What the evidence says
The sources here are mostly simulations, plus one experiment on public support. They measure who pays a carbon tax and who gains once revenue is returned. The measured burden of a carbon tax with no recycling depends on how welfare is measured. Hassett (2009) finds that carbon taxes appear more regressive when annual income is the welfare measure than when current or lifetime consumption is. The direct component, from fuel consumption, is more regressive than the indirect component embedded in goods, by a factor of roughly 2-6x across deciles depending on year. In 1987 the indirect component was mildly progressive, because higher-income groups paid a larger fraction of income.
Returning revenue as an equal per-person dividend changes the net result in the simulations. Douenne (2022) finds that lower-income households gain larger amounts and larger percentage gains than higher-income households under a tax and dividend. About 70 percent of households gain in monetary terms, and the average net gain is €24 per consumption unit per year. In the same study, only 14 percent of households think they would gain. Young-Brun (2025) finds the same direction globally. The poorest 50% of households within countries generally gain from a uniform global carbon tax with international transfers in 2030. The lowest decile is a net transfer recipient in almost all countries. The stated mechanism is that the poor receive larger transfers than the carbon tax and mitigation costs they pay. Global inequality, measured by the Gini index, decreases by up to around 1.75 percentage points in all scenarios compared to business-as-usual.
Kulkarni (2025) models the case without recycling. Uniform carbon pricing increases the energy burden of poorer groups, and inequality effects peak in 2025–2045 and decline toward near zero by 2100. Electricity price rises add to the burden of low-income households, mainly in regions with large capacity expansion (Russia, South Africa, Western Africa). Redistributive add-ons decrease the projected Gini increase. Averaged over 2025–2100 and excluding outliers such as Russia, exemptions cut it by about 20%, rooftop PV incentives by about 36%, and the combined policy by about 49%. Exemptions decrease energy burden in the short term, but their benefit becomes modest later as households exceed the income threshold. Exemptions also increase the risk of reinforcing fossil fuel reliance among exempt groups.
Perceived distribution also changes public support. In Bergquist (2025), a distributionally unfair carbon tax decreases support by 56.5 percentage points compared with a fair one. A procedurally unfair tax decreases support by 8.6 points. The pattern held in the USA, the UK and India.
Several results rest on model assumptions. Douenne uses assumed price elasticities. Young-Brun assumes equal per-capita transfers. Kulkarni assumes modeled income trajectories. The sources here contain no ex post measurement of an implemented dividend program.
Findings
A carbon tax with an equal dividend increases net gains for lower-income households relative to higher-income households
confidence — direction: highconfidence — magnitude: mediumconsensus: thin — one study, direction consistent with the global simulation below
- mechanism
- revenue is returned as a dividend, so lower-income households receive a transfer that exceeds their added carbon cost
- effect
- lower-income households gain more than richer households in both relative and absolute terms. About 70 percent of households gain in monetary terms. The average net gain is €24 per consumption unit per year (€18 from transport, €6 from housing). Only 14 percent of households think they would gain.
- conditions / moderators
- simulation on survey data. Depends on assumed elasticities (−0.4 for transport, −0.2 for housing) and on measured household energy consumption. Not established beyond the single study.
Global recycling of carbon tax revenues increases consumption of the poor and decreases global inequality
confidence — direction: highconfidence — magnitude: mediumconsensus: thin — one modeling study
- mechanism
- the poor receive larger transfers than the carbon tax and mitigation costs they pay
- effect
- the global Gini index decreases by up to around 1.75 percentage points in all 2°C scenarios compared to business-as-usual. The bottom 50% in low-income countries gain about 430 USD2017 per capita (around +68% of BAU consumption per capita) under global per-capita recycling. The lowest decile is a net transfer recipient in almost all countries, with a few exceptions in very rich or carbon-intensive countries. Under the Loss and Damage scenario, targeted low-income countries receive upward of 50 USD2017 per capita in 2030. Compensating all damages in low-income countries would require 100 billion USD2017 in 2030 (5% of revenues) and 500 billion USD2017 in 2050 (15% of revenues).
- conditions / moderators
- depends on the damage function calibration, equal per-capita transfers, a lower consumption elasticity of tax burden for the poor, and negligible redistribution cost
Measured regressivity of a carbon tax without recycling is larger on annual income than on consumption
confidence — direction: highconfidence — magnitude: highconsensus: thin — one study
- mechanism
- annual income is a noisier proxy for lifetime welfare than consumption, so low annual-income households look more burdened
- effect
- the tax appears more regressive when income is the welfare measure than when current or lifetime consumption is
- conditions / moderators
- cross-sectional analysis of three years using the Consumer Expenditure Survey and Input-Output tables. Results are sensitive to the choice of income or consumption measure. The lifetime correction procedure is not validated.
The direct component of a carbon tax is more regressive than the indirect component
confidence — direction: highconfidence — magnitude: mediumconsensus: thin — one study
- mechanism
- the direct component falls on fuel consumption. The indirect component falls on the carbon embedded in goods.
- effect
- by annual income, the direct burden is more regressive by a factor of roughly 2-6x across deciles depending on year. The indirect component was mildly progressive in 1987, because higher-income groups paid a larger fraction of income.
- conditions / moderators
- depends on how well Input-Output accounts separate direct from indirect effects and on the precision of price pass-through calculations. Measured by annual income.
Uniform carbon pricing without redistribution increases the burden on poorer households, mostly early
confidence — direction: highconfidence — magnitude: mediumconsensus: thin — one preprint simulation
- mechanism
- poorer groups carry front-loaded carbon-price burdens, and electricity price rises raise their energy spending despite low electricity demand
- effect
- poorer groups are disproportionately burdened. Inequality effects peak in 2025–2045 and decline toward near zero by 2100 as energy systems decarbonize. Electricity price effects are large in Russia, South Africa and Western Africa and small in Brazil, India and China.
- conditions / moderators
- depends on modeled technology diffusion, household energy-choice elasticities, and electricity supply curves and capacity expansion needs
Exemptions and rooftop PV incentives decrease the projected rise in inequality under carbon pricing
confidence — direction: highconfidence — magnitude: highconsensus: thin — one preprint simulation
- mechanism
- exemptions lower the energy costs of low-income households directly. PV incentives raise solar generation capacity among eligible income groups.
- effect
- projected Gini increases (2025–2100 average) fall by about 20% with exemptions, about 36% with PV incentives, and about 49% under the combined policy, excluding outliers such as Russia. Combined, the Gini increase falls by up to half on average across regions. Exemptions lower energy burden in the short term, with benefits becoming modest later as households exceed the income threshold. PV incentives are somewhat less effective in the short term and generally outperform exemptions in the medium (2046–2065) and long term (2066–2085). The PV effect is especially strong in South Africa, Brazil and India. PV incentives can reduce inequality peaks in the short term and lead to progressive outcomes in some regions.
- conditions / moderators
- depends on regional fossil fuel intensity, solar irradiation, roof availability, income trajectories, and household adoption of PV
Carbon price exemptions increase the risk of reinforced fossil fuel reliance among exempt low-income households
confidence — direction: highconfidence — magnitude: mediumconsensus: thin — one simulation
- mechanism
- exempt households face no carbon price signal toward switching away from fossil fuels
- effect
- increases fossil fuel dependence among exempt households. No magnitude is stated.
- conditions / moderators
- depends on whether modeled household energy choice reflects technological lock-in observed empirically. Otherwise not established.
Perceived unfair distribution of carbon tax costs decreases public support
confidence — direction: highconfidence — magnitude: highconsensus: thin — one paper, consistent across the USA, the UK and India
- mechanism
- distributional unfairness acts as a psychological barrier to support, stronger than procedural unfairness
- effect
- support is 56.5 percentage points lower for a distributionally unfair carbon tax than a fair one, and 8.6 points lower for a procedurally unfair one. In Experiment 3, the 75–25 unequal split decreased support for carbon tax by 47–54 percentage points compared with 50–50. When participants modified unfair proposals, they set distributions at about 30–70% between poor and rich and 28–72% between citizens and corporations.
- conditions / moderators
- pre-registered between-subjects experiments with 495 participants (three countries) and 1398 participants (Experiment 3). The modification task had 208 participants and stated preferences lack external validity.