry development (e.g., manufacturing, electricity, and transportation), which aggravates the greenhouse effect. Also, trade openness can boost the expansion of economic activities and increased energy consumption, inevitably accelerating CO₂ emissions (Dong et al. 2022b). On the contrary, technological innovation and industrial structure adjustment are valid accelerators to address greenhouse effect and facilitate carbon neutrality (Cheng et al. 2021; Zhou et al. 2013). With the increasing emphasis of national officials on ecological protection and green technologies, each country is committed to the research and development (R&D) of green and low-polluting technologies. Following the statistical data from BP (2020), the energy use per unit of GDP in China increased from 2.149 US dollars/kg of oil equivalent in 1995 to 6.060 US dollars/kg of oil equivalent in 2016, and that in Russia increased from 1.308 US dollars/kg of oil equivalent from 5.614 US dollars/kg of oil equivalent. The green revolution of Britain has led to the rapid development of green technologies, with energy use per unit of GDP increasing from 5.502 US dollars/kg of oil equivalent to 15.902 US dollars/kg of oil equivalent in 2016, an increase of approximately three times. This trend will significantly reduce the amount of energy used, decrease the energy intensity, and substantially increase the GDP created by unit energy consumption, which is conducive to reducing CO₂ emissions. Furthermore, to reduce production costs and maximize profits, high-polluting enterprises are beginning to pay attention to industrial upgrading and transformation. In addition to developing clean technologies, polluting enterprises are starting to innovate aggressively, gradually moving away from secondary industries to higher-value tertiary industries. This can promote green and sustainable development of manufacturing, and thus help alleviate CO₂ emissions.

5.3 Robustness analysis

We further conduct a robustness check by considering two different spatial weight matrices (i.e., W₁ and W₂) to detect whether the findings of SAE model is stable and robust. To be more specific, for W₁, ${W}_{{\text{ij}}}=1/{d}_{{\text{ij}}}^{2}$, where d_ij is the Euclidean distance between the centroids of two countries, s_i and s_j, and W_ij = 0. For W₂, units are considered as ${W}_{{\text{ij}}}=1/{e}^{{d}_{{\text{ij}}}}$, where d_ij is the Euclidean distance between the centroids of two countries, s_i and s_j, and W_ij = 0. Both spatial weight matrices are row-standardized weights matrices; the comparison results are presented in Table 4. As shown, the SAR models with different spatial weight matrices are basically consistent with the findings of the SAR model in Table 3. Consequently, the estimation coefficients of the SAR model are considered stable.

Table 4 Robustness check: SAR with different spatial weight matrices

5.4 Regional heterogeneous analysis

From the perspective of urban location, the global ICIE and CO₂ emissions exhibit substantial spatial differences (Zhao et al. 2024). To distinguish the differential effects of ICIE on CO₂ emissions with various economies levels, this study firstly calculates the mean GDP of each country (denoted as Mi) during the sample period (i.e., 1995–2016), and then further gauges the mean GDP of these countries from 1995 to 2016 (expressed as Mp). When Mi is greater than or equal to Mp, it is considered to be high-GDP economies; otherwise, it is regarded as low-GDP countries (the specific information can see Table A2). The relevant results are illustrated in Table 5.

Table 5 Results of regional heterogeneous analysis

As this table shows, in the high-GDP countries, a growth of ICIE by 1% will stimulate per capita carbon emissions by 0.0534%, while in the low-GDP countries, a 1% growth in ICIE can lead to a 0.0206% decrease in carbon emissions. Specifically, through data analysis, it can be found that countries with huge aggregates, especially developed countries, usually have a low level of income gap and a high level of CO₂ emissions. The continuous narrowing of residents’ income gap will promote the gradual equality of income distribution and increase the income level of the middle group. This can, on the one hand, raise awareness of environmental protection through widespread education, and on the other hand, provide financial support for the purchase of environmentally friendly products and the use of clean fuels.

On the contrary, countries with lower economic aggregates tend to have higher ICIE, and most countries with lower GDP are associated with lower per capita CO₂ emissions. Compared with the low-income group, the increase of ICIE will improve the income level of the high-income group (Chen et al. 2020; Zhao et al. 2020). Although the rich pay more attention to environmental quality, they do not necessarily invest in environmental protection. At the same time, low-income groups, on the one hand, have to rely on traditional biomass fuels due to their purchasing power; but on the other, they will destroy the environment and consume fossil energy to increase their income, which will exacerbate the greenhouse effect.

6 Further discussion on the causal mediating effect

6.1 Causal mediating analysis

Through the spatial econometric model (i.e., SAR model) discussed above, it is clear that an expanding income gap is a valid determinant of a country’s greenhouse effect. To this end, another interesting question is worth considering: What are the specific impact channels from income inequality to CO₂ emissions across the globe? A clear understanding of the influence mechanism in the ICIE-CO₂ nexus will not only help in-depth exploration of ways to significantly mitigate greenhouse effect from the perspective of income distribution, but also provide a reference for policymakers to effectively achieve carbon neutrality goal. Accordingly, following the conceptual framework of Copeland and Taylor (1994), this study assumes that ICIE affects CO₂ emissions by affecting three main mediating effects, that is the economic effect (i.e., economic growth), technical effect (i.e., technological innovation), and composition effect (i.e., industrial structure adjustment). To effectively analyze the spatial mediating effect of the ICIE-CO₂ nexus, the traditional mediating approaches are no longer applicable (Dong et al. 2023; Wang et al. 2022). To this end, we focus on the mediating effect based on the research of Imai et al. (2010). The specific spatial mediating effect models are presented as follows:

$${\text{LnCO}}_{{2{\text{it}}}} = \rho_{1} \mathop \sum \limits_{j = 1}^{N} W_{{{\text{ij}}}} {\text{LnCO}}_{{{\text{2it}}}} + b_{1} {\text{LnGini}}_{{{\text{it}}}} + \gamma_{1}{\prime} {\text{LnX}}_{{{\text{it}}}} + \alpha_{i} + \varepsilon_{{{\text{it}}}}$$

(13)

$${\text{Ln}}M_{{{\text{it}}}} = b_{2} {\text{LnGini}}_{{{\text{it}}}} + \gamma_{2}{\prime} {\text{Ln}}X_{{{\text{it}}}} + \varepsilon_{{{\text{it}}}}$$

(14)

$${\text{LnCO}}_{{2{\text{it}}}} = \rho_{2} \mathop \sum \limits_{j = 2}^{N} W_{{{\text{ij}}}} {\text{LnCO}}_{{2{\text{it}}}} + b_{3} {\text{LnGini}}_{{{\text{it}}}} + b_{4} {\text{Ln}}M_{{{\text{it}}}} + \gamma_{3}{\prime} {\text{Ln}}X_{{{\text{it}}}} + \alpha_{i} + \varepsilon_{{{\text{it}}}}$$

(15)

where M_it indicates the mediators in the ICIE-CO₂ nexus, which consist mainly of economic effect (i.e., economic growth), technical effect (i.e., technological innovation), and composition effect (i.e., industrial structure adjustment). X_it represents a vector of the observed preprocessing confounding factors. b₁ refers to the total treatment effect, and b₃ denotes the average direct effect; thus, ${b}_{1}-{b}_{3}$ describe the average causal mediating effect, that is the indirect impact in the ICIE-CO₂ nexus (Baron and Kenny 1986).

Notably, the process of estimating the causal mediating effect between the two core variables mainly includes the following two steps. First, we employ the bootstrapping method to examine the statistical significance of the average causal mediating effect (i.e.,${b}_{1}-{b}_{3}$), which emphasizes the existence of the indirect effect (Preacher and Hayes 2008). Second, if the average causal mediating effect is statistically significant, it is necessary to identify the directionality of the nexus between ICIE and three mediators (i.e., GDP, Ind, and Tec). This can provide effective reference for the specific impact channels in the ICIE-CO₂ nexus.

6.2 Results of causal mediating analysis

To empirically investigate the average causal mediating effect between the core variables, we first estimate the average causal mediating effects (i.e.,${ b}_{1}-{b}_{3}$) of the three mediators (i.e., GDP, Ind, and Tec) following the estimation procedure proposed in Sect. 6.1, and the empirical results are listed in Table 6. Clearly, the coefficient of the average causal mediating effects of economic growth (i.e., GDP) is negative, while that of technological innovation (i.e., Tec) is substantially positive; this implies that GDP and Tec are effective mediators between ICIE and CO₂ emissions across the globe. Furthermore, the insignificance of the coefficient of Ind denotes that Ind is not a valid mediator. In recent years, to reduce risks and increase returns, high-income groups have tended to invest in tertiary industries with low-pollutant features, which will in turn reduce the share of industrial manufacturing. Furthermore, with the gradual strengthening of environmental regulation of various countries, shutting down polluting enterprises will harm their interests. To achieve green and sustainable advancement, enterprises in secondary industries have focused on attaching importance to the R&D of energy-saving and low-polluting technologies. This can help reduce pollution emissions and facilitate the transformation of enterprises. However, there is no doubt that the continuous evolution of related technologies is a long-term process, and the widening income gap will restrict the process of technological innovation (Vona and Patriarca 2011); consequently, the coefficient of composition effect is negative, but not significant.

Table 6 Results of the average causal mediating effects for the mediators

After conducting the average causal mediating effect, how does ICIE affect the three mediators (i.e., GDP, Ind, and Tec) is further examined by estimating the coefficient of ICIE in Eq. (14) (i.e., b₂) to explore the complete information of the mediating effect between the core variables across the globe; the empirical results are reported in Table 7. It is obvious in this table that the estimated coefficients of ICIE are all negative, which suggests that the expanded income gap has worsening consequences in economic, technical, and structural dimensions. Specifically, as Kennedy et al. (2017) stressed, falling ICIE could substantially boost economic growth. It is clear that unbalanced income distribution concentrates resources mainly to promote the formation of growth poles and cause a mismatch of resources, leading to social instability and unsustainable economic development, which harms normal production activities, thereby hindering economic growth and alleviating CO₂ emissions to some extent. In terms of technical effect, it has been confirmed that excessive ICIE can harm the R&D of environmental technologies (Vona and Patriarca 2011). And the negative relationship between technical effect and CO₂ emissions denotes that reducing GDP per unit energy consumption can aggravate the greenhouse effect. This implies that ICIE can hinder the R&D of low-carbon and energy-saving technology, thereby promoting CO₂ emissions across the globe. Figure 2 shows the specific mediating impact mechanism of ICIE-CO₂ nexus.

Table 7 Results of the effect of inequality on the mediators

Fig. 2

Mediating impact mechanism between income inequality and CO₂ emissions around the globe

7 Conclusions and policy recommendations

To systematically examine the causal ICIE-CO₂ nexus, this study first derives the theoretical framework between ICIE and CO₂ emissions, and then investigates the spatial linkage between these two variables across the globe based on a global sample dataset from 1995 to 2016 by using the SAR model. Furthermore, we conduct regional heterogeneous analysis and discuss the three mediators to explore the transmission pathways. The main findings of this study are as follows:

• (1) The ESDA indicates that CO₂ emissions exhibit a significant spatial correlation. The primary conclusion of this paper indicates that ICIE affects global CO₂ emissions positively; by implication, the increasingly expanded income gap is a crucial obstacle and threshold restricting the realization of greenhouse effect mitigation and carbon neutrality.

• (2) The regression results of the regional heterogeneous analysis indicate that in high-GDP countries, ICIE is positively correlated with country’s carbon emissions, as huge economic aggregate can provide a solid economic basis for residents’ preference for environmentally friendly products and investment in environmental protection; in low-GDP countries, expanded income gap can significantly reduce CO₂ emissions. This may be that rising ICIE leads to inconsistent environmental protection intentions, with the rich unwilling to invest in greenery and the poor consuming large amounts of dirty fuel due to the economic stress.

• (3) The results of the causal mediating effect imply that accelerating the reduction of ICIE and promoting the realization of common prosperity indirectly contribute to the greenhouse effect by influencing economic and technological effects; in other words, economic scale and technological innovation are significant mediators of the ICIE-CO₂ nexus. However, the composition effect (i.e., industrial structure adjustment) imposed by ICIE on global greenhouse effect is not obvious. In other words, rising ICIE cannot significantly affect CO₂ emissions by adjusting industrial structure.

Following the above conclusions, we propose following policy recommendations. First, the positive ICIE-CO₂ nexus in the benchmark regression implies that narrowing the income gap and promoting residents’ income distribution play an important role in reducing CO₂ emissions. Notably, improving ICIE is a huge challenge facing all countries, and governments should accordingly improve the social security system and adjust the income distribution. Specifically, strengthening tax collection and the administration of high-income groups to ensure fairness and justice in social redistribution is necessary. Besides, measures such as compulsory education and training and financial subsidies should be adopted to raise the employment and income levels of low-income groups. This will contribute to the dissolution of the dual urban–rural economic system and removing barriers to the free flow of labor and production factors, and providing more job opportunities and better employment conditions for low-income groups. In addition, a significant spatial correlation of CO₂ emissions indicates that no country is an independent entity, and it is essential for various countries to enhance the cooperation and communication to achieve carbon neutrality. Accordingly, various countries should actively participate in environmental protection conventions, and further deepen and broaden cooperation with international or regional organizations to ensure environmental protection.

Second, the empirical outcomes of regional heterogeneous analysis with different GDP levels indicate that when planning policies to rationalize income distribution and reduce carbon emissions, governments and policymakers of various countries should formulate appropriate and targeted strategies based on their own characteristics and actual conditions. More specifically, in high-GDP countries, especially developed countries with prosperous economy, the income gap is relatively low and welfare benefits are relatively complete; thus, the government should continue to popularize the concept of ecological harmony and intensify the education and publicity of environmental protection, so as to effectively raise residents’ environmental awareness. On the contrary, in economically backward countries with low GDP, the relatively considerable income disparity and severe environmental pollution have become arduous tasks for the government to solve urgently. To this end, the governments of various countries should actively adjust the structure of household income distribution, improve the rationality of the country’s overall income distribution, and increase the income level of low-income groups. In addition, a country’s abundant economic aggregate is a solid foundation for realizing common prosperity; accordingly, the governments should combine multiple policies and measures to consolidate the national foundation as vigorously as possible.

Third, considering the significant causal mediating effect of the economic effect and technical effect, countries’ policymakers should formulate effective and suitable measures to promote carbon reduction by adjusting the scale and technical mediating effects. Rather than significantly reducing CO₂ emissions at the expense of decreasing economic scale, a better distribution of income can effectively boost economic growth and thus provide financial support to promote technological innovation and mitigate the greenhouse effect. Therefore, strengthening the smooth reform of income distribution system and optimizing the income equity of high and low income groups will not only contribute to the achievement of social equity, but will also make all strata stakeholders more capable and willing to protect the ecological environment, thus promoting a win–win situation of income equity and pollutant emission mitigation.

It is worth noting that our paper only provides preliminary study on the ICIE-CO₂ nexus across the globe, certain limitations still exist. First, other spatial econometric models (e.g., the spatial Durbin model (SDM) and the spatial error model (SEM)) are also used to analyze the spatial effect. However, given the research topic of this study, the SAR model is utilized, while the SDM and SEM are not considered. Second, the impact of energy consumption on greenhouse gas emissions is not assessed due to the potential multicollinearity in energy use and technological innovation gauged by energy consumption per unit of output. In future research, we should add the energy consumption structure or natural gas consumption into our econometric model for empirical analysis. Third, considering the underlying dynamic effect of CO₂ emissions, checking the dynamic impact of ICIE on alleviating greenhouse effect by employing the GMM technique become our future research direction.

Appendix A

See Tables 

Table 8 Definition and description of the variables

8,

Table 9 Specific countries included in the sub-samples

9 and

Table 10 List of abbreviations

10.