Professor Scheidegger is an Associate Professor of Economics at the University of Lausanne and a member of the Enterprise for Society (E4S). His research interests span several topics\, including macroeconomics\, computational economics\, climate change economics\, and machine learning.
Simon will be discussing the paper Using Machine Learning to compute all constrained Pareto optimal carbon tax rules in a stochastic OLG model.
Abstract
This work presents a novel computational framework to derive constrained Pareto-optimal carbon tax rules within a stochastic overlapping generations (OLG) model. By integrating deep reinforcement learning with advanced machine learning techniques\, the framework systematically identifies optimal carbon taxation policies for heterogeneous agents facing climate risks. The methodology utilizes Deep Equilibrium Nets to efficiently compute equilibrium policy functions and employs Gaussian Process regression with Bayesian active learning to approximate welfare outcomes across varying tax structures.
The economic model features a 12-period OLG framework\, where exogenous shocks impact carbon emissions\, the carbon-temperature relationship\, and climate-related damages\, leading to asymmetric risks such as potential climate disasters. The pursuit of constrained Pareto efficiency involves designing tax and revenue-sharing rules that improve welfare despite inherent market inefficiencies. Surrogate modelling techniques enable construction of the full constrained Pareto frontier in a single evaluation\, revealing nontrivial trade-offs between intergenerational welfare and carbon taxation.
The findings demonstrate that optimal carbon taxes can substantially mitigate climate risk and enhance welfare across generations\, offering scalable\, precise tools for macroeconomic policy design in complex stochastic environments.
We believe that this approach offers a powerful tool for analysing complex economic problems involving heterogeneous agents and could contribute to more informed policymaking in various domains.
The Research Seminar Series is open to all LSE researchers. No pre-registration required to attend the seminars. If you wish to attend the seminars or want to be kept informed about upcoming seminars\, please email Gri.Events@lse.ac.uk.
URL;VALUE=URI:https://www.lse.ac.uk/granthaminstitute/events/simon-scheidegger-research-seminar-series/ SUMMARY:Using Machine Learning to compute all constrained Pareto optimal carbon tax rules in a stochastic OLG model | Simon Scheidegger DTSTART;TZID=Europe/London:20250205T140000 END:VEVENT END:VCALENDAR