Most collective dilemmas-that is, situations in which private interests contrast with collective interests-have an embedded intertemporal component in that they often imply that the rewards from defection are immediate but the rewards from cooperation are delayed and often accrue to people in the future. This also applies to carbon taxes since they imply additional individual costs for benefits which will mostly be enjoyed by future generations, which undermines their political support. In an experiment on a representative sample of 1000 United States adults, we presented individuals with twelve alternative carbon tax formulations with varying start dates, temporal horizons of carbon abatement objectives and revenue uses. We find that public support is highest when individuals can pre-commit to policies that start a few years into the future and for policies that express their emission cuts objectives in more distant and ambitious terms -i.e. achieving carbon neutrality by 2050 as opposed to halving emissions by 2030. Individual temporal discounting, exogenously measured, account for large part of these preferences. These preferences are in contrast with the most efficient policy, which is the one that starts immediately and distributes equitably mitigation costs across time. We find two ways to realign preferences with it. First, the most efficient policy becomes politically feasible when the tax includes a dividend that is redistributed to citizens. Delivering an economic compensation at the same time of the individual costs of the tax neutralizes the effect of individual discounting and of the policy's temporal context on tax support. Secondly, when the price of the carbon tax is adjusted upward to compensate for the opportunity cost of delaying its introduction, individuals start trading off its delay with avoidance of tax increases.