A multi-model assessment of inequality and climate change

The study investigates how climate change and climate policies affect within-country economic inequality and overall welfare. Motivated by concerns that carbon pricing can be regressive while climate impacts disproportionately harm lower-income groups, the authors aim to quantify: (1) the effect of climate impacts on within-country inequality; (2) the distributional consequences of mitigation policies compatible with the Paris Agreement; (3) the role of revenue recycling—particularly equal per-capita transfers—in offsetting regressive effects; (4) the relative magnitude of these channels; and (5) the robustness of findings across diverse modeling paradigms. The context includes growing societal focus on inequality, mixed public support for redistribution, and limited integration of inequality metrics in mainstream climate-economy models. This work fills a key gap by conducting a coordinated multi-model comparison explicitly designed around inequality outcomes.

Prior literature documents regressive effects of carbon pricing and higher energy costs, with potential adverse impacts on policy acceptability. Empirical and modeling studies show both climate policy and climate impacts can disproportionately affect poorer households within and across countries. Carbon revenue recycling—through equal per-capita (EPC) or targeted transfers—has been proposed to mitigate regressive impacts and alleviate poverty, with mixed evidence on public support across income levels and countries. Advances in empirical damage estimation highlight large regional disparities in climate risks and an increasing within-country inequality effect, especially in low- and middle-income countries. Despite this evidence, major model databases (e.g., IPCC AR6) do not routinely report inequality indicators, and earlier multi-model efforts were limited in scope (e.g., focused on international emissions trading and CGE models).

Design: A coordinated multi-model ensemble of eight integrated assessment models (IAMs) was used, spanning four theoretical paradigms: computable general equilibrium (CGE), detailed process-based IAMs (DP-IAM), cost-benefit IAMs (CB-IAM), and a macroeconometric/post-Keynesian model. The analysis focuses on within-country distribution using income/consumption deciles where available, harmonized to consumption-based inequality when possible. A decile-level dataset of energy expenditures for housing and transport supports the distributional analysis in several models. Countries and Horizon: Ten large countries with representation in at least three models were analyzed: Brazil, Canada, China, France, India, Japan, Mexico, Russia, South Africa, and the United States. Most models run to 2100; E3ME runs to 2050. Scenarios: Six core scenarios were implemented, orthogonal along two dimensions—climate policy stringency and revenue recycling: - Reference: No new/strengthened climate policy after 2020 (business-as-usual or current policies to 2020), no global carbon pricing; includes a version with and without climate impacts. - Paris (650 GtCO2 budget, 2020–2100): A uniform global carbon price across sectors and countries without overshoot, also applied to non-CO2 GHGs via AR6 GWP; no international transfers or permit trade; includes versions with and without climate impacts. - Paris with EPC: Same carbon budget and pricing as Paris; all domestic carbon revenues redistributed equally per capita to households; negative transfers from potential net-negative emissions are precluded (set to zero). - Robustness: Weak Paris scenarios with a 1,150 GtCO2 budget were also run (with and without climate impacts). Overall, the matrix yields ten scenarios (five policy settings each with and without impacts where applicable). Inequality Measurement: Emphasis is on consumption inequality (preferred over income inequality due to data reliability). Models differ in representation: parametric distributions (e.g., log-normal in ReMIND), explicit decile-level income/consumption with detailed expenditure patterns in CGEs, and elasticity-based decile approaches in CB-IAMs. Climate Impacts: Three models include their own impact modules (NICE, RICE50°, ReMIND). For other models, a post-processing damage function at the decile level was applied based on recent empirical estimates, linking decile income growth to temperature (and controlling for precipitation) with country fixed effects, time effects, and trends. Damages have an income elasticity around 0.6, implying modestly regressive within-country impacts. Global-mean temperature is linked to cumulative CO2 emissions via a transient climate response to cumulative emissions of about 0.44 °C per 1,000 GtCO2, then downscaled to the country level using CMIP6-based linear downscaling. Policy Incidence Estimation: The regressivity or progressivity of mitigation costs was summarized by estimating the income elasticity of policy cost using ordinary least squares across deciles, time, and models for each country. Macroeconomic and Welfare Metrics: Macroeconomic impacts are reported as GDP deviations (PPP) from the Reference without impacts. Welfare is evaluated using the equally distributed equivalent (EDE) consumption measure under isoelastic (log) utility, integrating both efficiency (aggregate GDP/consumption effects) and equity (inequality) into a single metric. Uncertainty: Model agreement is quantified as the share of models agreeing on the sign of effects by country and scenario, and results are reported as model medians with agreement shading where relevant.

- Climate impacts and inequality: In the Reference pathway reaching roughly 2.75 °C (2.5–3.1 °C) by 2100, climate impacts increase within-country inequality by a median 1.4 Gini points by 2100; some region–model combinations show up to 2.5 points. The effect is small in the short run but substantial by century’s end. Estimated increase is about 0.4 Gini points per degree of country-level warming. Model agreement on the sign is high. - Paris policy without redistribution: Carbon pricing compatible with a 650 GtCO2 budget leads to a modest, temporary increase in inequality: around +0.1 Gini points in 2030. By 2100, inequality is higher by about 0.4 Gini points relative to a world without policies or impacts; however, compared with the damages scenario, Paris reduces inequality by roughly 1 Gini point, highlighting equality benefits of climate stabilization. CGE models show larger increases (up to +3.5 Gini points in GEM-E3 by 2100), DP-IAMs smaller, and the macroeconometric model the smallest. Model agreement on the sign exceeds 70%. - Policy cost incidence: The income elasticity of policy cost averages 1.04 (≈ proportional on average). Across the ten countries, elasticities range 0.75–1.05. Richer countries tend to show more regressive incidence; some poorer countries show progressive incidence. - Revenue recycling (EPC): Equal per-capita redistribution of carbon revenues makes overall effects inequality-reducing in the short term. In 2030, Gini decreases by about 2.4 points versus the Reference without impacts, with larger benefits in developing countries (notably India and South Africa). Benefits decline as emissions (and revenues) fall: by 2050, Gini is ~0.7 points below the Reference without impacts and roughly equal by 2100. Relative to a Reference with damages, combining Paris plus EPC yields Gini gains of about 2.4 (2030), 0.9 (2050), and 1.3 (2100) points. On average, a US$1,000 per-capita carbon dividend reduces Gini by about 0.6 points; largest reductions occur in South Africa, India, and Brazil; OECD countries see smaller reductions (≈1.0–1.5 points). Model classes with higher economic resolution (CGE) show larger short-term inequality reductions. - Macroeconomic outcomes: With climate impacts, Reference GDP losses average about 4.9% by 2100, with regional variation and some high-latitude gains. Under Paris, policy costs reduce global GDP by roughly 2–5% per year on average until 2050. By 2100, total GDP losses are about one-third lower than in the Reference due to avoided damages. Some models (especially CGEs) show costs up to ~20% in certain regions; others show occasional gains. Revenue recycling has minimal macro impact in most models; an exception is Imaclim due to interactions with pre-existing tax recycling and induced emissions feedbacks. Model agreement on GDP impacts is high (≥86%). - Welfare (EDE): In the Reference with damages, welfare losses accumulate to about 7.1% by 2100. In Paris without transfers, welfare losses are smaller (≈5%) by mid- and end-century, about one-quarter lower than Reference. With EPC transfers, welfare increases by about 1.1% in 2030 (redistribution outweighs GDP loss in welfare terms), halves welfare losses by 2050 to ≈2.5%, and retains long-term welfare gains from limiting warming to well below 2 °C. Country-specific outcomes vary, with consistently high but somewhat lower agreement once transfers are introduced.

The results jointly address the study’s questions: climate damages substantially raise within-country inequality over time, while Paris-aligned mitigation mitigates these inequality increases in the long term despite small short-term regressivity. Revenue recycling via equal per-capita transfers can more than offset short-term regressive effects of carbon pricing, delivering substantial near-term inequality reductions and measurable welfare gains without materially harming macroeconomic performance. Benefits are largest in lower- and middle-income countries, where poorer deciles gain most from uniform dividends. The analysis shows that the design of climate policy—particularly how revenues are recycled—critically shapes distributional and welfare outcomes. Across diverse modeling paradigms, sign agreement is high for key effects, underscoring the robustness of the central message: climate stabilization and inclusive redistribution can be complementary, not competing, objectives.

This study provides a first-of-its-kind, coordinated multi-model assessment explicitly integrating within-country inequality into climate mitigation and impact scenarios. It shows that: (i) unmitigated climate change increases inequality substantially by century’s end; (ii) Paris-consistent mitigation reduces long-term inequality relative to a damages world; and (iii) equal per-capita recycling of carbon revenues can offset short-term regressivity and raise welfare in the near term. Together, these insights suggest that well-designed climate policies can simultaneously stabilize climate and promote economic inclusion. Future work should extend to international redistribution and effort-sharing mechanisms, explore alternative and targeted transfer schemes, improve granularity of within-country impact functions, consider broader sets of countries and microdata integration, and assess sensitivity to alternative baselines, overshoot pathways, and inequality metrics.

- Policy stylization: Uniform global carbon price across all sectors and countries without international transfers or permit trade is politically unrealistic; the design isolates within-country distribution but omits cross-border redistribution effects. - Country coverage: Focus on ten large countries; results may not generalize to smaller or less-studied economies. - Model heterogeneity: Different structures and distributions (parametric vs explicit deciles) may affect comparability; E3ME only runs to 2050. - Climate damages: For several models, impacts are post-processed using empirically estimated damage functions with assumptions (e.g., income elasticity ≈0.6, temperature downscaling) that introduce uncertainty. - Revenue dynamics: Inequality benefits from EPC recycling diminish as emissions decline and revenues shrink; negative emissions years are set to zero transfers to avoid negative payouts. - Baselines and measures: Many models rely on SSP2-like inequality projections and primarily consumption-based inequality; alternative baselines or inequality metrics could yield different magnitudes. - Uncertainty: Some country–scenario combinations show less than two-thirds model agreement; CGE models sometimes produce high GDP losses, reflecting sensitivity to structural and policy assumptions.