Pollution abatement costs are central to understanding the welfare impacts of many environmental policies. In this paper, we take a revealed preference approach to estimating marginal abatement costs. For policies that use storable permits to achieve environmental goals, banking behavior reveals information about program participants’ expectations of future compliance costs. If firms believe that compliance will be more costly in the future, they will bank permits today to prepare for those future costs. Using dynamic models of the Acid Rain Program permit market and the Corporate Average Fuel Economy (CAFE) compliance credit program, we show how the data on program stringency and the size of the permit bank can be used to identify expectations of future marginal abatement costs. Application of our model to those programs brings novel results in both settings. In the Acid Rain Program, we find that firms expected relatively high abatement costs in the later phase of the program. For passenger cars CAFE we find that the compliance costs were low but rising quickly, providing a strong incentive to bank permits at the onset of the program.