To analyse the inequality of global environmental footprints, we calculated the expenditure-specific footprints of 6 environmental indicatorsâCO2 emissions, HANPP, intentional N fixation, P fertilizer use, blue-water consumption and MSA lossâfor 168 countries. This was achieved by linking detailed expenditure data with an EE-MRIO model. Following the principle that the resources and ecosystems of our planet are the global commons and all people on the planet are entitled to an equal sustainable share of them3,13, we downscaled the boundary allowances of these six proxy indicators to per capita equivalents. We compared these with footprint indicators and determined the contributions of different countries and consumption segments to the transgression of PBs.
EE-MRIO model
The EE-MRIO model was used to estimate the environmental footprints of different consumption segments in this study. The EE-MRIO model has been widely used to estimate the global environmental impact of consumption and trade31,32. A particularly frequent application is the analysis of environmental footprints of consumption, such as water, carbon and biodiversity footprints32,33,34,35. One great virtue of this method is that it can model both direct and indirect footprints of consumption, including direct environmental impacts stemming from consumption and indirect environmental impact across the supply chains36,37,38. The EE-MRIO model uses the economic multi-region inputâoutput (MRIO) table to describe the monetary flows and their correlation among economic sectors in international and regional economies39,40,41,42. The environmental satellite accounts add an environmental dimension, making it possible to quantify the environmental impacts of consumption. Combined with the MRIO table, the calculation equation can be expressed as:
$${{\rm{EF}}}_{i,q,c}={E}_{c}{(I-A)}^{-1}{y}_{i,q}$$
(1)
where (IâââA)â1 is the Leontief inverse matrix and yi,q is the final demand vector of consumption segment (q) in country (i). The environmental emissions and resource usage intensity (Ec) of environmental indicator (c) is a row vector by industry sector for each country, which can be obtained by dividing the environmental emissions and resource usage of the production industry sector and country by the total input of the industry sector and country. EFi,q,c is the environmental footprint of consumption segment (q) in country (i) for environmental indicator (c). Notably, it is also necessary to include direct household emissions when calculating the carbon footprint43,44.
PBs and the corresponding proxy indicators
Since the PBs framework was proposed in ref. 6, it has undergone intense debates and multiple updates6,9,10,11. Today, the concept of PBs, including the selection and quantification of proxy indicators, remains a hot topic in the literature45. This study does not aim to explore alternative boundaries or their limits. In other words, we do not attempt to revise the PB framework. Our study focuses on quantitative accounting based on the current PB framework. Five PBs were considered in this study: climate change, land-system change, biogeochemical cycles, freshwater use and biosphere integrity. The PB for novel entities was not selected because their impacts on the Earth system as a whole remain largely unstudied, and quantitative evaluation methods on a global scale are still lacking10,46. The ocean acidification PB was not included as a separate boundary as it is driven by climate change, and the corresponding pressure indicator has been included in our study. After the Montreal Protocol in 1987, many ozone-depleting substances were phased out. Owing to the decreased human perturbation of the stratospheric ozone depletion, the stratospheric ozone depletion PB was not selected in this study. Atmospheric aerosol loading is controlled by multiple factors and it is difficult to quantify from a consumption-based perspective. In addition, ref. 10 pointed out that this PB is still within the safe operating space. Thereby, it was not selected in this study.
The PB framework includes a series of proxy indicators that represent the âstateâ of PBs, such as atmospheric CO2 concentration. In this study, we define a proxy indicator and a global yearly budget for each selected PB. Notably, the selection of indicators for PBs is based on a comprehensive consideration of the existing PB literature, data availability and the computability of human environmental footprints. Therefore, the selected indicators and related global budgets are not necessarily identical to those in refs. 1,10,47,48. For five selected PBs in our study, proxy indicators and corresponding global budgets were set according to the literature. Specifically, CO2 emissions, HANPP, N fixation, application of P fertilizers, blue-water consumption and MSA loss, as well as their global yearly budget limits, were used to represent the PB categories of climate change, land-system change, biogeochemical cycles, freshwater use and biosphere integrity, respectively1,10,11,13. Owing to biogeochemical cycles being represented by two indicators (N and P), six indicators were included in our analysis. Notably, all global budgets of the six indicators were annual, and the cumulative budget of certain indicators was converted to the annual budget in a linear manner2,13. For example, we assumed that the budget consistent with 1.5â°C warming would be used up with an equal annual distribution of the CO2 emission budget over 2011â2100, in line with the common practice recommended in the literature1,49. The following sections provide a detailed description of the PBs and the corresponding proxy indicators. Supplementary Table 1 also presents the global performance of the six key environmental indicators concerning per capita PBs.
Climate change
The proposed measurements in the PB framework for climate change include anthropogenic radiative forcing threshold and the maximum atmospheric CO2 concentration. This is translated into maximum allowable global temperature increase in the documents of international policy and reporting11, with the goal set by Paris Agreement at 1.5â°C or 2â°C (ref. 50). Some literature has strengthened the target to 1â°C for fairness and local considerations9,10. The set of actionable targets for climate change mitigation is always one of the defining discourses in climate research and international policy51,52,53. In this study, the CO2 concentration consistent with the global budget on the strict Paris Agreement goals (1.5â°C) is selected. As there is an almost one-to-one link between the maximum allowable global temperature increase and the cumulative CO2 emissions, the latter is thus selected as the proxy indicator to represent the climate change boundary.
The literature puts the estimation of the global cumulative CO2 emission budget consistent with 1.5â°C at 860âGtCO2 equivalent from 2011 to 2100 (ref. 54). There are many methods to achieve the transformation from the cumulative budget to the annual budget, and each has specific pros and cons and relies on varying assumptions. We adopt the frequently used method that the CO2 emission budget would be used up with an equal annual distribution between 2011 and 2100 (refs. 1,13,49). In addition, previous accounting has shown that the budget of about 290âGtCO2 emissions has been used up globally from 2011 to the end of 2017 (ref. 55). Considering that our research year is 2017, a global budget of 570âGtCO2 from 2018 onwards is used in this study, resulting in an annual CO2 emission budget of approximately 7âGtCO2âyrâ1. This budget results in approximately 0.95âtCO2 per capita when given a population of 7.3âbillion.
It is worth noting that many factors, including political and technological, have the potential to either expand or reduce this budget. For example, the implementation of negative emission technologies could potentially increase this budget. However, such technologies also have inherent flaws, potentially leading to biophysical, technical and economic risks56. Numerous studies have also highlighted that substantially increasing the CO2 emission budget may be unrealistic, given current constraints and technological limitations50. However, the per capita boundary may contract owing to population growth. A recent study also pointed out that the remaining carbon budget for keeping warming to 1.5â°C was only around 250âGtCO2 as of 2023 with a 50% chance57. Although there are large uncertainties in the remaining carbon budgets, this estimate suggest that the budget we used in this research may be a very optimistic estimate and that the actual situation may be even worse. However, compared with the figure of 1.61âtCO2 per capita as suggested in a previous study13, the budget in this study is already strict.
Only energy- and industry-related CO2 emissions were considered, excluding other non-CO2 greenhouse gases. Therefore, the estimate of the consumption-based carbon footprint and the overshoot of climate change is optimistic, and the actual situation may be worse. Previous literature has indicated that the net emissions of land-use change over the 2010â2100 period is projected to be close to zero50; therefore, we did not consider emissions related to land use. The CO2 emissions data for 2017 for footprints estimation were obtained from the Global Trade Analysis Project (GTAP) 11 database.
Land-system change
Initially, the proxy indicator utilized in the PB framework for land-system change was the percentage of global land cover converted to cultivated land6. Subsequent updates have shifted the emphasis towards the biophysical processes within the land system, advocating for the amount of forest cover as a proxy measurement10,11. However, measuring the area of forested land associated with the consumption of goods and services is challenging. After ref. 6 proposed the PB framework, many studies have suggested that the HANPP could serve as an alternative PB. HANPP integrates various boundaries58,59,60, including land-system change, biosphere integrity61,62, freshwater use and biogeochemical cycles, owing to its comprehensive approach to assessing human impacts on ecosystems. Consequently, HANPP has been widely applied as an indicator for PBs, particularly for land-system change2,13,63,64,65.
Although ref. 10 recently assigned HANPP to biosphere integrity, we selected HANPP as the indicator for the land-system change PB. This decision is based on the multi-representativeness and consumption-based quantifiability of HANPP for PBs10,13,59,62. We argue that this selection does not undermine the main findings of our study. As emphasized, our aim is not to revise the PB framework but to make a comprehensive selection based on the literature and data availability. According to ref. 10, the global potential natural vegetation of HANPP was estimated to be 53.7â54.6âGtCâyrâ1 between 2000 and 2020, with a threshold set at 20% HANPP, equivalent to 10.8âGtCâyrâ1, to maintain Earthâs balance and sustainability. This translates to 1.47âtC per capita in 2017.
We have constructed a long-time-series HANPP environmental account based on the LundâPotsdamâJena dynamic global vegetation model (LPJ-DGVM), which has been well matched with the GTAP model in our previous studies. More details can be found in refs. 65,66.
Biosphere integrity
It is challenging to select indicators and set boundaries for biosphere integrity from an Earth system perspective. Initially, ref. 6 used the species extinction rate (rate of biodiversity loss) as a provisional indicator. Reference 11 later incorporated both the global extinction rate and the Biodiversity Intactness Index (BII) as interim indicators for this boundary. However, some studies have pointed out that the BII cannot be directly linked to establishing an Earth system state. Reference 10 retained the extinction rate and introduced HANPP as an alternative indicator to replace BII. Reference 9 also proposed two complementary indicators of biodiversity: the area of largely intact natural ecosystems and the functional integrity of all ecosystems.
In this study, we used the rate of biodiversity loss as the key indicator for biosphere integrity, in line with the PB framework. However, instead of using the extinctions per million species-years metric recommended by the original PB framework, we used MSA loss, as developed in ref. 67. MSA loss measures the average abundance of original species in a disturbed environment relative to their average abundance in an undisturbed reference environment. A notable advantage of the MSA indicator is its integration into footprint calculations, which help clarify the relationship between human activities and their impacts on biodiversity across various pressures67. We chose MSA loss for its quantifiability and practical application in assessing biodiversity impacts1,68.
Given the existence of better quantitative indicators for the biosphere integrity PB, HANPP was assigned to represent another PB: land-system change. According to ref. 1, the limit for MSA is set at 3,724âmillion MSA-loss ha per year, which can be converted to 0.51 MSA-loss ha per capita per year.
The MSA-loss environmental account, consistent with the GTAP model, was formulated using the methodology of ref. 67. We applied the MSA-loss account utilizing transparent methodologies and data provided by refs. 1,67, ensuring a comprehensive and accurate representation of MSA-loss within our study framework.
Biogeochemical cycles
The planetary framework outlines two sub-boundaries for biogeochemical flows, specifically focusing on the biogeochemical cycles for N and P (ref. 11). For the N cycle, the proposed proxy indicator is the intentional N fixation, which includes N fixation in fertilizer and from crop fixation. For the P cycle, the proposed proxy indicators include the P flow from fertilizers to erodible soils and the P flow from fresh water into the ocean. However, quantifying the amount of P transitioning from fresh water to the ocean is fraught with considerable uncertainty, and comparing this quantification with the consumption-based environmental footprints poses great challenges. Consequently, the intentional N fixation and P fertilizer use were selected as the proxy indicators for the N and P cycles, respectively1,2. According to the most recent research by ref. 9, the updated budgets are 62âTgNâyrâ1 and 4.5â9.0âTgPâyrâ1, respectively. By synthesizing the findings from refs. 9,10,11, this study adopted 62âTgNâyrâ1and 6.2âTgPâyrâ1 as the budgets for the N and P cycles, respectively. Correspondingly, the per capita N and P budgets were 0.85âkgPâyrâ1 and 8.5âkgNâyrâ1, respectively.
The intentional N fixation and P fertilizer use accounts were built with the bottom-up method. Both N fertilizer usage and biological N fixation were considered in N fixation. The national N fertilizer usage data were obtained from the Food and Agriculture Organization. The N fertilizer usage data for 13 crops in each country were obtained from the International Fertilizer Industry Association. We allocated the national N fertilizer usage values to eight agricultural sectors in GTAP. For biological N fixation, we obtained the N fixation coefficient (per kg of crop yield) for nitrogen-fixing crops from refs. 12,69. Finally, the biological N fixation was allocated to the specific agricultural sectors in GTAP.
Like the account-building process of N fertilizer usage, we obtained the P2O5 fertilizer usage from the Food and Agriculture Organization and allocated it to the eight agricultural sectors of GTAP based on the International Fertilizer Industry Association data. Multiplying the amount of P2O5 fertilizer usage by the chemical factor content of P (approximately 62/142) generated the quantity of P fertilizer use.
Freshwater use
The PB for freshwater use represents the maximum quantity of freshwater that can be appropriated by humans70. Typically, the available amount of consumptive runoff, or blue water, serves as the proxy indicator for freshwater use. Recently, ref. 10 proposed an alternative: the percentage of annual global ice-free land area with deviations in streamflow and root-zone soil moisture from preindustrial levels. Reference 9 refined this by introducing two sub-boundaries: a flow alteration boundary for surface water and a drawdown boundary for groundwater, each with its respective boundary threshold9.
AÂ substantial portion of blue water is inaccessible for human use23, and integrating this refined indicator into consumption accounting poses substantial challenges. Consequently, this study opted for global consumption of blue water as the proxy indicator for this PB11. We noticed that the threshold for freshwater use is a matter of ongoing debate, with discrepancies between estimates of global blue-water consumption derived from bottom-up and top-down methods, the former usually yielding lower estimates13. Given these limitations and adhering to the precautionary principle, this study adopted a more stringent budget, setting the threshold for blue-water consumption at 2,800âkm3âyrâ1, as defined by ref. 71, to navigate the complexities and uncertainties surrounding freshwater use and ensure a conservative approach to managing this critical resource11,13. This translated to a per capita budget of 384âm3âyrâ1.
The blue-water consumption account was also built with the bottom-up method. We considered crop farming, husbandry and other sectors separately. Blue-water consumption coefficients were adopted from ref. 72, and 2017 crop production data of 161 crops from the Food and Agriculture Organization were used to calculate blue-water consumption in crop farming. The results were allocated across the eight agricultural sectors defined in GTAP. For husbandry, country-specific blue-water consumption data provided by ref. 73 were used. For other sectors, the water use coefficient from GTAP-2014 was used. This allowed us to obtain the blue-water consumption estimates for husbandry and other sectors in 2017.
Discrepancies exist between estimates of global blue-water consumption derived from the bottom-up and top-down methods, with the former typically yielding lower estimates13. Consequently, our global pressure estimates of blue-water consumption were lower than those presented in ref. 11 but align closely with the findings of ref. 74. Regardless of the estimates considered, global blue-water usage is well within the PBs. However, regional water security issues continue to pose great challenges10,75,76.
PB downscaling
Various methods exist for downscaling PBs, each reflecting alternative views on distributive fairness77,78,79,80. Some studies have advocated for a multiscale method, arguing that considering regional background heterogeneity is more appropriate for PBs owing to the diverse ecological contexts found globally49,81,82. Other studies have supported a top-down allocation approach, asserting its appropriateness based on general concepts of distributive fairness13,48,83. Many top-down methods have been proposed in this context, including grandfathering (a right-based approach), equal per capita (emphasizing equal individual rights), ability to pay (a duty-based approach) and accounting for cumulative emissions (addressing historical responsibility)1,13. Although we recognize the practical suitability of the multiscale method for managing resource use, our study used a top-down approach, utilizing the equal-per-capita method to allocate the global budget. This choice aligns with our research focus on examining the contributions of various consumption segments to global transgressions of PBs. We operated under the premise that every individual has equal rights to access natural resources, and thus we allocate the global budget of PBs using the equal-per-capita approach3,13,84. Consequently, the PB for each proxy indicator is evenly distributed among the global population. In this way, we obtained the per capita equivalents of five PBs, providing a fair and equitable basis for analysing resource use and environmental impact across diverse populations.
Responsibilities quantifying
Our responsibility allocation is based on a 1-year scale in alignment with the current PB framework and the accounting feasibility. However, from the perspective of historical responsibility, high-end consumer groups and countries bear a greater responsibility for ecological breakdown51,52,85,86. More discussion in this regard can be found in Supplementary Information section 1.1.
The exceedance ratio measures the severity of transgressing the PBs, which is calculated as follows:
$${{\rm{Exceedance\; ratio}}}_{q}=\frac{{{\rm{EF}}}_{q}-{{\rm{Share}}}_{q}}{{{\rm{Share}}}_{q}}$$
(2)
in which EFq and Shareq refer to the environmental footprint and the fair share of consumption segment q. The fair share is determined according to the equal-per-capita approach discussed above.
The share of overshoot represents the relative responsibility of different groups for the transgressions of PBs. Following ref. 3, we posit that the undershoot of one group does not offset the overshoot of another. Consequently, the exceedance ratio for a group in undershoot is assigned a value of zero in the responsibility calculations. Thus, the share of overshoot for group q can be calculated as follows:
$${{\rm{Share\; of\; overshoot}}}_{q}=\frac{{{\rm{Exceedance\; ratio}}}_{q}}{{{\rm{Exceedance\; ratio}}}_{q\_{\rm{total}}}}$$
(3)
Moreover, we also quantify the responsibilities for PB transgressions associated with necessary versus discretionary consumptions based on the expenditure elasticity theory22,30,87. Discretionary goods are defined as having an expenditure elasticity greater than 1, whereas necessities have an expenditure elasticity less than 1. This approach helps identify the consumption attributes by distinguishing expenditure types (see Supplementary Information section 1.2 for details).
Inequality measurements
The Gini coefficient was used to measure expenditure and footprint inequalities. It ranges from 0 (perfect equality) to 1 (perfect inequality)88,89,90. The basic income Gini coefficient is calculated by:
$${\rm{G}}=\mathop{\sum }\limits_{i=1}^{n}{P}_{i}{Y}_{i}+2\mathop{\sum }\limits_{i=1}^{n}{P}_{i}(1-{C}_{i})-1$$
(4)
where G refers to income Gini coefficient, Pi, Yi and Ci are the population share, income share and cumulative income share of income group i, respectively, and n is the number of groups. Similarly, the environmental footprints inequality (EF-Gini) can be calculated by replacing the income with the environmental footprint in the equation89, and using Yi and Ci to represent environmental footprint and the accumulated footprints of consumption segment i.
We also used the Lorentz curve to show the expenditure and environmental footprint inequality, which is the ordered distribution of the cumulative share of population against the cumulative share of expenditure and environmental footprints.
Scenarios setting
The term âoverconsumptionâ is widely discussed in both the scientific literature and the mass media26,91,92,93,94, but there is no clear definition of the standards for overconsumption. Rather than attempting to define overconsumption, we set scenarios to quantify the mitigation effect of (1) reducing consumption by the affluent groups to a more sustainable level acceptable within their own group and (2) achieving the best consumer performance with existing technology and social norms within their group13,24,26,27,28,95. The global 10th percentile level of final demand is about US$27,000 per year, equivalent to the European average in 2017. The global 20th percentile level is about US$12,000 per year, comparable to the threshold of high-income countries defined by the United Nations in 2017. These two thresholds represent typical levels of affluent consumption where high living standards are maintained, as frequently referenced in the mass media, government reports and academic literature. Our analysis considers the lowest observed environmental impact intensity of consumption within each of these two groups as the âbest performanceâ achievable under existing technology and social norms within the group13,96,97. To explore the potential impact of these behavioural adjustments, we set six scenarios as detailed in Extended Data Table 1.
Data sources and process
The MRIO table was taken from the GTAP 11 database98,99. GTAP 11 is a global detailed MRIO database developed by harmonizing and detailing supply use and international trade tables for 141 countries and regions. It provides a detailed classification with 65 sectors and the corresponding household final demand. In this study, GTAP 11 covering the year 2017 was used.
The household expenditure data used in this study are a composite dataset100,101,102, sourced from the World Bank Global Consumption Database (WBGCD)5,103, the Eurostat Household Budget Survey (HBS), the Japanese Family Income and Expenditure Survey (FIES), the Canada Survey of Household Spending (SHS) and the Australia Household Expenditure Survey (HES). The WBGCD data provide a comprehensive description of household and consumption characteristics, featuring detailed information on 33 categories of consumption items and 201 expenditure levels across 116 countries for the year 2011. The HBS delineates household and consumption characteristics across 12 major categories and 47 sub-categories for 5 quintiles in 32 European countries for the year 2015. The FIES details household and consumption characteristics across 23 categories for 10 deciles in Japan for the year 2017. The SHS provides data on 358 consumption categories for 5 quintiles in 2017. The HES data cover 12 major categories and 46 sub-categories for 5 quintiles in 2015.
The household expenditure survey data have to be bridged and matched to GTAP to calculate the environmental footprints among expenditure groups. First, considering that WBGCD has the broadest geographical coverage, we used the consumption shares of each sector by expenditure bins in WBGCD as the basis, updating them with other national expenditure survey data where available. For countries lacking data, we approximated the expenditure distribution structure using neighbouring countries with comparable levels of development. Given the constraints of data availability, this approach was deemed appropriate87. As a result, the refined expenditure dataset encompasses 33 sectors, 201 bins and 168 countries. Next, we constructed a bridging matrix to link these 33 sectors to the 65 sectors in the GTAP MRIO table, adhering to the sector definitions provided in refs. 5,104. This matrix enabled the derivation of consumption shares in each sector by expenditure bins for the 65 sectors in the GTAP MRIO table. This means that our analysis was consistently based on basic prices (producer price), and the information we retrieved from the expenditure data pertained to the expenditure shares, not the monetary values of expenditure. This process yielded household final demand data, which are consistent with the GTAP classification and across different consumption segments105. In addition, we updated the 2011 population data in the expenditure survey data to 2017, using population statistics from the World Bank and maintaining the original datasetâs population distribution between expenditure bins. It is worth noting that to ensure comparability in discussing the PB, we assumed that the final demand from government and the investment of different consumption segments would follow the same distribution as the household consumption, in the absence of additional pertinent information5. Further discussion on the uncertainties and limitations of the methods and data are provided in Supplementary Information section 3.