Project-based mechanisms for emissions reductions credits, like the Clean Development Mechanism, pose important challenges for policy design because of several inherent characteristics. Participation is voluntary, so it will not occur without sufficient credits. Evaluating reductions requires assigning an emissions baseline for a counterfactual that cannot be measured. Some investments have both economic and environmental benefits and might occur anyway. Uncertainty surrounds both emissions and investment returns, and parties to the project are likely to have more information than the certifying authority. The certifying agent is limited in its ability to design a contract that would reveal investment intentions. As a result, rules for benchmarking emissions may be systematically biased to overallocate, and they also risk creating inefficient investment incentives. This paper evaluates, in a situation with asymmetric information, the efficacy of the main baseline rules currently under consideration: historical emissions, an average industry emissions standard, and expected emissions. (c) 2004 Elsevier Ltd. All rights reserved.