4

Climate policy claims a reflexion about public decisions under uncertainty. This work focuses on the case where a risk of catastrophe of unknown range is anticipated for pollution levels greater than a certain critical threshold. If constrained by the initial information, the best decision is to accept this risk when relevant expected welfare exceeds the certain welfare of preservation. In reality, the present best decision also bears with future information. The optimal strategy must take advantage of it. The question is then to know whether this incitates to pollute more or less today. Our analysis indicates that the answer depends upon the process of information arrival, and, therefore, upon the social planner ability to influence it. If the signals are emitted by the environment, the social planner can shorten the time until the arrival of future information, simply by accelerating the rate of accumulation of greenhouse gases. In such a case, learning incitates to pollute more. If the arrival of information follows an autonomous process, the social planner loses his ability to influence this delay and decides only the level of pollution when the information comes. In this case, it is impossible to determine whether or not an irreversibility effect exists. Classification JEL: Q28