Experiments on damage-based ambient taxes for nonpoint source polluters

This article presents experimental tests of a linear and a nonlinear ambient tax mechanism that involve modest information requirements for the regulator. When agents are not allowed to communicate, both tax mechanisms result in emission levels that approximate the social optimum. When agents can communicate, emissions are considerably below the optimum, but we show that the tax function can be scaled to achieve social efficiency. Finally, by disaggregating the overall efficiency measure, we show that changing the pollution threshold that triggers the tax increases the inefficiency resulting from variation in agent-level decisions, but does not affect average emissions.