Active learning and optimal climate policy

This paper develops a climate-economy model with uncertainty, irreversibility and active learning. Whereas previous papers assume passive learning from one observation per period, or experiment with control variables to gain additional information, this paper considers active learning from research investment in improved observations. We restrict ourselves to improving observations of the global mean temperature. We find that the decision maker invests a significant amount of money in climate research, far more than the current level, in order to increase the rate of learning about climate change. This helps the decision maker take improved decisions. The level of uncertainty decreases more rapidly in the active learning model with research investment than in the passive learning model only with temperature observations. As a result, active learning reduces the optimal carbon tax. The greater the risk, the larger is the effect of learning. The method proposed here is applicable to any dynamic control problem where the quality of monitoring is a choice variable.