Optimally differentiated carbon prices for unilateral climate policy

- A typical CGE model (WorldScan) can produce a pattern of highly differentiated optimal carbon prices.
- The range of this pattern is from almost prohibitive taxes to high subsidies.
- The welfare gain from switching from a uniform price to optimally differentiated prices is equivalent to a 27% emission reduction for free.
- The most important drivers of carbon price differentiation are market power in export markets and pre-existing taxes.

Economic thought on climate policy as an instance of environmental regulation is strongly influenced by the principle of a uniform carbon price. Economists acknowledge that this principle breaks down in a “second-best” world with other distortions, such as taxes and market power in domestic and international markets. However, systematic analysis of this point in the economic climate policy literature is scarce. In the present paper, a computable general equilibrium (CGE) set-up is chosen in order to examine what pattern of differentiated carbon prices emerges as optimal in a second-best world.The CGE model WorldScan, which is considered to be representative of the class of models routinely used for numerical climate policy analysis, produces three main results: First, the optimal pattern of carbon prices is highly differentiated, ranging from almost prohibitive taxes to high subsidies (with a range of more than 1700 euros per ton of CO2). Second, the welfare gain from switching from a uniform price to optimally differentiated prices is enormous, equivalent to a 27% emission reduction for free. Third, the most important drivers of carbon price differentiation are market power in export markets as well as taxes on consumption, intermediate inputs and domestic output. This shows that carbon price differentiation cannot be dismissed as a policy option lightly. However, before translating these findings into concrete policy advice, the relevant features of modelling pre-existing distortions in CGE models need close revision.