| This dissertation investigates the optimal design and efficiency of different environmental policies such as taxes, quotas, and tradable permits in regulating a stock externality under alternative situations, including uncertainty, asymmetric information, endogenous learning, technical change, and rational expectations.; The first essay studies a dynamic regulation model where firms' actions contribute to a stock externality, considering both open-loop and feedback policies. The regulator and firms have asymmetric information about serially correlated abatement costs. If the regulator uses price-based policies such as taxes, or if firms trade quotas efficiently, the regulator learns the evolution of both stock and costs. This ability to learn is important in determining the ranking of taxes and quotas, and in determining the value of a feedback rather than an open-loop policy.; The second essay considers endogenous technical change. With expectations on the regulator's future policies, a polluting firm chooses its technical improvement by investing in abatement capital. The regulator learns about a firm's private cost schedule by observing its emission response and/or investment response. Even though the representative firm does not internalize the environmental damage in its cost minimization, higher pollutant stock induces higher investment because these policies affect the shadow value of abatement capital. By calibrating the model with climate change studies, I assess different policies for controlling greenhouse gasses. The results suggest that the stock effect in the damage side favors the use of quotas, and the stock effect in the cost side favors the use of taxes.; The third essay investigates how the uncertainty about environmental damages as well as abatement costs, together with anticipated endogenous learning in both damages and costs, influences optimal regulation of the stock externality and the ranking of policies. The regulator learns actively and updates his belief about damages using Bayesian inference. I calibrate the model with climate change studies and solve the optimal control problem numerically with the embedded stochasticity and endogenous learning. The solution approximates the value function with a flexible function form using neural networks. |
