The inequality impacts of the carbon tax in China

The study investigates how carbon tax policies in China affect income inequality and subjective well-being (proxied by relative utility). Prior work shows the poor spend a larger share of income on pollution-intensive goods, making many climate policies regressive; however, inequality effects are often omitted in policy design and evaluation despite their potential to weaken policy effectiveness. The paper links inequality impacts of climate policy to behavioral insights on relative income and subjective well-being, measuring welfare changes using relative utility that depends on income relative to a societal reference. The research asks: how do the distribution of climate damages, tax payment incidence, and recycling of tax revenues under carbon taxes influence inequality and relative utility across household income groups in China? The paper contributes by (1) linking climate-policy-induced inequality to behavioral studies of relative utility, (2) measuring welfare impacts of tax-induced inequality via relative utility, and (3) analyzing how alternative revenue recycling schemes affect inequality.

The paper draws on literature showing complex links between inequality and emissions, and the regressive potential of climate policies as the poor allocate more income to carbon-intensive goods. Studies document negative effects of reference income on happiness and worker satisfaction, supporting the use of relative utility as a welfare metric. Prior evidence often finds regressive impacts of carbon pricing absent compensatory measures, though redistribution (e.g., lump-sum transfers) can mitigate regressivity. Inequality in climate damages is emphasized, with poorer populations disproportionately exposed. The paper positions its contribution by integrating these strands: distribution of damages, tax incidence, and revenue recycling as key channels for inequality effects, and subjective well-being effects captured through relative utility.

A dynamic recursive, top-down CGE model (Johansen-style; Walrasian general equilibrium) with two regions (China, Rest of World) and four economic entities (household, enterprise, government, foreign) is developed for 2015–2030. The Social Accounting Matrix is calibrated to the 2015 China Input-Output Table. Sectoral detail follows Guo et al. (2014), aggregating to 29 sectors after disaggregating the electricity sector into nine subsectors (following Lindner et al., 2013) to distinguish renewables (largely untaxed) from fossil-based generation. Production nests use a Leontief top level and CES lower levels; elasticity parameters are from Guo et al. (2014), with electricity subsectors inheriting the sector’s elasticities. Sensitivity analysis varies elasticities by ±10%, ±20%, ±50% to assess robustness. Income-expenditure: A representative household and enterprise are used initially; the household receives labor, capital, and transfer income and consumes domestic/imported goods; the enterprise earns capital income and pays wages, taxes (including carbon), and transfers. Government collects consumption, production, trade, and carbon taxes, and allocates spending across consumption, transfers, and savings via a CES utility. Trade follows the Armington assumption with balanced trade. Dynamics: Exogenous paths include population (UN WPP 2017, medium variant), prices (OECD 2014), energy consumption growth (EIA 2017), sectoral output growth (energy sectors track energy demand; other sectors track OECD GDP projection), capital accumulation (Long and Herrera, 2016), and human capital (CHLR 2018). Details are in Supplementary Information. Household disaggregation: The representative household is split into three income groups using CHIP 2013: low-income (bottom 40%), mid-income (middle 50%), high-income (top 10%). Shares of income sources (labor, capital, transfers) and consumption by commodity are assumed time-invariant (Tables 1–2). Group gross income and consumption are computed by applying these shares to factor and commodity flows. Distribution of climate damages: Household climate damage shares vary by the income elasticity of damage parameter ξ. Three cases are analyzed: positive ξ (damages proportional to income/consumption, high-income lose more), zero ξ (damages independent of income, proportional to population shares), and negative ξ (damages inversely related to income, low-income lose more). Scenarios: Twelve scenarios vary tax incidence and revenue recycling: recipients of tax revenues are governments, households evenly, low-income households only, or enterprises; tax payments are borne by high-income households only, proportional to income, or independent from income (12 combinations, SCRO1–SCRO12). Relative utility (RU): Group RU is defined following Johansson-Stenman et al. (2002) and Howarth and Kennedy (2016): HGRU = 1 − (HGANI/ANYT)^(−γ2), with γ1 the weight on relative income (assumed 0.35) and γ2 the rate at which RU falls as income rises (assumed 1.72). Overall RU is a population-weighted sum across groups. With γ2>1, RU values are negative and larger absolute values indicate more negative feelings about inequality for a given national income. Carbon tax design: Three ad valorem carbon taxes at fixed rates of 1%, 2%, and 3% are levied on the monetary value of non-renewable energy consumption, directly affecting fossil electricity generation and end-use fossil fuels; electricity end use is indirectly affected via pass-through pricing by enterprises. The implied rates correspond approximately to 2.4–4.9, 3.6–7.8, and 4.4–9.8 USD/tCO2 over 2015–2030, comparable to Chinese guidance levels. ETS interactions are not modeled; the current ETS price (~9 USD/tCO2) is noted for context.

- Baseline inequality: The Palma ratio in the baseline (no tax) under the zero-damage-correlation assumption is 2.63. - Climate damage distribution: A positive ξ (damages proportional to income) yields the lowest inequality; a negative ξ (damages inversely related to income) yields the highest. Across ξ assumptions, introducing carbon taxes reduces inequality (Palma ratio declines) because high-income households earn more from energy sectors and bear larger net income losses. - Relative utility and tax rate: Despite inequality reductions, higher tax rates decrease national income and increase the absolute value of RU (more negative feelings about inequality), indicating absolute income dominates RU relative to inequality effects. Under a fixed tax rate, RU absolute value is lowest with positive ξ and highest with negative ξ. - Tax payment incidence: When only high-income households pay the tax, inequality is lowest (progressive). When payments are proportional to income, impacts on inequality are minimal (neutral). When payments are independent of income, inequality is highest and exceeds the baseline (regressive). RU absolute values follow the same ordering: lowest when high-income households pay, higher when proportional, highest when independent. - Revenue recycling: Recycling to households lowers inequality more than recycling to governments or enterprises (similar to no-recycling). Targeting transfers to low-income households further reduces inequality relative to even household rebates. RU absolute values are lowest when revenues are recycled to low-income households, higher for even household recycling, and highest for enterprises or governments. - Robustness: Sensitivity tests varying CES elasticities by up to ±50% change the Palma ratio by less than 0.1% and RU by less than 10%, indicating result robustness.

The findings demonstrate that inequality impacts of carbon taxes hinge on three channels: (1) how climate damages are distributed across income groups (ξ), (2) who pays the tax, and (3) how revenues are recycled. A negative ξ exacerbates inequality, consistent with evidence that poorer populations are more exposed to climate damages. If tax payments are independent of income, carbon taxes are regressive, aligning with prior studies; conversely, concentrating tax burdens on high-income households makes taxes progressive. Even though carbon taxes can lower measured inequality, higher tax rates reduce national income and increase the absolute value of RU, implying absolute income exerts a stronger influence on subjective well-being than relative income differentials. Revenue recycling to households, especially targeting low-income households, reduces inequality and improves RU relative to recycling to governments or enterprises, corroborating literature on progressive recycling designs. Overall, the design of tax incidence and recycling can align emissions mitigation with equity and subjective welfare objectives.

The paper quantifies how the design of carbon taxes in China affects income inequality and relative utility using a dynamic CGE model with household disaggregation. Inequality outcomes are primarily shaped by (i) the correlation between income and climate damages, (ii) tax payment incidence, and (iii) revenue recycling. Policies that (a) place tax burdens on high-income households and (b) recycle revenues to households—preferably targeted to low-income groups—most effectively reduce inequality and mitigate negative effects on relative utility. Although carbon taxes can reduce measured inequality, declines in national income at higher tax rates raise the absolute magnitude of relative utility losses, indicating the dominant role of absolute income in subjective welfare. The study links climate-policy-induced inequality to behavioral measures of relative utility and provides policy-relevant guidance on revenue recycling to achieve progressive outcomes.

- Household heterogeneity is limited to three income groups based on CHIP 2013; intra-group variation and more granular distributions are not modeled. - Ratios of income sources and commodity consumption shares by household group are assumed time-invariant over 2015–2030. - Electricity subsector elasticities are assumed equal to the aggregate electricity sector elasticity from Guo et al. (2014). - Dynamic parameters (population, prices, energy and output growth, capital and human capital) are exogenous and drawn from external projections. - Climate damage distribution is stylized via three ξ cases (positive, zero, negative) rather than empirically estimated household-specific damages. - Only carbon taxes are modeled; interactions with China’s ETS and other policy instruments are excluded. - Relative utility parameters (γ1, γ2) are taken from experimental medians and held constant; alternative preference heterogeneity is not explored.