Despite its status with economists, carbon taxes rarely win the public support afforded to green subsidies. While political and motivational factors are often blamed, we find evidence of a cognitive explanation. We isolate a novel mechanism: underweighting of policy effects attributable to others' behavior. In addition, we find significant underestimation of resposiveness to prices and a failure of many subjects to arrive at simple, useful heuristics for evaluating Pigouvian policies. These errors in analysis are measured in a carefully designed lab experiment, removed of scope for confounds like time and social preferences, tax and subsidy framing, uncertainty, and environmental context, as well as separately using a survey on climate policy attitudes. The lab and survey findings shed light on puzzling biases in the climate political economy literature, such as pessimism about carbon pricing's impact on pollution and on inequality. We modify the classic carbon fee-and-dividend policy to create a financially equivalent proposal that neatly circumvents many important cognitive obstacles. We measure stronger support toward the modified policy than the classic framing among a representative U.S. sample.