Many integrated energy–economy–climate models have been developed to address climate change policy. While these models are quite varied in scope, most share a common core of economic optimization and equilibrium assumptions. By contrast, system dynamics models of energy–economy interactions focus on disequilibrium dynamics, with behavioral decision rules and explicit stocks and flows of capital, labor, resources and money. This article tests climate policies using a system dynamics model that includes many features missing from economic models. Among these are endogenous technological change and boundedly rational decision making. Energy requirements are embodied in capital, and energy production capacity depends on explicit capital stocks. The search for optimal policies is decoupled from other decisions, and uses criteria that are fair across generations. Earlier experiments with the model, briefly reported here, indicate that these features greatly alter policy outcomes. The model is used to test a family of emissions permit and tax policies like the Kyoto Protocol under a range of assumptions. Uncertainty is included in the analysis through Monte Carlo simulation. Results suggest that nearly all policy options are a net benefit, and that the landscape for policy choice is more forgiving than generally supposed. However, implementation remains a critical issue, and the viability of tradeable permits is questionable. Carbon tax policies are found to outperform fixed emissions permits in nearly all circumstances. The model used in this article is available at http://www.sd3.info. Copyright © 2002 John Wiley & Sons, Ltd.