In our discussion to this point, we have assumed that there is perfect and symmetric information among all agents involved in the design and execution of environmental policy. This has allowed us to refer to damage and cost functions that are fully observed by the regulator, firms, and households alike. Among other things, the full information assumption helped us establish the equivalency of auctioned permits and emission taxes, and more generally suggested that a particular pollution reduction goal can be obtained using one of several economic incentive instruments – all of which have similar efficiency properties. Though it is pedagogically useful, a perfect information assumption is a clear departure from reality. In this chapter we begin to consider how imperfect information may alter the effectiveness and efficiency properties of the policy instruments.

Imperfect information plays a large role in environmental economics, and as reviewed by Pindyck (2007), it comes in many guises. In this chapter we focus on a particular type of uncertainty: the regulator's inability to fully observe aggregate pollution damage and abatement cost functions. Importantly, we assume for this analysis that firms know their own abatement cost functions with certainty. Once we acknowledge that these functions must be estimated by the regulator (inevitably with error), it is natural to ask about the extent to which estimation error influences the performance of different regulatory approaches. Casual intuition suggests estimation errors will result in imprecise policy, but will not cause us to systematically favor one approach over another. Investigation of the extent to which this intuition holds began with Weitzman's (1974) classic paper, and has continued in various forms to this day. The unifying theme in this large literature is a search for policy instruments that minimize the ex post inefficiencies that occur due to the regulator's uncertainty. A related literature beginning with Kwerel (1977) examines ways that the regulator can design mechanisms that resolve the uncertainty, thereby eliminating ex post inefficiencies. In this chapter we examine these two strands of literature.