International financial support to achieve the net-zero emissions goal could help resolve equity trade-off between developing and developed countries

Overall methodological framework

We used AIM and integrated a computable general equilibrium model (AIM-Hub), a household consumption and income distribution tool (AIM-PHI), and a hunger projection tool (AIM-Hunger). The core scenarios were developed based on the NDC and NZE targets, aiming to achieve 2 °C goals along with a baseline scenario, which followed the Shared Socioeconomic Pathway 2 (SSP2) baseline projections. Scenarios were also constructed to determine whether key conclusions would be affected by our key assumptions. We assessed the impacts of decarbonization on macroeconomic indicators such as consumption loss, poverty, hunger risk, and CDR.

Models

AIM-Hub

AIM-Hub is a 1-year-step recursive dynamic general equilibrium model that covers all regions globally. AIM-Hub includes 17 regions and 58 industrial classifications (Supplementary Table 1 and Supplementary Table 2). The energy supply technologies were disaggregated to a finer resolution for an appropriate assessment of the energy system. Multiple crop and livestock sectors were explicitly represented to consider bioenergy and land use appropriately⁴⁷. Production sectors were assumed to maximize profits through multi-nested constant elasticity substitution (CES) functions and input prices. Input energy and value added for the energy transformation sector were treated as fixed coefficients of the output, maintaining the energy conversion efficiency of the energy transformation sector as technologically feasible. Power generation associated with multiple energy sources was combined using a logit function to ensure energy balance, unlike the CES function. As reported previously, curtailment and battery storage are represented within this model using a simplified exponential function for the share of variable renewable energy, with parameter values based on a previous study⁴⁸. A linear expenditure system function was used to describe household expenditures on each commodity type, for which the adopted parameters were recursively updated based on income elasticity assumptions⁴⁹. Land use was determined by logit selection with multiple nodes⁴⁷. AIM-Hub includes representations of the major GHGs and air pollutants, as described previously⁵⁰ which is finally used to derive climate information. The non-CO₂ abatement is driven by abatement cost, carbon, and carbon prices

CDR options were key for this study; AIM-Hub represents six CDR options: BECCS, afforestation, DACCS, soil carbon, biochar, and enhanced weathering. BECCS and afforestation have been applied in past numerical studies using AIM-Hub⁵⁰. For DACCS, the model assumes that an industry that provides DACCS services for each government follows a Leontief production function, using previously reported technological parameters⁵¹; further details are presented elsewhere⁵². The remaining CDR approaches, including soil carbon, biochar, and enhanced weathering, are similarly represented, and new industries are assumed for each CDR service, with cost assumptions described in a previous study⁵³. The agricultural land is considered a potential supply constraint.

The base year of AIM-Hub is 2005. Because recent energy information was used as it became available, the model results regarding energy supply and consumption generally followed International Energy Agency (IEA) World Energy Balances until 2019. Although IEA data were available until 2022, we stopped the parameter adjustment because we were unable to judge whether the effects of the COVID-19 pandemic and the Ukraine conflict should be fully reflected in this process. The same methodology used for this historical calibration was used for model integration with an energy system model, in which the final energy, transport energy share, and power energy technological share were obtained exogenously. In contrast, corresponding parameters in the production function and household consumption were endogenized^52,54.

Regarding the data for AIM-Hub, the Global Trade Analysis Project (GTAP) and World Energy Balances were used as the basis for the social accounting matrix. These data were reconciled with other available data, such as national accounting statistics⁵⁵. The concept underlying the reconciliation method has been described previously⁵⁶. GHGs and air pollutant emissions were calibrated using the U.S. Securities and Exchange Commission Electronic Data Gathering, Analysis, and Retrieval dataset v4.2⁵⁷. For land-use and agricultural sectors, agricultural statistics⁵⁸, land-use representative concentration pathway data⁵⁹, and GTAP data⁶⁰ were used as physical data. Agricultural consumption was converted into caloric intake using a conversion factor derived from agricultural statistics⁵⁸. Data on solar and wind energy, as energy resource potentials, were obtained from previous studies⁶¹ and calculated using high-spatial-resolution data (0.5 arcmin or approximately 1 km at the Equator). Techno-economic information related to energy supply facilities, such as capital and operation costs, was based on information available in 2020, including the IEA World Energy Outlook.

AIM-PHI

The AIM-Poverty, Household, and Income distribution (AIM-PHI) model projects the income distribution, poverty headcount, and household consumption at 5-year intervals with socioeconomic assumptions, and policy impacts on household expenditures and commodity prices from the upstream model. AIM-PHI contains a nonlinear demand system based on a lognormal assumption of disposable income^62,63,64. The lognormal distribution captures national average household consumption in the upstream model and matches Gini projections of the corresponding SSPs⁶⁵ for 184 modelling countries.

To reflect distributional impacts, we discretized each modelling country into hundreds of income segments and ran a numerical consumer demand model, AIDADS⁶⁶, encompassing 14 commodities. Household consumption patterns in each country were calibrated to global and national household consumption survey databases in a two-step process. The counterfactual household expenditure was decided using commodity prices projected in upstream models. Price elasticity in household demand at each income level was differentiated to properly represent the poor income class and poverty projections. The model was validated according to whether the behavior of the AIDADS model reproduced patterns of goods consumption that were consistent with the development stage; further details are presented elsewhere^62,63,64.

Finally, we derived an analytical income distribution function from which the poverty headcount and domestic Gini coefficient could be calculated under various policy impacts. The AIM-PHI model accommodates poverty headcount assessments based on different poverty lines; we focused on an extreme poverty line of USD $2.15 per capita per day.

We used multiple data sources for household expenditure: the World Bank Global Consumption Database⁶⁷ (GCD) and EUROSTAT household budget surveys⁶⁸. These databases cover 90 and 34 countries, respectively. In addition to them, we collected household consumption surveys for developed countries, including Japan, the USA, Canada, and Australia, for the parameter estimation. Using these data sets, we first identify a set of parameters using cross-country data, which enables us to apply this model to countries that do not have observations in household expenditure.

In the second estimation step, we incorporated multi-household sector information and derived country-specific parameters so that we can properly reflect individual countries’ household expenditure structure. GCD contains four income segments and distinguishes between urban and rural households. EUROSTAT classifies income into quintile segments. The GCD contains single-year information that has been reconciled by the World Bank for 2010, whereas EUROSTAT contains information for multiple years. Although the International Consumption Project (ICP) database is often used for global-scale household studies⁶⁹, we did not use it because the poverty headcount reported in the World Development Indicators (WDI) is based on the PovcalNet Database⁷⁰ household survey that is made by GCD, which provides good reproducibility, as noted above, but makes no assessment of actual consumption from other sectors (e.g., governmental education services). The ICP includes this assessment, which leads to inconsistency in the poverty headcount, household income, and household expenditure data. In principle, the household consumption estimate should account for real consumption, which is covered in the System of National Accounts assessment. However, we prioritized consistency instead and used the poverty headcount statistics. The model also used a purchasing power parity–market exchange rate conversion factor for each country; these were obtained from the WDI and kept constant over time. Similarly, the market exchange rate was implicitly assumed to be identical to that of the base year.

AIM-Hunger

To determine the population at risk of hunger, we used the AIM hunger-risk projection tool (AIM-Hunger) developed by Hasegawa et al.⁴², which calculates the population at risk of hunger according to mean dietary energy availability, the mean minimum dietary energy requirement (MDER), and the coefficient of variation of the domestic food consumption distribution, which is assumed to have a lognormal distribution. AIM-Hub provides calorie-based food consumption as the mean dietary energy availability, based on the national average price elasticity in food demand. The coefficient of variation of food consumption is derived from base year information and its relationship with income growth. The population living under the domestic MDER, which is adjusted for the future demographic changes from the base year, is defined as being at risk of hunger.

The price elasticity in aggregated food demand at the household level, which was used for poverty projection, was often larger in absolute terms than the item-wise price elasticity at the national level, which was used for hunger-risk projections. This discrepancy resulted in differential sensitivity between the hunger-risk and poverty projections in response to agricultural productivity assumptions. Furthermore, because we did not consider direct and indirect subsidies, the climate policy impacts for lower-income segments may have been overstated. For example, Iran and Libya had average subsidization rates, which were calculated as subsidies or reference fuel prices⁷¹, exceeding 90%. Carbon tax revenues can also represent a source of subsidies. Subsidies play important roles in moderating price shocks and income loss, particularly among lower-income segments of the population. Therefore, actual price responses may be weaker than those determined by our model.

Scenarios

The core scenarios were implemented by differentiating emissions constraints imposed for each region, financial transfer, and CDR options (Table 2). The emissions target represented as NDC is based on emissions targets for 2030; any improvement in emissions intensity observed from 2020 to 2030 based on this NDC target was extended thereafter. If the emissions targets exceeded the baseline emissions for a given country, then the emissions constraints in that country were ignored. The emissions target represented as NZE indicates that each region reaches the net zero target after 2030; however, the year in which NZE is realized depends on each country’s announced target. For example, most developed countries anticipate reaching NZE by 2050, whereas China and India have target years of 2060 and 2070, respectively. Under the NZE assumption, countries without an NZE target are also assumed to achieve NZE by 2070, which is the latest NZE target year among specified target years in long-term strategy announcements. In this study, the NZE condition is harmonized as CO₂ only (including land-use emissions) despite ambiguity and variation in the coverage of GHGs of each country’s long-term strategy announcement, because this approach allowed us to treat emissions constraints consistently. The assumptions of the NZE–EC scenario originated from the NZEOnlyDeveloped scenario. We determined additional emissions reductions from NZEOnlyDeveloped for developed regions to have approximately the same global carbon budget as the NZE–Def scenario based on two metrics: the lowest emissions and the time at which that level was reached. There were multiple possibilities for determining these two metrics; using a trial-and-error method, we obtained –25 Gt CO₂/year and 2080, respectively. Further deep negative emissions or early realization of the lowest emissions could alter the quantitative results, but would not greatly affect the conclusion. The total additional emissions reduction for developed countries was distributed in proportion to the 2020 GHG emissions within developed regions.

The amount of financial support applied in the NZE–FS scenario was assumed to be the difference in macroeconomic household consumption between the NDC and NZE–Def scenarios, which would imply that consumption losses in developing countries are approximately compensated by financial support. This assumption forms the basis of one of the CBDR principles, and as is confirmed in the results, the level of emissions reduction is more or less in line with the greenhouse gas development rights. Meanwhile, the distribution of donor regions was proportional to GDP. The burden share among the developed regions is rationale based on each country’s ability to pay.

We conducted sensitivity analyses consisting of three sets of two scenarios, comparing those with and without financial support. The three analyses consisted of changing the degree of emissions reductions in developed countries, regional classifications, and BECCS availability. The amount of financial support was determined in a manner similar to that of the NZE–Def scenario, by calculating the difference in consumption loss for the corresponding NDC scenario, and transferring these differences from developed to developing regions.

For sensitivity scenarios including financial support, the amount of financial support was determined in the same manner as in the NZE–Def scenario; i.e., we calculated the differences in consumption loss for the corresponding NDC scenarios and transferred them from developed to developing regions.

NZE goals and NDCs

We collected data on NDCs and the NZE targets that were announced by each country, as input for our model. Note that this study included only emissions reduction targets indicated in NDCs, and excluded targets related to amounts of renewable energy introduced or other factors. Information on the NZE targets is available via Net Zero Tracker⁷². We used data on NZE targets available as of November 2021, which included data for 103 countries and regions (including the 27 European Union countries) and accounted for 86.0% of global GHG emissions in 2018. The target years for NZE were 2030 for four countries, 2040 for two countries, 2050 for 85 countries, 2060 for nine countries, and other for three countries. The target gases were GHG for 58 countries, CO₂ for five countries, and unknown for 40 countries.

The NZE targets submitted or announced by each country differed in terms of the format of their announcement, depending on whether they were clearly stated in the NDC or as long-term strategies, whether they were clearly stated in the country’s own laws, or whether they were simply announced in the media or elsewhere. In this study, we treated the NZE targets announced by each country as being implemented, regardless of the format.

The NZE targets differed among countries in terms of the specific emissions covered; some countries did not specify which gases were covered. Therefore, we assumed that all NZE targets covered GHGs, and treated NZE targets that covered CO₂ as achieving GHG net zero. When we estimated the emissions reduction targets of each country using the AIM-Hub model, the emissions reduction targets and emissions inventories of each country were first used to estimate target emissions values when the emissions reduction targets were implemented. These values were aggregated according to the 17 regional divisions of the AIM-Hub model. The aggregated emission targets were then imposed as emission constraints. Because most countries indicated emissions reduction targets for GHGs, information on the GHG emission targets of each country in 2030 was available. Additionally, information on GHG emissions by 17 regions and CO₂ emissions by 17 regions in 2030 could be obtained from the estimation results of the AIM-Hub model. However, the CO₂ emission targets for 2030 for each country were unavailable, such that even if NZE was indicated to cover CO₂, it was impossible to set the CO₂ emission pathway for each country from 2030 to 2050. Therefore, we treated all NZE targets as being for GHGs. Note that this approach may have led to somewhat stricter assumptions regarding emissions reductions compared to the NZE indicated by each country.

The year in which an NZE target was implemented differed among countries, and in many cases, the year was not specified. Therefore, we assumed that NZE targets would be implemented from 2030, after the implementation of the NDC, in a uniform manner. We also assumed that emissions constraints would decrease linearly from 2030 to the year in which the NZE target was achieved. When setting the emission constraints, the AIM-Hub model first estimated the 2050 emissions targets for each country, at which the 2030 emissions targets for each country under the NDC implementation would be linearly reduced to zero, and then aggregated the 2050 emission targets into 17 regions. Next, emissions constraints were imposed such that emissions were reduced linearly from 2030 levels in the 17 regions when the NDCs were implemented, as estimated by the AIM-Hub model, to the emissions targets in 2050, which were aggregated into 17 regions. For countries that have not announced an NZE target, we assumed that the rate of emissions reduction from the baseline scenario in 2030 (1−Emissions_NDC/Emissions_Baseline) would be maintained after 2030.

Carbon budget of each effort-sharing scheme

Carbon budgets were estimated as previously described²¹. Baseline emissions projections were sourced from the AIM-Hub results, and GDP and population data were obtained from the SSP2 scenario. The responsibility–capacity Index was determined using a climate equity reference calculator⁷³, and historical GHG emissions data were obtained from the PRIMAP emissions database⁷⁴. The equations for each effort-sharing scheme were as follows:

where \(t\) is the year; \(i\) is the region; \(G{F}_{i}\) is the carbon budget for region \(i\) in the grandfathering scheme; \({e}_{i,t=2020}\) represents the emissions for region \(i\) for \(t=2020\); \({E}_{t=2020}\) represents the global emissions for \(t=2020\); \(B\) is the global carbon budget; \({IEP}{C}_{i}\) is the carbon budget for region \(i\) in the immediate per capita convergence scheme; \({po}{p}_{i,t}\) is the population for region \(i\) in year \(t\); \({{POP}}_{t}\) is the global population in year \(t\); \({PC}{C}_{i}\) is the carbon budget for region \(i\) in the per capita convergence scheme; \(w\) is a weighting factor ( = 0.5); \({ECP}{C}_{i}\) is the carbon budget for region \(i\) in the equal cumulative per capita emissions scheme; \(d\) is a discount factor ( = 0.02); \({{AP}}_{i}\) is the carbon budget for region \(i\) in the ability-to-pay scheme; \({{bau}}_{i,t}\) represents the regional baseline emissions for region \(i\) in year \(t\); \({{rAP}}_{i}\) is the carbon budget reduction target from baseline for region \(i\) in the ability-to-pay scheme, defined in Eq. (7); \({corr\_r}\) is the global correction factor, defined in Eq. (8); \({GD}{R}_{i}\) is the carbon budget for region \(i\) in the greenhouse development rights scheme; \({{BAU}}_{t}\) represents the global baseline emissions for year \(t\); and \({{rci}}_{i}\) is the responsibility–capacity Index for region i. We define \({{rAP}}_{i}\) and \({corr\_r}\) as follows:

where \({{gdp}}_{i,t}\) is the GDP for region \(i\) in year \(t\); \({{GDP}}_{t}\) is the global GDP in year \(t\); and \(I\) represents all regions.