Environmental policies in competitive electricity markets

In this thesis we model and analyze several environmental policies in an existing mathematical representation of a perfectly competitive electricity market. We contribute to the literature by theoretically and numerically establishing a number of effects of environmental policies on investment strategies and prices. We provide a theoretical benchmark for environmental regulators aiming to achieve certain policy goals, and present a way to use numerical tools in case a complete theoretical analysis cannot be obtained. Two policies that charge firms for their carbon emissions, namely cap-and-trade and carbon taxation, are modeled into both a stylized deterministic and a two-stage stochastic framework. In the former we characterize equilibria, leading to key results on the dispatching order of technologies and identification of unused technologies. The latter framework is analyzed through a sampling study and focuses on the effectiveness of the policies in the presence of network limitations. We successively study a renewable energy obligation, which indirectly subsidizes electricity production from renewable resources through green certificates. We additionally explore the effects of technology banding, meaning that different renewable technologies are eligible for a different number of certificates. To account for some of the drawbacks of the existing UK technology banding system, we introduce an alternative banding policy. Finally, a feed-in tariff (FIT) is a direct subsidy on electricity production from renewable resources. In a stochastic framework we derive analytically that under linear cost assumptions, this price based instrument cannot guarantee that quantity based policy targets are met. Assuming non-linear convex cost, we find that the opposite holds and that a regulator has the freedom to set FITs in such a way that any desired mixture of renewable technologies can be attained at equilibrium. These FITs are derived analytically or, when necessary, estimated using the numerical tools that we propose.