[Paper Review] Modelling Distributional Impacts of Carbon Taxation: a Systematic Review and Meta-Analysis

Jules Linden, Cathal O'Donoghue+1 more|arXiv (Cornell University)|Jan 12, 2026

Climate Change Policy and Economics0 citations

TL;DR

A systematic review and meta-analysis of microsimulation studies on carbon taxation, showing how modelling choices influence estimated distributional impacts and regressive vs. progressive outcomes across countries.

ABSTRACT

Carbon taxes are increasingly popular among policymakers but remain politically contentious. A key challenge relates to their distributional impacts; the extent to which tax burdens differ across population groups. As a response, a growing number of studies analyse their distributional impact ex-ante, commonly relying on microsimulation models. However, distributional impact estimates differ across models due to differences in simulated tax designs, assumptions, modelled components, data sources, and outcome metrics. This study comprehensively reviews methodological choices made in constructing microsimulation models designed to simulate the impacts of carbon taxation and discusses how these choices affect the interpretation of results. It conducts a meta-analysis to assess the influence of modelling choices on distributional impact estimates by estimating a probit model on a sample of 217 estimates across 71 countries. The literature review highlights substantial diversity in modelling choices, with no standard practice emerging. The meta-analysis shows that studies modelling carbon taxes on imported emissions are significantly less likely to find regressive results, while indirect emission coverage has ambiguous effects on regressivity, suggesting that a carbon border adjustment mechanism may reduce carbon tax regressivity. Further, we find that estimates using older datasets, using explicit tax progressivity or income inequality measures, and accounting for household behaviour are associated with a lower likelihood of finding regressive estimates, while the inclusion of general equilibrium effects increases this likelihood.

Motivation & Objective

• Assess how microsimulation models are built to simulate carbon taxation on households and individuals.
• Identify key modelling choices that drive differences in distributional impact estimates.
• Quantify how implementation decisions affect the likelihood of finding regressive versus progressive impacts across countries.
• Discuss implications for interpreting results and for designing carbon tax policies with distributional concerns.

Proposed method

• Systematic review following PRISMA to select bottom-up microsimulation studies on carbon taxation.
• Categorization of modelling choices into data, tax coverage, and behavioural/GE integration.
• Meta-analysis using a probit model on 217 estimates from 71 countries to assess the influence of modelling choices on regressive outcomes.
• Distinction between conceptual modelling choices (e.g., indirect emissions) and implementation choices (e.g., IO database, multi-regional vs single-region).
• Comparison with prior literature to isolate the effects of specific modelling decisions on distributional results.

Figure 1: Modelling steps and choices.

Experimental results

Research questions

• RQ1 How do different microsimulation modelling choices affect the estimated distributional impact of carbon taxes?
• RQ2 Do implementation choices (e.g., indirect emissions modeling, IO data, regional scope) systematically alter the likelihood that carbon taxes appear regressive or progressive?
• RQ3 What role do tax coverage design and revenue recycling play in distributional outcomes?
• RQ4 Does including general equilibrium effects change the estimated regressivity when accounting for data sources and coverage of imported emissions?
• RQ5 What are the cross-country patterns in regressive versus progressive findings in relation to modelling approaches?

Key findings

• There is substantial diversity in modelling choices with no standard practice emerging.
• Modeling imported emissions and broader indirect emissions can affect the likelihood of progressive results, with imported-emissions focus linked to less regressive findings.
• Older datasets, explicit tax progressivity or income-inequality measures, and incorporating household behaviour are associated with a higher likelihood of progressive estimates.
• Inclusion of general equilibrium effects increases the likelihood of regressive outcomes when accounting for IO database and imported-emissions coverage.
• Using MRIO/EE-IO approaches and broader emission coverage tends to influence distributional results differently than single-region or direct-emission-only models.
• The meta-analysis shows that implementation-related predictors substantially increase explanatory power for regressive vs. progressive findings (up to about 55% improvement in the probit model).

Figure 2: Comparison of carbon tax burdens by welfare concept and ranking variable for Hungary.