This paper considers the implications of an important cognitive bias in information processing, confirmation bias, in a political agency setting. In the baseline two-period case where only the politician’s actions are observable before the election, we show that when voters have this bias, it decreases pandering by the incumbent, and can raise voter welfare as a consequence. This result is robust in several directions, including to the case where the voter can also observe payoffs with some probability before the election (Maskin and Tirole’s “feedback” case). In the three-period case, with two elections, the dynamic evolution of confirmation bias can lead to more pandering before the first election. Finally, we show that when confirmation bias is present, other things equal, the case for decision-making by an elected rather than an appointed official is greater.