When people receive information about the world around them, the information often evokes both what they currently believe and what they are motivated to believe. This paper theoretically and experimentally explores motivated reasoning, the impact of motivation on inference. I analyze a model of motivated reasoning in which people distort their updating process towards particular beliefs that they are more motivated to hold. To test predictions of the model, I create a new experimental design in which people make inferences about the veracity of customized news sources. These sources are tailored to subjects' current beliefs such that there is nothing to infer from the signal sent, but motivated reasoning would lead to directional distortions. Using a large online experiment, I find evidence for politically-driven motivated reasoning on eight topics: immigration, upward mobility, racial discrimination, crime, gender-based math ability, climate change, gun laws, and beliefs about other subjects. There is evidence for ego-driven motivated reasoning on beliefs about own performance, but only for men. I also find that motivated reasoning leads people to become more polarized, less accurate, and overconfident in their beliefs.