Two office-motivated politicians compete in an election by proposing policies. There are two possible states of the world: climate change is either mild, causing no lasting effect on welfare if the appropriate policy is implemented, or climate change is severe, leading to decreased welfare even if appropriate measures are taken. Voters receive signals about the state of the world but may choose to hold motivated beliefs, interpreting these signals in a non-Bayesian manner to increase anticipatory utility. If voters a priori believe that severe climate change is unlikely, an equilibrium always exists where they ignore any signal suggesting severe consequences, thereby incentivizing politicians to also disregard their information and offer policies only for mild climate change---even when the politicians are certain of the contrary. If severe climate change results in catastrophic welfare losses, this equilibrium is also unique. However, if severe climate change only leads to moderate welfare losses, an efficient equilibrium exists where politicians choose the optimal policies based on the actual state of the world. In this case, voters' beliefs about the implemented policies become self-fulfilling: if voters trust that politicians will choose optimal policies, this indeed happens in equilibrium. Conversely, if voters expect politicians to ignore information and adhere to a popular policy platform, politicians will follow suit and deliver precisely those policy platforms. I discuss the model's implications for political rhetoric and the importance of trust in government. A first look at the data supports the model's predictions.