This paper studies, theoretically and experimentally, the effects of overconfidence and fake news on information aggregation and the quality of democratic choice in a common interest setting. We theoretically show that overconfidence exacerbates the adverse effects of widespread misinformation (i.e., fake news). We study extensions that allow for partisan biases, targeted misinformation intended to move public opinion in a specific direction, and correlated news signals (due to media ownership concentration or censure). In our experiment, voters are exposed to correct news or misinformation depending on their cognitive ability. Absent overconfidence, more cognitively able subjects are predicted to vote while less able subjects are predicted to abstain, and information is predicted to aggregate well. We provide evidence that overconfidence induces misinformed subjects to vote excessively, thereby severely undermining information aggregation.