Welfare Economics and Climate Policy Credibility: Reading the 2025 National Academies and DOE Reports

What welfare economics demands that climate reports omit

Welfare economics provides a formal structure for comparing policies: specify a social welfare function, identify feasible actions, and characterize what is known about consequences under uncertainty [1]. When uncertainty prevents optimization, the standard fallback is expected welfare maximization; when subjective probabilities are not credible, criteria such as maximin or minimax regret apply uniformly over feasible unknown values [1]. Integrated assessment models operationalize this by coupling emissions to atmospheric concentrations and temperature, specifying a damage function, and specifying an abatement cost function [1]. The planning problem then becomes an optimal-control tradeoff: moderate policies have lower abatement costs but greater warming damages, while aggressive policies reverse that pattern [1].

The anchor paper's central claim is that this framework is absent from the 2025 NASEM and DOE reports [1]. The NASEM report's five conclusions are qualitative, and it does not discuss abatement costs at all, so beyond endorsing the binary EPA Endangerment Finding it provides no basis for evaluating alternative regulations [1]. The DOE report does discuss uncertainty quantitatively in places, including equilibrium climate sensitivity and sea level rise, and it states that excessively aggressive mitigation could prove more detrimental than beneficial [1]. But it then argues against vehicle emissions regulation on scale grounds without quantifying the tradeoff [1]. Both reports, in other words, assert conclusions that welfare economics would require them to demonstrate.

Dueling certitudes: how the two reports diverge and what they share

The NASEM and DOE reports disagree sharply on empirical claims. The DOE report states that elevated CO2 enhances plant growth and agricultural productivity, while NASEM reports negative climate-driven impacts on crops and livestock [1]. The DOE report claims mainstream research overestimates emissions and warming trends, while NASEM states that CO2 emissions estimates have the lowest uncertainties because they are based on accurately tracked fuel consumption data [1]. On extreme weather, the DOE report finds no long-term trends in most U.S. events, while NASEM reports increases in hot extremes and extreme single-day precipitation [1]. These are not differences in values but in factual characterization.

The anchor paper's diagnostic contribution is to show that despite this dissonance, both reports share a methodological flaw: neither quantifies the opposing tendencies that welfare economics identifies as central [1]. Mainstream reports emphasize potential damages while downplaying abatement costs; contrarian reports do the opposite [1]. The DOE report's own concluding thoughts call for weighing risks and benefits and acknowledging uncertainties, yet the report's anti-regulatory conclusion does not follow from a quantified comparison [1]. The NASEM report mentions uncertainty but almost always in a binary manner, and sustained quantification appears only in an appendix that is a 2009 EPA document rather than a product of the 2025 committee [1].

The prior frontier: what integrated assessment had already established

The anchor paper builds on a lineage of integrated assessment work that had already formalized climate policy tradeoffs. Manski, Sanstad, and DeCanio (2021) argued that multi-model ensemble analysis be abandoned and proposed framing climate model uncertainty as partial identification, applying minimax regret without weighting climate model forecasts [1]. DeCanio, Manski, and Sanstad (2022) generalized that analysis to include uncertainty about the correct climate model and the appropriate intergenerational discount rate, showing that low discount rates favor aggressive and rapid emissions reduction while high rates favor modest and slow action [1]. These papers established that the tradeoff between abatement costs and climate damages can be evaluated coherently even under deep uncertainty [1].

Competing and complementary evidence extends the evaluative space in directions the anchor paper acknowledges but does not fully incorporate. Young-Brun et al. (2025) develop a global integrated assessment model with subnational income distribution and find that carbon taxation with revenue recycling can improve global welfare and reduce inequality, with a Loss and Damage policy requiring only about 15% of global carbon tax revenues by 2050 to compensate low-income countries [7]. Their results show that welfare evaluation depends on distributional assumptions that the anchor paper's framework notes as a limitation of conventional integrated assessment modeling [1][7]. Welgryn (2026) compares carbon contracts for difference and finds that once earmarking and fiscal constraints are included, the state's exposure to risk becomes an essential determinant of policy efficiency and welfare, with the German scheme's budget driven mostly by risk provisions [2]. This demonstrates that welfare evaluation of specific policy instruments can overturn rankings that ignore public risk [2].

Platz (2025) offers a competing perspective on cost-benefit analysis in climate policy, though the supplied material is limited to a fragment noting a critique of Lomborg's use of cost-benefit analysis [4]. The National Academies' own 2016 report on updating the social cost of carbon provides validation that the Academies have previously engaged with welfare-economic quantification, making the 2025 report's qualitative framing a departure rather than an institutional norm [5]. Lait et al. (2025) examine strategies for assessing co-benefits in energy and climate policy and compare quantitative and qualitative approaches, highlighting that the choice between them carries strengths and limitations [6]. This is directly relevant to the anchor paper's critique: the NASEM and DOE reports chose qualitative comparison, and Lait et al.'s framework would predict that this choice limits the policy guidance the reports can provide [6].

Boundaries of the diagnosis and open questions for credible discourse

The anchor paper's evidence boundary is explicit: it uses two 2025 reports as case studies and does not claim that all climate policy analysis suffers from the same failure [1]. The diagnosis is also limited by what the reports were asked to do. The NASEM Statement of Task requested a binary yes/no conclusion on whether climate change endangers public health and welfare, not an assessment of severity or a comparison of regulatory alternatives [1]. The DOE report was commissioned by the Secretary of Energy with the stated objective of encouraging a more thoughtful and science-based conversation [1]. Neither report was tasked with welfare-economic evaluation, which means the absence of tradeoff analysis may reflect institutional design rather than analytical failure alone.

The broader research frontier suggests that welfare-economic evaluation of climate policy is feasible and informative when applied. Choi et al. (2026) use an integrated assessment model to show that Korea can achieve a 58.9% emissions reduction by 2035 through coordinated sectoral policies, with coal phase-out timing as a critical determinant [3]. Their scenario design explicitly embeds policy instruments and feasibility constraints, demonstrating that quantitative policy evaluation can be policy-grounded rather than abstract [3]. Aksoy and Arlı (2025) estimate a social cost of carbon of $123.25 per ton for E7 economies, significantly higher than conventional estimates due to demographic trends and structural constraints [8]. Estrada et al. (2025) estimate the social cost of carbon at $187 per ton globally, rising to $490 when urban heat island warming is included, with urban areas representing 78–93% of the global SCC [13]. These estimates show that the quantitative inputs to welfare analysis are available and consequential.

Other work extends the tradeoff space in ways that the anchor paper's framework does not fully capture. Younis et al. (2025) project Canadian water use under decarbonization scenarios and find potential tradeoffs and synergies with water resources at provincial and river-basin scales [9]. Li et al. (2026) find that lower-consumption futures are associated with substantially lower warming, 0.3–1.1°C across carbon price trajectories, and that global income convergence has a small influence on temperature outcomes [10]. Zheng et al. (2026) develop a minimax-regret framework for solar radiation modification and find that policies designed for adverse cooling-response states avoid temperature-target failure but generate welfare losses from over-deployment [11]. Schreyer et al. (2025) show that a full EU fossil phase-out requires marginal abatement costs rising from 460 to 630 EUR per ton CO2 compared to a least-cost net-zero scenario [12]. These studies demonstrate that welfare-relevant tradeoffs extend beyond the abatement-cost-versus-damage framework to include water, consumption, equity, and technology-specific risks.

The open question is whether the anchor paper's proposed remedy—using welfare economics to recognize tradeoffs and face up to uncertainty—can bridge the gap between mainstream and contrarian positions [1]. The paper argues it should enable more productive communication, but it does not claim it will eliminate scientific disagreements or differences in societal values [1]. The DOE report's own concluding passage calls for weighing risks and benefits and acknowledging uncertainties, suggesting that the rhetorical materials for a welfare-economic approach are present even in contrarian reports [1]. Whether that potential can be realized depends on whether institutions commission and publish analyses that quantify tradeoffs rather than assert conclusions.

About These Sources

This research page is built on 13 studies (12 peer-reviewed, 1 preprint) — published from 2016 to 2026, 12 from 2024 or later — selected as the most relevant from 13 studies that passed quality screening, drawn from 56 papers retrieved from a database of over 500 million.

Sources used in this answer