Bayesian learning, growth, and pollution

Abstract

Environmental problems, such as climate change, have great uncertainties. Current expectations are that uncertainties about climate change will be resolved quickly. We examine this hypothesis theoretically and computationally. We consider Bayesian learning about the relationship between greenhouse gas levels and global mean temperature changes, a key uncertainty. Learning is non-trivial because of a stochastic shock to the realized temperature. We find theoretically the expected learning time, which is related to the variance of the shock and the emissions policy, implying a tradeoff between the benefits of controlling emissions and information. We imbed the learning model into an optimal growth model with a climate sector and solve the resulting dynamic program. We find computationally that learning takes on average over 90 yr, far longer than currently believed.