Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

Using Machine Learning to Compute Constrained Optimal Carbon Tax Rules

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A simple linear carbon tax plus intergenerational transfers is shown to be Pareto-improving and to capture most of the available welfare gain in a climate OLG model.

desk verdict Useful framework with a real methodological trick, but the Pareto-improvement headline needs fresh-model validation before it can be trusted. read the letter →

arxiv 2507.01704 v1 pith:CKWAMZE5 submitted 2025-07-02 econ.GN q-fin.EC

Felix Kübler , Simon Scheidegger , Oliver Surbek This is my paper
classification econ.GNq-fin.EC
keywords climatepolicycarbontaxationoverlappinggenerationsParetoimprovementoptimalGaussianprocessesdeepequilibriumnetsRamsey
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper sets out to show that a politically usable carbon policy can be found by systematic search rather than by guesswork: in a stochastic overlapping-generations model with climate risk, a simple tax that is linear in cumulative emissions, paired with an intergenerational transfer scheme, can be made Pareto-improving and delivers a 0.42% aggregate welfare gain in consumption-equivalent terms. The paper also claims that enriching the tax base with carbon intensity and tipping-point distance raises the gain only to 0.45%, so complexity buys little in this calibration. The point of the exercise is to demonstrate a computational route to welfare-improving taxes that no current generation has to oppose, and to offer that route as a template for other heterogeneous-agent policy problems.

What carries the argument

The method chains three surrogates. First, Deep Equilibrium Nets — neural networks trained so that the model's first-order equilibrium conditions hold along simulated paths — solve the OLG economy for a continuum of tax-and-transfer rules by treating the rule coefficients as pseudo-state variables. Second, a vector-valued Gaussian-process surrogate (a statistical emulator with closed-form predictive means and variances) is fitted to each cohort's expected utility as a function of those coefficients, using roughly 500–800 costly simulations selected partly by Bayesian active learning. Third, a constrained optimizer maximizes the social-welfare surrogate subject to per-cohort constraints that every generation be no worse off than in business as usual. This machinery is what makes the search over many candidate policies cheap enough to find a Pareto optimum.

What would settle it

Re-run 10,000 fresh simulations of the full DEQN equilibrium at the optimal parameters of Sections 5.3 and 5.4, computing each of the 40 cohorts' consumption-equivalent gain relative to business as usual; the central claim fails if any cohort is worse off or if the aggregate gain measured from these fresh simulations falls short of 0.42% (or 0.45% for the richer rule). The claim also fails if the GP's posterior standard deviation at the optimum is on the order of the 0.03 percentage point gap between the two policies.

Watch

Extended reading notes

Core claim

Within the calibrated 12-period stochastic OLG model, the unconstrained welfare-maximizing linear cumulative-emissions tax stabilizes mean warming at about 2.7°C but costs the initial cohorts up to 5% in consumption equivalents. Adding an optimized 12-cohort revenue-sharing rule and imposing that every cohort be at least as well off as under business-as-usual turns this into a Pareto-improving policy: the aggregate gain is 0.42%, the oldest cohorts stay at their BAU utility, and later cohorts gain up to about 1.4%. Extending the tax to be linear in carbon intensity and in distance to a climate tipping point yields 0.45%, a gain of only about 7% over the simpler scheme. The paper reads this as evidence that the transfer design is at least as important as the tax base, and that simple observable-based rules capture most of the feasible welfare improvement.

Load-bearing premise

The Pareto-improvement guarantee rests on the accuracy of the surrogate models at the final policy, because the per-generation constraints are checked on Gaussian-process predictions rather than on fresh full-model simulations.

Editorial extensions

If this is right

  • A Pareto-improving carbon policy exists in the model: a tax linear in cumulative emissions plus a revenue-sharing rule that keeps all generations at or above their business-as-usual utility.
  • The aggregate welfare gain of this policy is 0.42% in consumption-equivalent terms, and adding carbon-intensity and tipping-point tax terms raises it only to 0.45%.
  • The Pareto-improving policy mainly truncates the upper tail of damages, reducing the 99th percentile of GDP damages from about 9% to about 7%, rather than shifting the mean path dramatically.
  • The computational pipeline—DEQN global solution, GP welfare surrogates, surrogate-based constrained optimization—runs in hours on a laptop and is presented as a template for other stochastic heterogeneous-agent Ramsey problems.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The 0.42%-versus-0.45% comparison is a point estimate from one calibration; the same pipeline could be re-run with different damage curvature or discounting to see whether complexity becomes valuable when tail risk is more severe.
  • The fixed, state-independent transfer shares in Sections 5.3 and 5.4 may understate what a fully state-dependent transfer scheme could achieve; testing time-varying or temperature-contingent transfers is a direct extension the paper does not pursue.
  • The GP's posterior variance could itself be used to design a robustly Pareto-improving policy that satisfies constraints under worst-case surrogate error; the paper optimizes the posterior mean only.
  • Beyond climate, any heterogeneous-agent economy with a slow aggregate state and a welfare-relevant externality—such as unemployment risk, health shocks, or financial frictions—fits the same three-step search for constrained optimal simple rules.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes a three-step computational method for constrained optimal carbon-tax design in a stochastic OLG model. Step 1 embeds tax and transfer parameters as pseudo-states in a Deep Equilibrium Net (DEQN) that approximates equilibrium policy functions; Step 2 fits Gaussian-process (GP) surrogates to cohort-level expected utilities under the parametrized policies; Step 3 optimizes a social-welfare objective subject to Pareto-improvement and tax-bounding constraints, using the GP predictive means. The application to a 12-period OLG model with stochastic carbon intensity and tipping points produces a BAU path with mean warming near 3°C, a welfare-maximizing linear cumulative-emissions tax that raises aggregate welfare by 1.6% while hurting initial cohorts by up to 5%, and a Pareto-improving tax-and-transfer rule with a 0.42% aggregate consumption-equivalent gain; extending the tax base to carbon intensity and tipping-point distance raises the gain to 0.45%.

Significance. If the numerical results are correct, the paper makes a useful methodological contribution: the pseudo-state DEQN approach solves the model across a continuum of policies in one pass, and GP-based Bayesian active learning makes the constrained optimization tractable. The headline policy conclusion — that a simple observable-based cumulative-emissions tax with revenue sharing captures most of the available welfare gain — is economically interesting and testable. The paper is unusually transparent about algorithm details, hyperparameters, and accuracy metrics, and it states the Pareto constraints precisely in Eqs. (54) and (59). The main caveat is that the Pareto-improvement guarantee and the 0.42% versus 0.45% comparison are certified only on surrogate predictions, not on fresh full-model simulations; this determines the revision.

major comments (3)
  1. [Section 5.3, Eq. (54); Section 5.4, Eq. (59); Section 4.2.3] The Pareto constraints are enforced on the GP posterior means μ_{*,t}(ϑ) computed from DEQN-based simulations, and Step 3 explicitly optimizes the surrogate 'without the need for further model simulations.' The paper reports no independent, fresh full-model evaluation at the optimal ϑ* in Tables 5 and 8, so the central claim that no cohort is made worse off is not verified. This is load-bearing because the two headline welfare gains are 0.42% and 0.45% — a 0.03 percentage-point difference — while the DEQN value-function errors in Tables 2–4 and 7 are of order 10^−3 in normalized units; after the consumption-equivalent transform, surrogate and policy-function errors of this size could affect both feasibility and the ordering. I ask the authors to add a validation step: simulate fresh Monte Carlo paths at the reported ϑ* using the trained DEQN policy, report per-cohort CE gains with Monte Carlo standard errors, and check that all 40 constraints hold. This is the minimal additional evidence needed to support the Pareto-improvement claim.
  2. [Appendix B.2, Table 14] The Dirichlet concentration parameters α used to sample transfer shares are described as 'informed by preliminary optimization runs to focus the search on regions of the policy space likely to yield Pareto improvements.' Because these parameters determine the density of training points for both the DEQN and the GP surrogates, they can influence which regions are approximated accurately and hence which optimum is found. The paper asserts that this does not bias the optimum, but no supporting test is given. I recommend a sensitivity check: re-run the GP construction and optimization with a different α (for example, a uniform simplex design augmented by validation points) and show that the resulting ϑ* and welfare gains are unchanged.
  3. [Sections 5.3 and 5.4, Eqs. (54) and (59)] The constrained optimization uses only the GP predictive mean in the 40 per-cohort constraints, with no uncertainty penalty or constraint tightening. Because Figs. 5b and 6b show early cohorts with CE gains near zero, the reported solution lies at or very close to the constraint boundary, so a small GP misspecification there could place the true optimum in the infeasible region. I ask the authors either to impose a chance constraint that uses σ_{*,t}(ϑ), or to demonstrate through the fresh simulation requested above that the constraint slack at ϑ* is positive for all cohorts.
minor comments (5)
  1. [Section 4.1, after Eq. (36)] The text reads 'an variant of stochastic gradient descent'; this should be 'a variant'.
  2. [Section 3, Table 1 and paragraph on damages] The damage-function parameter ψ1 is listed as 13.16 in Table 1 but the text says 'ψ1 = 13.17'; please reconcile the two values.
  3. [Section 5.2, after Eq. (49)] The sentence reporting the leave-one-out cross-validation error is grammatically incomplete as printed ('The resulting leave-one-out cross-validation error, which is an excellent 3.2 · 10−5.'). Please rewrite it and also report the leave-one-out errors for the 40-GP Pareto surrogates used in Sections 5.3 and 5.4.
  4. [Section 2.4 and Eq. (63)] The conversion factor C2CO2 is given as 3.666; the correct CO2-to-C mass ratio is 44/12 ≈ 3.667. Please correct this and use consistent CO2 notation.
  5. [Section 4.2.3, Eq. (113)] The symbol αUCB is used both for the acquisition function and for the exploitation weight in the UCB rule; please rename one of the two to avoid ambiguity.

Circularity Check

0 steps flagged · score 1.0 of 10

No meaningful circularity: the headline 0.42% vs 0.45% welfare gains are simulation outputs from the calibrated model, not read back from the fitted surrogate or from a self-cited theorem.

full rationale

The paper's derivation chain is: (i) solve the stochastic OLG model with DEQN, reporting Euler and value-function errors; (ii) fit GP surrogates to cohort utilities from DEQN simulations; (iii) optimize the surrogate subject to Pareto constraints; (iv) simulate the economy at the resulting policy parameters and report welfare gains. The final gains in Table 10 follow from those model simulations, so they are not the GP objective evaluated at the optimum by construction. The Pareto constraints are imposed on GP predictive means, but the paper then re-simulates outcomes, which makes surrogate accuracy a numerical-validation concern rather than a circular reduction. Self-citations for DEQN and for the stochastic-tipping damage specification are methodological or modeling choices with independent content; no load-bearing conclusion rests on an unverified self-cited uniqueness theorem. The only mild self-referential element is that the BAU baseline and the policy counterfactual use the same DEQN approximation, which could bias small differences (0.42 vs 0.45 percentage points), but that is a correctness risk, not circularity. Calibration targets (RCP4.5 emissions and roughly 3 degrees C BAU warming) are reported as BAU properties, but they are disclosed calibration choices, not the paper's central claim.

Assumptions & free parameters 6 free parameters · 7 assumptions · 0 invented entities

The model is a calibrated numerical laboratory rather than a closed-form derivation. The central outputs depend on standard preference and production parameters, a linear carbon-climate response, a damage function with stochastic tipping, and on numerical choices such as DEQN loss thresholds, tax-domain bounds, and Dirichlet concentration parameters. No new physical entities are postulated.

free parameters (6)
  • Carbon intensity AR(1) parameters (rho_0, rho_inf, delta_rho) = rho_0=1.08, rho_inf=0.91, delta_rho=0.04
    Calibrated in Section 3 so that BAU emissions approximately follow RCP4.5 and mean warming is about 3C; the targeted outcomes therefore cannot be treated as independent predictions.
  • Damage function curvature psi_1 = 13.16 in Table 1, 13.17 in text
    Chosen to match damage assumptions following Kotlikoff et al. (2021b); the small inconsistency between Table 1 and Section 3 is a documentation issue.
  • Abatement cost parameters theta_1, theta_2 = theta_1=0.7, theta_2=2.6
    Taken from Nordhaus (2017); theta_1 is deliberately set high so that full abatement is never optimal, which shapes the tax results.
  • Climate sensitivity sigma_CCR = 1.7
    Exogenous climate parameter from the literature; temperature is linear in cumulative emissions by Eq. (11).
  • Dirichlet concentration parameters alpha for transfer sampling = Table 14 vector
    Chosen in Appendix B.2 from preliminary optimization runs to focus the search on Pareto-improving regions; the reported optimum may depend on this choice.
  • Welfare weights gamma_t = 0.025 for t=-10,...,29
    Uniform weights assumed for all cohorts; aggregate welfare gain would change under alternative weights.
assumptions (7)
  • domain assumption Cumulative emissions map linearly to temperature, T_AT,t = sigma_CCR E_t (Eq. 11).
    This linear carbon-climate response is taken from Matthews et al. (2009) and Dietz and Venmans (2019), not derived; it omits other climate feedbacks.
  • domain assumption Climate damages follow Weitzman (2012) with stochastic tipping (Eq. 3), with tipping point dynamics specified by Eqs. (23)-(24).
    The damage function shape and the tipping process are imposed model inputs, not outcomes of the analysis.
  • domain assumption Households maximize expected CRRA utility over a fixed 12-period life, with no lifetime uncertainty and no intra-cohort heterogeneity (Section 2.2).
    This standard OLG structure defines the welfare comparisons; it abstracts from within-cohort inequality and mortality risk.
  • domain assumption Pareto improvement is defined as time-zero expected utility per cohort (Eq. 40), not ex-post state-by-state welfare.
    This weaker notion of Pareto improvement is used for the policy guarantee; individual realized states can still be worse off.
  • ad hoc to paper DEQN training converges to an approximate equilibrium with loss threshold 1e-6, and Euler/value function errors of order 1e-3 are treated as negligible for welfare comparisons (Tables 2-4, 7).
    The numerical tolerance is chosen by the authors and is load-bearing for comparing welfare gains of a few tenths of a percent.
  • ad hoc to paper Heuristic tax-domain bounds informed by BAU simulations adequately restrict the search without excluding optima (Appendix B.1).
    The bounds in Table 13 are derived from BAU scenario states and are validated only post-hoc, not proven.
  • ad hoc to paper Dirichlet concentration parameters informed by preliminary runs focus sampling on Pareto-improving regions without biasing the optimum (Appendix B.2).
    The concentration parameters alpha are chosen from earlier optimization runs, so the final sampling distribution is itself an artifact of the search process.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Using Machine Learning to Compute Constrained Optimal Carbon Tax Rules." pith.science (2026). https://pith.science/paper/CKWAMZE5

@misc{pith2026250701704,
  author       = {Pith},
  title        = {Pith review of: Using Machine Learning to Compute Constrained Optimal Carbon Tax Rules},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CKWAMZE5}},
  note         = {Machine review of arXiv:2507.01704}
}
read the original abstract

We develop a computational framework for deriving Pareto-improving and constrained optimal carbon tax rules in a stochastic overlapping generations (OLG) model with climate change. By integrating Deep Equilibrium Networks for fast policy evaluation and Gaussian process surrogate modeling with Bayesian active learning, the framework systematically locates optimal carbon tax schedules for heterogeneous agents exposed to climate risk. We apply our method to a 12-period OLG model in which exogenous shocks affect the carbon intensity of energy production, as well as the damage function. Constrained optimal carbon taxes consist of tax rates that are simple functions of observables and revenue-sharing rules that guarantee that the introduction of the taxes is Pareto improving. This reveals that a straightforward policy is highly effective: a Pareto-improving linear tax on cumulative emissions alone yields a 0.42% aggregate welfare gain in consumption-equivalent terms while adding further complexity to the tax provides only a marginal increase to 0.45%. The application demonstrates that the proposed approach produces scalable tools for macro-policy design in complex stochastic settings. Beyond climate economics, the framework offers a template for systematically analyzing welfare-improving policies in various heterogeneous-agent problems.

Figures

Figures reproduced from arXiv: 2507.01704 by the authors.

Figure 1
Figure 1. provides a schematic overview of the three-step procedure outlined above. Step 1: DEQN Step 2: GP(θ) Step 3: Optimizer P(xstate,t, ϑ) QoI( y ϑ) arg max QoI( y ϑ) [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗
Figure 2
Figure 2. Projected Climate and Economic Outcomes under the BAU Scenario. [PITH_FULL_IMAGE:figures/full_fig_p017_2.png] view at source ↗
Figure 3
Figure 3. Contour plot of the welfare surrogate. Given this accurate surrogate, the third step of our method is to find the tax parameters by maximizing it, and thus, the SWF. To ensure the optimizer does not search outside the surrogate (that is, does not extrapolate), we enforce a set of linear constraints on the tax rate [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Projected Climate and Economic Outcomes under the Optimal Tax on Cumu [PITH_FULL_IMAGE:figures/full_fig_p020_4.png]
Figure 5
Figure 5. Figure 5: Projected Climate and Economic Outcomes under the Optimal Tax on Cumu [PITH_FULL_IMAGE:figures/full_fig_p023_5.png]
Figure 6
Figure 6. Figure 6: Projected Climate and Economic Outcomes under the Optimal Linear Tax on [PITH_FULL_IMAGE:figures/full_fig_p026_6.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Equilibrium World Models

    econ.GN 2026-06 unverdicted novelty 7.0 of 10

    Equilibrium World Models are a deep-learning solver that enforces exact equilibrium conditions on broad model-generated state distributions to globally solve dynamic stochastic models featuring rare disasters, binding...

  2. Bayesian Optimization on the Equilibrium Manifold

    econ.GN 2026-06 unverdicted novelty 6.0 of 10

    Bayesian optimization on low-dimensional Negishi-weight equilibrium manifolds enables reliable computation and probabilistic certification of optimal carbon taxes in a calibrated heterogeneous-agent climate economy.

Reference graph

Works this paper leans on

48 extracted references · 40 canonical work pages · cited by 2 Pith papers

  1. [1]

    DEEP EQUILIBRIUM NETS

    Marlon Azinovic, Luca Gaegauf, and Simon Scheidegger. DEEP EQUILIBRIUM NETS . International Economic Review, 63 0 (4): 0 1471--1525, November 2022. ISSN 0020-6598, 1468-2354. doi:10.1111/iere.12575

    work page doi:10.1111/iere.12575 2022

  2. [2]

    Pricing uncertainty induced by climate change

    Michael Barnett, William Brock, and Lars Peter Hansen. Pricing uncertainty induced by climate change . Review of Financial Studies, 33 0 (3): 0 1024--1066, 2020. ISSN 14657368. doi:10.1093/rfs/hhz144

    doi:10.1093/rfs/hhz144 2020

  3. [3]

    R. Bellman. Adaptive Control Processes: A Guided Tour. 'Rand Corporation. Research studies. Princeton University Press, 1961

    work page 1961

  4. [4]

    Simulating russia's and other large economies' challenging and interconnected transitions

    Seth G Benzell, Eugene Goryunov, Maria Kazakova, Laurence J Kotlikoff, Guillermo LaGarda, Kristina Nesterova, and Andrey Zubarev. Simulating russia's and other large economies' challenging and interconnected transitions. Technical report, National Bureau of Economic Research, 2015

    work page 2015

  5. [5]

    Taxes, debts, and redistributions with aggregate shocks

    Anmol Bhandari, David Evans, Mikhail Golosov, and Thomas J Sargent. Taxes, debts, and redistributions with aggregate shocks. Technical report, National Bureau of Economic Research, 2013

    work page 2013

  6. [6]

    Computing Equilibria in Dynamic Stochastic Macro-Models with Heterogeneous Agents, volume 2 of Econometric Society Monographs, pages 185--230

    Johannes Brumm, Felix Kubler, and Simon Scheidegger. Computing Equilibria in Dynamic Stochastic Macro-Models with Heterogeneous Agents, volume 2 of Econometric Society Monographs, pages 185--230. Cambridge University Press, 2017

    work page 2017

  7. [7]

    Peter S. Burton. Intertemporal preferences and intergenerational equity considerations in optimal resource harvesting . Journal of Environmental Economics and Management, 24 0 (2): 0 119--132, 1993. ISSN 0095-0696. doi:10.1006/jeem.1993.1008. URL http://www.sciencedirect.com/science/article/B6WJ6-45P13FR-1D/2/3df3a733c83612ef9a6056ec25a0c47c

    arXiv 1993

  8. [8]

    The social cost of carbon with economic and climate risks

    Yongyang Cai and Thomas Lontzek. The social cost of carbon with economic and climate risks. Journal of Political Economy, 2018. doi:10.1086/701890. URL https://doi.org/10.1086/701890

    work page doi:10.1086/701890 2018

Show all 48 references
  1. [9]

    Deep Surrogates for Finance : With an Application to Option Pricing

    Hui Chen, Antoine Didisheim, and Simon Scheidegger. Deep Surrogates for Finance : With an Application to Option Pricing . Journal of Financial Economics, 2025. doi:10.2139/ssrn.3782722. Forthcoming

    2025 doi

  2. [10]

    Chapter 1 - introduction to integrated assessment modeling of climate change

    Simon Dietz. Chapter 1 - introduction to integrated assessment modeling of climate change. volume 1 of Handbook of the Economics of Climate Change, pages 1--51. North-Holland, 2024. doi:https://doi.org/10.1016/bs.hesecc.2024.10.002. URL https://www.sciencedirect.com/science/ar...

    2024 doi

  3. [11]

    Simon Dietz and Geir B. Asheim. Climate policy under sustainable discounted utilitarianism. Journal of Environmental Economics and Management, 63 0 (3): 0 321--335, 2012. ISSN 0095-0696. doi:https://doi.org/10.1016/j.jeem.2012.01.003. URL https://www.sciencedirect.com/science/...

    2012 doi

  4. [12]

    Cumulative carbon emissions and economic policy: in search of general principles

    Simon Dietz and Frank Venmans. Cumulative carbon emissions and economic policy: in search of general principles. Journal of Environmental Economics and Management, 96: 0 108--129, 2019

    2019

  5. [13]

    Are economists getting climate dynamics right and does it matter? Journal of the Association of Environmental and Resource Economists, 8 0 (5): 0 895--921, 2021

    Simon Dietz, Frederick van der Ploeg, Armon Rezai, and Frank Venmans. Are economists getting climate dynamics right and does it matter? Journal of the Association of Environmental and Resource Economists, 8 0 (5): 0 895--921, 2021. doi:10.1086/713977. URL https://doi.org/10.10...

    2021 doi

  6. [14]

    Optimal fiscal policy in a climate-economy model with heterogeneous households

    Thomas Douenne, Albert Jan Hummel, and Marcelo Pedroni. Optimal fiscal policy in a climate-economy model with heterogeneous households. CEPR Discussion Paper 19151, 2024. URL https://cepr.org/publications/dp19151

    2024

  7. [15]

    Optimal fiscal policy in a model with uninsurable idiosyncratic income risk

    Sebastian Dyrda and Marcelo Pedroni. Optimal fiscal policy in a model with uninsurable idiosyncratic income risk. The Review of Economic Studies, 90 0 (2): 0 744--780, 2023

    2023

  8. [16]

    Building interpretable climate emulators for economics

    Aryan Eftekhari, Doris Folini, Aleksandra Friedl, Felix K \"u bler, Simon Scheidegger, and Olaf Schenk. Building interpretable climate emulators for economics. arXiv preprint arXiv:2411.10768, 2024

    2024 arXiv

  9. [17]

    Neural architecture search: A survey

    Thomas Elsken, Jan Hendrik Metzen, and Frank Hutter. Neural architecture search: A survey. Journal of Machine Learning Research, 20 0 (55): 0 1--21, 2019

    2019

  10. [18]

    Optimal taxation with incomplete markets--an exploration via reinforcement learning

    Zhigang Feng, Jiequn Han, and Shenghao Zhu. Optimal taxation with incomplete markets--an exploration via reinforcement learning. Available at SSRN 4758552, 2024

    2024

  11. [19]

    Taming the curse of dimensionality: quantitative economics with deep learning

    Jes \'u s Fern \'a ndez-Villaverde, Galo Nu \ n o, and Jesse Perla. Taming the curse of dimensionality: quantitative economics with deep learning. Technical report, National Bureau of Economic Research, 2024

    2024

  12. [20]

    Gillingham, and Simon Scheidegger

    Jesús Fernández-Villaverde, Kenneth T. Gillingham, and Simon Scheidegger. Climate change through the lens of macroeconomic modeling. Annual Review of Economics, 2025. ISSN 1941-1383. doi:https://doi.org/10.1146/annurev-economics-091124-045357. URL https://www.annualreviews.org...

    2025 doi

  13. [21]

    The Climate in Climate Economics

    Doris Folini, Aleksandra Friedl, Felix Kübler, and Simon Scheidegger. The Climate in Climate Economics . The Review of Economic Studies, forthcoming, 2024

    2024

  14. [22]

    Deep uncertainty quantification: With an application to integrated assessment models

    Aleksandra Friedl, Felix K \"u bler, Simon Scheidegger, and Takafumi Usui. Deep uncertainty quantification: With an application to integrated assessment models. 2023

    2023

  15. [23]

    Optimal taxes on fossil fuel in general equilibrium

    Mikhail Golosov, John Hassler, Per Krusell, and Aleh Tsyvinski. Optimal taxes on fossil fuel in general equilibrium. Econometrica, 82 0 (1): 0 41--88, 2014

    2014

  16. [24]

    Social security and risk sharing

    Piero Gottardi and Felix Kubler. Social security and risk sharing. Journal of Economic Theory, 146 0 (3): 0 1078--1106, 2011

    2011

  17. [25]

    Uncertainty quantification and global sensitivity analysis for economic models

    Daniel Harenberg, Stefano Marelli, Bruno Sudret, and Viktor Winschel. Uncertainty quantification and global sensitivity analysis for economic models. Quantitative Economics, 10 0 (1): 0 1--41, 2019. doi:https://doi.org/10.3982/QE866. URL https://onlinelibrary.wiley.com/doi/abs...

    2019 doi

  18. [26]

    Heijdra, Jan Peter Kooiman, and Jenny E

    Ben J. Heijdra, Jan Peter Kooiman, and Jenny E. Ligthart. Environmental quality, the macroeconomy, and intergenerational distribution . Resource and Energy Economics, 28 0 (1): 0 74--104, 2006. ISSN 09287655. doi:10.1016/j.reseneeco.2005.05.001

    2006 doi

  19. [27]

    Optimal Development Policies With Financial Frictions

    Oleg Itskhoki and Benjamin Moll. Optimal Development Policies With Financial Frictions . Econometrica, 87 0 (1): 0 139--173, 2019. ISSN 0012-9682. doi:10.3982/ECTA13761

    2019 doi

  20. [28]

    The political economy of environmental policy with overlapping generations

    Larry Karp and Armon Rezai. The political economy of environmental policy with overlapping generations . International Economy Review, 55 0 (3): 0 711--733, 2014

    2014

  21. [29]

    Selfish incentives for climate policy: Empower the young! Journal of the Association of Environmental and Resource Economists, 11 0 (5): 0 1165--1200, 2024

    Larry Karp, Alesandro Peri, and Armon Rezai. Selfish incentives for climate policy: Empower the young! Journal of the Association of Environmental and Resource Economists, 11 0 (5): 0 1165--1200, 2024. doi:10.1086/728740. URL https://doi.org/10.1086/728740

    2024 doi

  22. [30]

    Estimating nonlinear heterogeneous agents models with neural networks

    Hanno Kase, Leonardo Melosi, and Matthias Rottner. Estimating nonlinear heterogeneous agents models with neural networks. Centre for Economic Policy Research, 2022

    2022

  23. [31]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  24. [32]

    Making carbon taxation a generational win win

    Laurence Kotlikoff, Felix Kubler, Andrey Polbin, Jeffrey Sachs, and Simon Scheidegger. Making carbon taxation a generational win win. International Economic Review, 62 0 (1): 0 3--46, 2021 a

    2021

  25. [33]

    Pareto-Improving Carbon-Risk Taxation

    Laurence Kotlikoff, Felix Kubler, Andrey Polbin, and Simon Scheidegger. Pareto-Improving Carbon-Risk Taxation . Economic Policy, 02 2021 b . ISSN 0266-4658. doi:10.1093/epolic/eiab008. URL https://doi.org/10.1093/epolic/eiab008. eiab008

    2021 doi

  26. [34]

    Can today's and tomorrow's world uniformly gain from carbon taxation? European Economic Review, 168: 0 104819, September 2024

    Laurence Kotlikoff, Felix Kubler, Andrey Polbin, and Simon Scheidegger. Can today's and tomorrow's world uniformly gain from carbon taxation? European Economic Review, 168: 0 104819, September 2024. ISSN 00142921. doi:10.1016/j.euroecorev.2024.104819

    2024

  27. [35]

    Pareto-improving social security reform when financial markets are incomplete!? American Economic Review, 96 0 (3): 0 737--755, 2006

    Dirk Krueger and Felix Kubler. Pareto-improving social security reform when financial markets are incomplete!? American Economic Review, 96 0 (3): 0 737--755, 2006

    2006

  28. [36]

    Tipping elements in the earth's climate system

    Timothy M Lenton, Hermann Held, Elmar Kriegler, Jim W Hall, Wolfgang Lucht, Stefan Rahmstorf, and Hans Joachim Schellnhuber. Tipping elements in the earth's climate system. Proceedings of the national Academy of Sciences, 105 0 (6): 0 1786--1793, 2008

    2008

  29. [37]

    The proportionality of global warming to cumulative carbon emissions

    H Damon Matthews, Nathan P Gillett, Peter A Stott, and Kirsten Zickfeld. The proportionality of global warming to cumulative carbon emissions. Nature, 459 0 (7248): 0 829--832, 2009

    2009

  30. [38]

    Kevin P. Murphy. Probabilistic Machine Learning: Advanced Topics. MIT Press, 2022. URL probml.ai

    2022

  31. [39]

    Nordhaus

    William D. Nordhaus. Revisiting the social cost of carbon. Proceedings of the National Academy of Sciences, 114 0 (7): 0 1518 LP -- 1523, February 2017. doi:10.1073/pnas.1609244114

    2017 doi

  32. [40]

    Optimal Monetary Policy with Heterogeneous Agents , 2020

    Galo Nu \ n o and Carlos Thomas. Optimal Monetary Policy with Heterogeneous Agents , 2020

    2020

  33. [41]

    Monetary policy with persistent supply shocks

    Galo Nu \ n o, Philipp Renner, and Simon Scheidegger. Monetary policy with persistent supply shocks. 2024

    2024

  34. [42]

    Carl Edward Rasmussen and Christopher K. I. Williams. Gaussian Processes for Machine Learning (Adaptive Computation and Machine Learning). The MIT Press, 2005. ISBN 026218253X

    2005

  35. [43]

    Machine learning for dynamic incentive problems

    Philipp Renner and Simon Scheidegger. Machine learning for dynamic incentive problems. Working paper. Available at SSRN: http://dx.doi.org/10.2139/ssrn.3282487, 2018

    2018 doi

  36. [44]

    Scheidegger, D

    S. Scheidegger, D. Mikushin, F. Kubler, and O. Schenk. Rethinking large-scale economic modeling for efficiency: Optimizations for gpu and xeon phi clusters. In 2018 IEEE International Parallel and Distributed Processing Symposium (IPDPS), pages 610--619, May 2018. doi:10.1109/...

    2018

  37. [45]

    Machine learning for high-dimensional dynamic stochastic economies

    Simon Scheidegger and Ilias Bilionis. Machine learning for high-dimensional dynamic stochastic economies. Journal of Computational Science, 33: 0 68 -- 82, 2019. ISSN 1877-7503. doi:https://doi.org/10.1016/j.jocs.2019.03.004. URL http://www.sciencedirect.com/science/article/pi...

    2019 doi

  38. [46]

    Christian P. Traeger. A 4- Stated DICE : Quantitatively Addressing Uncertainty Effects in Climate Change . Environmental and Resource Economics, 59 0 (1): 0 1--37, September 2014. ISSN 0924-6460. doi:10.1007/s10640-014-9776-x

    2014 doi

  39. [47]

    On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming

    Andreas W \"a chter and Lorenz T Biegler. On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming. Mathematical programming, 106: 0 25--57, 2006

    2006

  40. [48]

    Ghg targets as insurance against catastrophic climate damages

    Martin L Weitzman. Ghg targets as insurance against catastrophic climate damages. Journal of Public Economic Theory, 14 0 (2): 0 221--244, 2012

    2012

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.