Towards Consumption-Based Carbon Inequality Metrics: Socioeconomic and Demographic Insights from Chinese Households

Abstract

The choice of carbon inequality metrics can significantly influence demand-side mitigation policies and their equity outcomes. We propose integrated carbon inequality metrics, including juxtaposing carbon inequality with economic inequality, disparity ratios across income and age groups, and structural income–urbanization inequality patterns. We then apply these new metrics and use the household expenditure survey data from China Family Panel Studies as a case study to examine household consumption-based carbon emissions in China. We assess the extent to which household consumption patterns, household expenditure, age, and urbanization contribute to the gap in per-capita household carbon footprints (CF) across income groups. We find that in relative terms, the top 20% income group accounts for 38% of total emissions, whereas the bottom 20% emit about 8% in China. Per-capita CFs vary slightly widely in their inequality than expenditure. The CF disparity ratios of all eight consumption categories across provinces concentrate around 4.5. CF disparity ratios of households with elderly members range from 1 to 3 and decrease with increasing household size. Rural CF-Gini exhibit a slightly wider range (0.15 to 0.52) than urban CF-Gini (0.16 to 0.42). Per capita CF of urban inhabitants was substantially larger than that of the rural ones, with 8.83 tCO₂ per capita in urban regions vs. 2.68 tCO₂ in rural regions. This study provides a nuanced understanding of within-country disparities to inform equitable demand-side mitigation solutions.

1. Introduction

The influence of socioeconomic and sociodemographic inequalities trends on household consumption and, as a result, carbon inequality ensues [1,2,3]. Climate-relevant consumer behavior choices are not consciously made but are highly impacted by the socioeconomic and sociodemographic drivers [4]. Behavior changes and mitigation intervention policies require recomposing consumption, which necessitates an accurate and detailed understanding of current household carbon emission profiles [2,5,6]. Consumption-based carbon inequality research is explicitly addressed in the Sixth Assessment Report (AR6) of the United Nations Intergovernmental Panel on Climate Change (IPCC), Chapter 2.6.2 ‘Factors affecting household consumption patterns and behavioral choices. The chapter provides an in-depth discussion of various socioeconomic and sociodemographic factors affecting household consumption-related carbon emissions. Four factors are particularly emphasized, including income, age, household size, and urbanization rate [7]. The chapter advocates that advancing emission mitigation efforts should not exacerbate existing inequalities. Policies should aim to recompose consumption patterns to enhance the inclusion of disproportionately burdened populations.

The wide uptake of the environmentally extended multi-regional input-output model (EEMRIO) in consumption-based carbon inequality studies reflects its methodological advances in the past decade. The key strengths of EEMRIO are its ability, in a world of trade, to link the carbon emissions with the household final demand of goods and services via global supply chains [8,9]. Household consumption-based responsibility can thus be attributed and intervened (for instance, high-income emitters’ high-carbon lifestyle) [10]. Driven by the need to examine intra-country carbon inequalities, a recent exercise is linking sub-national household expenditure survey (HES) data to the final demand category in sub-national EEMRIO models. For a highly locally relevant carbon mitigation intervention policy, a much greater spatial resolution of local consumption dynamics is required [11]. For instance, Zheng, Long [12] analyzed the impact of aging on carbon inequality by coupling sub-national HES data from 32 developed countries, and found that the US and Australia seniors have the highest per capita footprint, twice the Western average. Feng, Hubacek [13] examined the impact of income on carbon inequality by coupling US HES data from the US Bureau of Labor Statistics, which found that the per-capita footprint of the highest income group is 2.6 times that of the lowest income group. In China, prior HES dataset is sourced from the Chinese provincial statistical yearbooks, which differentiate per capita annual expenditure by eight consumption categories across five rural income groups and eight urban income groups [14,15]. Another HES data from China Family Panel Studies (CFPS) covered 15,000 households and has only recently received research attention [16,17].

The purpose of this paper is to introduce an integrated carbon inequality analytical metric that aligns closely with the most recent environmental inequality research methodologies in the IPCC AR6 Chapter 2.6 [7]. We place emphasis on the insights drawn from these integrated analytical indicators and comparable international studies. An integrated approach is needed because the choice of carbon inequality metrics can significantly influence the conclusions drawn. Most studies that consider carbon footprints and inequality of households focus on absolute carbon footprint values. For instance, Ivanova and Barrett [18] provide a systematic review of the mitigation potential of consumption options for global countries by a meta-review. International and consumption-granular comparisons remain restricted to absolute values instead of relative indicators that benchmark against the most disadvantaged social strata. The merits of using integrated analytical metrics instead of absolute conventional footprints support social equity policies and behavior intervention.

Firstly, conventional indicators such as the environmental Gini coefficient are widely adopted to analyze carbon inequality, yet the Gini index is used to measure overall unequal distribution across a population and is less sensitive at the tails of income distribution. The carbon footprint Gini (CF-Gini) coefficient may be the same for two entities, but differ in the income gap between the richest and the poorest strata. Therefore, superimposing environmental inequality with economic inequality, i.e., CF-Gini on expenditure-Gini, is the next logical and necessary analytical step. This approach is increasingly used to analyze intra-country inequality and thus obtain a higher granularity [6,19]. We assess the relationship between CF-Gini and expenditure-Gini.

Second, disparity ratios evaluate carbon differences between distinct social stratification groups. Social equity policies target the most disadvantaged strata in a society; therefore, a reference social group from which to measure disparity is critical [20]. Actionable disparity ratios complement aggregate metrics by highlighting absolute differences that broader Carbon-Gini indicators may obscure. For example, while a declining Gini coefficient signal reduced overall inequality, a widening disparity ratio between rural and urban households may indicate persistent structural challenges. Disparity ratios by income and race has become research focus in utility level household energy use, which are computed as the ratio of average energy attribute in the lowest 20% income (or the top 20% most racially diverse) group versus the highest 20% (or the 20% least racially diverse) [21,22].

Third, the combined impact of urbanization and income interaction on carbon efficiency in China is worth researching. Herein, we define carbon efficiencies as carbon intensity within a specific income stratum. For the same level of consumption of goods and services, an average urban dweller tends to emit less carbon than their rural counterparts, due to a dense and compact urban form that enables economies of scale, such as sharing appliances and heated or cooled living spaces with family members [23]. However, there is growing evidence that urbanized living with smaller households, high-income-associated consumption will eventually offset the benefits of higher population densities [24,25,26]. We assess whether a higher population density, in reducing per capita urban carbon emissions, is offset by higher consumption levels.

Studies directly comparing within-country household-level carbon inequality are still limited. The impact of income on Chinese household carbon emissions is confirmed. Income-based carbon inequality was extensively analyzed using CF-Gini. Wiedenhofer, Guan [15] and Mi, Zheng [14] found that CF-Gini declined between 2007 (0.44) and 2012 in China (0.37). Recent studies reported an overall decrease in CF-Gini in China [17,27]. Contradictory findings were found in aging societies. The impact of age on emissions is contested in China. The demographic shift to small and aging households will increase [28] or decrease [16] household CF. Globally, the CF of the aged 60+ group increased from 25.2% to 32.7% between 2005 and 2015 [12]. The impact of the urban-rural divide on emissions is also unclear. Urban-rural gaps in household emissions narrow as household size expands [29]. Prior research shows that in China, urbanization and carbon emission efficiency exhibit an inverted U-shaped relationship from 2000 to 2016 [30]. In developing countries, urban areas have a higher CF than rural areas for low and middle-income groups, but the trend reverses for high-income strata [31].

We propose integrated carbon inequality metrics that can advance carbon equity research. The remainder of the paper is structured as follows. In Section 2, we employ the household expenditure survey data from China Family Panel Studies as a case study to examine household consumption-based carbon emissions in China. Section 3 presents the analytical metrics, including juxtaposing carbon inequality with economic inequality, disparity ratios across age groups, and structural income–urbanization inequality patterns. By considering within-country and intra-provincial inequality, along with several related metrics, we assess the extent to which household consumption patterns, household expenditure, age, and urbanization contribute to the gap in per capita household carbon footprints across income groups. In Section 4.1, we contextualize our findings through comparison with other subnational carbon inequalities. In Section 4.2, we propose a research agenda to advance the modeling of carbon inequality in demand-side climate mitigation frameworks.

2. Materials and Methods

2.1. Environmentally Extended Input-Output Model

We employ the most recent available 2017 provincial-level EEMRIO model and its emission inventories from the Carbon Emission Accounts and Datasets (CEADs) database [32]. The calculations can be denoted as in Equation (1)

$${\mathit{F}}_n=\widehat{m_n^{DCI}{\mathit{x}}^{-1}}{(\mathit{I}-\mathit{A})}^{-1}{\mathit{y}}_n$$

(1)

${\mathit{F}}_n$ is a carbon map specific to one province (n). $\widehat{m_n^{DCI}{\mathit{x}}^{-1}}$ is the diagonalized vector of direct carbon intensities, calculated as the product of a vector of carbon and the inverse matrix ${\mathit{x}}^{-1}$ of diagonalized total industry output. I is an identity matrix with ones on the diagonal and zeroes elsewhere. A is the technology coefficient matrix, calculated as the product of the input-output transaction matrix T and the inverse matrix ${\mathit{x}}^{-1}$ of the diagonalized total industry output. ${\mathit{y}}_n$ is a vector of the total final demand for province n. Note that the ${\mathit{F}}_n$ for each province (n) can only be calculated once at a time. The ${\mathit{y}}_n$ final demand matrix featured five categories: rural household consumption, urban household consumption, government consumption, fixed capital formation, and inventory changes.

2.2. Linking Household Expenditure Survey Data

We linked the household expenditure survey data with the CEADs EEMRIO. The 2018 China Family Panel Studies (CFPS) dataset describes micro-level household consumption in 31 provinces using approximately 15,000 household interviews and approximately 44,000 individual questionnaires. Since no CFPS survey was conducted for 2017, we assume that the 2018 data serve as a representative proxy for 2017 expenditures to align with the MRIO model. To avoid a small sample size and lack of representativeness in Tibet, Hainan, Qinghai, Xinjiang, Ningxia, and Inner Mongolia, we excluded them. Eight household consumption categories are surveyed, including food, clothing, residence, household facilities, transport, education, health care, and others. The CFPS consumption expenditure category is different from CEADs. We construct a concordance matrix to translate 42 sectors from CEADs into eight sectors from CFPS. For example, a mapping exercise of sectors as ‘Agriculture, Forestry, Animal Husbandry, and Fishery’, ‘Food and tobacco processing’ is all determined as ‘Food’ in the CFPS consumption database.

We decompose the original urban and rural household consumption into three socioeconomic and sociodemographic dimensions: income, age, and household size. These two final demands were further disaggregated into five quintile income groups, three household sizes, and two age groups, respectively. The quintile income groups are 0–20%, 20–40%, 40–60%, 60–80%, and 80–100%. Based on household size and structure [28], we classified all households into six groups according to household size (A for single family, AA for couple family, AAA for all the remaining family types) and with or without a family member who is 65 or older. Direct household carbon emission is not included in this study as we focus on supply chain emissions and relative disparity ratios. Household demographic attributes such as age, household size, and urbanization rate were found in the 2018 China Statistical Yearbook.

2.3. Carbon Inequality Metrics

We adopt two carbon inequality indicators. First, the Gini coefficient, ranging from 0 (perfect equality) to 1 (perfect inequality), was employed to assess expenditure and footprint disparities. The calculations can be denoted as in Equation (2)

$$G=\sum _{i=1}^n{P}_i{Y}_i+2\sum _{i=1}^n{P}_i{(1-C}_i)-1$$

(2)

where $G$ refers to the carbon Gini coefficient, $P_i$, $Y_i$ and $C_i$ are the population share, carbon footprint share, and cumulative carbon footprint share of income group i, respectively, and n is the number of groups. We also used the Lorenz curve to show the expenditure and carbon footprint inequality across consumption categories, which is the ordered distribution of the cumulative share of population against the cumulative share of expenditure and carbon footprints. The Gini coefficients and the distributional Lorenz curves quantify sample dispersion without accounting for social stratification.

Second, we assess the extent to which the most advantaged social strata are compared to the least disadvantaged social strata across income group consumption categories and age. The disparity ratio calculates the ratio of the difference in each household’s average annual CF by various socioeconomic and sociodemographic groups. Take income as an example, the form of expression is to take the statistics of household carbon footprints in different regions from the quintile ratio, by dividing the average household CF reported in income groups with average household income higher than the 80th percentile (P80) with the average household CF in income groups with average household income lower than the 20th percentile (P20). The disparity ratio across the two income quintiles is shown below.

$$D{R}_{CF}={\displaystyle \frac{Avg.C{F}_{inc>P80}}{Avg.C{F}_{inc<P20}}}$$

(3)

The larger the absolute value of the ratio, the more significant the disparity between the two groups. For example, if it is higher than 1.5, it can be called significant. Disparity ratios are closely linked to differences across income groups; for instance, a disparity ratio of 2 between the bottom and top income groups indicates a 100% difference relative to the top income group.

2.4. Index Decomposition Index

This study identified three main driving factors that can affect the carbon footprint. The driving factors included the affluence effect, household structure effect, and GHG intensity effect as follows:

$$C={\sum }_i^n({EXP}_i/{POP}_i)\times ({POP}_i/POP)\times ({\mathrm{C}\mathrm{O}}_{2,i}/{EXP}_i)$$

(4)

where the subject i represented specific expenditure class sectors (i.e., <4 K, 4 K–7 K, 7 K–11 K, 11 K–17 K > 17 K). C refers to the total carbon footprint per capita. POP is population; EXP is average household expenditure; ${\mathrm{C}\mathrm{O}}_2$ is carbon footprint. The equation can be shortened to

$$C=A\times HS\times GHG$$

(5)

$A$ referred to household expenditure (the affluence effect), $HS$ referred to household size (household structure effect), and $GHG$ referred to household consumption patterns (GHG intensity effect).

3. Results

3.1. Overview

Juxtaposing expenditure inequality with carbon inequality reveals a contrast in the degree of disparity (Figure 1). We found that while economic inequalities are sizable, the gaps in carbon emissions between different income-differentiated segments of the population are even more pronounced. Urban CF-Gini exhibit a slightly wider range (0.16 to 0.42) than Gini coefficients of expenditure (0.15 to 0.41). Rural CF-Gini ranges from 0.15 to 0.52, while Gini-expenditure ranges from 0.14 to 0.53. These indicate a moderate level of disparity in household spending and consumption patterns, and a more pronounced level of carbon inequality. Shanghai, one of the wealthiest provinces in China, has a relatively high urban CF-Gini of around 0.37, while some less developed provinces, such as Guangxi, have a lower CF-Gini of around 0.16.

When the expenditure-Gini of urban residents is very unequal (i.e., with a high Gini coefficient, in the higher range, above 0.36), the associated inequality in carbon footprints for its rural residents tends to be even greater. For example, urban expenditure-Gini in Jiangsu is 0.41, and rural expenditure-Gini is 0.53. Its urban CF-Gini is 0.42, and rural CF-Gini is 0.53. This clear urban-rural inequality is reflected in regions with high expenditure-Gini and CF-Gini. High-income provinces such as Jiangsu, Shanghai, and Zhejiang fall into this category. Their CF-Gini are 0.36–0.42 for urban and 0.44–0.53 for rural areas.

For regions where the Gini expenditure is distributed in the middle range (0.21–0.31), a more mixed picture can be observed. Generally, metrics are more inclined to be alike and the same at the middle expenditure and CF disparity (for instance, the Gini coefficient in Guangdong is 0.28 for both expenditure and CF). A slightly more pronounced carbon inequality is found in urban residents than rural (for example, urban CF-Gini in Henan is 0.27 versus rural CF-Gini 0.24). There are also exceptions. Urban CF-Gini is found around 1.7 times higher than rural in exceptionally high-income provinces (for instance, Beijing, 0.28 urban versus 0.48 rural, Fujian, 0.26 urban versus 0.38 rural). On the contrary, rural CF-Gini can be 1.7 times higher than urban in both high and low-income provinces (for instance, Sichuan, 0.31 urban versus 0.19 rural, Gansu, 0.32 urban versus 0.14 rural).

For regions where the Gini expenditure is distributed in the lower range (below 0.2), the same urban-rural inequality can be reached for low-income regions (for example, CF-Gini in Yunnan is 0.18 for urban residents versus 0.34 for rural, and Guangxi is 0.16 for urban residents versus 0.34 for rural).

3.2. Carbon Footprints Inequality Across Income Groups

For most provinces, the Gini coefficients of the consumption categories range from 0.2 to 0.4 (Figure 2). The sharpest incline is observed in transport-related consumption in Tianjin, Heilongjiang, Zhejiang, and Henan, signifying the most unequal distribution, with the poorest 50% contributing less than 10% of the total, and the top 10% accounting for 45% of the carbon footprint. While Food, Household Facilities, and Education (0.4) are more equally distributed among the Chinese population. This may be attributed to a high similarity in daily dietary structure across all income levels. We find higher inequality for carbon footprints of Residence (0.45) and Transport (0.48) in Shanghai. Although these consumption categories exhibit the least inequality, with Gini coefficients of 0.5, they can still be described as moderately unequal.

Disparity ratio is frequently used in conjunction with the Gini coefficient to offer a complete picture of income distribution within a social stratification. Carbon footprint disparity ratios by income are calculated as the ratio of average carbon footprint reported in cohorts within the highest income quintile block groups (20% highest) compared to those in the lowest income quintile (20% lowest). We found that in most provinces, the CF disparity ratios of all eight consumption categories concentrate around 4.5, indicating that the highest quintile is approximately 350% higher than the lowest quintile (Figure 3). Jiangsu, Chongqing, and Zhejiang are at the higher end, with an average disparity ratio of 16.5, 11.6, and 10.7.

However, the explicit degree of the carbon inequality measured by carbon disparity does not vary substantially across consumption categories, nor across provinces. Within-country differences are not overlooked. Compared with the huge variability measured by CF-Gini, a consistent disparity ratio around four means that a universal nationwide policy (with few exceptions in Jiangsu, Zhejiang, Hubei, and Chongqing) can be implemented without disproportionately burdening the lowest quintile income population. Each province has its own economic development stage and income distributions, and across-province disparities will be persistent for a relatively long period of time. Interventions can only be effective when emission gaps across social stratification are visualized and internal distributional equity is guaranteed. Such insights are vital for designing tailored interventions.

We quantified the contributions of household expenditure, household size and household consumption patterns (GHG intensity) across different income groups: lowest (4 K), lower-middle (4–7 K), middle (7–11 K), higher-middle (11–17 K), and highest (>17 K) (Figure 4). Income increases more than fourfold from the lowest to the highest income category, and per capita carbon footprints follow a similar but less pronounced pattern (on average, 313%). The per capita carbon footprint of the highest income group (on average 6.4 tons/capita) is 4.6 times that of the lowest income group (1.4 tons per capita). For instance, in Beijing, the carbon footprint of the lowest income group is approximately 5.2 tCO₂ per capita, while for the highest income group, it reaches around 18.2 tCO₂ per capita—a more than 350% increase.

This growth is primarily driven by household expenditure, which accounts for the largest share of the gap. For example, if the two groups had the same household size and household consumption patterns (GHG intensity), consumption per household alone would result in a 60% average increase in per capita CF between the lowest and lower-middle income groups. This would be up to 81% and 86% between the lower-middle and middle, middle and higher-middle income cohorts, respectively. The highest increase, 95%, occurs between the higher-middle and the high-income groups.

Compared to household expenditure, the household size only would result in a marginal increase (on average 39%, 17%, 22%, and 10%, respectively) per capita CF across all income groups. In Fujian, where carbon footprints are highest, household size contributes 48% for the lowest income group (<4 k) but steadily decreases to 8% for the highest income group (>17 k). This trend suggests that as income increases, the mitigating effect of household size diminishes, likely due to smaller household sizes in higher-income groups, which reduce shared resource use.

The impact of household consumption patterns (GHG intensity) on the variation in per capita carbon footprints across all income groups is relatively negligible compared to the previous two factors. It would only result in an incremental decrease (−8%, −11%, −11%, and 5%, respectively) per capita CF across all income groups. This suggests that household consumption patterns have a smaller mitigating effect in high-income groups. Household consumption patterns play a variable role across provinces and income groups. Shanghai and Zhejiang are exceptions; GHG intensity contributes around 74% and 79% of the gap between the lower-middle and middle income groups, highlighting the carbon-intensive nature of consumption in wealthier households. This suggests that consumption patterns in these groups are associated with goods and services that are more carbon-intensive. The contribution of consumption patterns (GHG intensity) in driving the carbon footprint gap is one-eighth smaller than household expenditure.

3.3. Carbon Footprints Inequality Across Age Groups

Per capita household CF with larger household size is significantly higher than households with smaller sizes (

Table 1. Disparity ratios of carbon footprints by age category in 2017.

Consumption Category	Carbon-Gini	AAA65 Share	AAA Share	Disparity Ratio	AA65 Share	AA Share	Disparity Ratio	A65 Share	A Share	Disparity Ratio
Food	0.54	8.1%	71.0%	1.06	1.3%	15.0%	1.46	0.2%	4.4%	3.05
Clothing	0.71	8.1%	70.0%	1.04	1.4%	17.0%	1.53	0.2%	3.3%	2.45
Residence	0.50	8.4%	70.3%	1.01	1.5%	15.8%	1.38	0.2%	3.9%	2.54
Household facilities	0.47	8.2%	70.2%	1.03	1.4%	16.1%	1.49	0.2%	4.0%	2.59
Transport	0.59	8.7%	70.4%	0.97	1.3%	15.6%	1.47	0.2%	3.7%	2.60
Education	0.50	8.6%	71.0%	1.00	1.4%	15.2%	1.42	0.2%	3.7%	2.62
Health Care	0.69	7.8%	71.4%	1.10	1.6%	15.5%	1.26	0.2%	3.6%	2.57
Others	0.63	7.6%	70.7%	1.11	1.3%	15.2%	1.48	0.2%	5.0%	3.09

). Specifically, among households with elderly members, families consisting of more than three individuals exhibit the largest CF of 15.7 tCO₂, which is approximately 3 times greater than that of a two-person household and 12 times greater than that of a single-person household.

In general, the per capita CF of households with elderly members is lower than that for the same household size without elderly members. In households with one or two members, those with elderly members have a much smaller per capita CF compared to those without elders (5.5 tCO₂ of AA65 vs. 8.2 tCO₂ of AA, 1.3 tCO₂ of A65 vs. 3.3 tCO₂ of A). This could potentially be explained by the distinct lifestyles and consumption patterns between elderly and younger individuals.

Carbon footprint disparity ratios by age are calculated as the ratio of the average carbon footprint reported in strata of the same household size compared to that of a household with an elder member over 65 years. Our CF disparity ratio by age range is 1 to 3, which is different in magnitude compared to the disparities by income (4.3 to 5.2) (Table 1). Disparity ratios with elderly members decrease with increasing household size. Households with two or more family members are characterized by comparable consumption habits, and no apparent differences are observed. The disparity ratio was significant only for single-person households, at around 2.6 among eight consumption categories. For the single household stratum, the carbon footprint of food and other goods and services is 3.05 times larger than that of an age 65+ household. Differences in lifestyle and household income might explain the discrepancy. For example, in Shanghai, a single-person household spends $1884 on average, whereas the age 65+ households spend $1170 per year on average. When considering a single individual, a limited income restricts more consumption possibilities.

3.4. Carbon Footprints Inequality Across Urban-Rural Divide

A clear urban–rural divide in carbon footprint is observed. The per capita CF of urban inhabitants was substantially larger than that of the rural ones, with 8.83 tCO₂ per capita in urban regions vs. 2.68 tCO₂ in rural regions. This may be attributed to differences in part because of the stronger spending power of the residents and the choice of lifestyle. When comparing the disparity in carbon emissions between two cities, typically, if one city’s population is twice that of the other, the difference in their carbon emissions is more than twice. For example, urban areas in Guangdong have a per capita CF of 2.7 tCO₂, which is 2.64 times higher than that in rural areas, while their expenditure is 2.23 times greater.

We found that household expenditure emerged as the primary driver of elevating per capita CF between rural and urban groups, experiencing a surge of 157.1%. In contrast, household size, accounting for a reduction of 50.2%, is largely attributable to the decrease in urban family members. In 2017, the average number of family members in a Chinese urban household was 2.87, compared to a rural household of 4.28. This contradicts the findings of research on US families, where larger household size often contributes to lower per capita CF due to the sharing of household facilities [13]. Furthermore, the impact of GHG intensity on the disparity in per capita CF between rural and urban areas can be considered negligible. It leads to a reduction of 6.9% in per capita CF.

Few prior studies investigate urbanization disparities in carbon footprint, separated from expenditure. Structural income–urbanization inequality pattern is observed, where all five income quintiles have similar expenditure regardless of the urbanization rate (ranging from 40% to 90%). This indicates that the urbanization effect was not observed for carbon footprint within any income strata. We then shifted our focus to carbon intensity as a measure of carbon efficiency (Figure 5). We found that carbon efficiency increased across income strata and exhibits a negative slope in relation to increased urbanization level. This implies that higher levels of urbanization lead to higher carbon efficiency compared to areas with lower levels of urbanization. The degree of urbanization has a significant impact on the CF intensity across all income quintiles.

4. Discussion

This paper examines disparities in household carbon footprint, which is typically affected by four socioeconomic and sociodemographic factors, including income, age, household size, and urbanization rate. First, we present the findings on the inequalities in consumption-based household carbon footprints within China. A comparison with the latest international research is provided, where we align with the research findings from the Sixth Assessment Report of the IPCC Chapter 2.6 [7]. Second, we proposed recommendations and future outlooks on how carbon inequality can be better modelled in demand-side mitigation policy. Then we discuss the integration of social equity into household-level mitigation policies in China.

4.1. Implications of Within-Country Carbon Inequality Research

Prior research indicates that income level is generally positively correlated with household carbon emissions [33,34]. In absolute terms, we found that in China, the CF of the top 20% income strata (6.9 tons) is 4.6 times that of the bottom 20% income group. In the U.S., the carbon emissions of the top 6% households (32.2 tons per year) are 2.6 times those of the lowest 9% households (12.3 tons per year) [13]. In high-income countries, the average per capita resource consumption is 60% higher than that in middle-income and low-income countries, and at least 10 times higher than that in low-income countries [8].

In relative terms, we found the top 20% income strata account for 38% of total emissions, whereas the bottom 20% emit about 8% in China. Chancel and Piketty [35] found globally that the top 10% income level accounts for 45% of emissions, while the bottom 50% emit only about 13%. Hubacek, Baiocchi [36] found that in the EU, Japan, Australia, and Canada, and the elite in developing countries and transition economies, the top 10% account for 26% of global carbon emissions while the bottom 50% account for only 13% of emissions. In the U.S., the top 6% (earning over $200,000 per year) account for 11% of the total carbon emissions, while the bottom 9% (earning less than $15,000 per year) account for only 6% of carbon emissions [35]. Mi, Zheng [14] found in China, the top 5% of income earners induced 17–19% of the national total, whereas the bottom 50% caused only 25%. This suggests that carbon emissions across income portfolios in China are less unequal compared to the global average.

By superimposing environmental inequality with economic inequality, we found that China’s CF-Gini exhibits a slightly wider range than that of expenditure-Gini, implying that carbon footprints vary slightly widely in their inequality than expenditure. This trend is consistent with other resource footprint studies, such as energy footprint [6]. The impact of wealth inequality on carbon disparity is often overlooked, despite resource and environmental footprints diverging more significantly than income or expenditure gaps [37]. In absolute terms, within-country China CF-Gini ranges from 0.17 to 0.51, with an average of 0.34. This implies a moderate level of disparity compared to the global level. Semieniuk and Yakovenko [38] find that global CF-Gini consistently remained above 0.6, from 0.73 in 1970 to 0.62 in 2013. The carbon inequality metric, measured by CF-Gini, suggests China is less unequal than the global average despite greater consumption. Nevertheless, statistics only reflect the macro-level tendency.

The within-country differentiated CF-Gini could inform sustainable consumption policy design, in particular for policies focusing on high-income super emitters. Excessive consumption (beyond decent living) can be redistributed and effectively intervened by, for instance, carbon taxing carbon-intensive lifestyles. In contrast to the much larger variability observed in CF-Gini coefficients, our carbon disparity ratios remain consistently around four across provinces and consumption categories. For interventions to be effective, emission disparities across social strata must first be made visible, while ensuring equitable internal distribution. The carbon disparity ratios can serve as complementary metrics that shift policy focus toward addressing social equity, particularly for low-income households, rather than high-income emitters.

The most unequal consumption category is transportation, in which the top 10% of households account for 45% of the total, which is 4.5 times that of the bottom 50%. Transportation often emerges as a highly unequal domain in emerging economies [39], and this is also applicable to China. High-income households are more inclined to use private vehicles and air transport, whereas low-income households predominantly rely on more affordable public transportation due to financial constraints. The second and third unequal categories are clothing and residence, in which the top 20% households are 5.2 times and 4.7 times that of the bottom 20%. This is partially similar to other countries, where higher-income households tend to consume emission-intensive products. Kerkhof and Benders [40] found that in Sweden and Norway, with increasing household income, carbon emissions intensity increased by 22% and 24% from 1997 to 2005, respectively, whereas in the Netherlands and the UK, it decreased by 12% and 22%. Multiple studies predict that the middle class in emerging economies tends to emulate the high-carbon-intensive lifestyle of the wealthy, and that advances in low-carbon technologies cannot offset the rapidly growing consumption demands [41,42]. Oswald, Steinberger [19] found that luxuries (transportation, education, and medical care) are more likely to exhibit higher energy, thus carbon footprint inequality than income inequality. Subsistence, such as food, clothing, and housing, is systematically more equal than income. As income inequality increases, the inequality in luxury categories grows more rapidly than the inequality in subsistence basics. Mitigation intervention policies may target high-carbon lifestyles of high-income emitters, informed by our consumption category findings. High-income emitters already consume far above the decent living thresholds, whereas low-income groups still face severe energy deprivation [6,19,33]. Future research should explore how an emerging economy can avoid carbon- and resource-intensive consumer lifestyles and track the way of high-income countries. Regulatory or common and control measures, such as fuel carbon taxes [43], could restrict access to carbon-intensive consumption. This should be balanced by making the low-carbon alternatives more affordable and accessible [44].

Prior research indicates that the impact of population aging on household carbon emissions remains unclear [45]. We found that the per capita CF of households with elderly members is lower than or the same as for the same household size without elderly members (1.3 tCO₂ of A65 vs. 3.3 tCO₂ of A, 5.5 tCO₂ of AA65 vs. 8.2 tCO₂ of AA, 15.7 tCO₂ of AAA65 vs. 15.7 tCO₂ of AAA). Zheng, Long [12] found that in 32 developed countries, the emissions of elderly individuals at the global level increased by 25% from 2005 to 32% in 2015. The per capita footprint of older adults in the United States and Australia is the highest, twice the average level of Western countries. Dalton, O’Neill [46] also found that population aging can reduce long-term emissions by up to 40% in low population scenarios, and in some cases, the impact of aging on emissions can be as significant as or greater than the effects of technological change. For example, in Japan, a country with a similar population structure to China, changes in Japan’s population structure, such as aging, declining birth rates, and reductions in the working-age population, are not significantly correlated with carbon footprints Shigetomi, Nansai [47]. In Japan, elder households generally produce higher per capita emissions due to a decrease in temperature, indicating inefficient energy usage among elderly citizens [48]. In China, the opposite trend has been observed. Liu and Zhang [16] found that households with elderly members tend to emit less. Our disparity ratio confirmed this trend, as the size of the household increases, households with elderly members lower their CF versus those with younger households. A significant disparity ratio (2.45–3.09) was observed exclusively among single-person households, but no clear trend was seen in larger household sizes.

We found that the urban-rural divide in China exacerbates household CF disparity. The per capita carbon footprint of Chinese urban households is substantially larger (8.83 tCO₂) than that of rural ones (2.68 tCO₂), which is 3.29 times higher. Rural CF-Gini exhibit a slightly wider range (0.15 to 0.52) than urban CF-Gini (0.16 to 0.42). Jones and Kammen [23] found that in the U.S., household carbon emissions in the largest 50 cities ranged from 25 to 80 tons per year, with urban center households (40 tons/yr) emitting 20% less carbon than suburban (50 tons/yr). Feng, Davis [49] suggested it was the dispersed settlement structure, auto-centric infrastructure, and less social and cultural preferences for mass transit systems that led to high heating and cooling, and private transport demand. Connolly, Shan [31] found that in developing countries, the average per capita carbon footprint of urban residents is 1.49 t CO₂, which is 45% higher than the rural average of 0.82 t CO₂. This suggests that per capita CF in rural regions in China is different from that of Western countries but shares some similarity with developing countries.

It is confirmed that at earlier phases, urbanization can drive emissions because income-induced consumption increases. However, at mature phases with increasing urbanization rate, the impact of urbanization on household carbon emissions is contested. Whether and to what extent emission reduction potentials can be realized depends on urban form and urban infrastructure. For the same level of consumption, an average urban dweller often emits less carbon than their rural counterparts, due to higher population densities that enable sharing of infrastructure and services [23]. We have investigated the combined impact of urbanization and income interaction on carbon efficiency in China. We found that households in urbanized regions are more carbon efficient than those in less urbanized regions. Albeit limited, there is also research that suggests larger regions are less efficient than smaller cities. Fragkias, Lobo [50] investigated the effect of population on carbon efficiency and found that in the US, larger populated cities are less carbon efficient than smaller population cities from 1998 to 2008, as CO₂ emissions scaled proportionally with urban population size. Ribeiro, Rybski [51] found in the US that increasing a city’s population has a greater impact on emissions than changes in population density. Overall, Chinese regions show economies of scale in CO₂ emissions savings. Within the same income strata in China, higher levels of urbanization result in higher carbon efficiency compared to areas with lower levels of urbanization. This implies that the benefits of higher population densities in reducing per capita urban carbon emissions are not offset by higher consumption levels. A counter-effect was not created to increase carbon emissions with urbanization in China [18,52].

4.2. Policy Implications of Carbon Equity Policies and Future Outlooks

The goal of carbon inequality research is to reduce disparities, particularly for the most disadvantaged groups. The reference social group from which to measure disparity is critical [18]. To this end, normative criteria are involved in defining a convergence goal that ensures distributive fairness. The new global demand-side mitigation agenda calls for a new referencing and converging group. We propose that future household-level carbon inequality and equity research should follow three steps: integrating decent living standards (DLS) as a reference group or sufficiency floor, modelling ‘fair’ carbon inequality within each country, and converging towards a global just sufficiency ceiling [53].

First, the growing attention to the DLS framework reflects its conceptual advances in addressing carbon inequality and promoting social justice [54]. Rooted in the principle of egalitarian justice, DLS ensures each individual has an equal entitlement to the basic material conditions necessary for a decent life [55]. DLS has evolved into a normative social benchmark that could serve as a common convergent goal for demand-side per capita resource use [56]. It can be interpreted as a threshold of social foundation [57,58]. This framework offers a fair and transparent set of reference groups to quantify needs-based resource requirements and is emerging as a globally applicable standard beyond context-specific constraints. Prior research suggests that no country has yet succeeded in meeting the DLS for its citizens within its fair share of planetary boundaries, with cross-national disparities ranging from two to six times [59]. Its policy relevance was examined in depth in the Sixth Assessment Report of the IPCC, featured in Chapter 5 on ‘Demand, Services, and Social Aspects of Mitigation’ [7]. To this end, to reduce carbon inequality, future disparity ratios could use DLS as a minimum reference and benchmark group whilst accounting for local heterogeneities.

The next logical step is to estimate current within-country carbon footprint inequalities and how ‘fair’ levels may vary across scenarios. The key questions to be addressed are how big carbon inequality can be. How would income inequality affect carbon inequality? Carbon equity research is criticized for unrealistically high distributive equality; therefore, a fair inequality scenario is proposed to allow for some degrees of space available for carbon inequality [53]. A ‘fair’ inequality scenario was recently introduced where within-country inequalities are reduced to ‘fair’ levels while also ensuring all are above a minimum threshold [60]. Converging an average consumption to the DLS reference group floor without addressing carbon inequality would lead to the least disadvantaged social strata falling below DLS [61]. The question of how to construct fair, redistributed emission scenarios across income groups follows. Quantifying ‘fairness’ is difficult because the levels of inequality people consider or perceive as fair vary considerably across countries [62]. The underestimation of household carbon footprint inequality is repeatedly reported [63]. Wiedenhofer, Guan [15] proposed that a fair income redistribution among the rich and poor households could reduce aggregate carbon footprints while improving DLS. Oswald, Steinberger [19] simulated ‘varying the global income distribution’ and the ‘paying for the poor’ scenarios, which focus on narrowing the income inequality gaps and the associated consequences on carbon footprint. Millward-Hopkins [64] model fair inequality and fairly large inequality scenarios and conclude that this requires an additional 40% and 115% final energy use on top of the DLS baseline. Kikstra, Mastrucci [65] estimate that by 2040, providing decent living energy to all in developing countries would account for only one-quarter of the additional global consumption.

Finally, we call for the implementation of a fully distributionally just global scenario. The key questions are how carbon disparities can be closed by redistribution across countries from an equity perspective. Decreasing carbon inequality while elevating the least disadvantaged social strata will not necessarily lead to an overall increase in emissions. The global carbon cost of equity is small. On a global scale, achieving the DLS sufficiency floor for all countries is fully compatible with a 1.5° world [66]. Bruckner et al. [1] estimate that lifting more than one billion people out of poverty leads to only small relative increases in global carbon emissions of 1.6–2.1% or less. Consensus on closing the inequity gap between the Global North and Global South has been reached [67]. In specific, the affluent Global North could adopt degrowth pathways while the Global South continues to pursue green growth [68]. This strategy would enable the absolute reductions in resource use in the Global North to offset the necessary increases in the Global South [69]. High-income countries can achieve prosperity with reduced material and energy use, and less-necessary production should be scaled down [70]. This approach frees up resources for low- and middle-income countries, where growth may still be necessary for development.

Social equity is rarely featured in household-level mitigation policies in China. At the household level, among a wide array of types of policy instruments available for governments, persuasion policy tools are prevalent, for instance, nudging green consumerism and a low-carbon lifestyle. Climate-friendly choice architecture includes green defaults, the salient positioning of green options, forms of framing, and communication of social norms [71,72]. In China, examples like carbon offsetting flights and forgoing cutlery are commonly practiced [73]. Yet, prior research found the effect diminishes when decision-makers face high financial costs [74]. Exposure to simple nudges can reduce support for more aggressive climate policies (e.g., carbon tax) by creating the false impression that problems can be tackled without considerable costs [75,76]. Market-based mechanisms in China, such as the personal carbon permit trading scheme, are not in policy discussion. Leveraging more effective carbon pricing tools, such as a mandatory carbon tax, is completely absent in current policy discourse. Command and control or regulatory policies that support behavior change are more favored by academics. This is largely due to the consensus reached that short-term voluntary efforts will not be sufficient by themselves to reach the drastic reductions needed to achieve the 1.5 °C goal changes [77]. Moreover, there are concerns about whether short-term, incremental voluntary climate mitigation efforts are sufficient to address large-scale climate change [78,79]. At the national level, equity is only marginally reflected in provincial-level carbon intensity reduction targets. Reducing urban-rural and regional inequality is the primary policy objective in the ‘Common Prosperity’, ‘Rural Revitalization’, and ‘Poverty Eradication’ initiatives. Yet, tackling carbon inequality was not among the policy goals. We suggest that future social and redistributive policies in China need to be analyzed in terms of their interplay with climate and energy policy.

The key strengths of EEMRIO are its ability to map carbon inequality by disaggregating final demand and coupling with HES data, thus providing a consumption-based accounting perspective. Yet, the model itself inherits issues. The provincial-level MRIO model is static, and the most up-to-date version is from 2017. The lack of genuine granular consumption expenditure data poses problems. Publicly available Chinese household survey datasets are limited. Other datasets include the National Residential Consumption Expenditure Database (NRCED, compiled by the National Bureau of Statistics), the China Household Finance Survey (CHFS), and the Chinese General Social Survey (CGSS) [80]. The NRCED data samples around 800,000 households; however, only province-aggregated statistics are released. Other household expenditure survey data samples are limited to approximately 15,000 households, which might not be representative of the whole Chinese population. Capital formation-related emissions, rather than household consumption, are increasingly recognized as the predominant contributors to China’s current carbon inequality among income groups [81,82]. Besides, other important socioeconomic and sociodemographic factors, such as mitigation capabilities of residents and time use, are not modelled in this study due to data unavailability.

5. Conclusions

Demand-side mitigation requires a nuanced understanding of how reducing inequality can improve climate outcomes. The disproportionate carbon footprints across households demand policy frameworks that simultaneously address within-country carbon and income inequalities. Our new integrated metrics become much more meaningful and relevant when devising context-specific climate equity policies. We answered questions like How can carbon disparities of Chinese households be better measured? How does the choice of carbon inequality metrics influence the conclusions drawn? Which consumption items contribute the most to this variation? How can these metrics help to inform policies aimed at lessening carbon inequality? Our study shows that carbon inequality research needs to move beyond simple absolute footprint value and instead recognize the importance of relative carbon inequality metrics that benchmark against the most disadvantaged social strata.

The future of carbon inequality research is to reduce disparities, particularly for the most disadvantaged groups. DLS emerges as a new normative criterion to guarantee distributive fairness. We envisage that future household-level carbon inequality and equity research will follow three steps: integrating DLS as a sufficiency floor, modeling ‘fair’ within-country carbon inequality, and converging toward a global just sufficiency ceiling.

In addition to the conceptual contribution of introducing integrated carbon inequality metrics, this study unravels socioeconomic and demographic inequality trends in household consumption. By coupling the most recent HES dataset with a multi-regional input-output model, we found that the top 20% income strata account for 38% of total emissions, whereas the bottom 20% emit about 8% in China. Per capita CF varies slightly widely in its inequality than expenditure. Inequality is reinforced by superimposing environmental inequality with economic inequality. CF disparity ratios of households with elderly members range from 1 to 3 and decrease with increasing household size. Rural CF-Gini exhibit a slightly wider range (0.15 to 0.52) than urban CF-Gini (0.16 to 0.42). Our study provides valuable insights into equitable demand-side mitigation scenarios and underscores the imperative for a multifaceted approach to carbon inequality research.